National Semiconductor DP83816 Technical data

September 2005
DP83816 10/100 Mb/s Integrated PCI Ethernet Media Access Controller and Physical Layer (MacPHYTER-II
)
DP83816 10/100 Mb/s Integrated PCI Ethernet Media Access Controller and Physical Layer (MacPHYTER-II™)
General Description
The DP83816 device is an integration of an enhanced version of the National Semiconductor PCI MAC/BIU (Media Access Controller/Bus Interface Unit) and a 3.3V CMOS physical layer interface.
Features
— IEEE 802.3 Compliant, PCI V2.2 MAC/BIU supports
traditional data rates of 10 Mb/s Ethernet and 100 Mb/s
Fast Ethernet (via internal phy) — Bus master - burst sizes of up to 128 dwords (512 bytes) — BIU compliant with PC 97 and PC 98 Hardware Design
Guides, PC 99 Hardware Design Guide draft, ACPI v1.0,
PCI Power Management Specification v1.1, OnNow
Device Class Power Management Reference
Specification - Network Device Class v1.0a — Wake on LAN (WOL) support compliant with PC98,
PC99, SecureOn, and OnNow, including directed
packets, Magic Packet, VLAN packets, ARP packets,
pattern match packets, and Phy status change — Clkrun function for PCI Mobile Design Guide — Virtual LAN (VLAN) and long frame support
— Support for IEEE 802.3x Full duplex flow control — Extremely flexible Rx packet filtration including: single
address perfect filter with MSb masking, broadcast, 512 entry multicast/unicast hash table, deep packet pattern matching for up to 4 unique patterns
— Statistics gathered for support of RFC 1213 (MIB II),
RFC 1398 (Ether-like MIB), IEEE 802.3 LME, reducing
CPU overhead for management — Internal 2 KB Transmit and 2 KB Receive data FIFOs — Serial EEPROM port with auto-load of configuration data
from EEPROM at power-on — Flash/PROM interface for remote boot support — Fully integrated IEEE 802.3/802.3u 3.3V CMOS physical
layer — IEEE 802.3 10BASE-T transceiver with integrated filters — IEEE 802.3u 100BASE-TX transceiver — Fully integrated ANSI X3.263 compliant TP-PMD
physical sublayer with adaptive equalization and
Baseline Wander compensation — IEEE 802.3u Auto-Negotiation - advertised features
configurable via EEPROM — Full Duplex support for 10 and 100 Mb/s data rates — Single 25 MHz reference clock — 144-pin LQFP package — Low power 3.3V CMOS design with typical consumption
of 383 mW operating, 297 mW during WOL and 53 mW
during sleep mode — IEEE 802.3u MII for connecting alternative external
Physical Layer Devices — 3.3V signalling with 5V tolerant I/O.
System Diagram
PCI Bus
10/100 Twisted Pair
DP83816
BIOS ROM
(optional)
MacPHYTER-II is a trademark of National Semiconductor Corporation. Magic Packet is a trademark of Advanced Micro Devices, Inc.
© 2005 National Semiconductor Corporation www.national.com
EEPROM (optional)
Isolation
Table of Contents
1.0 Connection Diagram . . . . . . . . . . . . . . . . . . 4
1.1 144 LQFP Package (VNG) . . . . . . . . . . . . 4
2.0 Pin Description . . . . . . . . . . . . . . . . . . . . . . 5
3.0 Functional Description . . . . . . . . . . . . . . . 11
3.1 MAC/BIU . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1.1 PCI Bus Interface . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1.2 Tx MAC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.3 Rx MAC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 Buffer Management . . . . . . . . . . . . . . . . . 13
3.2.1 Tx Buffer Manager . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2.2 Rx Buffer Manager . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2.3 Packet Recognition . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2.4 MIB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Interface Definitions . . . . . . . . . . . . . . . . . 14
3.3.1 PCI System Bus . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3.2 Boot PROM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3.3 EEPROM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3.4 Clock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4 Physical Layer . . . . . . . . . . . . . . . . . . . . . 16
3.4.1 Auto-Negotiation . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.4.2 Auto-Negotiation Register Control . . . . . . . . . . . . . 16
3.4.3 Auto-Negotiation Parallel Detection . . . . . . . . . . . . 16
3.4.4 Auto-Negotiation Restart . . . . . . . . . . . . . . . . . . . . 17
3.4.5 Enabling Auto-Negotiation via Software . . . . . . . . 17
3.4.6 Auto-Negotiation Complete Time . . . . . . . . . . . . . . 17
3.5 LED Interfaces . . . . . . . . . . . . . . . . . . . . . 17
3.6 Half Duplex vs. Full Duplex . . . . . . . . . . . 18
3.7 Phy Loopback . . . . . . . . . . . . . . . . . . . . . 18
3.8 Status Information . . . . . . . . . . . . . . . . . . 18
3.9 100BASE-TX TRANSMITTER . . . . . . . . . 18
3.9.1 Code-group Encoding and Injection . . . . . . . . . . . 19
3.9.2 Scrambler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.9.3 NRZ to NRZI Encoder . . . . . . . . . . . . . . . . . . . . . . 20
3.9.4 Binary to MLT-3 Convertor / Common Driver . . . . 20
3.10 100BASE-TX Receiver . . . . . . . . . . . . . . 21
3.10.1 Input and Base Line Wander Compensation . . . . 21
3.10.2 Signal Detect . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.10.3 Digital Adaptive Equalization . . . . . . . . . . . . . . . . 23
3.10.4 Line Quality Monitor . . . . . . . . . . . . . . . . . . . . . . . 24
3.10.5 MLT-3 to NRZI Decoder . . . . . . . . . . . . . . . . . . . . 24
3.10.6 Clock Recovery Module . . . . . . . . . . . . . . . . . . . . 25
3.10.7 NRZI to NRZ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.10.8 Serial to Parallel . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.10.9 De-scrambler . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.10.10 Code-group Alignment . . . . . . . . . . . . . . . . . . . . 25
3.10.11 4B/5B Decoder . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.10.12 100BASE-TX Link Integrity Monitor . . . . . . . . . . 25
3.10.13 Bad SSD Detection . . . . . . . . . . . . . . . . . . . . . . 25
3.11 10BASE-T Transceiver Module . . . . . . . . 26
3.11.1 Operational Modes . . . . . . . . . . . . . . . . . . . . . . . . 26
3.11.2 Smart Squelch . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.11.3 Collision Detection . . . . . . . . . . . . . . . . . . . . . . . . 26
3.11.4 Normal Link Pulse Detection/Generation . . . . . . .26
3.11.5 Jabber Function . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.11.6 Automatic Link Polarity Detection . . . . . . . . . . . . . 27
3.11.7 10BASE-T Internal Loopback . . . . . . . . . . . . . . . . 27
3.11.8 Transmit and Receive Filtering . . . . . . . . . . . . . . . 27
3.11.9 Transmitter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.11.10 Receiver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.11.11 Far End Fault Indication . . . . . . . . . . . . . . . . . . . 27
3.12 802.3u MII . . . . . . . . . . . . . . . . . . . . . . . . 27
3.12.1 MII Access Configuration . . . . . . . . . . . . . . . . . . . 27
3.12.2 MII Serial Management . . . . . . . . . . . . . . . . . . . . 27
DP83816
3.12.3 MII Serial Management Access . . . . . . . . . . . . . 28
3.12.4 Serial Management Access Protocol . . . . . . . . . 28
3.12.5 Nibble-wide MII Data Interface . . . . . . . . . . . . . . 28
3.12.6 Collision Detection . . . . . . . . . . . . . . . . . . . . . . . 29
3.12.7 Carrier Sense . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.0 Register Set . . . . . . . . . . . . . . . . . . . . . . . . 30
4.1 Configuration Registers . . . . . . . . . . . . . . 30
4.1.1 Configuration Identification Register . . . . . . . . . . . 30
4.1.2 Configuration Command and Status Register . . . 31
4.1.3 Configuration Revision ID Register . . . . . . . . . . . 32
4.1.4 Configuration Latency Timer Register . . . . . . . . . 33
4.1.5 Configuration I/O Base Address Register . . . . . . . 33
4.1.6 Configuration Memory Address Register . . . . . . . 34
4.1.7 Configuration Subsystem Identification Register . 34
4.1.8 Boot ROM Configuration Register . . . . . . . . . . . . 35
4.1.9 Capabilities Pointer Register . . . . . . . . . . . . . . . . 35
4.1.10 Configuration Interrupt Select Register . . . . . . . . 36
4.1.11 Power Management Capabilities Register . . . . . 36
4.1.12 Power Management Control and Status Register 37
4.2 Operational Registers . . . . . . . . . . . . . . . 38
4.2.1 Command Register . . . . . . . . . . . . . . . . . . . . . . . . 39
4.2.2 Configuration and Media Status Register . . . . . . . 40
4.2.3 EEPROM Access Register . . . . . . . . . . . . . . . . . . 42
4.2.4 EEPROM Map . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.2.5 PCI Test Control Register . . . . . . . . . . . . . . . . . . . 43
4.2.6 Interrupt Status Register . . . . . . . . . . . . . . . . . . . . 44
4.2.7 Interrupt Mask Register . . . . . . . . . . . . . . . . . . . . 45
4.2.8 Interrupt Enable Register . . . . . . . . . . . . . . . . . . . 47
4.2.9 Interrupt Holdoff Register . . . . . . . . . . . . . . . . . . . 47
4.2.10 Transmit Descriptor Pointer Register . . . . . . . . . 48
4.2.11 Transmit Configuration Register . . . . . . . . . . . . . 48
4.2.12 Receive Descriptor Pointer Register . . . . . . . . . . 50
4.2.13 Receive Configuration Register . . . . . . . . . . . . . 51
4.2.14 CLKRUN Control/Status Register . . . . . . . . . . . . 52
4.2.15 Wake Command/Status Register . . . . . . . . . . . . 54
4.2.16 Pause Control/Status Register . . . . . . . . . . . . . . 56
4.2.17 Receive Filter/Match Control Register . . . . . . . . 57
4.2.18 Receive Filter/Match Data Register . . . . . . . . . . 58
4.2.19 Receive Filter Logic . . . . . . . . . . . . . . . . . . . . . . 59
4.2.20 Boot ROM Address Register . . . . . . . . . . . . . . . . 63
4.2.21 Boot ROM Data Register . . . . . . . . . . . . . . . . . . 63
4.2.22 Silicon Revision Register . . . . . . . . . . . . . . . . . . 63
4.2.23 Management Information Base Control Register 64
4.2.24 Management Information Base Registers . . . . . . 65
4.3 Internal PHY Registers . . . . . . . . . . . . . . . 66
4.3.1 Basic Mode Control Register . . . . . . . . . . . . . . . . 66
4.3.2 Basic Mode Status Register . . . . . . . . . . . . . . . . . 67
4.3.3 PHY Identifier Register #1 . . . . . . . . . . . . . . . . . . 68
4.3.4 PHY Identifier Register #2 . . . . . . . . . . . . . . . . . . 68
4.3.5 Auto-Negotiation Advertisement Register . . . . . . 68
4.3.6 Auto-Negotiation Link Partner Ability Register . . . 69
4.3.7 Auto-Negotiate Expansion Register . . . . . . . . . . . 70
4.3.8 Auto-Negotiation Next Page Transmit Register . . 70
4.3.9 PHY Status Register . . . . . . . . . . . . . . . . . . . . . . . 71
4.3.10 MII Interrupt Control Register . . . . . . . . . . . . . . . 73
4.3.11 MII Interrupt Status and Misc. Control Register . 73
4.3.12 False Carrier Sense Counter Register . . . . . . . . 74
4.3.13 Receiver Error Counter Register . . . . . . . . . . . . . 74
4.3.14 100 Mb/s PCS Configuration and Status Register 74
4.3.15 PHY Control Register . . . . . . . . . . . . . . . . . . . . . 75
4.3.16 10BASE-T Status/Control Register . . . . . . . . . . . 76
5.0 Buffer Management . . . . . . . . . . . . . . . . . . 77
5.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.1.1 Descriptor Format . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.1.2 Single Descriptor Packets . . . . . . . . . . . . . . . . . . 79
2 www.national.com
5.1.3 Multiple Descriptor Packets . . . . . . . . . . . . . . . . . . 80
5.1.4 Descriptor Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.2 Transmit Architecture . . . . . . . . . . . . . . . 81
5.2.1 Transmit State Machine . . . . . . . . . . . . . . . . . . . . . 81
5.2.2 Transmit Data Flow . . . . . . . . . . . . . . . . . . . . . . . . 83
5.3 Receive Architecture . . . . . . . . . . . . . . . . 84
5.3.1 Receive State Machine . . . . . . . . . . . . . . . . . . . . . 84
5.3.2 Receive Data Flow . . . . . . . . . . . . . . . . . . . . . . . . . 86
6.0 Power Management and Wake-On-LAN. . 87
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 87
6.2 Definitions (for this document only) . . . . . 87
6.3 Packet Filtering . . . . . . . . . . . . . . . . . . . . 87
6.4 Power Management . . . . . . . . . . . . . . . . 87
6.4.1 D0 State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.4.2 D1 State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.4.3 D2 State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.4.4 D3hot State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.4.5 D3cold State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.5 Wake-On-LAN (WOL) Mode . . . . . . . . . . 88
6.5.1 Entering WOL Mode . . . . . . . . . . . . . . . . . . . . . . . 88
6.5.2 Wake Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.5.3 Exiting WOL Mode . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.6 Sleep Mode . . . . . . . . . . . . . . . . . . . . . . . 89
6.6.1 Entering Sleep Mode . . . . . . . . . . . . . . . . . . . . . . 89
6.6.2 Exiting Sleep Mode . . . . . . . . . . . . . . . . . . . . . . . . 89
6.7 Pin Configuration for Power Management 89
7.0 DC and AC Specifications . . . . . . . . . . . . . 90
7.1 DC Specifications . . . . . . . . . . . . . . . . . . . 90
7.2 AC Specifications . . . . . . . . . . . . . . . . . . . 91
7.2.1 PCI Clock Timing . . . . . . . . . . . . . . . . . . . . . . . . . 91
7.2.2 X1 Clock Timing . . . . . . . . . . . . . . . . . . . . . . . . . . 91
7.2.3 Power On Reset (PCI Active) . . . . . . . . . . . . . . . . 92
7.2.4 Non Power On Reset . . . . . . . . . . . . . . . . . . . . . . 92
7.2.5 POR PCI Inactive . . . . . . . . . . . . . . . . . . . . . . . . . 93
7.2.6 PCI Bus Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
7.2.7 EEPROM Auto-Load . . . . . . . . . . . . . . . . . . . . . . 99
7.2.8 Boot PROM/FLASH . . . . . . . . . . . . . . . . . . . . . . 100
7.2.9 100BASE-TX Transmit . . . . . . . . . . . . . . . . . . . 101
7.2.10 10BASE-T Transmit End of Packet . . . . . . . . . 102
7.2.11 10 Mb/s Jabber Timing . . . . . . . . . . . . . . . . . . 102
7.2.12 10BASE-T Normal Link Pulse . . . . . . . . . . . . . 103
7.2.13 Auto-Negotiation Fast Link Pulse (FLP) . . . . . . 103
7.2.14 Media Independent Interface (MII) . . . . . . . . . . 104
List of Figures
Figure 3-1 DP83816 Functional Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11
Figure 3-2 MAC/BIU Functional Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .12
Figure 3-3 Ethernet Packet Format. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .14
Figure 3-4 DSP Physical Layer Block Diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .15
Figure 3-5 LED Loading Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17
Figure 3-6 100BASE-TX Transmit Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .19
Figure 3-7 Binary to MLT-3 conversion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .20
Figure 3-8 100 M/bs Receive Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .22
Figure 3-9 100BASE-TX BLW Event Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .23
Figure 3-10 EIA/TIA Attenuation vs. Frequency for 0, 50, 100, 130 & 150 meters of CAT V cable . . . . . . . . 24
Figure 3-11 MLT-3 Signal Measured at AII after 0 meters of CAT V cable. . . . . . . . . . . . . . . . . . . . . . . . . . .24
Figure 3-12 MLT-3 Signal Measured at AII after 50 meters of CAT V cable. . . . . . . . . . . . . . . . . . . . . . . . . .24
Figure 3-13 MLT-3 Signal Measured at AII after 100 meters of CAT V cable. . . . . . . . . . . . . . . . . . . . . . . . .24
Figure 3-14 10BASE-T Twisted Pair Smart Squelch Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .26
Figure 3-15 Typical MDC/MDIO Read Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28
Figure 3-16 Typical MDC/MDIO Write Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .29
Figure 4-1 Pattern Buffer Memory - 180h words (word = 18bits) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .60
Figure 4-2 Hash Table Memory - 40h bytes addressed on word boundaries . . . . . . . . . . . . . . . . . . . . . . . .62
Figure 5-1 Single Descriptor Packets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .79
Figure 5-2 Multiple Descriptor Packets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .80
Figure 5-3 List and Ring Descriptor Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .80
Figure 5-4 Transmit Architecture. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .81
Figure 5-5 Transmit State Diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .82
Figure 5-6 Receive Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .84
Figure 5-7 Receive State Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .86
List of Tables
Table 3-1 4B5B Code-Group Encoding/Decoding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .20
Table 3-2 Typical MDIO Frame Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28
Table 4-1 Configuration Register Map. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .30
Table 4-2 Operational Register Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .38
Table 4-3 MIB Registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .65
Table 5-1 DP83816 Descriptor Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .77
Table 5-2 cmdsts Common Bit Definitions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .77
Table 5-3 Transmit Status Bit Definitions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .78
Table 5-4 Receive Status Bit Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .79
Table 5-5 Transmit State Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .82
Table 5-6 Receive State Tables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .85
Table 6-1 Power Management Modes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .87
Table 6-2 PM Pin Configuration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .89
DP83816
3 www.national.com
1.0 Connection Diagram
1.1 144 LQFP Package (VNG)
NC
VSS
AUXVDD
NC
VSS
TXEN
TXCLK
CRS
DP83816
MA3/EEDI
MA4/EECLK
VSS
TXD3/MA15
COL/MA16
AUXVDD
TXD2/MA14
TXD1/MA13
VSS
C1
TXD0/MA12
VSS
AUXVDD
X2
X1
RXDV/MA11
RXER/MA10
RXOE
RXD2/MA8
RXD3/MA9
RXD0/MA6
RXD1/MA7
VSS
AUXVDD
MDIO
MDC
RXCLK
MA5
NC
VSS
IAUXVDD
VREF
RESERVED
NC NC
VSS
TPRDM
TPRDP
IAUXVDD
REGEN
VSS
RESERVED
VSS VSS
TPTDM
TPTDP
VSS
AUXVDD
VSS
AUXVDD
PMEN/CLKRUNN
PCICLK
INTAN
RSTN GNTN REQN
VSS AD31 AD30 AD29
PCIVDD
AD28 AD27 AD26
105
106
107
123456789
108
144 143 142 141 140 139 138 137 136 135 134 133 132 131 130 129 128 127 126 125 124 123 122 121 120 119 118 117 116 115 114 113 112 111 110 109
MA2/LED100N MA1/LED10N MA0/LEDACTN MD7 MD6 MD5 MD4/EEDO AUXVDD VSS MD3 MD2 MD1/CFGDISN MD0 MWRN MRDN MCSN EESEL RESERVED NC NC NC PWRGOOD 3VAUX AD0 AD1 AD2 AD3 PCIVDD AD4 AD5 VSS AD6 AD7 CBEN0 AD8 AD9
363534
37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72
73
323329
Pin1
Identification
DP83816
99
98
97
96
95
94
93
92
91
90
89
88
87
86
85
84
83
82
81
80
79
78
77
76
75
74
100
101
102
103
104
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
30
31
AD24
AD25
VSS
IDSEL
CBEN3
AD22
AD23
PCIVDD
AD20
AD21
NC
AD19
AD18
AD16
AD17
VSS
CBEN2
FRAMEN
VSS
PAR
AD15
AD14
AD13
AD12
AD11
IRDYN
TRDYN
STOPN
PCIVDD
DEVSELN
PERRN
SERRN
CBEN1
AD10
PCIVDD
NC
For Normal Operating Temperature - Order Number DP83816AVNG
See NS Package Number VNG144A
4 www.national.com
2.0 Pin Description
PCI Bus Interface
DP83816
Symbol
AD[31-0] 66, 67, 68, 70,
CBEN[3-0] 75,
PCICLK 60 I Clock: This PCI Bus clock provides timing for all bus phases. The rising edge
DEVSELN 95 I/O Device Select: As a bus master, the DP83816 samples this signal to insure that the
FRAMEN 91 I/O Frame: As a bus master, this signal is asserted low to indicate the beginning and
LQFP Pin
No(s) Dir Description
I/O Address and Data: Multiplexed address and data bus. As a bus master, the 71, 72, 73, 74, 78, 79, 81, 82, 83, 86, 87, 88, 101, 102, 104, 105, 106, 108, 109, 110, 112,
113, 115, 116,
118, 119, 120,
121
89,
100,
111
DP83816 will drive address during the first bus phase. During subsequent phases, the DP83816 will either read or write data expecting the target to increment its address pointer. As a bus target, the DP83816 will decode each address on the bus and respond if it is the target being addressed.
I/O Bus Command/Byte Enable: During the address phase these signals define the
“bus command” or the type of bus transaction that will take place. During the data phase these pins indicate which byte lanes contain valid data. CBEN[0] applies to byte 0 (bits 7-0) and CBEN[3] applies to byte 3 (bits 31-24) in the Little Endian Mode. In Big Endian Mode, CBEN[3] applies to byte 0 (bits 31-24) and CBEN[0] applies to byte 3 (bits 7-0).
defines the start of each phase. The clock frequency ranges from 0 to 33 MHz.
destination address for the data transfer is recognized by a PCI target. As a target, the DP83816 asserts this signal low when it recognizes its address after FRAMEN is asserted.
duration of a bus transaction. Data transfer takes place when this signal is asserted. It is de-asserted before the transaction is in its final phase. As a target, the device monitors this signal before decoding the address to check if the current transaction is addressed to it.
GNTN 63 I Grant: This signal is asserted low to indicate to the DP83816 that it has been
IDSEL 76 I Initialization Device Select: This pin is sampled by the DP83816 to identify when
INTAN 61 O Interrupt A: This signal is asserted low when an interrupt condition occurs as
IRDYN 92 I/O Initiator Ready: As a bus master, this signal will be asserted low when the
PAR 99 I/O Parity: This signal indicates even parity across AD[31-0] and CBEN[3-0] including
PERRN 97 I/O Parity Error: The DP83816 as a master or target will assert this signal low to
REQN 64 O Request: The DP83816 will assert this signal low to request ownership of the bus
RSTN 62 I Reset: When this signal is asserted all PCI bus outputs of DP83816 will be in TRI-
granted ownership of the bus by the central arbiter. This input is used when the DP83816 is acting as a bus master.
configuration read and write accesses are intended for it.
defined in the Interrupt Status Register, Interrupt Mask, and Interrupt Enable registers.
DP83816 is ready to complete the current data phase transaction. This signal is used in conjunction with the TRYDN signal. Data transaction takes place at the rising edge of PCICLK when both IRDYN and TRDYN are asserted low. As a target, this signal indicates that the master has put the data on the bus.
the PAR pin. As a master, PAR is asserted during address and write data phases. As a target, PAR is asserted during read data phases.
indicate a parity error on any incoming data (except for special cycles). As a bus master, it will monitor this signal on all write operations (except for special cycles).
from the central arbiter.
®
STATE
and the device will be put into a known state.
5 www.national.com
2.0 Pin Description (Continued)
PCI Bus Interface
DP83816
Symbol
SERRN 98 I/O System Error: This signal is asserted low by DP83816 during address parity errors
STOPN 96 I/O Stop: This signal is asserted low by the target device to request the master device
TRDYN 93 I/O Target Ready: As a master, this signal indicates that the target is ready for the data
PMEN/
CLKRUNN
3VAUX 122 I PCI Auxiliary Voltage Sense: This pin is used to sense the presence of a 3.3V
LQFP Pin
No(s) Dir Description
and system errors if enabled.
to stop the current transaction.
during write operation and with the data during read operation. As a target, this signal will be asserted low when the (target) device is ready to complete the current data phase transaction. This signal is used in conjunction with the IRDYN signal. Data transaction takes place at the rising edge of PCICLK when both IRDYN and TRDYN are asserted low.
59 I/O Power Management Event/Clock Run Function: This pin is a dual function pin.
The function of this pin is determined by the CLKRUN_EN bit 0 of the CLKRUN Control and Status register (CCSR). Default operation of this pin is PMEN.
Power Management Event: This signal is asserted low by the DP83816 to indicate that a power management event has occurred. For pin connection please refer to Section 6.7.
Clock Run Function: In this mode, this pin is used to indicate when the PCICLK will be stopped.
auxiliary supply in order to define the PME Support available. For pin connection please refer to Section 6.7.
This pin has an internal weak pull down.
PWRGOOD 123 I PCI bus power good: Connected to PCI bus 3.3V power, this pin is used to sense
the presence of PCI bus power during the D3 power management state.
This pin has an internal weak pull down.
6 www.national.com
2.0 Pin Description (Continued)
Media Independent Interface (MII)
DP83816
Symbol
COL 28 I Collision Detect: The COL signal is asserted high asynchronously by the external
CRS 29 I Carrier Sense: This signal is asserted high asynchronously by the external PMD
MDC 5 O Management Data Clock: Clock signal with a maximum rate of 2.5 MHz used to
MDIO 4 I/O Management Data I/O: Bidirectional signal used to transfer management
RXCLK 6 I Receive Clock: A continuous clock, sourced by an external PMD device, that is
RXD3/MA9, RXD2/MA8, RXD1/MA7, RXD0/MA6
RXDV/MA11 15 IOReceive Data Valid: Indicates that the external PMD is presenting recovered and
LQFP Pin
No(s) Dir Description
PMD upon detection of a collision on the medium. It will remain asserted as long as the collision condition persists.
upon detection of a non-idle medium.
transfer management data for the external PMD on the MDIO pin.
information for the external PMD. (See Section 3.12.4 for details on connections when MII is used.)
recovered from the incoming data. During 100 Mb/s operation RXCLK is 25 MHz and during 10 Mb/s this is 2.5 MHz.
12, 11, 10,
7
IOReceive Data: Sourced from an external PMD, that contains data aligned on nibble
boundaries and are driven synchronous to RXCLK. RXD[3] is the most significant bit and RXD[0] is the least significant bit.
BIOS ROM Address: During external BIOS ROM access, these signals become part of the ROM address.
decoded nibbles on the RXD signals, and that RXCLK is synchronous to the recovered data in 100 Mb/s operation. This signal will encompass the frame, starting with the Start-of-Frame delimiter (JK) and excluding any End-of-Frame delimiter (TR).
BIOS ROM Address: During external BIOS ROM access, this signal becomes part of the ROM address.
RXER/MA10 14 IOReceive Error: Asserted high synchronously by the external PMD whenever it
detects a media error and RXDV is asserted in 100 Mb/s operation.
BIOS ROM Address: During external BIOS ROM access, this signal becomes part of the ROM address.
RXOE 13 O Receive Output Enable: Used to disable an external PMD while the BIOS ROM is
being accessed.
TXCLK 31 I Transmit Clock: A continuous clock that is sourced by the external PMD. During
100 Mb/s operation this is 25 MHz +/- 100 ppm. During 10 Mb/s operation this clock is 2.5 MHz +/- 100 ppm.
TXD3/MA15, TXD2/MA14, TXD1/MA13, TXD0/MA12
TXEN 30 O Transmit Enable: This signal is synchronous to TXCLK and provides precise
Note: MII is normally in TRI-STATE, unless enabled by CFG:EXT_PHY. See Section 4.2.2.
25, 24, 23,
22
OOTransmit Data: Signals which are driven synchronous to the TXCLK for
transmission to the external PMD. TXD[3] is the most significant bit and TXD[0] is the least significant bit.
BIOS ROM Address: During external BIOS ROM access, these signals become part of the ROM address.
framing for data carried on TXD[3-0] for the external PMD. It is asserted when TXD[3-0] contains valid data to be transmitted.
7 www.national.com
2.0 Pin Description (Continued)
100BASE-TX/10BASE-T Interface
DP83816
Symbol
TPTDP, TPTDM 54, 53 A-O Transmit Data: Differential common output driver. This differential common output
TPRDP, TPRDM 46, 45 A-I Receive Data: Differential common input buffer. This differential common input can
LQFP Pin
No(s) Dir Description
is configurable to either 10BASE-T or 100BASE-TX signaling:
10BASE-T: Transmission of Manchester encoded 10BASE-T packet data as well as Link Pulses (including Fast Link Pulses for Auto-Negotiation purposes).
100BASE-TX: Transmission of ANSI X3T12 compliant MLT-3 data.
The DP83816 will automatically configure this common output driver for the proper signal type as a result of either forced configuration or Auto-Negotiation.
be configured to accept either 100BASE-TX or 10BASE-T signaling:
10BASE-T: Reception of Manchester encoded 10BASE-T packet data as well as normal Link Pulses and Fast Link Pulses for Auto-Negotiation purposes.
100BASE-TX: Reception of ANSI X3T12 compliant scrambled MLT-3 data.
The DP83816 will automatically configure this common input buffer to accept the proper signal type as a result of either forced configuration or Auto-Negotiation.
BIOS ROM/Flash Interface
Symbol
MCSN 129 O BIOS ROM/Flash Chip Select: During a BIOS ROM/Flash access, this signal is
LQFP Pin
No(s) Dir Description
used to select the ROM device.
MD7, MD6, MD5, MD4/EEDO, MD3, MD2, MD1/CFGDISN, MD0
MA5, MA4/EECLK, MA3/EEDI, MA2/LED100LNK, MA1/LED10LNK, MA0/LEDACT
MWRN 131 O BIOS ROM/Flash Write: During a BIOS ROM/Flash access, this signal is used to
MRDN 130 O BIOS ROM/Flash Read: During a BIOS ROM/Flash access, this signal is used to
Note: DP83816 supports NM27LV010 for the BIOS ROM interface device.
141, 140, 139,
138, 135,
134, 133,
132
3, 2,
1, 144, 143,
142
I/O BIOS ROM/Flash Data Bus: During a BIOS ROM/Flash access these signals are
used to transfer data to or from the ROM/Flash device.
MD[5:0] pins have internal weak pull ups.
MD6 and MD7 pins have internal weak pull downs.
O BIOS ROM/Flash Address: During a BIOS ROM/Flash access, these signals are
used to drive the ROM/Flash address.
enable data to be written to the Flash device.
enable data to be read from the Flash device.
8 www.national.com
2.0 Pin Description (Continued)
Clock Interface
DP83816
Symbol
X1 17 I Crystal/Oscillator Input: This pin is the primary clock reference input for the
X2 18 O Crystal Output: This pin is used in conjunction with the X1 pin to connect to an
LQFP Pin
No(s) Dir Description
DP83816 and must be connected to a 25 MHz 0.005% (50ppm) clock source. The DP83816 device supports either an external crystal resonator connected across pins X1 and X2, or an external CMOS-level oscillator source connected to pin X1 only.
external 25 MHz crystal resonator device. This pin must be left unconnected if an external CMOS oscillator clock source is utilized. For more information see the definition for pin X1.
LED Interface
Symbol
LEDACTN/MA0 142 O TX/RX Activity: This pin is an output indicating transmit/receive activity. This pin is
LED100N/MA2 144 O 100 Mb/s Link: This pin is an output indicating the 100 Mb/s Link status. This pin is
LQFP Pin
No(s) Dir Description
driven low to indicate active transmission or reception, and can be used to drive a low current LED (<6 mA). The activity event is stretched to a minimum duration of approximately 50 ms.
driven low to indicate Good Link status for 100 Mb/s operation, and can be used to drive a low current LED (<6 mA).
LED10N/MA1 143 O 10 Mb/s Link: This pin is an output indicating the 10 Mb/s Link status. This pin is
driven low to indicate Good Link status for 10 Mb/s operation, and can be used to drive a low current LED (<6 mA).
Serial EEPROM Interface
Symbol
EESEL 128 O EEPROM Chip Select: This signal is used to enable an external EEPROM device.
EECLK/MA4 2 O EEPROM Clock: During an EEPROM access (EESEL asserted), this pin is an
EEDI/MA3 1 O EEPROM Data In: During an EEPROM access (EESEL asserted), this pin is an
EEDO/MD4 138 I EEPROM Data Out: During an EEPROM access (EESEL asserted), this pin is an
MD1/CFGDISN 133 I/O Configuration Disable: When pulled low at power-on time, disables load of
Note: DP83816 supports NMC93C46 for the EEPROM device.
LQFP Pin
No(s) Dir Description
output used to drive the serial clock to an external EEPROM device.
output used to drive opcode, address, and data to an external serial EEPROM device.
input used to retrieve EEPROM serial read data.
This pin has an internal weak pull up.
configuration data from the EEPROM. Use 1 K to ground to disable configuration load.
9 www.national.com
2.0 Pin Description (Continued)
External Reference Interface
DP83816
Symbol
VREF 40 I Bandgap Reference: External current reference resistor for internal Phy bandgap
LQFP Pin
No(s) Dir Description
circuitry. The value of this resistor is 10K 1% metal film (100 ppm/ be connected from the VREF pin to analog ground.
o
C) which must
No Connects and Reserved
Symbol
NC 34, 42, 43, 36,
RESERVED 41, 50, 127 These pins are reserved and cannot be connected to any external logic or net.
REGEN 48 Reserved and cannot be connected to any external logic or net..
LQFP Pin
No(s) Dir Description
No Connect
37, 84, 85,
124, 125, 126
This pin has an internal weak pull down.
Supply Pins
Symbol
LQFP Pin
No(s) Dir Description
C1 19 S Connect to GND through 10uF and 0.1uF external capacitors in parallel.
IAUXVDD 39, 47 S Connect to isolated Aux 3.3V supply VDD
AUXVDD 9, 21, 27, 33,
56, 58, 137
PCIVDD 69, 80, 94,
107, 117
VSS 8, 16, 20, 26,
32, 35, 38, 44, 49, 51, 52, 55, 57, 65, 77, 90,
103, 114, 136
S Connect to Aux 3.3V supply VDD
S PCI VDD - connect to PCI bus 3.3V VDD
S VSS
10 www.national.com
3.0 Functional Description
DP83816
DP83816 consists of a MAC/BIU (Media Access Controller/Bus Interface Unit), a physical layer interface, SRAM, and miscellaneous support logic. The MAC/BIU includes the PCI bus, BIOS ROM and EEPROM interfaces,
TPRDP/M
25 MHz Clk
3V DSP Physical Layer
SRAM
RX-2 KB
RAM
SRAM
RXFilter
.5 KB
SRAM
TX-2 KB
BIST
Logic
and an 802.3 MAC. The physical layer interface used is a single-port version of the 3.3V DsPhyterII. Internal memory consists of one - 0.5 KB and two - 2 KB SRAM blocks.
TPTDP/M
MII TX
MII RX
MII Mgt
Test data in
Test data out
MII RX MII TX
Interface
Logic
MII Mgt
BIOS ROM Cntl BIOS ROM Data EEPROM/LEDs
PCI CLK
PCI CNTL
PCI AD
Rx rd data
Tx rd data
Tx Addr
Tx wr data
Rx Addr
Rx wr data
MII Mgt
BROM/EE
MII TX
MII RX
MAC/BIU
DP83816
Figure 3-1 DP83816 Functional Block Diagram
11 www.national.com
3.0 Functional Description (Continued)
DP83816
PCI Bus
32
Interface
PCI Bus
32
32
32
32
32
32
15
Data FIFO
Tx Buffer Manager
Data FIFO
Rx Buffer Manager
MIB
Rx Filter
Pkt Recog
Logic
SRAM
32
4
Tx MAC
32
4
Rx MAC
16
Physical Layer Interface
93C46
Serial
EEPROM
Figure 3-2 MAC/BIU Functional Block Diagram
Boot ROM/
Flash
3.1 MAC/BIU
The MAC/BIU is a derivative design from the DP83810 (Euphrates). The original MAC/BIU design has been optimized to improve logic efficiency and enhanced to add features consistent with current market needs and specification compliance. The MAC/BIU design blocks are discussed in this section.
3.1.1 PCI Bus Interface
This block implements PCI v2.2 bus protocols, and configuration space. Supports bus master reads and writes to CPU memory, and CPU access to on-chip register space. Additional functions provided include: configuration
MAC/BIU
control, serial EEPROM access with auto configuration load, interrupt control, power management control with support for PME or CLKRUN function.
3.1.1.1 Byte Ordering
The DP83816 can be configured to order the bytes of data on the AD[31:0] bus to conform to little endian or big endian ordering through the use of the Configuration Register, bit 0 (CFG:BEM). By default, the device is in little endian ordering. Byte ordering only affects data FIFOs. Register information remains bit aligned (i.e. AD[31] maps to bit 31 in any register space, AD[0] maps to bit 0, etc.).
12 www.national.com
3.0 Functional Description (Continued)
Little Endian (CFG:BEM=0): The byte orientation for
receive and transmit data in system memory is as follows:
232431
Byte 0Byte 1Byte 2Byte 3
MSB
Big Endian (CFG:BEM=1): The byte orientation for receive and transmit data in system memory is as follows:
232431
LSB
3.1.1.2 PCI Bus Interrupt Control
PCI bus interrupts for the DP83816 are asynchronously performed by asserting pin INTAN. This pin is an open drain output. The source of the interrupt can be determined by reading the Interrupt Status Register (ISR). One or more bits in the ISR will be set, denoting all currently pending interrupts. Caution: Reading of the ISR clears ALL bits. Masking of specified interrupts can be accomplished by using the Interrupt Mask Register (IMR).
3.1.1.3 Timer
The Latency Timer described in CFGLAT:LAT defines the minimum number of bus clocks that the device will hold the bus. Once the device gains control of the bus and issues FRAMEN, the Latency Timer will begin counting down. If GNTN is de-asserted before the DP83816 has finished with the bus, the device will maintain ownership of the bus until the timer reaches zero (or has finished the bus transfer). The timer is an 8-bit counter.
3.1.2 Tx MAC
This block implements the transmit portion of 802.3 Media Access Control. The Tx MAC retrieves packet data from the Tx Buffer Manager and sends it out through the transmit portion. Additionally, the Tx MAC provides MIB control information for transmit packets.
3.1.3 Rx MAC
This block implements the receive portion of 802.3 Media Access Control. The Rx MAC retrieves packet data from the receive portion and sends it to the Rx Buffer Manager. Additionally, the Rx MAC provides MIB control information and packet address data for the Rx Filter.
LSB
C/BE[0]C/BE[1]C/BE[2]C/BE[3]
Byte 3Byte 2Byte 1Byte 0
C/BE[0]C/BE[1]C/BE[2]C/BE[3]
3.2 Buffer Management
The buffer management scheme used on the DP83816 allows quick, simple and efficient use of the frame buffer memory. Frames are saved in similar formats for both
0781516
0781516
MSB
DP83816
transmit and receive. The buffer management scheme also uses separate buffers and descriptors for packet information. This allows effective transfers of data from the receive buffer to the transmit buffer by simply transferring the descriptor from the receive queue to the transmit queue.
The format of the descriptors allows the packets to be saved in a number of configurations. A packet can be stored in memory with a single descriptor per single packet, or multiple descriptors per single packet. This flexibility allows the user to configure the DP83816 to maximize efficiency. Architecture of the specific system’s buffer memory, as well as the nature of network traffic, will determine the most suitable configuration of packet descriptors and fragments. Refer to the Buffer Management Section (Section 5.0) for more information.
3.2.1 Tx Buffer Manager
This block DMAs packet data from PCI memory space and places it in the 2 KB transmit FIFO, and pulls data from the FIFO to send to the Tx MAC. Multiple packets (4) may be present in the FIFO, allowing packets to be transmitted with minimum interframe gap. The way in which the FIFO is emptied and filled is controlled by the FIFO threshold values in the TXCFG register: FLTH (Tx Fill Threshold) and DRTH (Tx Drain Threshold). These values determine how full or empty the FIFO must be before the device requests the bus. Additionally, once the DP83816 requests the bus, it will attempt to empty or fill the FIFO as allowed by the MXDMA setting in the TXCFG register.
3.2.2 Rx Buffer Manager
This block retrieves packet data from the Rx MAC and places it in the 2 KB receive data FIFO, and pulls data from the FIFO for DMA to PCI memory space. The Rx Buffer Manager maintains a status FIFO, allowing up to 4 packets to reside in the FIFO at once. Similar to the transmit FIFO, the receive FIFO is controlled by the FIFO threshold value in the RXCFG register: DRTH (Rx Drain Threshold). This value determines the number of long words written into the FIFO from the MAC unit before a DMA request for system memory access occurs. Once the DP83816 gets the bus, it will continue to transfer the long words from the FIFO until the data in the FIFO is less than one long word, or has reached the end of the packet, or the max DMA burst size is reached (RXCFG:MXDMA).
3.2.3 Packet Recognition
The Receive packet filter and recognition logic allows software to control which packets are accepted based on destination address and packet type. Address recognition logic includes support for broadcast, multicast hash, and unicast addresses. The packet recognition logic includes support for WOL, Pause, and programmable pattern recognition.
The standard 802.3 Ethernet packet consists of the following fields: Preamble (PA), Start of Frame Delimiter (SFD), Destination Address (DA), Source Address (SA), Length (LEN), Data and Frame Check Sequence (FCS). All fields are fixed length except for the data field. During reception, the PA, SFD and FCS are stripped. During transmission, the DP83816 generates and appends the PA, SFD and FCS.
13 www.national.com
3.0 Functional Description (Continued)
FCSDataLENSADAPA6BSFD
60b 4b 6B 2B 46B-1500B
Note: B = Bytes
b = bits
Figure 3-3 Ethernet Packet Format
3.2.4 MIB
The MIB block contains counters to track certain media events required by the management specifications RFC 1213 (MIB II), RFC 1398 (Ether-like MIB), and IEEE 802.3 LME. The counters provided are for events which are either difficult or impossible to be intercepted directly by software. Not all counters are implemented, however required counters can be calculated from the counters provided.
3.3 Interface Definitions
3.3.1 PCI System Bus
This interface allows direct connection of the DP83816 to a 33 MHz PCI system bus. The DP83816 supports zero wait state data transfers with burst sizes up to 128 dwords. The
4B
DP83816
DP83816 conforms to 3.3V AC/DC specifications, but has 5V tolerant inputs.
3.3.2 Boot PROM
The BIOS ROM interface allows the DP83816 to read from and write data to an external ROM/Flash device.
3.3.3 EEPROM
The DP83816 supports the attachment of an external EEPROM. The EEPROM interface provides the ability for the DP83816 to read from and write data to an external serial EEPROM device. The DP83816 will auto-load values from the EEPROM to certain fields in PCI configuration space and operational space and perform a checksum to verify that the data is valid. Values in the external EEPROM allow default fields in PCI configuration space and I/O space to be overridden following a hardware reset. If the EEPROM is not present, the DP83816 initialization uses default values for the appropriate Configuration and Operational Registers. Software can read and write to the EEPROM using “bit-bang” accesses via the EEPROM Access Register (MEAR).
3.3.4 Clock
The clock interface provides the 25 MHz clock reference input for the DP83816 IC. The X1 input signal amplitude should be approximately 1V. This interface supports operation from a 25 MHz, 50 ppm CMOS oscillator, or a 25 MHz, 50 ppm, parallel, 20 pF load, < 40 ESR crystal resonator. A 20pF crystal resonator would require C1 and
14 www.national.com
3.0 Functional Description (Continued)
POWER ON
CONFIGURATION
PINS
MANAGEMENT
DP83816
MAC INTERFACE
SERIAL
TX_DATA
TRANSMIT CHANNELS &
STATE MACHINES
100 MB/S 10 MB/S
4B/5B
ENCODER
SCRAMBLER
PARALLEL TO
SERIAL
NRZ TO NRZI
ENCODER
BINARY TO
MLT-3
ENCODER
10/100 COMMON
OUTPUT DRIVER
TXCLK
TXD(3:0)
TX_DATA
NRZ TO
MANCHESTER
ENCODER
LINK PULSE
GENERATOR
TRANSMIT
FILTER
TXER
TXCLK
TXEN
MDIO
MDC
REGISTERS
MII
PHY ADDRESS
AUTO
NEGOTIATION
BASIC MODE
CONTROL
PCS CONTROL
10BASE-T
100BASE-X
FAR-END-FAULT
STATE MACHINE
AUTO-NEGOTIATION
STATE MACHINE
COL
CRS
RXEN
RXER
RXDV
RXD(3:0)
RX_DATARXCLK
RECEIVE CHANNELS &
STATE MACHINES
100 MB/S 10 MB/S
4B/5B
DECODER
CODE GROUP
ALIGNMENT
DESCRAMBLER
SERIAL TO
PARALLEL
NRZI TO NRZ
DECODER
CLOCK
RECOVERY
MLT-3 TO
BINARY
DECODER
ADAPTIVE
EQ AND BLW
COMP.
RXCLK
NCLK_50M
RX_DATA
RXCLK
MANCHESTER
TO NRZ
DECODER
CLOCK
RECOVERY
LINK PULSE
DETECTOR
RECEIVE
FILTER
SMART
SQUELCH
TD±
CLOCK
GENERATION
LED
DRIVERS
LEDS
SYSTEM CLOCK
REFERENCE
Figure 3-4 DSP Physical Layer Block Diagram
15 www.national.com
10/100 COMMON
INPUT BUFFER
RD±
(ALSO FX_RD±)
3.0 Functional Description (Continued)
3.4 Physical Layer
The DP83816 has a full featured physical layer device with integrated PMD sub-layers to support both 10BASE-T and 100BASE-TX Ethernet protocols. The physical layer is designed for easy implementation of 10/100 Mb/s Ethernet home or office solutions. It interfaces directly to twisted pair media via an external transformer. The physical layer utilizes on chip Digital Signal Processing (DSP) technology and digital PLLs for robust performance under all operating conditions, enhanced noise immunity, and lower external component count when compared to analog solutions.
3.4.1 Auto-Negotiation
The Auto-Negotiation function provides a mechanism for exchanging configuration information between two ends of a link segment and automatically selecting the highest performance mode of operation supported by both devices. Fast Link Pulse (FLP) Bursts provide the signalling used to communicate Auto-Negotiation abilities between two devices at each end of a link segment. For further detail regarding Auto-Negotiation, refer to Clause 28 of the IEEE
802.3u specification. The DP83816 supports four different Ethernet protocols (10 Mb/s Half Duplex, 10 Mb/s Full Duplex, 100 Mb/s Half Duplex, and 100 Mb/s Full Duplex), so the inclusion of Auto-Negotiation ensures that the highest performance protocol will be selected based on the advertised ability of the Link Partner. The Auto-Negotiation function within the DP83816 is controlled by internal register access. Auto-Negotiation will be set at power­up/reset, and also when a link status (up/valid) change occurs.
3.4.2 Auto-Negotiation Register Control
When Auto-Negotiation is enabled, the DP83816 transmits the abilities programmed into the Auto-Negotiation Advertisement register (ANAR) via FLP Bursts. Any combination of 10 Mb/s, 100 Mb/s, Half-Duplex, and Full Duplex modes may be selected. The default setting of bits [8:5] in the ANAR and bit 12 in the BMCR register are determined at power-up.
The BMCR provides software with a mechanism to control the operation of the DP83816. Bits 1 & 2 of the PHYSTS register are only valid if Auto-Negotiation is disabled or after Auto-Negotiation is complete. The Auto-Negotiation protocol compares the contents of the ANLPAR and ANAR registers and uses the results to automatically configure to the highest performance protocol common to the local and far-end port. The results of Auto-Negotiation may be accessed in register C0h (PHYSTS), bit 4: Auto­Negotiation Complete, bit 2: Duplex Status and bit 1: Speed Status.
Auto-Negotiation Priority Resolution: — (1) 100BASE-TX Full Duplex (Highest Priority) — (2) 100BASE-TX Half Duplex
— (3) 10BASE-T Full Duplex — (4) 10BASE-T Half Duplex (Lowest Priority) The Basic Mode Control Register (BMCR) provides control
for enabling, disabling, and restarting the Auto-Negotiation process. When Auto-Negotiation is disabled the Speed Selection bit in the BMCR (bit 13) controls switching between 10 Mb/s or 100 Mb/s operation, and the Duplex Mode bit (bit 8) controls switching between full duplex operation and half duplex operation. The Speed Selection and Duplex Mode bits have no effect on the mode of
DP83816
operation when the Auto-Negotiation Enable bit (bit 12) is set.
The Basic Mode Status Register (BMSR) indicates the set of available abilities for technology types, Auto-Negotiation ability, and Extended Register Capability. These bits are permanently set to indicate the full functionality of the DP83816 (only the 100BASE-T4 bit is not set since the DP83816 does not support that function).
The BMSR also provides status on: — Auto-Negotiation complete (bit 5) — Link Partner advertising that a remote fault has occurred
(bit 4) — Valid link has been established (bit 2) — Support for Management Frame Preamble suppression
(bit 6) The Auto-Negotiation Advertisement Register (ANAR)
indicates the Auto-Negotiation abilities to be advertised by the DP83816. All available abilities are transmitted by default, but any ability can be suppressed by writing to the ANAR. Updating the ANAR to suppress an ability is one way for a management agent to change (force) the technology that is used.
The Auto-Negotiation Link Partner Ability Register (ANLPAR) is used to receive the base link code word as well as all next page code words during the negotiation. Furthermore, the ANLPAR will be updated to either 0081h or 0021h for parallel detection to either 100 Mb/s or 10 Mb/s respectively.
The Auto-Negotiation Expansion Register (ANER) indicates additional Auto-Negotiation status. The ANER provides status on:
— Parallel Detect Fault occurrence (bit 4) — Link Partner support of the Next Page function (bit 3) — DP83816 support of the Next Page function (bit 2). The
DP83816 supports the Next Page function. — Current page being exchanged by Auto-Negotiation has
been received (bit1) — Link Partner support of Auto-Negotiation (bit 0)
3.4.3 Auto-Negotiation Parallel Detection
The DP83816 supports the Parallel Detection function as defined in the IEEE 802.3u specification. Parallel Detection requires both the 10 Mb/s and 100 Mb/s receivers to monitor the receive signal and report link status to the Auto-Negotiation function. Auto-Negotiation uses this information to configure the correct technology in the event that the Link Partner does not support Auto-Negotiation yet is transmitting link signals that the 100BASE-TX or 10BASE-T PMAs (Physical Medium Attachments) recognize as valid link signals.
If the DP83816 completes Auto-Negotiation as a result of Parallel Detection, bits 5 and 7 within the ANLPAR register will be updated to reflect the mode of operation present in the Link Partner. Note that bits 4:0 of the ANLPAR will also be set to 00001 based on a successful parallel detection to indicate a valid 802.3 selector field. Software may determine that negotiation completed via Parallel Detection by reading the ANER (98h) register with bit 0, Link Partner Auto-Negotiation Able bit, being reset to a zero, once the Auto-Negotiation Complete bit, bit 5 of the BMSR (84h) register is set to a one. If configured for parallel detect
16 www.national.com
3.0 Functional Description (Continued)
mode, and any condition other than a single good link occurs, then the parallel detect fault bit will set to a one, bit 4 of the ANER register (98h).
3.4.4 Auto-Negotiation Restart
Once Auto-Negotiation has completed, it may be restarted at any time by setting bit 9 (Restart Auto-Negotiation) of the BMCR to one. If the mode configured by a successful Auto­Negotiation loses a valid link, then the Auto-Negotiation process will resume and attempt to determine the configuration for the link. This function ensures that a valid configuration is maintained if the cable becomes disconnected.
A renegotiation request from any entity, such as a management agent, will cause the DP83816 to halt any transmit data and link pulse activity until the break_link_timer expires (~1500 ms). Consequently, the Link Partner will go into link fail and normal Auto­Negotiation resumes. The DP83816 will resume Auto­Negotiation after the break_link_timer has expired by issuing FLP (Fast Link Pulse) bursts.
3.4.5 Enabling Auto-Negotiation via Software
It is important to note that if the DP83816 has been initialized upon power-up as a non-auto-negotiating device (forced technology), and it is then required that Auto­Negotiation or re-Auto-Negotiation be initiated via software, bit 12 (Auto-Negotiation Enable) of the Basic Mode Control Register must first be cleared and then set for any Auto­Negotiation function to take effect.
3.4.6 Auto-Negotiation Complete Time
Parallel detection and Auto-Negotiation take approximately 2-3 seconds to complete. In addition, Auto-Negotiation with next page should take approximately 2-3 seconds to complete, depending on the number of next pages sent.
Refer to Clause 28 of the IEEE 802.3u standard for a full description of the individual timers related to Auto­Negotiation.
3.5 LED Interfaces
The DP83816 has parallel outputs to indicate the status of Activity (Transmit or Receive), 100 Mb/s Link, and 10 Mb/s Link.
The LEDACTN pin indicates the presence of transmit or receive activity. The standard CMOS driver goes low when RX or TX activity is detected in either 10 Mb/s or 100 Mb/s operation.
The LED100N pin indicates a good link at 100 Mb/s data rate. The standard CMOS driver goes low when this occurs. In 100BASE-T mode, link is established as a result of input receive amplitude compliant with TP-PMD specifications which will result in internal generation of signal detect. This signal will assert after the internal Signal Detect has remained asserted for a minimum of 500 us. The signal will de-assert immediately following the de­assertion of the internal signal detect.
The LED10N pin indicates a good link at 10 Mb/s data rate. The standard CMOS driver goes low when this occurs. 10 Mb/s Link is established as a result of the reception of at least seven consecutive normal Link Pulses or the reception of a valid 10BASE-T packet. This will cause the assertion of this signal. the signal will de-assert in accordance with the Link Loss Timer as specified in IEEE
802.3. The DP83816 LED pins are capable of 6 mA. Connection
of these LED pins should ensure this is not overloaded. Using 2 mA LED devices the connection for the LEDs could be as shown in Figure 3-5.
LEDACTN
453
LED100N
453
LED10N
453
V
DD
Figure 3-5 LED Loading Example
DP83816
17 www.national.com
3.0 Functional Description (Continued)
3.6 Half Duplex vs. Full Duplex
The DP83816 supports both half and full duplex operation at both 10 Mb/s and 100 Mb/s speeds.
Half-duplex is the standard, traditional mode of operation which relies on the CSMA/CD protocol to handle collisions and network access. In Half-Duplex mode, CRS responds to both transmit and receive activity in order to maintain compliance with IEEE 802.3 specification.
Since the DP83816 is designed to support simultaneous transmit and receive activity it is capable of supporting full­duplex switched applications with a throughput of up to 200 Mb/s per port when operating in 100BASE-TX mode. Because the CSMA/CD protocol does not apply to full­duplex operation, the DP83816 disables its own internal collision sensing and reporting functions.
It is important to understand that while full Auto-Negotiation with the use of Fast Link Pulse code words can interpret and configure to support full-duplex, parallel detection can not recognize the difference between full and half-duplex from a fixed 10 Mb/s or 100 Mb/s link partner over twisted pair. Therefore, as specified in 802.3u, if a far-end link partner is transmitting forced full duplex 100BASE-TX for example, the parallel detection state machine in the receiving station would be unable to detect the full duplex capability of the far-end link partner and would negotiate to a half duplex 100BASE-TX configuration (same scenario for 10 Mb/s).
For full duplex operation, the following register bits must also be set:
— TXCFG:CSI (Carrier Sense Ignore) — TXCFG:HBI (HeartBeat Ignore) — RXCFG:ATX (Accept Transmit Packets) Additionally, the Auto-Negotiation Select bits in the
Configuration register must show full duplex support: — CFG:ANEG_SEL
3.7 Phy Loopback
The DP83816 includes a Phy Loopback Test mode for easy board diagnostics. The Loopback mode is selected through bit 14 (Loopback) of the Basic Mode Control Register (BMCR). Writing 1 to this bit enables transmit data to be routed to the receive path early in the physical layer cell. Loopback status may be checked in bit 3 of the PHY Status Register (C0h). While in Loopback mode the data will not be transmitted onto the media. This is true for either 10 Mb/s as well as 100 Mb/s data.
In 100BASE-TX Loopback mode the data is routed through the PCS and PMA layers into the PMD sublayer before it is
DP83816
looped back. Therefore, in addition to serving as a board diagnostic, this mode serves as quick functional verification of the device.
Note: A Mac Loopback can be performed via setting bit 29 (Mac Loopback) in the Tx Configuration Register.
3.8 Status Information
There are 3 pins that are available to convey status information to the user through LEDs to indicate the speed (10 Mb/s or 100 Mb/s) link status and receive or transmit activity.
10 Mb/s Link is established as a result of the reception of at least seven consecutive Normal Link Pulses or the reception of a valid 10BASE-T packet. LED10N will de­assert in accordance with the Link Loss Timer specified in IEEE 802.3.
100BASE-T Link is established as a result of an input receive amplitude compliant with TP-PMD specifications which will result in internal generation of Signal Detect. LED100N will assert after the internal Signal Detect has remained asserted for a minimum of 500 µs. LED100N will de-assert immediately following the de-assertion of the internal Signal Detect.
Activity LED status indicates Receive or Transmit activity.
3.9 100BASE-TX TRANSMITTER
The 100BASE-TX transmitter consists of several functional blocks which convert synchronous 4-bit nibble data, to a scrambled MLT-3 125 Mb/s serial data stream. Because the 100BASE-TX TP-PMD is integrated, the differential output pins, TD±, can be directly routed to the magnetics.
The block diagram in Figure 3-6 provides an overview of each functional block within the 100BASE-TX transmit section.
The Transmitter section consists of the following functional blocks:
— Code-group Encoder and Injection block (bypass option) — Scrambler block (bypass option) — NRZ to NRZI encoder block — Binary to MLT-3 converter / Common Driver The bypass option for the functional blocks within the
100BASE-TX transmitter provides flexibility for applications such as 100 Mb/s repeaters where data conversion is not always required. The DP83816 implements the 100BASE­TX transmit state machine diagram as specified in the IEEE 802.3u Standard, Clause 24.
18 www.national.com
3.0 Functional Description (Continued)
TXCLK
FROM CGM
DP83816
TXD(3:0)/TXER
4B5B CODE-
GROUP ENABLER
BP_4B5B
BP_SCR
100BASE-TX
LOOPBACK
MUX
5B PARALLEL
TO SERIAL
SCRAMBLER
MUX
NRZ TO NRZI
BINARY TO MLT-3/ COMMON
DRIVER
MUX
ENCODER
Figure 3-6 100BASE-TX Transmit Block Diagram
3.9.1 Code-group Encoding and Injection
The code-group encoder converts 4-bit (4B) nibble data generated by the MAC into 5-bit (5B) code-groups for transmission. This conversion is required to allow control data to be combined with packet data code-groups. Refer to Table 3-1 for 4B to 5B code-group mapping details.
The code-group encoder substitutes the first 8-bits of the MAC preamble with a J/K code-group pair (11000 10001) upon transmission. The code-group encoder continues to replace subsequent 4B preamble and data nibbles with corresponding 5B code-groups. At the end of the transmit packet, upon the de-assertion of Transmit Enable signal from the MAC, the code-group encoder injects the T/R code-group pair (01101 00111) indicating the end of frame.
After the T/R code-group pair, the code-group encoder continuously injects IDLEs into the transmit data stream until the next transmit packet is detected (re-assertion of Transmit Enable).
TD +/-
3.9.2 Scrambler
The scrambler is required to control the radiated emissions at the media connector and on the twisted pair cable (for 100BASE-TX applications). By scrambling the data, the total energy launched onto the cable is randomly distributed over a wide frequency range. Without the scrambler, energy levels at the PMD and on the cable could peak beyond FCC limitations at frequencies related to repeating 5B sequences (i.e., continuous transmission of IDLEs).
The scrambler is configured as a closed loop linear feedback shift register (LFSR) with an 11-bit polynomial. The output of the closed loop LFSR is X-ORd with the serial NRZ data from the code-group encoder. The result is a scrambled data stream with sufficient randomization to decrease radiated emissions at certain frequencies by as much as 20 dB.
19 www.national.com
3.0 Functional Description (Continued)
3.9.3 NRZ to NRZI Encoder
After the transmit data stream has been serialized and scrambled, the data must be NRZI encoded in order to comply with the TP-PMD standard for 100BASE-TX transmission over Category-5 un-shielded twisted pair cable. There is no ability to bypass this block within the DP83816.
binary_plus
D
binary_in
Q
Q
binary_minus
differential MLT-3
3.9.4 Binary to MLT-3 Convertor / Common Driver
The Binary to MLT-3 conversion is accomplished by converting the serial binary data stream output from the NRZI encoder into two binary data streams with alternately phased logic one events. These two binary streams are then fed to the twisted pair output driver which converts the voltage to current and alternately drives either side of the transmit transformer primary winding, resulting in a minimal current (20 mA max) MLT-3 signal. Refer to Figure 3-7
binary_in
binary_plus
COMMON
DRIVER
binary_minus
MLT-3
DP83816
Figure 3-7 Binary to MLT-3 conversion
Table 3-1 4B5B Code-Group Encoding/Decoding
Name PCS 5B Code-group Description/4B Value
DATA CODES
0 11110 0000 1 01001 0001 2 10100 0010 3 10101 0011 4 01010 0100 5 01011 0101 6 01110 0110 7 01111 0111 8 10010 1000 9 10011 1001 A 10110 1010
B 10111 1011 C 11010 1100 D 11011 1101 E 11100 1110 F 11101 1111
IDLE AND CONTROL CODES
H 00100 HALT code-group - Error code
I 11111 Inter-Packet IDLE - 0000
J 11000 First Start of Packet - 0101 K 10001 Second Start of Packet - 0101 T 01101 First End of Packet - 0000 R 00111 Second End of Packet - 0000
20 www.national.com
3.0 Functional Description (Continued)
Table 3-1 4B5B Code-Group Encoding/Decoding
Name PCS 5B Code-group Description/4B Value
INVALID CODES
V 00000 V 00001 V 00010 V 00011 V 00101 V 00110 V 01000 V 01100 V 10000 V 11001
DP83816
The 100BASE-TX MLT-3 signal sourced by the TD± common driver output pins is slew rate controlled. This should be considered when selecting AC coupling magnetics to ensure TP-PMD Standard compliant transition times (3 ns < Tr < 5 ns).
The 100BASE-TX transmit TP-PMD function within the DP83816 is capable of sourcing only MLT-3 encoded data. Binary output from the TD± outputs is not possible in 100 Mb/s mode.
3.10 100BASE-TX Receiver
The 100BASE-TX receiver consists of several functional blocks which convert the scrambled MLT-3 125 Mb/s serial data stream to synchronous 4-bit nibble data that is provided to the MAC. Because the 100BASE-TX TP-PMD is integrated, the differential input pins, RD±, can be directly routed from the AC coupling magnetics.
See Figure 3-8 for a block diagram of the 100BASE-TX receive function. This provides an overview of each functional block within the 100BASE-TX receive section.
The Receive section consists of the following functional blocks:
—ADC — Input and BLW Compensation — Signal Detect — Digital Adaptive Equalization — MLT-3 to Binary Decoder — Clock Recovery Module — NRZI to NRZ Decoder — Serial to Parallel — De-scrambler (bypass option) — Code Group Alignment — 4B/5B Decoder (bypass option) — Link Integrity Monitor — Bad SSD Detection The bypass option for the functional blocks within the
100BASE-TX receiver provides flexibility for applications such as 100 Mb/s repeaters where data conversion is not always required.
3.10.1 Input and Base Line Wander Compensation
Unlike the DP83223V Twister, the DP83816 requires no external attenuation circuitry at its receive inputs, RD+/−. It accepts TP-PMD compliant waveforms directly, requiring only a 100 termination plus a simple 1:1 transformer.
The DP83816 is completely ANSI TP-PMD compliant and includes Base Line Wander (BLW) compensation. The BLW compensation block can successfully recover the TP­PMD defined “killer” pattern and pass it to the digital adaptive equalization block.
BLW can generally be defined as the change in the average DC content, over time, of an AC coupled digital transmission over a given transmission medium. (i.e. copper wire).
BLW results from the interaction between the low frequency components of a transmitted bit stream and the frequency response of the AC coupling component(s) within the transmission system. If the low frequency content of the digital bit stream goes below the low frequency pole of the AC coupling transformers then the droop characteristics of the transformers will dominate resulting in potentially serious BLW.
The digital oscilloscope plot provided in Figure 3-9 illustrates the severity of the BLW event that can theoretically be generated during 100BASE-TX packet transmission. This event consists of approximately 800 mV of DC offset for a period of 120 us. Left uncompensated, events such as this can cause packet loss.
3.10.2 Signal Detect
The signal detect function of the DP83816 is incorporated to meet the specifications mandated by the ANSI FDDI TP­PMD Standard as well as the IEEE 802.3 100BASE-TX Standard for both voltage thresholds and timing parameters.
Note that the reception of normal 10BASE-T link pulses and fast link pulses per IEEE 802.3u Auto-Negotiation by the 100BASE-TX receiver do not cause the DP83816 to assert signal detect.
21 www.national.com
3.0 Functional Description (Continued)
DP83816
LINK INTEGRITY
MONITOR
RX_DATA VALID
SSD DETECT
RXCLK
BP_SCR
BP_RX
BP_4B5B
RXD(3:0)/RXER
MUX
4B/5B DECODER
SERIAL TO
PARALLEL
CODE GROUP
ALIGNMENT
MUX
SD
MUX
CLOCK
RECOVERY
CLOCK
MODULE
DESCRAMBLER
NRZI TO NRZ
DECODER
MLT-3 TO BINARY
DECODER
DIGITAL
ADAPTIVE
EQUALIZATION
AGC
INPUT BLW
COMPENSATION
ADC
SIGNAL
DETECT
RD +/-
Figure 3-8 100 M/bs Receive Block Diagram
22 www.national.com
3.0 Functional Description (Continued)
DP83816
Figure 3-9 100BASE-TX BLW Event Diagram
3.10.3 Digital Adaptive Equalization
When transmitting data at high speeds over copper twisted pair cable, frequency dependent attenuation becomes a concern. In high-speed twisted pair signalling, the frequency content of the transmitted signal can vary greatly during normal operation based primarily on the randomness of the scrambled data stream. This variation in signal attenuation caused by frequency variations must be compensated for to ensure the integrity of the transmission.
In order to ensure quality transmission when employing MLT-3 encoding, the compensation must be able to adapt to various cable lengths and cable types depending on the installed environment. The selection of long cable lengths for a given implementation, requires significant compensation which will over-compensate for shorter, less attenuating lengths. Conversely, the selection of short or intermediate cable lengths requiring less compensation will cause serious under-compensation for longer length cables. Therefore, the compensation or equalization must be adaptive to ensure proper conditioning of the received signal independent of the cable length.
The DP83816 utilizes an extremely robust equalization scheme referred to herein as ‘Digital Adaptive Equalization’. Traditional designs use a pseudo adaptive equalization scheme that determines the approximate cable length by monitoring signal attenuation at certain frequencies. This attenuation value was compared to the internal receive input reference voltage. This comparison would indicate the amount of equalization to use. Although this scheme is used successfully on the DP83223V twister,
it is sensitive to transformer mismatch, resistor variation and process induced offset. The DP83223V also required an external attenuation network to help match the incoming signal amplitude to the internal reference.
The Digital Equalizer removes ISI (Inter Symbol Interference) from the receive data stream by continuously adapting to provide a filter with the inverse frequency response of the channel. When used in conjunction with a gain stage, this enables the receive 'eye pattern' to be opened sufficiently to allow very reliable data recovery.
Traditionally 'adaptive' equalizers selected 1 of N filters in an attempt to match the cables characteristics. This approach will typically leave holes at certain cable lengths, where the performance of the equalizer is not optimized. The DP83816 equalizer is truly adaptive.
The curves given in Figure 3-10 illustrate attenuation at certain frequencies for given cable lengths. This is derived from the worst case frequency vs. attenuation figures as specified in the EIA/TIA Bulletin TSB-36. These curves indicate the significant variations in signal attenuation that must be compensated for by the receive adaptive equalization circuit.
Figure 3-11 represents a scrambled IDLE transmitted over zero meters of cable as measured at the AII (Active Input Interface) of the receiver. Figure 3-12 and Figure 3-13 represent the signal degradation over 50 and 100 meters of category V These plots show the extreme degradation of signal integrity and indicate the requirement for a robust adaptive equalizer.
cable respectively, also measured at the AII.
23 www.national.com
3.0 Functional Description (Continued)
DP83816
2ns/div
Figure 3-10 EIA/TIA Attenuation vs. Frequency for 0, 50,
100, 130 & 150 meters of CAT V
cable
2ns/div
Figure 3-11 MLT-3 Signal Measured at AII after 0 meters
of CAT V cable
3.10.4 Line Quality Monitor
It is possible to determine the amount of Equalization being used by accessing certain test registers with the DSP engine. This provides a crude indication of connected cable length. This function allows for a quick and simple verification of the line quality in that any significant deviation from an expected register value (based on a known cable length) would indicate that the signal quality has deviated from the expected nominal case.
Figure 3-12 MLT-3 Signal Measured at AII after 50
meters of CAT V
cable
2ns/div
Figure 3-13 MLT-3 Signal Measured at AII after 100
meters of CAT V
3.10.5 MLT-3 to NRZI Decoder
The DP83816 decodes the MLT-3 information from the Digital Adaptive Equalizer block to binary NRZI data.
cable
24 www.national.com
3.0 Functional Description (Continued)
3.10.6 Clock Recovery Module
The Clock Recovery Module (CRM) accepts 125 Mb/s MLT3 data from the equalizer. The DPLL locks onto the 125 Mb/s data stream and extracts a 125 MHz recovered clock. The extracted and synchronized clock and data are used as required by the synchronous receive operations as generally depicted in Figure 3-8.
The CRM is implemented using an advanced all digital Phase Locked Loop (PLL) architecture that replaces sensitive analog circuitry. Using digital PLL circuitry allows the DP83816 to be manufactured and specified to tighter tolerances.
3.10.7 NRZI to NRZ
In a typical application, the NRZI to NRZ decoder is required in order to present NRZ formatted data to the de­scrambler (or to the code-group alignment block, if the de­scrambler is bypassed, or directly to the PCS, if the receiver is bypassed).
3.10.8 Serial to Parallel
The 100BASE-TX receiver includes a Serial to Parallel converter which supplies 5-bit wide data symbols to the PCS Rx state machine.
3.10.9 De-scrambler
A serial de-scrambler is used to de-scramble the received NRZ data. The de-scrambler has to generate an identical data scrambling sequence (N) in order to recover the original unscrambled data (UD) from the scrambled data (SD) as represented in the equations:
SD UD N()= UD SD N()=
Synchronization of the de-scrambler to the original scrambling sequence (N) is achieved based on the knowledge that the incoming scrambled data stream consists of scrambled IDLE data. After the de-scrambler has recognized 12 consecutive IDLE code-groups, where an unscrambled IDLE code-group in 5B NRZ is equal to five consecutive ones (11111), it will synchronize to the receive data stream and generate unscrambled data in the form of unaligned 5B code-groups.
In order to maintain synchronization, the de-scrambler must continuously monitor the validity of the unscrambled data that it generates. To ensure this, a line state monitor and a hold timer are used to constantly monitor the synchronization status. Upon synchronization of the de­scrambler the hold timer starts a 722 µs countdown. Upon detection of sufficient IDLE code-groups (58 bit times) within the 722 µs period, the hold timer will reset and begin a new countdown. This monitoring operation will continue indefinitely given a properly operating network connection with good signal integrity. If the line state monitor does not
DP83816
recognize sufficient unscrambled IDLE code-groups within the 722 µs period, the entire de-scrambler will be forced out of the current state of synchronization and reset in order to re-acquire synchronization.
3.10.10 Code-group Alignment
The code-group alignment module operates on unaligned 5-bit data from the de-scrambler (or, if the de-scrambler is bypassed, directly from the NRZI/NRZ decoder) and converts it into 5B code-group data (5 bits). Code-group alignment occurs after the J/K code-group pair is detected. Once the J/K code-group pair (11000 10001) is detected, subsequent data is aligned on a fixed boundary.
3.10.11 4B/5B Decoder
The code-group decoder functions as a look up table that translates incoming 5B code-groups into 4B nibbles. The code-group decoder first detects the J/K code-group pair preceded by IDLE code-groups and replaces the J/K with MAC preamble. Specifically, the J/K 10-bit code-group pair is replaced by the nibble pair (0101 0101). All subsequent 5B code-groups are converted to the corresponding 4B nibbles for the duration of the entire packet. This conversion ceases upon the detection of the T/R code­group pair denoting the End of Stream Delimiter (ESD) or with the reception of a minimum of two IDLE code-groups.
3.10.12 100BASE-TX Link Integrity Monitor
The 100 Base-TX Link monitor ensures that a valid and stable link is established before enabling both the Transmit and Receive PCS layer.
Signal detect must be valid for 395 µs to allow the link monitor to enter the 'Link Up' state, and enable the transmit and receive functions.
Signal detect can be forced active by setting Bit 1 of the PCSR.
Signal detect can be optionally ANDed with the de­scrambler locked indication by setting bit 8 of the PCSR. When this option is enabled, then De-scrambler 'locked' is required to enter the Link Up state, but only Signal detect is required to maintain the link in the link Up state.
3.10.13 Bad SSD Detection
A Bad Start of Stream Delimiter (Bad SSD) is any transition from consecutive idle code-groups to non-idle code-groups which is not prefixed by the code-group pair J/K.
If this condition is detected, the DP83816 will assert RXER and present RXD[3:0] = 1110 to the MAC for the cycles that correspond to received 5B code-groups until at least two IDLE code groups are detected. In addition, the False Carrier Event Counter will be incremented by one.
Once at least two IDLE code groups are detected, the error is reported to the MAC.
25 www.national.com
3.0 Functional Description (Continued)
3.11 10BASE-T Transceiver Module
The 10BASE-T Transceiver Module is IEEE 802.3 compliant. It includes the receiver, transmitter, collision, heartbeat, loopback, jabber, and link integrity functions, as defined in the standard. An external filter is not required on the 10BASE-T interface since this is integrated inside the DP83816. This section focuses on the general 10BASE-T system level operation.
3.11.1 Operational Modes
The DP83816 has two basic 10BASE-T operational modes:
— Half Duplex mode - functions as a standard IEEE 802.3
10BASE-T transceiver supporting the CSMA/CD protocol.
— Full Duplex mode - capable of simultaneously
transmitting and receiving without reporting a collision. The DP83816's 10 Mb/s ENDEC is designed to encode and decode simultaneously.
3.11.2 Smart Squelch
The smart squelch is responsible for determining when valid data is present on the differential receive inputs (RD±). The DP83816 implements an intelligent receive squelch to ensure that impulse noise on the receive inputs will not be mistaken for a valid signal. Smart squelch operation is independent of the 10BASE-T operational mode.
DP83816
The squelch circuitry employs a combination of amplitude and timing measurements (as specified in the IEEE 802.3 10BASE-T standard) to determine the validity of data on the twisted pair inputs (refer to Figure 3-14).
The signal at the start of packet is checked by the smart squelch and any pulses not exceeding the squelch level (either positive or negative, depending upon polarity) will be rejected. Once this first squelch level is overcome correctly, the opposite squelch level must then be exceeded within 150 ns. Finally the signal must again exceed the original squelch level within a 150 ns to ensure that the input waveform will not be rejected. This checking procedure results in the loss of typically three preamble bits at the beginning of each packet.
Only after all these conditions have been satisfied will a control signal be generated to indicate to the remainder of the circuitry that valid data is present. At this time, the smart squelch circuitry is reset.
Valid data is considered to be present until the squelch level has not been generated for a time longer than 150 ns, indicating the End of Packet. Once good data has been detected the squelch levels are reduced to minimize the effect of noise causing premature End of Packet detection.
V
V
SQ+(reduced)
V
SQ-(reduced)
SQ+
V
<150 ns
SQ-
<150 ns
start of packet
Figure 3-14 10BASE-T Twisted Pair Smart Squelch Operation
3.11.3 Collision Detection
When in Half Duplex, a 10BASE-T collision is detected when the receive and transmit channels are active simultaneously. Collisions are reported to the MAC. Collisions are also reported when a jabber condition is detected.
If the ENDEC is receiving when a collision is detected it is reported immediately (through the COL signal).
When heartbeat is enabled, approximately 1 µs after the transmission of each packet, a Signal Quality Error (SQE) signal of approximately 10 bit times is generated to indicate successful transmission.
The SQE test is inhibited when the physical layer is set in full duplex mode. SQE can also be inhibited by setting the HEARTBEAT_DIS bit in the TBTSCR register.
>150 ns
end of packet
3.11.4 Normal Link Pulse Detection/Generation
The link pulse generator produces pulses as defined in the IEEE 802.3 10BASE-T standard. Each link pulse is nominally 100 ns in duration and transmitted every 16 ms in the absence of transmit data.
Link pulses are used to check the integrity of the connection with the remote end. If valid link pulses are not received, the link detector disables the 10BASE-T twisted pair transmitter, receiver and collision detection functions.
When the link integrity function is disabled (FORCE_LINK_10 of the TBTSCR register), good link is forced and the 10BASE-T transceiver will operate regardless of the presence of link pulses.
26 www.national.com
3.0 Functional Description (Continued)
3.11.5 Jabber Function
The jabber function monitors the DP83816's output and disables the transmitter if it attempts to transmit a packet of longer than legal size. A jabber timer monitors the transmitter and disables the transmission if the transmitter is active for approximately 20-30 ms.
Once disabled by the jabber function, the transmitter stays disabled for the entire time that the ENDEC module's internal transmit enable is asserted. This signal has to be de-asserted for approximately 400-600 ms (the “unjab” time) before the jabber function re-enables the transmit outputs.
The Jabber function is only meaningful in 10BASE-T mode.
3.11.6 Automatic Link Polarity Detection
The DP83816's 10BASE-T transceiver module incorporates an automatic link polarity detection circuit. When seven consecutive link pulses or three consecutive receive packets with inverted End-of-Packet pulses are received, bad polarity is reported.
A polarity reversal can be caused by a wiring error at either end of the cable, usually at the Main Distribution Frame (MDF) or patch panel in the wiring closet.
The bad polarity condition is latched. The DP83816's 10BASE-T transceiver module corrects for this error internally and will continue to decode received data correctly. This eliminates the need to correct the wiring error immediately.
3.11.7 10BASE-T Internal Loopback
When the LOOPBACK bit in the BMCR register is set, 10BASE-T transmit data is looped back in the ENDEC to the receive channel. The transmit drivers and receive input circuitry are disabled in transceiver loopback mode, isolating the transceiver from the network.
Loopback is used for diagnostic testing of the data path through the transceiver without transmitting on the network or being interrupted by receive traffic. This loopback function causes the data to loopback just prior to the 10BASE-T output driver buffers such that the entire transceiver path is tested.
3.11.8 Transmit and Receive Filtering
External 10BASE-T filters are not required when using the DP83816, as the required signal conditioning is integrated into the device.
Only isolation/step-up transformers and impedance matching resistors are required for the 10BASE-T transmit and receive interface. The internal transmit filtering ensures that all the harmonics in the transmit signal are attenuated by at least 30 dB.
3.11.9 Transmitter
The encoder begins operation when the transmit enable input to the physical layer is asserted and converts NRZ data to pre-emphasized Manchester data for the transceiver. For the duration of assertion, the serialized transmit data is encoded for the transmit-driver pair (TD±). The last transition is always positive; it occurs at the center of the bit cell if the last bit is a one, or at the end of the bit cell if the last bit is a zero.
3.11.10 Receiver
The decoder consists of a differential receiver and a PLL to separate a Manchester encoded data stream into internal
DP83816
clock signals and data. The differential input must be externally terminated with a differential 100 termination network to accommodate UTP cable. The internal impedance of RD± (typically 1.1K) is in parallel with two
54.9 resistors to approximate the 100 termination. The decoder detects the end of a frame when no more mid-
bit transitions are detected.
3.11.11 Far End Fault Indication
Auto-Negotiation provides a mechanism for transferring information from the Local Station to the Link Partner that a remote fault has occurred for 100BASE-TX.
A remote fault is an error in the link that one station can detect while the other cannot. An example of this is a disconnected fiber at a station’s transmitter. This station will be receiving valid data and detect that the link is good via the Link Integrity Monitor, but will not be able to detect that its transmission is not propagating to the other station.
If three or more FEFI IDLE patterns are detected by the DP83816, then bit 4 of the Basic Mode Status register is set to one until read by management, additionally bit 7 of the PHY Status register is also set.
The first FEFI IDLE pattern may contain more than 84 ones as the pattern may have started during a normal IDLE transmission which is actually quite likely to occur. However, since FEFI is a repeating pattern, this will not cause a problem with the FEFI function. It should be noted that receipt of the FEFI IDLE pattern will not cause a Carrier Sense error to be reported.
If the FEFI function has been disabled via FEFI_EN (bit 3) of the PCSR Configuration register, then the DP83816 will not send the FEFI IDLE pattern.
3.12 802.3u MII
The DP83816 incorporates the Media Independent Interface (MII) as specified in Clause 22 of the IEEE 802.3u standard. This interface may be used to connect PHY devices. This section describes the MII configuration steps as well as the serial MII management interface and nibble wide MII data interface.
3.12.1 MII Access Configuration
The DP83816 must be specifically configured for accessing the MII. This is done by first connecting pin 133 (MD1/CFGDISN) to GND through a 1K resistor. Then setting bit 12 (EXT_PHY) of the CFG register (offset 04h) to 1. See Section 4.2.2. When this bit is set, the internal Phy is automatically disabled, as reported by bit 9 (PHY_DIS) of the CFG register. The MII must then be initialized, as described in Section 3.12.4, before the external PHY can be detected.
If external MII is not selected through the register setting as described, then the internal Phy is used and the MII pins of the MacPhyter-II can be left unconnected.
3.12.2 MII Serial Management
The MII serial management interface allows for the configuration and control of PHY registers, gathering of status, error information, and the determination of the type and capabilities of the attached PHY(s).
The MII serial management specification defines a set of thirty-two 16-bit status and control registers that are accessible through the management interface pins MDC and MDIO. A description of the serial management interface access and access protocol follows.
27 www.national.com
3.0 Functional Description (Continued)
3.12.3 MII Serial Management Access
Management access to the PHY(s) is done via Management Data Clock (MDC) and Management Data Input/Output (MDIO). MDC has a maximum clock rate of 25 MHz and no minimum rate. The MDIO line is bi-directional and may be shared by up to 32 devices. The internal PHY counts as one of these 32 devices.
The internal PHY has the advantage of having direct register access but can also be controlled exactly like a PHY, with a default address of 1Fh, connected to the MII.
Access and control of the MDC and MDIO pins is done via the MII/EEPROM Access Register (MEAR). The clock (MDC) is created by alternating writes of 0 then 1 to the MDC bit (bit 6). Control of data direction is done by the MDDIR bit (bit 5). Data is either recorded or written by the MDIO bit (bit 4). Setting the MDDIR bit to a 1 allows the DP83816 to drive the MDIO pin. Setting the MDDIR bit to a 0 allows the MDIO bit to reflect the value of the MDIO pin. See Section 4.2.3
This bit-bang access of the MDC and MDIO pins thus requires 64 accesses to the MEAR register to complete a single PHY register transaction. Since a PHY device is typically self configuring and adaptive this serial management access is usually only required at initialization time and therefore is not time critical.
DP83816
3.12.4 Serial Management Access Protocol
The serial control interface clock (MDC) has a maximum clock rate of 25 MHz and no minimum rate. The MDIO line is bi-directional and may be shared by up to 32 devices. The MDIO frame format is shown in Table 3-2.
If external PHY devices may be attached and removed from the MII there should be a 15 K pull-down resistor on the MDIO signal. If the PHY will always be connected then there should be a 1.5 kpull-up resistor which, during IDLE and turnaround, will pull MDIO high. In order to initialize the MDIO interface, the DP83816 sends a sequence of 32 contiguous logic ones on MDIO provides the PHY(s) with a sequence that can be used to establish synchronization. This preamble may be generated either by driving MDIO high for 32 consecutive MDC clock cycles, or by simply allowing the MDIO pull-up resistor to pull the MDIO pin high during which time 32 MDC clock cycles are provided. In addition 32 MDC clock cycles should be used to re-sync the device if an invalid start, opcode, or turnaround bit is detected.
Table 3-2 Typical MDIO Frame Format
MII Management
<idle><start><op code><device addr><reg addr><turnaround><data><idle>
Serial Protocol
Read Operation <idle><01><10><AAAAA><RRRRR><Z0><xxxx xxxx xxxx xxxx><idle>
Write Operation <idle><01><01><AAAAA><RRRRR><10><xxxx xxxx xxxx xxxx><idle>
The Start code is indicated by a <01> pattern. This assures the MDIO line transitions from the default idle line state.
Turnaround is defined as an idle bit time inserted between the Register Address field and the Data field. To avoid contention during a read transaction, no device shall actively drive the MDIO signal during the first bit of
Turnaround. The addressed PHY drives the MDIO with a zero for the second bit of turnaround and follows this with the required data. Figure 3-15 shows the timing relationship between MDC and the MDIO as driven/received by the DP83816 and a PHY for a typical register read access.
MDC
MDIO
(STA)
MDIO
(PHY)
Z
Z
00011 110000 000
Idle Start
Opcode
(Read)
PHY Address
(PHYAD = 0Ch)
Register Address
(00h = BMCR)
Z
Z
Z
0 0 011000100000000
TA
Register Data
Figure 3-15 Typical MDC/MDIO Read Operation
Z
Z
Idle
For write transactions, the DP83816 writes data to the addressed PHY thus eliminating the requirement for MDIO Turnaround. The Turnaround time is filled by the DP83816 by inserting <10>. Figure 3-16 shows the timing relationship for a typical MII register write access.
3.12.5 Nibble-wide MII Data Interface
Clause 22 of the IEEE 802.3u specification defines the Media Independent Interface. This interface include separate dedicated receive and transmit busses. These two data buses, along with various control and indication signals, allow for the simultaneous exchange of data between the DP83816 and PHY(s).
28 www.national.com
3.0 Functional Description (Continued)
MDC
DP83816
MDIO
(STA)
Z
00011110000000
Idle Start
Opcode (Write)
PHY Address
(PHYAD = 0Ch)
Register Address
(00h = BMCR)
Figure 3-16 Typical MDC/MDIO Write Operation
The receive interface consists of a nibble wide data bus RXD[3:0], a receive error signal RXER, a receive data valid flag RXDV, and a receive clock RXCLK for synchronous transfer of the data. The receive clock can operate at 2.5 MHz to support 10 Mb/s operation modes or at 25 MHz to support 100 Mb/s operational modes.
The transmit interface consists of a nibble wide data bus TXD[3:0], a transmit enable control signal TXEN, and a transmit clock TXCLK which runs at 2.5 MHz or 25 MHz.
Additionally, the MII includes the carrier sense signal CRS, as well as a collision detect signal COL. The CRS signal asserts to indicate the reception of data from the network or as a function of transmit data in Half Duplex mode. The COL signal asserts as an indication of a collision which can occur during half-duplex operation when both a transmit and receive operation occur simultaneously.
3.12.6 Collision Detection
For Half Duplex, a 10BASE-T or 100BASE-TX collision is detected when the receive and transmit channels are active simultaneously. Collisions are reported by the COL signal on the MII.
If the PHY is transmitting in 10 Mb/s mode when a collision is detected, the collision is not reported until seven bits
ZZ
0 0 0 000 00000000
1000
TA
Register Data
have been received while in the collision state. This prevents a collision being reported incorrectly due to noise on the network. The COL signal remains set for the duration of the collision.
If a collision occurs during a receive operation, it is immediately reported by the COL signal.
When heartbeat is enabled (only applicable to 10 Mb/s operation), approximately 1µs after the transmission of each packet, a Signal Quality Error (SQE) signal of approximately 10 bit times is generated (internally) to indicate successful transmission. SQE is reported as a pulse on the COL signal of the MII.
3.12.7 Carrier Sense
Carrier Sense (CRS) is asserted due to receive activity, once valid data is detected, during 10 Mb/s operation. During 100 Mb/s operation CRS is asserted when a valid link (SD) and two non-contiguous zeros are detected.
For 10 or 100 Mb/s Half Duplex operation, CRS is asserted during either packet transmission or reception.
For 10 or 100 Mb/s Full Duplex operation, CRS is asserted only due to receive activity.
CRS is de-asserted following an end of packet.
Z
Idle
29 www.national.com
4.0 Register Set
4.1 Configuration Registers
The DP83816 implements a PCI version 2.2 configuration register space. This allows a PCI BIOS to "soft" configure the DP83816. Software Reset has no effect on configuration registers. Hardware Reset returns all configuration registers to their hardware reset state. For all unused registers, writes are ignored, and reads return 0.
Table 4-1 Configuration Register Map
Offset Tag Description Access
00h CFGID Configuration Identification Register RO
04h CFGCS Configuration Command and Status Register R/W
08h CFGRID Configuration Revision ID Register RO
0Ch CFGLAT Configuration Latency Timer Register RO
10h CFGIOA Configuration IO Base Address Register R/W
14h CFGMA Configuration Memory Address Register R/W
18h-28h Reserved (reads return zero)
2Ch CFGSID Configuration Subsystem Identification Register RO
30h CFGROM Boot ROM configuration register R/W
34h CAPPTR Capabilities Pointer Register RO
38h Reserved (reads return zero)
3Ch CFGINT Configuration Interrupt Select Register R/W
40h PMCAP Power Management Capabilities Register RO
44h PMCSR Power Management Control and Status Register R/W
48-FFh Reserved (reads return zero)
DP83816
4.1.1 Configuration Identification Register
This register identifies the DP83816 Controller to PCI system software.
Tag: CFGID Size: 32 bits Hard Reset: 0020100Bh
Offset: 00h Access: Read Only Soft Reset: Unchanged
Bit Bit Name Description
31-16 DEVID Device ID
This field is read-only and is set to the device ID assigned by National Semiconductor to the DP83816, which is 0020h.
15-0 VENID Vendor ID
This field is read-only and is set to a value of 100Bh which is National Semiconductor's PCI Vendor ID.
30 www.national.com
4.0 Register Set (Continued)
4.1.2 Configuration Command and Status Register
The CFGCS register has two parts. The upper 16-bits (31-16) are devoted to device status. A status bit is reset whenever the register is written, and the corresponding bit location is a 1. The lower 16-bits (15-0) are devoted to command and are used to configure and control the device.
Tag: CFGCS Size: 32 bits Hard Reset: 02900000h
Offset: 04h Access: Read Write Soft Reset: Unchanged
Bit Bit Name Description
31 DPERR Detected Parity Error
Refer to the description in the PCI V2.2 specification.
30 SSERR Signaled SERR
Refer to the description in the PCI V2.2 specification.
29 RMABT Received Master Abort
Refer to the description in the PCI V2.2 specification.
28 RTABT Received Target Abort
Refer to the description in the PCI V2.2 specification.
27 STABT Sent Target Abort
Refer to the description in the PCI V2.2 specification.
26-25 DSTIM DEVSELN Timing
This field will always be set to 01 indicating that DP83816 supports “medium” DEVSELN timing.
24 DPD Data Parity Detected
Refer to the description in the PCI V2.2 specification.
23 FBB Fast Back-to-Back Capable
DP83816 will set this bit to 1.
22-21 unused
(reads return 0)
20 NCPEN New Capabilities Enable
When set, this bit indicates that the Capabilities Pointer contains a valid value and new capabilities such as power management are supported. When clear, new capabilities (CAPPTR, PMCAP, PMCS) are disabled. This bit is loaded from a strap option, MD0 pin 132. A subsequent load of the configuration data from the EEPROM will overwrite any pre-existing value.
19-16 Unused
(reads return 0)
15-10 Unused
(reads return 0)
9 FBBEN Fast Back-to-Back Enable
Set to 1 by the PCI BIOS to enable the DP83816 to do Fast Back-to-Back transfers (FBB transfers as a master is not implemented in the current revision).
8 SERREN SERRN Enable
When SERREN and PERRSP are set, DP83816 will generate SERRN during target cycles when an address parity error is detected from the system. Also, when SERREN and PERRSP are set and CFG:PESEL is reset, master cycles detecting data parity errors will generate SERRN.
7 Unused
(reads return 0)
DP83816
31 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
6 PERRSP Parity Error Response
When set, DP83816 will assert PERRN on the detection of a data parity error when acting as the target, and will sample PERRN when acting as the initiator. Also, setting PERRSP allows SERREN to enable the assertion of SERRN. When reset, all address and data parity errors are ignored and neither SERRN nor PERRN are asserted.
5-3 Unused
(reads return 0)
2 BMEN Bus Master Enable
When set, DP83816 is allowed to act as a PCI bus master. When reset, DP83816 is prohibited from acting as a PCI bus master.
1 MSEN Memory Space Address
When set, DP83816 responds to memory space accesses. When reset, DP83816 ignores memory space accesses.
0 I/OSEN I/O Space Access
When set, DP83816 responds to I/O space accesses. When reset, DP83816 ignores I/O space accesses.
4.1.3 Configuration Revision ID Register
This register stores the silicon revision number, revision number of software interface specification and lets the configuration software know that it is an Ethernet controller in the class of network controllers.
DP83816
Tag: CFGRID Size: 32 bits Hard Reset: 02000000h
Offset: 08h Access: Read Only Soft Reset: Unchanged
Bit Bit Name Description
31-24 BASECL Base Class
Returns 02h which specifies a network controller.
23-16 SUBCL Sub Class
Returns 00h which specifies an Ethernet controller.
15-8 PROGIF Programming IF
Returns 00h which specifies the first release of the DP83816 Software Interface Specification.
7-0 REVID Silicon Revision
Returns 00h which specifies the silicon revision.
32 www.national.com
4.0 Register Set (Continued)
4.1.4 Configuration Latency Timer Register
This register gives status and controls such miscellaneous functions as BIST, Latency timer and Cache line size.
Tag: CFGLAT Size: 32 bits Hard Reset: 00000000h
Offset: 0Ch Access: Read Write Soft Reset: Unchanged
Bit Bit Name Description
31 BISTCAP BIST Capable
Reads will always return 0.
30 BISTEN BIST Enable
Reads will return a 0, writes are ignored.
29-16 Reserved
Reads will return a 0, writes are ignored.
15-8 LAT Latency Timer
Set by software to the number of PCI clocks that DP83816 may hold the PCI bus.
7-0 CLS Cache Line Size
Ignored by DP83816.
DP83816
DP83816 Bus Master Operations: Independent of cache line size, the DP83816 will use the following PCI commands for bus mastered transfers:
0110 - Mem Read for all read cycles, 0111 - Mem Write for all write cycles.
4.1.5 Configuration I/O Base Address Register
This register specifies the Base I/O address which is required to build an address map during configuration. It also specifies the number of bytes required as well as an indication that it can be mapped into I/O space.
Tag: CFGIOA Size: 32 bits Hard Reset: 00000001h
Offset: 10h Access: Read Write Soft Reset: Unchanged
Bit Bit Name Description
31-8 IOBASE Base I/O Address
This is set by software to the base I/O address for the Operational Register Map.
7-2 IOSIZE Size indication
Read back as 0. This allows the PCI bridge to determine that the DP83816 requires 256 bytes of I/O space.
1 Unused
(reads return 0).
0 IOIND I/O Space Indicator
Set to 1 by DP83816 to indicate that DP83816 is capable of being mapped into I/O space. Read Only.
33 www.national.com
4.0 Register Set (Continued)
4.1.6 Configuration Memory Address Register
This register specifies the Base Memory address which is required to build an address map during configuration. It also specifies the number of bytes required as well as an indication that it can be mapped into memory space.
Tag: CFGMA Size: 32 bits Hard Reset: 00000000h
Offset: 14h Access: Read Write Soft Reset: unchanged
Bit Bit Name Description
31-12 MEMBASE Memory Base Address
This is set by software to the base address for the Operational Register Map.
11- 4 MEMSIZE Memory Size
These bits return 0, which indicates that the DP83816 requires 4096 bytes of Memory Space (the minimum recommended allocation).
3 MEMPF Prefetchable
Set to 0 by DP83816. Read Only.
2-1 MEMLOC Location Selection
Set to 00 by DP83816. This indicates that the base register is 32-bits wide and can be placed anywhere in the 32-bit memory space. Read Only.
0 MEMIND Memory Space Indicator
Set to 0 by DP83816 to indicate that DP83816 is capable of being mapped into memory space. Read Only.
DP83816
4.1.7 Configuration Subsystem Identification Register
The CFGSID allows system software to distinguish between different subsystems based on the same PCI silicon. The values in this register can be loaded from the EEPROM if configuration is enabled.
Tag: CFGSID Size: 32 bits Hard Reset: 00000000h
Offset: 2Ch Access: Read Only Soft Reset: unchanged
Bit Bit Name Description
31-16 SDEVID Subsystem Device ID
Set to 0 by DP83816.
15-0 SVENID Subsystem Vendor ID
Set to 0 by DP83816.
34 www.national.com
4.0 Register Set (Continued)
4.1.8 Boot ROM Configuration Register
Tag: CFGROM Size: 32 bits Hard Reset: 00000000h
Offset: 30h Access: Read Write Soft Reset: unchanged
Bit Bit Name Description
31-16 ROMBASE ROM Base Address
Set to the base address for the boot ROM.
15-11 ROMSIZE ROM Size
Set to 0 indicating a requirement for 64K bytes of Boot ROM space. Read only.
10-1 unused
(reads return 0)
0 ROMEN ROM Enable
This is used by the PCI BIOS to enable accesses to boot ROM. This allows the DP83816 to share the address decode logic between the boot ROM and itself. The BIOS will copy the contents of the boot ROM to system RAM before executing it. Set to 1 enables the address decode for boot ROM disabling access to operational target registers.
DP83816
4.1.9 Capabilities Pointer Register
This register stores the capabilities linked list offset into the PCI configuration space.
Tag: CAPPTR Size: 32 bits Hard Reset: 00000040h
Offset: 34h Access: Read Only Soft Reset: unchanged
Bit Bit Name Description
31-8 unused
(reads return 0)
7-0 CLOFS Capabilities List Offset
Offset into PCI configuration space for the location of the first item in the Capabilities Linked List, set to 40h to point to the PMCAP register.
35 www.national.com
4.0 Register Set (Continued)
4.1.10 Configuration Interrupt Select Register
This register stores the interrupt line number as identified by the POST software that is connected to the interrupt controller as well as DP83816 desired settings for maximum latency and minimum grant. Max latency and Min latency can be loaded from the EEPROM.
Tag: CFGINT Size: 32 bits Hard Reset: 340b0100h
Offset: 3Ch Access: Read Write Soft Reset: unchanged
Bit Bit Name Description
31-24 MXLAT Maximum Latency
The DP83816 desired setting for Max Latency. The DP83816 will initialize this field to 52d (13 µsec). The value in this register can be loaded from the EEPROM.
23-16 MNGNT Minimum Grant
The DP83816 desired setting for Minimum Grant. The DP83816 will initialize this field to 11d (2.75 µsec). The value in this register can be loaded from the EEPROM.
15-8 IPIN Interrupt Pin
Read Only, always return 0000 0001 (INTA).
7-0 ILINE Interrupt Line
Set to which line on the interrupt controller that the DP83816's interrupt pin is connected to.
DP83816
4.1.11 Power Management Capabilities Register
This register provides information on the capabilities of the functions related to power management. This register also contains a pointer to the next item in the capabilities list and the capability ID for Power Management. This register is only visible if CFGCS[4] is set.
Tag: PMCAP Size: 32 bits Hard Reset: FF820001
Offset: 40h Access: Read Only Soft Reset: unchanged
Bit Bit Name Description
31-27 PMES PME Support
This 5 bit field indicates the power states in which DP83816 may assert PMEN. A 1 indicates PMEN is enabled for that state, a 0 indicates PMEN is inhibited in that state.
XXXX1 - PMEN can be asserted from state D0
XXX1X - PMEN can be asserted from state D1
XX1XX - PMEN can be asserted from state D2
X1XXX - PMEN can be asserted from state D3hot
1XXXX - PMEN can be asserted from state D3cold
The DP83816 will only report PME support for D3cold if auxiliary power is detected on the 3VAUX pin, in addition this value can be loaded from the EEPROM when in the D3cold state.
26 D2S D2 Support
This bit is set to a 1 when the DP83816 supports the D2 state.
25 D1S D1 Support
This bit is set to a 1 when the DP83816 supports the D1 state.
36 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
24-22 AUX_CURRENT Aux_Current
This 3 bit field reports the 3.3Vaux auxiliary current requirements for the PCI function.
If PMEN generation from D3cold is not supported by the function(PMCAP[31]), this field returns a value of "000b" when read.
Bit 3.3Vaux
24 23 22
1 1 0 320 mA
0 0 0 0 (self powered)
21 DSI Device Specific Initialization
This bit is set to 1 to indicate to the system that initialization of the DP83816 device is required (beyond the standard PCI configuration header) before the generic class device driver is able to use it. A 1 indicates that DP83816 requires a DSI sequence following transition to the D0 uninitialized state. This bit can be loaded from the EEPROM.
20 Reserved
(reads return 0)
19 PMEC PME Clock
Returns 0 to indicate PCI clock not needed for PMEN.
18-16 PMV Power Management Version
This bit field indicates compliance to a specific PM specification rev level. Currently set to 010b.
15-8 NLIPTR Next List Item Pointer
Offset into PCI configuration space for the location of the next item in the Capabilities Linked List. Returns 00h as no other capabilities are offered.
7-0 CAPID Capability ID
Always returns 01h for Power Management ID.
Max. Current Required
DP83816
4.1.12 Power Management Control and Status Register
This register contains PM control and status information.
Tag: PMCSR Size: 32 bits Hard Reset: 00000000h
Offset: 44h Access: Read Write Soft Reset: unchanged
Bit Bit Name Description
31-24 Reserved
(reads return 0)
23-16 BSE Bridge Support Extensions
unused (reads return 0)
15 PMESTS PME Status
Sticky bit which represents the state of the PME logic, regardless of the state of the PMEEN bit.
14-9 Reserved
(reads return 0)
8 PMEEN PME Enable
When set to 1, this bit enables the assertion of the PME function on the PMEN pin. When 0, the PMEN pin is forced to be inactive. This value can be loaded from the EEPROM.
7-2 Unused
(reads return 0)
1-0 PSTATE Power State
This 2 bit field is used to determine the current power state of DP83816, and to set a new power state.
00 - D0 10 - D2 01 - D1 11 - D3hot/cold
37 www.national.com
4.0 Register Set (Continued)
4.2 Operational Registers
The DP83816 provides the following set of operational registers mapped into PCI memory space or I/O space. Writes to reserved register locations are ignored. Reads to reserved register locations return undefined values.
Table 4-2 Operational Register Map
Offset Tag Description Access
MAC/BIU Registers
00h CR Command Register R/W
04h CFG Configuration Register R/W
08h MEAR EEPROM Access Register R/W
0Ch PTSCR PCI Test Control Register R/W
10h ISR Interrupt Status Register RO
14h IMR Interrupt Mask Register R/W
18h IER Interrupt Enable Register R/W
1Ch IHR Interrupt Holdoff Register R/W
20h TXDP Transmit Descriptor Pointer Register R/W
24h TXCFG Transmit Configuration Register R/W
28-2Ch Reserved Reserved
30h RXDP Receive Descriptor Pointer Register R/W
34h RXCFG Receive Configuration Register R/W
38 Reserved Reserved
3Ch CCSR CLKRUN Control/Status Register R/W
40h WCSR Wake on LAN Control/Status Register R/W
44h PCR Pause Control/Status Register R/W
48h RFCR Receive Filter/Match Control Register R/W
4Ch RFDR Receive Filter/Match Data Register R/W
50h BRAR Boot ROM Address R/W
54h BRDR Boot ROM Data R/W
58h SRR Silicon Revision Register RO
5Ch MIBC Management Information Base Control Register R/W
60-78h MIB Management Information Base Data Registers RO
7Ch Reserved Reserved
Internal Phy Registers
80h BMCR Basic Mode Control Register R/W
84h BMSR Basic Mode Status Register RO
88h PHYIDR1 PHY Identifier Register #1 RO
8Ch PHYIDR2 PHY Identifier Register #2 RO
90h ANAR Auto-Negotiation Advertisement Register R/W
94h ANLPAR Auto-Negotiation Link Partner Ability Register R/W
98h ANER Auto-Negotiation Expansion Register R/W
9Ch ANNPTR Auto-Negotiation Next Page TX R/W
A0-BCh Reserved Reserved
C0h PHYSTS PHY Status Register RO
C4h MICR MII Interrupt Control Register R/W
C8h MISR MII Interrupt Status Register R/W
CCh Reserved Reserved
D0h FCSCR False Carrier Sense Counter Register R/W
D4h RECR Receive Error Counter Register R/W
D8h PCSR 100 Mb/s PCS Configuration and Status Register R/W
DCh-E0h Reserved Reserved
E4h PHYCR PHY Control Register R/W
E8h TBTSCR 10Base-T Status/Control Register R/W
ECh-FCh Reserved Reserved
DP83816
38 www.national.com
4.0 Register Set (Continued)
4.2.1 Command Register
This register is used for issuing commands to DP83816. These commands are issued by setting the corresponding bits for the function. A global software reset along with individual reset and enable/disable for transmitter and receiver are provided here.
Tag: CR Size: 32 bits Hard Reset: 00000000h
Offset: 0000h Access: Read Write Soft Reset: 00000000h
Bit Bit Name Description
31-9 unused
8 RST Reset
Set to 1 to force the DP83816 to a soft reset state which disables the transmitter and receiver, reinitializes the FIFOs, and resets all affected registers to their soft reset state. This operation implies both a TXR and a RXR. This bit will read back a 1 during the reset operation, and be cleared to 0 by the hardware when the reset operation is complete. EEPROM configuration information is not loaded here.
7 SWI Software Interrupt
Setting this bit to a 1 forces the DP83816 to generate a hardware interrupt. This interrupt is mask-able via the IMR.
6 unused
5 RXR Receiver Reset
When set to a 1, this bit causes the current packet reception to be aborted, the receive data and status FIFOs to be flushed, and the receive state machine to enter the idle state (RXE goes to 0). This is a write-only bit and is always read back as 0.
4 TXR Transmit Reset
When set to a 1, this bit causes the current transmission to be aborted, the transmit data and status FIFOs to be flushed, and the transmit state machine to enter the idle state (TXE goes to 0). This is a write-only bit and is always read back as 0.
3 RXD Receiver Disable
Disable the receive state machine after any current packets in progress. When this operation has been completed the RXE bit will be cleared to 0. This is a write-only bit and is always read back as 0. The driver should not set both RXD and RXE in the same write, the RXE will be ignored, and RXD will have precedence.
2 RXE Receiver Enable
When set to a 1, and the receive state machine is idle, then the receive machine becomes active. This bit will read back as a 1 whenever the receive state machine is active. After initial power-up, software must insure that the receiver has completely reset before setting this bit (See ISR:RXRCMP).
1 TXD Transmit Disable
When set to a 1, halts the transmitter after the completion of the current packet. This is a write-only bit and is always read back as 0. The driver should not set both TXD and TXE in the same write, the TXE will be ignored, and TXD will have precedence.
0 TXE Transmit Enable
When set to a 1, and the transmit state machine is idle, then the transmit state machine becomes active. This bit will read back as a 1 whenever the transmit state machine is active. After initial power-up, software must insure that the transmitter has completely reset before setting this bit (See ISR:TXRCMP).
DP83816
39 www.national.com
4.0 Register Set (Continued)
4.2.2 Configuration and Media Status Register
This register allows configuration of a variety of device and phy options, and provides phy status information.
Tag: CFG Size: 32 bits Hard Reset: 00000000h
Offset: 0004h Access: Read Write Soft Reset: 00000000h
Bit Bit Name Description
31 LNKSTS Link Status
Link status of the internal phy. Asserted when link is good. RO
30 SPEED100 Speed 100 Mb/s
Speed 100 Mb/s indicator for internal phy. Asserted when speed is set or has negotiated to 100 Mb/s. De-asserted when speed has been set or negotiated to 10 Mb/s. RO
29 FDUP Full Duplex
Full Duplex indicator for internal phy. Asserted when duplex mode is set or has negotiated to FULL. De­asserted when duplex mode has been set or negotiated to HALF. RO
28 POL 10 Mb/s Polarity Indication
Twisted pair polarity indicator for internal phy. Asserted when operating and 10 Mb/s and the polarity has been detected as reversed. De-asserted when polarity is normal or phy is operating at 100 Mb/s. RO
27 ANEG_DN Auto-negotiation Done
Auto-negotiation done indicator from internal phy. Asserted when auto-negotiation process has completed or is not active. RO
26-24 unused
23-18 PHY_CFG Phy Configuration
Miscellaneous internal phy Power-On-Reset configuration control bits.
17 PINT_ACEN Phy Interrupt Auto Clear Enable
When set to a 1, this bit allows the phy interrupt source to be automatically cleared whenever the ISR is read. When this bit is 0, the phy interrupt source must be manually cleared via access of the phy registers. R/W
16 PAUSE_ADV Pause Advertise
This bit is loaded from EEPROM at power-up and is used to configure the internal phy to advertise the capability of 802.3x pause during auto-negotiation. Setting this bit to 1 will cause the pause function to be advertised if the phy has also been configured to advertise full duplex capability (See ANEG_SEL). R/W
15-13 ANEG_SEL Auto-negotiation Select
These bits are loaded from EEPROM at power-up and are used to define the default state of the internal phy auto-negotiation logic. R/W These bits are encoded as follows:
000 Auto-negotiation disabled, force 10 Mb/s half duplex
010 Auto-negotiation disabled, force 100 Mb/s half duplex
100 Auto-negotiation disabled, force 10 Mb/s full duplex
110 Auto-negotiation disabled, force 100 Mb/s full duplex
001 Auto-negotiation enabled, advertise 10 Mb/s half & full duplex
011 Auto-negotiation enabled, advertise 10/100 Mb/s half duplex
101 Auto-negotiation enabled, advertise 100 Mb/s half & full duplex
111 Auto-negotiation enabled, advertise 10/100 Mb/s half & full duplex
12 EXT_PHY External Phy Support
Act as a stand-alone MAC. When set, this bit enables the MII and disables the internal Phy (sets bit 9). R/W
11 Reserved
DP83816
40 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
10 PHY_RST Reset internal Phy
Asserts reset to internal phy. Can be used to cause phy to reload options from the CFG register. This bit does not self clear when set. R/W
9 PHY_DIS Disable internal Phy
When set to a 1, this bit forces the internal phy to its low-power state. R/W
8 EUPHCOMP DP83810 Descriptor Compatibility
When set, DP83816 will use DP83810 compatible (but single fragment) descriptor format. Descriptors are four 32-bit words in length, but the fragment count field is ignored. When clear, DP83816 will only fetch 3 32-bit words in descriptor fetches with the third word being the fragment pointer. R/W
7 REQALG PCI Bus Request Algorithm
Selects mode for making requests for the PCI bus. When set to 0 (default), DP83816 will use an aggressive Request scheme. When set to a 1, DP83816 will use a more conservative scheme. R/W
6 SB Single Back-off
Setting this bit to 1 forces the transmitter back-off state machine to always back-off for a single 802.3 slot time instead of following the 802.3 random back-off algorithm. A 0 (default) allows normal transmitter back-off operation. R/W
5 POW Program Out of Window Timer
This bit controls when the Out of Window collision timer begins counting its 512 bit slot time. A 0 causes the timer to start after the SFD is received. A 1 causes the timer to start after the first bit of the preamble is received. R/W
4 EXD Excessive Deferral Timer disable
Setting this bit to 1 will inhibit transmit errors due to excessive deferral. This will inhibit the setting of the ED status, and the logging of the TxExcessiveDeferral MIB counter. R/W
3 PESEL Parity Error Detection Action
This bit controls the assertion of SERR when a data parity error is detected while the DP83816 is acting as the bus master. When set, parity errors will not result in the assertion of SERR. When reset, parity errors will result in the assertion of SERR, indicating a system error. This bit should be set to a one by software if the driver can handle recovery from and reporting of data parity errors. R/W
2 BROM_DIS Disable Boot ROM interface
When set to 1, this bit inhibits the operation of the Boot ROM interface logic. R/W
1 Reserved
(reads return 0)
0 BEM Big Endian Mode
When set, DP83816 will perform bus-mastered data transfers in “big endian” mode. Note that access to register space is unaffected by the setting of this bit. R/W
DP83816
41 www.national.com
4.0 Register Set (Continued)
4.2.3 EEPROM Access Register
The EEPROM Access Register provides an interface for software access to the NMC9306 style EEPROM The default values given assume that the EEDO line has a pullup resistor to VDD.
Tag: MEAR Size: 32 bits Hard Reset: 00000002h
Offset: 0008h Access: Read Write Soft Reset: 00000002h
Bit Bit Name Description
31-7 unused
6 MDC MII Management Clock
Controls the value of the MDC pin. When set, the MDC pin is 1; when clear the MDC pin is 0. R/W
5 MDDIR MII Management Direction
Controls the direction of the MDIO pin. When set, DP83816 drives the MDIO pin. When clear MDIO bit reflects the current state of the MDIO pin. R/W
4 MDIO MII Management Data
Software access to the MDIO pin (see MDDIR above). R/W
3 EESEL EEPROM Chip Select
Controls the value of the EESEL pin. When set, the EESEL pin is 1; when clear the EESEL pin is 0. R/W
2 EECLK EEPROM Serial Clock
Controls the value of the EECLK pin. When set, the EECLK pin is 1; when clear the EECLK pin is 0. R/W
1 EEDO EEPROM Data Out
Returns the current state of the EEDO pin. When set, the EEDO pin is 1; when clear the EEDO pin is 0. RO
0 EEDI EEPROM Data In
Controls the value of the EEDI pin. R/W
DP83816
4.2.4 EEPROM Map
EEPROM
Address
0000h CFGSID[0:15] D008h
0001h CFGSID[16:31] 0400h
0002h CFGINT[24:31],CFGINT[16:23] 2CD0h
0003h CFGCS[20],PMCAP[31],PMCAP[21],PMCSR[8],
0004h SOPAS[1:16] 0000h
0005h SOPAS[17:32] 0000h
0006h SOPAS[33:47],PMATCH[0] 000Nh
0007h PMATCH[1:16] NNNNh
0008h PMATCH[17:32] NNNNh
0009h PMATCH[33:47],WCSR[0] NNNNh
000Ah WCSR[1:4],WCSR[9:10],RFCR[20],RFCR[22],
000Bh checksum value XX55
Configuration/Operation Register Bits
CFG[13:16],CFG[18:23],CR[2], SOPAS[0]
RFCR[27:31],000b (3 bits)
In the above table: N denotes the value is dependent on the ethernet MAC ID Number. X denotes the value is dependent on the checksum value.
Default Value
(16 bits)
CF82h
A098h
42 www.national.com
4.0 Register Set (Continued)
PMATCH[47:0] can be accessed via the combination of the RFCR (offset 0048h) and RFDR (offset 004Ch) registers. PMATCH holds the Ethernet address info. See Section 3.3.3.
The lower 8 bits of the checksum value should be 55h. For the upper 8 bits, add the top 8 data bits to the lower 8 data bits for each address. Sum the resultant 8 bit values for all addresses and then add 55h. Take the 2’s complement of the final sum. This 2’s complement number should be the upper 8 bits of the checksum value in the last address.
As an example, consider an EEPROM with two addresses. EEPROM address 0000h contains the data 1234h. EEPROM address 0001h contains the data 5678h.
12h + 34h = 46h 56h + 78h = CEh 46h + CEh + 55h = 69h The 2’s complement of 69h is 97h so the checksum value entered into EEPROM address 0002h would be 9755h.
4.2.5 PCI Test Control Register
Tag: PTSCR Size: 32 bits Hard Reset: 00000000h
Offset: 000Ch Access: Read Write Soft Reset: 00000000h
DP83816
Bit Bit Name Description
31-13 unused
12 Reserved for NSC internal use only.
Must be written as a 0 otherwise. R/W
11 Reserved
10 RBIST_RST SRAM BIST Reset
Setting this bit to 1 allows the SRAM BIST engine to be reset. R/W
9-8 Reserved for NSC internal use only.
Must be written as a 00 otherwise. R/W
7 RBIST_EN SRAM BIST Enable
Setting this bit to 1 starts the SRAM BIST engine. R/W
6 RBIST_DONE SRAM BIST Done
This bit is set to one when the BIST has completed its current test. It is cleared when either the BIST is active or disabled. RO
5 RBIST_RXFAIL RX FIFO BIST Fail
This bit is set to 1 if the SRAM BIST detects a failure in the RX FIFO SRAM. RO
4 RBIST_TXFAIL TX FIFO Fail
This bit is set to 1 if the SRAM BIST detects a failure in the TX FIFO SRAM. RO
3 RBIST_RXFFAIL RX Filter RAM BIST Fail
This bit is set to 1 if the SRAM BIST detects a failure in the RX Filter SRAM. RO
2 EELOAD_EN Enable EEPROM Load
This bit is set to a 1 to manually initiate a load of configuration information from EEPROM. A 1 is returned while the configuration load from EEPROM is active (approx. 1500 us). R/W
1 EEBIST_EN Enable EEPROM BIST
This bit is set to a 1 to initiate EEPROM BIST, which verifies the EEPROM data and checksum without reloading configuration values to the device. A 1 is returned while the EEPROM BIST is active. R/W
0 EEBIST_FAIL EE BIST Fail indication
This bit is set to a 1 upon completion of the EEPROM BIST (EEBIST_EN returns 0) if the BIST logic encountered an invalid checksum. RO
43 www.national.com
4.0 Register Set (Continued)
4.2.6 Interrupt Status Register
This register indicates the source of an interrupt when the INTA pin goes active. Enabling the corresponding bits in the Interrupt Mask Register (IMR) allows bits in this register to produce an interrupt. When an interrupt is active, one or more bits in this register are set to a “1”. The Interrupt Status Register reflects all current pending interrupts, regardless of the state of the corresponding mask bit in the IMR. Reading the ISR clears all interrupts. Writing to the ISR has no effect.
Tag: ISR Size: 32 bits Hard Reset: 03008000h
Offset: 0010h Access: Read Only Soft Reset: 03008000h
Bit Bit Name Description
31-26 Reserved
25 TXRCMP Transmit Reset Complete
Indicates that a requested transmit reset operation is complete.
24 RXRCMP Receive Reset Complete
Indicates that a requested receive reset operation is complete.
23 DPERR Detected Parity Error
This bit is set whenever CFGCS:DPERR is set, but cleared (like all other ISR bits) when the ISR register is read.
22 SSERR Signaled System Error
The DP83816 signaled a system error on the PCI bus.
21 RMABT Received Master Abort
The DP83816 received a master abort generated as a result of target not responding.
20 RTABT Received Target Abort
The DP83816 received a target abort on the PCI bus.
19-17 unused
16 RXSOVR Rx Status FIFO Overrun
Set when an overrun condition occurs on the Rx Status FIFO.
15 HIBERR High Bits Error Set
A logical OR of bits 25-16.
14 PHY Phy interrupt
Set to 1 when internal phy generates an interrupt.
13 PME Power Management Event
Set when WOL conditioned detected.
12 SWI Software Interrupt
Set whenever the SWI bit in the CR register is set.
11 MIB MIB Service
Set when one of the enabled management statistics has reached its interrupt threshold. (See Section 4.2.24)
10 TXURN Tx Underrun
Set when a transmit data FIFO underrun condition occurs.
9 TXIDLE Tx Idle
This event is signaled when the transmit state machine enters the idle state from a non-idle state. This will happen whenever the state machine encounters an "end-of-list" condition (NULL link field or a descriptor with OWN clear).
8 TXERR Tx Packet Error
This event is signaled after the last transmit descriptor in a failed transmission attempt has been updated with valid status.
DP83816
44 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
7 TXDESC Tx Descriptor
This event is signaled after a transmit descriptor when the INTR bit in the CMDSTS field has been updated.
6 TXOK Tx Packet OK
This event is signaled after the last transmit descriptor in a successful transmission attempt has been updated with valid status.
5 RXORN Rx Overrun
Set when a receive data FIFO overrun condition occurs.
4 RXIDLE Rx Idle
This event is signaled when the receive state machine enters the idle state from a running state. This will happen whenever the state machine encounters an "end-of-list" condition (NULL link field or a descriptor with OWN set).
3 RXEARLY Rx Early Threshold
Indicates that the initial Rx Drain Threshold has been met by the incoming packet, and the transfer of the number of bytes specified by the DRTH field in the RXCFG register has been completed by the receive DMA engine. This interrupt condition will occur only once per packet.
2 RXERR Rx Packet Error
This event is signaled after the last receive descriptor in a failed packet reception has been updated with valid status.
1 RXDESC Rx Descriptor
This event is signaled after a receive descriptor with the INTR bit set in the CMDSTS field has been updated.
0 RXOK Rx OK
Set by the receive state machine following the update of the last receive descriptor in a good packet.
DP83816
4.2.7 Interrupt Mask Register
This register masks the interrupts that can be generated from the ISR. Writing a “1” to the bit enables the corresponding interrupt. During a hardware reset, all mask bits are cleared. Setting a mask bit allows the corresponding bit in the ISR to cause an interrupt. ISR bits are always set to 1, however, if the condition is present, regardless of the state of the corresponding mask bit.
Tag: IMR Size: 32 bits Hard Reset: 00000000h
Offset: 0014h Access: Read Write Soft Reset: 00000000h
Bit Bit Name Description
31-26 unused
25 TXRCMP Transmit Reset Complete
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
24 RXRCMP Receive Reset Complete
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
23 DPERR Detected Parity Error
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
22 SSERR Signaled System Error
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
21 RMABT Received Master Abort
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
45 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
20 RTABT Received Target Abort
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
19-17 unused
16 RXSOVR Rx Status FIFO Overrun
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
15 HIERR High Bits Error
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
14 PHY Phy interrupt
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
13 PME Power Management Event
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
12 SWI Software Interrupt
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
11 MIB MIB Service
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
10 TXURN Tx Underrun
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
9 TXIDLE Tx Idle
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
8 TXERR Tx Packet Error
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
7 TXDESC Tx Descriptor
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
6 TXOK Tx Packet OK
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
5 RXORN Rx Overrun
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
4 RXIDLE Rx Idle
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
3 RXEARLY Rx Early Threshold
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
2 RXERR Rx Packet Error
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
1 RXDESC Rx Descriptor
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
0 RXOK Rx OK
When this bit is 0, the corresponding bit in the ISR will not cause an interrupt.
DP83816
46 www.national.com
4.0 Register Set (Continued)
4.2.8 Interrupt Enable Register
The Interrupt Enable Register controls the hardware INTR signal.
Tag: IER Size: 32 bits Hard Reset: 00000000h
Offset: 0018h Access: Read Write Soft Reset: 00000000h
Bit Bit Name Description
31-1 unused
0 IE Interrupt Enable
When set to 1, the hardware INTR signal is enabled. When set to 0, the hardware INTR signal will be masked, and no interrupts will be generated. The setting of this bit has no effect on the ISR or IMR. This provides the ability to disable the hardware interrupt to the host with a single access (eliminating the need for a read-modify-write cycle). The actual enabling of interrupts can be delayed based on the Interrupt Holdoff Register defined in the following section. If IE = 0, the interrupt holdoff timer will not start.
4.2.9 Interrupt Holdoff Register
The Interrupt Holdoff Register provides interrupt holdoff support. This allows interrupts to be delayed based on a programmable delay timer.
Tag: IHR Size: 32 bits Hard Reset: 00000000h
Offset: 001Ch Access: Read Write Soft Reset: 00000000h
DP83816
Bit Bit Name Description
31-9 unused
8 IHCTL Interrupt Holdoff Control
If this bit is set, the interrupt holdoff timer will start when the first interrupt event occurs and interrupts are enabled. When this bit is not set, the interrupt holdoff timer will start as soon as the timer is loaded and interrupts are enabled. In other words, when not set, the timer will delay the interrupt enable.
7-0 IH Interrupt Holdoff
The register contains a counter value for use in preventing interrupt assertion for a programmed amount of time. When the ISR is read, the interrupt holdoff timer is loaded with this value. It begins to count down to 0 based on the setting of the IHCTL bit. Once it reaches 0, interrupts will be enabled. The counter value is in units of 100 µs.
When Interrupts are enabled IE = 1, and IH contains a value other than 00h, IHCTL determines when the Interrupt Holdoff timer will begin its countdown as such:
IHCTL = 1: The timer does not begin until an interrupt event occurs.
The reporting of an interrupt event is delayed for a fixed amount of time from when the interrupt occurs.
IHCTL = 0: The timer begins immediately without waiting for an interrupt event.
The reporting of an interrupt event is delayed for a non-fixed amount of time from when the interrupt occurs.
When IH = 00h (default), there is no delay applied regardless of what IHCTL is set to.
47 www.national.com
4.0 Register Set (Continued)
4.2.10 Transmit Descriptor Pointer Register
This register points to the current Transmit Descriptor.
Tag: TXDP Size: 32 bits Hard Reset: 00000000h
Offset: 0020h Access: Read Write Soft Reset: 00000000h
Bit Bit Name Description
31-2 TXDP Transmit Descriptor Pointer
The current value of the transmit descriptor pointer. When the transmit state machine is idle, software must set TXDP to the address of a completed transmit descriptor. While the transmit state machine is active, TXDP will follow the state machine as it advances through a linked list of active descriptors. If the link field of the current transmit descriptor is NULL (signifying the end of the list), TXDP will not advance, but will remain on the current descriptor. Any subsequent writes to the TXE bit of the CR register will cause the transmit state machine to reread the link field of the current descriptor to check for new descriptors that may have been appended to the end of the list. Transmit descriptors must be aligned on an even 32-bit boundary in host memory (A1-A0 must be 0).
1-0 unused
4.2.11 Transmit Configuration Register
This register defines the Transmit Configuration for DP83816. It controls such functions as Loopback, Heartbeat, Auto Transmit Padding, programmable Interframe Gap, Fill & Drain Thresholds, and maximum DMA burst size.
DP83816
Tag: TXCFG Size: 32 bits Hard Reset: 00040102h
Offset: 0024h Access: Read Write Soft Reset: 00040102h
Bit Bit Name Description
31 CSI Carrier Sense Ignore
Setting this bit to 1 causes the transmitter to ignore carrier sense activity, which inhibits reporting of CRS status to the transmit status register. When this bit is 0 (default), the transmitter will monitor the CRS signal during transmission and reflect valid status in the transmit status register and MIB counter block. This bit must be set to enable full-duplex operation.
30 HBI HeartBeat Ignore
Setting this bit to 1 causes the transmitter to ignore the heartbeat (CD) pulse which follows the packet transmission and inhibits logging of TXSQEErrors in the MIB counter block. When this bit is set to 0 (default), the transmitter will monitor the heartbeat pulse and log TXSQEErrors to the MIB counter block. This bit must be set to enable full-duplex operation
29 MLB MAC Loopback
Setting this bit to a 1 places the DP83816 MAC into a loopback state which routes all transmit traffic to the receiver, and disables the transmit and receive interfaces of the MII. A 0 in this bit allows normal MAC operation. The transmitter and receiver must be disabled before enabling the loopback mode. (Packets received during MLB mode will reflect loopback status in the receive descriptor’s
28 ATP Automatic Transmit Padding
Setting this bit to 1 causes the MAC to automatically pad small (runt) transmit packets to the Ethernet minimum size of 64 bytes. This allows driver software to transfer only actual packet data. Setting this bit to 0 disables the automatic padding function, forcing software to control runt padding.
cmdsts.LBP field.)
48 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
27-26 IFG Interframe Gap Time
This field allows the user to adjust the interframe gap time below the standard 9.6 960ns @100 Mb/s. The time can be programmed from 9.6 @100 Mb/s. Note that any value other than zero may violate the IEEE 802.3 standard. The formula for the interframe gap is:
9.6µs - 0.4(IFG[1:0]) µs @10 Mb/s and
960ns - 40(IFG[1:0])ns @100 Mb/s
25-24 Reserved for NSC internal use only.
Must be written as a 00 otherwise. R/W
23 ECRETRY Excessive Collision Retry Enable
This bit enables automatic retries of excessive collisions. If set, the transmitter will retry the packet up to 4 excessive collision counts, for a total of 64 attempts. If the packet still does not complete successfully, then the transmission will be aborted after the 64th attempt. If this bit is not set, then the transmit will be aborted after the 16th attempt. Note that setting this bit will change how collisions are reported in the status field of the transmit descriptor.
22-20 MXDMA Max DMA Burst Size per Tx DMA Burst
This field sets the maximum size of transmit DMA data bursts according to the following table:
000 = 128 32-bit words (512 bytes)
001 = 1 32-bit word (4 bytes)
010 = 2 32-bit words (8 bytes)
011 = 4 32-bit words (16 bytes)
100 = 8 32-bit words (32 bytes)
101 = 16 32-bit words (64 bytes)
110 = 32 32-bit words (128 bytes)
111 = 64 32-bit words (256 bytes)
NOTE: The MXDMA setting value MUST not be greater than the TXCFG:FLTH (Tx Fill Threshold) value.
19 unused
18 Reserved for NSC internal use only.
Must be set to 1. Setting this bit to 0 selects a non-standard back-off algorithm that could increase the likelihood of excessive collisions.
17-14 unused
13-8 FLTH Tx Fill Threshold
Specifies the fill threshold in units of 32 bytes. When the number of available bytes in the transmit FIFO reaches this level, the transmit bus master state machine will be allowed to request the PCI bus for transmit packet fragment reads. A value of 0 in this field will produce unexpected results and must not be used.
Note: The FLTH value should be greater than the TXCFG:MXDMA value, but less than (txFIFOsize ­TXCFG:DRTH). In order to prevent FIFO pointer overlap internal to the device, the sum of the FLTH and TXCFG:DRTH values should not exceed 2016 Bytes.
7-6 unused
5-0 DRTH Tx Drain Threshold
Specifies the drain threshold in units of 32 bytes. When the number of bytes in the FIFO reaches this level (or the FIFO contains at least one complete packet) the MAC transmit state machine will begin the transmission of a packet.
NOTE: In order to prevent a deadlock condition from occurring, the DRTH value should always be less than (txFIFOsize - TXCFG:FLTH). A value of 0 in this field will produce unexpected results and must not be used. Also, in order to prevent FIFO pointer overlap internal to the device, the sum of the DRTH and TXCFG:FLTH values should not exceed 2016 Bytes.
DP83816
µs @10 Mb/s and
µs to 8.4µs @10 Mb/s and 960ns to 840ns
49 www.national.com
4.0 Register Set (Continued)
4.2.12 Receive Descriptor Pointer Register
This register points to the current Receive Descriptor.
Tag: RXDP Size: 32 bits Hard Reset: 00000000h
Offset: 0030h Access: Read Write Soft Reset: 00000000h
Bit Bit Name Description
31-2 RXDP Receive Descriptor Pointer
The current value of the receive descriptor pointer. When the receive state machine is idle, software must set RXDP to the address of an available receive descriptor. While the receive state machine is active, RXDP will follow the state machine as it advances through a linked list of available descriptors. If the link field of the current receive descriptor is NULL (signifying the end of the list), RXDP will not advance, but will remain on the current descriptor. Any subsequent writes to the RXE bit of the CR register will cause the receive state machine to reread the link field of the current descriptor to check for new descriptors that may have been appended to the end of the list. Software should not write to this register unless the receive state machine is idle. Receive descriptors must be aligned on 32-bit boundaries (A1-A0 must be zero). A 0 written to RXDP followed by a subsequent write to RXE will cause the receiver to enter silent RX mode, for use during WOL. In this mode packets will be received and buffered in FIFO, but no DMA to system memory will occur. The packet data may be recovered from the FIFO by writing a valid descriptor address to RXDP and then strobing RXE.
1-0 unused
DP83816
50 www.national.com
4.0 Register Set (Continued)
4.2.13 Receive Configuration Register
This register is used to set the receive configuration for DP83816. Receive properties such as accepting error packets, runt packets, setting the receive drain threshold etc. are controlled here.
Tag: RXCFG Size: 32 bits Hard Reset: 00000002h
Offset: 0034h Access: Read Write Soft Reset: 00000002h
Bit Bit Name Description
31 AEP Accept Errored Packets
When set to 1, all packets with CRC, alignment, and/or collision errors will be accepted. When set to 0, all packets with CRC, alignment, and/or collision errors will be rejected if possible. Note that depending on the type of error, some packets may be received with errors, regardless of the setting of AEP. These errors will be indicated in the CMDSTS field of the last descriptor in the packet.
30 ARP Accept Runt Packets
When set to 1, all packets under 64 bytes in length without errors are accepted. When this bit is 0, all packets less than 64 bytes in length will be rejected if possible.
29 unused
28 ATX Accept Transmit Packets
When set to 1, data received simultaneously to a local transmission (such as during a PMD loopback or full duplex operation) will be accepted as valid received data. Additionally, when set to 1, the receiver will ignore collision activity. When set to 0 (default), all data receive simultaneous to a local transmit will be rejected. This bit must be set to 1 for PMD loopback and full duplex operation.
27 ALP Accept Long Packets
When set to 1, all packets > 1518 bytes in length and <= 2046 bytes will be treated as normal receive packets, and will not be tagged as long or error packets. All packets > 2046 bytes in length will be truncated at 2046 bytes and either rejected from the FIFO, or tagged as long packets. Care must be taken when accepting long packets to ensure that buffers provided are of adequate length. When ALP is set to 0, packets larger than 1518 bytes (CRC inclusive) will be truncated at 1514 bytes, and rejected if possible.
26 unused
25-23 unused
Writes are ignored, reads return 000b.
22-20 MXDMA Max DMA Burst Size per Rx DMA Burst
This field sets the maximum size of receive DMA data bursts according to the following table:
000 = 128 32-bit words (512 bytes)
001 = 1 32-bit word (4 bytes)
010 = 2 32-bit words (8 bytes)
011 = 4 32-bit words (16 bytes)
100 = 8 32-bit words (32 bytes)
101 = 16 32-bit words (64 bytes)
110 = 32 32-bit words (128 bytes)
111 = 64 32-bit words (256 bytes)
19-6 unused
DP83816
51 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
5-1 DRTH Rx Drain Threshold
Specifies the drain threshold in units of 8 bytes. When the number of bytes in the receive FIFO reaches this value (times 8), or the FIFO contains a complete packet, the receive bus master state machine will begin the transfer of data from the FIFO to host memory. Care must be taken when setting DRTH to a value lower than the number of bytes needed to determine if packet should be accepted or rejected. In this case, the packet might be rejected after the bus master operation to begin transferring the packet into memory has begun. When this occurs, neither the OK bit or any error status bit in the descriptor’s cmdsts will be set. A value of 0 is illegal, and the results are undefined.
This value is also used to compare with the accumulated packet length for early receive indication. When the accumulated packet length meets or exceeds the DRTH value, the RXEARLY interrupt condition is generated.
0 Reserved
4.2.14 CLKRUN Control/Status Register
This register mirrors the read/write control of the PMESTS and PMEEN from the PCI Configuration register PMCSR and controls whether the chip is in the CLKRUNN or PMEN mode.
Tag: CCSR Size: 32 bits Hard Reset: 00000000h
Offset: 003Ch Access: Read Write Soft Reset: unchanged
DP83816
Bit Bit Name Description
31-16 reserved
(reads return 0)
15 PMESTS PME Status
Sticky bit which represents the state of the PME/CLKRUN logic, regardless of the state of the PMEEN bit. Mirrored from PCI configuration register PMCSR. Writing a 1 to this bit clears it.
14-9 reserved
(reads return 0)
8 PMEEN PME Enable
When set to 1, this bit enables the assertion of the PMEN/CLKRUNN pin. When 0, the PMEN/CLKRUNN pin is forced to be inactive. This value can be loaded from the EEPROM. Mirrored from PCI configuration register PMCSR.
7-1 unused
(reads return 0)
0 CLKRUN_EN Clkrun Enable
When set to 1, this bit enables the CLKRUNN functionality of the PMEN/CLKRUNN pin. When 0, normal PMEN functionality is active.
52 www.national.com
4.0 Register Set (Continued)
4.2.14.1 CLKRUNN Function
CLKRUNN is a dual-function optional signal. It is used by the central PCI clock resource to indicate clock status (i.e. PCI clock running normally or slowed/stopped), and it is used by PCI devices to request that the central resource restart the PCI clock or keep it running normally.
In the DP83816, CLKRUNN shares a pin with PMEN (pin
59). This means the chip cannot be simultaneously PCI Power Management and PCI Mobile Design Guide­compliant; however, it is unlikely that a system would use both of these functions simultaneously. The function of the PMEN/CLKRUNN pin is selected with the CLKRUN_EN bit of CCSR.
DP83816
Situation 1 is a “clock continue” event and can occur if the DP83816 has not completed a pending packet transmit or receive. Situation 2 is a “clock start” event and can occur if the DP83816 has been programmed to a WOL state and it receives a wake packet, or the PCI clock has simply been stopped and the receiver has data ready to DMA. In either of these situations, the DP83816 asserts CLKRUNN until it detects two rising edges of the PCI clock; it then releases assertion of CLKRUNN. At this point, the central resource is driving CLKRUNN low, and cannot drive it high again until at least four rising edges of the PCI clock have occurred since the initial CLKRUNN assertion by the DP83816. Also in either situation, the DP83816 must have detected CLKRUNN de-asserted for two consecutive rising edges of the PCI clock before it is allowed to assert CLKRUNN.
CCSR bits 15 and 8 (PMESTS and PMEEN) are mirrored from PCI configuration space to allow them to be accessed by software. The functionality of these bits is the same as in the PCI configuration register PMCSR.
As an output, CLKRUNN is open-drain like PMEN, i.e. it can only drive low. CLKRUNN is an input unless one of the following two conditions occurs:
1. the system drives CLKRUNN high but the DP83816 is not ready for the PCI clock to be stopped or
2. the PCI clock is stopped or slowed (CLKRUNN is pulled high by the system) and the DP83816 requires the use of the PCI bus.
NOTES: * If a clock start or continue event has completed but a PCI
interrupt has not been serviced yet, the CLKRUN logic will not prevent the system from stopping the PCI clock.
* If PMEEN is not set, the DP83816 cannot assert CLKRUNN to request a clock start or continue. In this case, if the system is going to stop the PCI clock, software must shut down the internal PHY to prevent receive errors.
* If another CLKRUN-enabled device in the system encounters a clock start or continue event, the cycle of assertions and de-assertions of CLKRUNN will cause the DP83816 clock mux to switch the clock to the RX block back and forth between the PCI clock and the X1 clock until the event completes.
53 www.national.com
4.0 Register Set (Continued)
4.2.15 Wake Command/Status Register
The WCSR register is used to configure/control and monitor the DP83816 Wake On LAN logic. The Wake On LAN logic is used to monitor the incoming packet stream while in a low-power state, and provide a wake event to the system if the desired packet type, contents, or Link change are detected.
Tag: WCSR Size: 32 bits Hard Reset: 00000000h
Offset: 0040h Access: Read Write Soft Reset: 00000000h
Bit Bit Name Description
31 MPR Magic Packet Received
Set to 1 if a Magic Packet has been detected and the WKMAG bit is set. RO, cleared on read.
30 PATM 3 Pattern 3 match
Associated bit set to 1 if a pattern 3 match is detected and the WKPAT3 bit is set. RO, cleared on read.
29 PATM 2 Pattern 2 match
Associated bit set to 1 if a pattern 2 match is detected and the WKPAT2 bit is set. RO, cleared on read.
28 PATM 1 Pattern 1 match
Associated bit set to 1 if a pattern 1 match is detected and the WKPAT1 bit is set. RO, cleared on read.
27 PATM 0 Pattern 0 match
Associated bit set to 1 if a pattern 0 match is detected and the WKPAT0 bit is set. RO, cleared on read.
26 ARPR ARP Received
Set to 1 if an ARP packet has been detected and the WKARP bit is set. RO, cleared on read.
25 BCASTR Broadcast Received
Set to 1 if a broadcast packet has been detected and the WKBCP bit is set. RO, cleared on read.
24 MCASTR Multicast Received
Set to 1 if a multicast packet has been detected and the WKMCP bit is set. RO, cleared on read.
23 UCASTR Unicast Received
Set to 1 if a unicast packet has been detected the WKUCP bit is set. RO, cleared on read.
22 PHYINT Phy Interrupt
Set to 1 if a Phy interrupt was detected and the WKPHY bit is set. RO, cleared on read.
21 Reserved Reserved
RO, cleared on read.
20 SOHACK SecureOn Hack Attempt
Set to 1 if the MPSOE and WKMAG bits are set, and a Magic Packet is receive with an invalid SecureOn password value. RO, Cleared on read.
19-11 unused
returns 0
10 MPSOE Magic Packet SecureOn Enable
Enable Magic Packet SecureOn feature. Only applicable when bit 9 is set. R/W
9 WKMAG Wake on Magic Packet
Enable wake on Magic Packet detection. R/W
8 WKPAT 3 Wake on Pattern 3 match
Enable wake on match of pattern 3. R/W
7 WKPAT 2 Wake on Pattern 2 match
Enable wake on match of pattern 2. R/W
6 WKPAT 1 Wake on Pattern 1 match
Enable wake on match of pattern 1. R/W
DP83816
54 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
5 WKPAT 0 Wake on Pattern 0 match
Enable wake on match of pattern 0. R/W
4 WKARP Wake on ARP
Enable wake on ARP packet detection. R/W
3 WKBCP Wake on Broadcast
Enable wake on broadcast packet detection. R/W
2 WKMCP Wake on Multicast
Enable wake on multicast packet detection. R/W
1 WKUCP Wake on Unicast
Enable wake on unicast packet detection. R/W
0 WKPHY Wake on Phy Interrupt
Enable wake on Phy Interrupt. The Phy interrupt can be programmed for Link Change and a variety of other Physical Layer events. R/W
4.2.15.1 Wake on LAN
The Wake on LAN logic provides several mechanisms for bringing the DP83816 out of a low-power state. Wake on ARP, Wake on Broadcast, Wake on Multicast Hash and Wake on Phy Interrupt are enabled by setting the corresponding bit in the Wake Command/Status Register, WCSR. Before the hardware is programmed to a low power state, the software must write a null receive descriptor pointer to the Receive Descriptor Pointer Register (RXDP) to ensure wake packets will be buffered in the RX fifo. Please refer to the description of the RXDP register for this procedure.
When a qualifying packet is received, the Wake on LAN logic generates a Wake event and pulses the PMEN PCI signal to request a Power Management state change. The software must then bring the hardware out of low power mode and, if the Power Management state was D3hot, reinitialize Configuration Register space. A Wake interrupt can also be generated which alerts the software that a Wake event has occurred and a packet was received. The software must then write a valid receive descriptor pointer to RXDP. The incoming packet can then be transferred into
host memory for processing. Note that the wake packet is retained for processing - this is a feature of the DP83816. In addition to the above Wake on LAN features, DP83816 also provides Wake on Pattern Matching, Wake on DA match and Wake on Magic Packet
Wake on Pattern Matching
Wake on Pattern Matching is an extension of the Pattern Matching feature provided by the Receive Filter Logic. When one or more of the Wake on Pattern Match bits are set in the WCSR, a packet will generate a wake event if it matches the associated pattern buffer. The pattern count and the pattern buffer memory are accessed in the same way as in Pattern Matching for packet acceptance. The minimum pattern count is 2 bytes and the maximum pattern count is 64 bytes for patterns 0 and 1, and 128 bytes for patterns 2 and 3. Packets are compared on a byte by byte basis and bytes may be masked in pattern memory, thus allowing for don’t cares. Refer to Section 4.2.19 Receive Filter Logic for programming examples.
DP83816
.
55 www.national.com
4.0 Register Set (Continued)
4.2.16 Pause Control/Status Register
The PCR register is used to control and monitor the DP83816 Pause Frame reception logic. The Pause Frame reception Logic is used to accept 802.3x Pause Frames, extract the pause length value, and initiate a TX MAC pause interval of the specified number of slot times.
Tag: PCR Size: 32 bits Hard Reset: 00000000h
Offset: 0044h Access: Read Write Soft Reset: 00000000h
Bit Bit Name Description
31 PSEN Pause Enable
Manually enables reception of 802.3x pause frames This bit is ORed with the PSNEG bit to enable pause reception. If pause reception has been enabled via PSEN bit (PSEN=1), setting this bit to 0 will cause any active pause interval to be terminated. R/W
30 PS_MCAST Pause on Multicast
When set to 1, this bit enables reception of 802.3x pause frames which use the 802.3x designated multicast address in the DA (01-80-C2-00-00-01). When this mode is enabled, the RX filter logic performs a perfect match on the above multicast address. No other address filtration modes (including multicast hash) are required for pause frame reception. R/W
29 PS_DA Pause on DA
When set to 1, this bit enables reception of a pause frame based on a DA match with either the perfect match register, or one of the pattern match buffers. R/W
28-24 unused
returns 0
23 PS_ACT Pause Active
This bit is set to a 1 when the TX MAC logic is actively timing a pause interval. RO
22 PS_RCVD Pause Frame Received
This bit is set to a 1 when a pause frame has been received. This bit will remain set until the TX MAC has completed the pause interval. RO
21 PSNEG Pause Negotiated
Status bit indicating that the 802.3x pause function has been enabled via auto-negotiation. This bit will only be set if DP83816 advertises pause capable by setting bit 16 in the CFG register. RO
20-17 unused
returns 0
16 MLD_EN Manual Load Enable
Setting this bit to a 1 will cause the value of bits 15-0 to be written to the pause count register. This write operation causes pause count interval to be manually initiated. This bit is not sticky, and reads will always return 0. WO
15-0 PAUSE_CNT Pause Counter Value
READ: These bits represent the current real-time value of the TX MAC pause counter register.
WRITE: If no pause count interval is in progress (PS_RCVD=0, PS_ACT=0), and MLD_EN=1 this value is written to the pause count register, and causes pause count interval to be manually initiated.
DP83816
56 www.national.com
4.0 Register Set (Continued)
4.2.17 Receive Filter/Match Control Register
The RFCR register is used to control and configure the DP83816 Receive Filter Control logic. The Receive Filter Control Logic is used to configure destination address filtering of incoming packets.
Tag: RFCR Size: 32 bits Hard Reset: 00000000h
Offset: 0048h Access: Read Write Soft Reset: 00000000h
Bit Bit Name Description
31 RFEN Rx Filter Enable
When this bit is set to 1, the Rx Filter is enabled to qualify incoming packets. When set to a 0, receive packet filtering is disabled (i.e. all receive packets are rejected). This bit must be 0 for the other bits in this register to be configured.
30 AAB Accept All Broadcast
When set to a 1, this bit causes all broadcast address packets to be accepted. When set to 0, no broadcast address packets will be accepted.
29 AAM Accept All Multicast
When set to a 1, this bit causes all multicast address packets to be accepted. When set to 0, multicast destination addresses must have the appropriate bit set in the multicast hash table mask in order for the packet to be accepted.
28 AAU Accept All Unicast
When set to a 1, this bit causes all unicast address packets to be accepted. When set to 0, the destination address must match the node address value specified through some other means in order for the packet to be accepted.
27 APM Accept on Perfect Match
When set to 1, this bit allows the perfect match register to be used to compare against the DA for packet acceptance. When this bit is 0, the perfect match register contents will not be used for DA comparison.
26-23 APAT Accept on Pattern Match
When one or more of these bits is set to 1, a packet will be accepted if the first n bytes (n is the value defined in the associated pattern count register) match the associated pattern buffer memory contents. When a bit is set to 0, the associated pattern buffer will not be used for packet acceptance.
22 AARP Accept ARP Packets
When set to 1, this bit allows all ARP packets (packets with a TYPE/LEN field set to 806h) to be accepted, regardless of the DA value. When set to 0, ARP packets are treated as normal packets and must meet other DA match criteria for acceptance.
21 MHEN Multicast Hash Enable
When set to 1, this bit allows hash table comparison for multicast addresses, i.e. a hash table hit for a multicast addressed packet will be accepted. When set to 0, multicast hash hits will not be used for packet acceptance.
20 UHEN Unicast Hash Enable
When set to 1, this bit allows hash table comparison for unicast addresses, i.e. a hash table hit for a unicast addressed packet will be accepted. When set to 0, unicast hash hits will not be used for packet acceptance.
19 ULM U/L bit Mask
When set to 1, this bit will cause the U/L bit (2nd MSb) of the DA to be ignored during comparison with the perfect match register.
18-10 Unused
returns 0
DP83816
57 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
9-0 RFADDR Receive Filter Extended Register Address
Selects which internal receive filter register is accessible via RFDR:
Perfect Match Register (PMATCH)
000h - PMATCH octets 1-0 002h - PMATCH octets 3-2 004h - PMATCH octets 5-4
Pattern Count Registers (PCOUNT)
006h - PCOUNT1, PCOUNT0 008h - PCOUNT3, PCOUNT2
SecureOn Password Register (SOPAS)
00Ah - SOPAS octets 1-0 00Ch - SOPAS octets 3-2 00Eh - SOPAS octets 5-4
Filter Memory
200h-3FE - Rx filter memory (Hash table/pattern buffers)
4.2.18 Receive Filter/Match Data Register
The RFDR register is used for reading from and writing to the internal receive filter registers, the pattern buffer memory, and the hash table memory.
.
Tag: RFDR Size: 32 bits Hard Reset: 00000000h
Offset: 004Ch Access: Read Write Soft Reset: 00000000h
DP83816
Bit Bit Name Description
31-18 unused
17-16 BMASK Byte mask
Used as byte mask values for pattern match template data.
15-0 RFDATA Receive Filter Data
58 www.national.com
4.0 Register Set (Continued)
4.2.19 Receive Filter Logic
The Receive Filter Logic supports a variety of techniques for qualifying incoming packets. The most basic filtering options include Accept All Broadcast, Accept All Multicast and Accept All Unicast packets. These options are enabled by setting the corresponding bit in the Receive Filter Control Register, RFCR. Accept on Perfect Match, Accept on Pattern Match, Accept on Multicast Hash and Accept on Unicast Hash are more robust in their filtering capabilities, but require additional programming of the Receive Filter registers and the internal filter RAM.
Accept on Perfect Match
When enabled, the Perfect Match Register is used to compare against the DA for packet acceptance. The Perfect Match Register is a 6-byte register accessed indirectly through the RFCR. The address of the internal receive filter register to be accessed is programmed through bits 8:0 of the RFCR. The Receive Filter Data Register, RFDR, is used for reading/writing the actual data.
RX Filter Address: 000h - Perfect Match octets 1-0
002h - Perfect Match octets 3-2 004h - Perfect Match octets 5-4
Octet 0 of the Perfect Match Register corresponds to the first octet of the packet as it appears on the wire. Octet 5 corresponds to the last octet of the DA as it appears on the wire.
The following steps are required to program the RFCR to accept packets on a perfect match of the DA.
Example: Destination Address of 08-00-17-07-28-55
DP83816
Accept on Pattern Match
The Receive Filter Logic provides access to 4 separate internal RAM-based pattern buffers to be used as additional perfect match address registers. Pattern buffers 0 and 1 are 64 bytes deep, allowing perfect match on the first 64 bytes of a packet, and pattern buffers 2 and 3 are 128 bytes deep, allowing perfect match on the first 128 bytes of a packet.
When one or more of the Pattern Match enable bits are set in the RFCR, a packet will be accepted if it matches the associated pattern buffer. As indicated above, the pattern buffers are 64 and 128 bytes deep organized as 32 or 64 words, where a word is 18 bits. Bits 17 and 18 of a respective word are mask bits for byte 0 and byte 1 of the 16-bit data word (bits 15:0). An incoming packet is compared to each enabled pattern buffer on a byte by byte basis for a specified count. Masking a pattern byte results in a byte match regardless of its value (a don’t care). A count value must be programmed for each pattern buffer to be used for comparison. The minimum valid count is 2 (2 bytes) and the maximum valid count is 32 for pattern buffers 0 and 1, and 64 for pattern buffers 2 and 3. The pattern count registers are internal receive filter registers accessed through the RFCR and the RFDR The Receive Filter memory is also accessed through the RFCR and the RFDR. A memory map of the internal pattern RAM is shown in Figure 4-1.
iow l $RFCR (0000) perfect match register, octets 1-0 iow l $RFDR (0008) write address, octets 1-0 iow l $RFCR (0002) perfect match register, octets 3-2 iow l $RFDR (0717) write address, octets 3-2 iow l $RFCR (0004) perfect match register, octets 5-4 iow l $RFDR (5528) write address, octets 5-4 iow l $RFDR
($RFEN|$APM) enable filtering, perfect match
59 www.national.com
4.0 Register Set (Continued)
Pattern3Word7F byte1 byte0 3FE
Pattern2Word7F byte1 byte0 3FC
Pattern3Word7E byte1 byte0 3FA
Pattern2Word7E byte1 byte0 3F8
DP83816
Byte1
Mask Bit
Byte0
Mask Bit
....................
Pattern3Word1 byte1 byte0 306
Pattern2Word1 byte1 byte0 304
Pattern3Word0 byte1 byte0 302
Pattern2Word0 byte1 byte0 300
Pattern1Word3F byte1 byte0 2FE
Pattern0Word3F byte1 byte0 2FC
Pattern1Word3E byte1 byte0 2FA
Pattern0Word3E byte1 byte0 2F8
....................
Pattern1Word1 byte1 byte0 286
Pattern0Word1 byte1 byte0 284
Pattern1Word0 byte1 byte0 282
Pattern0Word0 byte1 byte0 280
Bit# 17 16 15
....................
....................
....................
....................
87 0
....................
....................
Figure 4-1 Pattern Buffer Memory - 180h words (word = 18bits)
60 www.national.com
4.0 Register Set (Continued)
Example: Pattern match on the following destination addresses:
02-00-03-01-04-02 12-10-13-11-14-12 22-20-23-21-24-22 32-30-33-31-34-32
set $PATBUF01 = 280 set $PATBUF23 = 300
# write counts iow l $RFCR (0006) # pattern count registers 1, 0 iow l $RFDR (0406) # count 1 = 4, count 0= 6 iow l $RFCR (0008) # pattern count registers 3, 2 iow l $RFDR (0406) # count 3 = 4, count 2 = 6
# write data pattern into buffer 0 iow l $RFCR ($PATBUF01) iow l $RFDR (0002) iow l $RFCR ($PATBUF01 + 4) iow l $RFDR (0103) iow l $RFCR ($PATBUF01 + 8) iow l $RFDR (0204) # write data pattern into buffer 1 iow l $RFCR ($PATBUF01 + 2) iow l $RFDR (1012) iow l $RFCR ($PATBUF01 + 6) iow l $RFDR (1113) iow l $RFCR ($PATBUF01 + a) iow l $RFDR (1214) # write data pattern into buffer 2 iow l $RFCR ($PATBUF23) iow l $RFDR (2022) iow l $RFCR ($PATBUF23 + 4) iow l $RFDR (2123) iow l $RFCR ($PATBUF23 + 8) iow l $RFDR (2224) # write data pattern into buffer 3 iow l $RFCR ($PATBUF23 +2) iow l $RFDR (3032) iow l $RFCR ($PATBUF23 + 6) iow l $RFDR (3133) iow l $RFCR ($PATBUF23 + a) iow l $RFDR (3234)
DP83816
#enable receive filter on all patterns iow l $RFCR ($RFEN|$APAT0|$APAT1|$APAT2|$APAT3)
Example of how to mask out a byte in a pattern:
# write data pattern into buffer 0 iow l $RFCR ($PATBUF01) iow l $RFDR (10002) #mask byte 0 (value = 02) iow l $RFCR ($PATBUF01 + 4) iow l $RFDR (20103) #mask byte 1 (value = 01) iow l $RFCR ($PATBUF01 + 8) iow l $RFDR (30204) #mask byte 0 and 1
61 www.national.com
4.0 Register Set (Continued)
Accept on Multicast or Unicast Hash
Multicast and Unicast addresses may be further qualified by use of the receive filter hash functions. An internal 512 bit (64 byte) RAM-based hash table is used to perform imperfect filtering of multicast or unicast packets. By enabling either Multicast Hashing or Unicast Hashing in the RFCR, the receive filter logic will use the 9 least significant bits of the destination addresses’ CRC as an index into the
Unused
Unused
X X byte63 byte62 23E
X X byte61 byte60 23C
X X byte5 byte4 204
X X byte3 byte2 202
DP83816
Hash Table memory. The upper 4 bits represent the word address and the lower 5 bits select the bit within the word. If the corresponding bit is set, then the packet is accepted, otherwise the packet is rejected. The hash table memory is accessed through the RFCR and the RFDR. Refer to Figure 4-2 for a memory map. Below is example code for setting/clearing a bit in the hash table.
..............
X X byte1 byte0 200
Bit# 17 16 15
87 0
Figure 4-2 Hash Table Memory - 40h bytes addressed on word boundaries
set HASH_TABLE = 200
crc $DA # compute the CRC of the destination address set index = ($crc >> 3) set bit = ($crc & 01f) # lower 5 bits select which bit in 32 bit word
# write word address into RFCR iow l $RFCR ($HASH_TABLE + $index)
# select bit to set/clear if ($bit > f) set bit = ($bit - 010h) # use 16 bit register interface into 32bit RAM set hash_bit = (0001 << $bit)
# read indexed word from table ior l $RFDR if ($SetBit) then set hash_word = ($rc | $hash_bit) iow l $RFDR ($hash_word) else set hash_bit = (~$hash_bit) set hash_word = ($rc & $hash_bit) iow l $RFDR ($hash_word)‘ endif
iow l $RFCR ($RFEN|$MHEN|$UHEN)# enable multicast and/or unicast
# address hashing
62 www.national.com
4.0 Register Set (Continued)
4.2.20 Boot ROM Address Register
The BRAR is used to setup the address for an access to an external ROM/FLASH device.
Tag: BRAR Size: 32 bits Hard Reset: FFFFFFFFh
Offset: 0050h Access: Read Write Soft Reset: unchanged
Bit Bit Name Description
31 AUTOINC Auto-Increment
When set, the contents of ADDR will auto increment with every 32-bit access to the BRDR register.
30-16 unused
15-0 ADDR Boot ROM Address
16-bit address used to access the external Boot ROM.
4.2.21 Boot ROM Data Register
The BRDR is used to read and write ROM/FLASH data from the data from/to an external ROM/FLASH device.
Tag: BRDR Size: 32 bits Hard Reset: undefined
Offset: 0054h Access: Read Write Soft Reset: undefined
DP83816
Bit Bit Name Description
31-0 DATA Boot ROM Data
Access port to external Boot ROM. Software can use BRAR and BRDR to read (and write if FLASH memory is used) the external Boot ROM. All accesses must be 32-bits wide and aligned on 32-bit boundaries.
4.2.22 Silicon Revision Register
Tag: SRR Size: 32 bits Hard Reset: as defined
Offset: 0058h Access: Read Only Soft Reset: unchanged
Bit Bit Name Description
31-16 unused
(reads return 0)
15-0 Rev Revision Level
SRR register value for the DP83816 silicon.
DP83816AVNG 00000505h
63 www.national.com
4.0 Register Set (Continued)
4.2.23 Management Information Base Control Register
The MIBC register is used to control access to the statistics block and the warning bits and to control the collection of management information statistics.
Tag: MIBC Size: 32 bits Hard Reset: 00000002h
Offset: 005ch Access: Read Write Soft Reset: 00000002h
Bit Bit Name Description
31-4 unused
3 MIBS MIB Counter Strobe
Writing a 1 to this bit location causes the counters in all enabled blocks to increment by 1, providing a single-step test function. The MIBS bit is always read back as 0. This bit is used for test purposes
only and should be set to 0 for normal counter operation.
2 ACLR Clear all counters
When set to a 1, this bit forces all counters to be reset to 0. This bit is always read back as 0.
1 FRZ Freeze all counters
When set to a 1, this bit forces count values to be frozen such that a read of the statistic block will represent management statistics at a given instant in time. When set to 0, the counters will increment normally and may be read individually while counting. While frozen events will not be recorded.
0 WRN Warning Test Indicator
This field is read only. This bit is set to 1 when statistic counters have reached their respective overflow warning condition. WRN will be cleared after one or more of the statistic counters have been cleared.
DP83816
64 www.national.com
4.0 Register Set (Continued)
4.2.24 Management Information Base Registers
The counters provide a set of statistics compliant with the following management specifications: MIB II, Ether-like MIB, and IEEE MIB. The values provided are accessed through the various registers as shown below. All MIB counters are cleared to 0 when read.
Due to cost and space limitations, the counter bit widths provided in the DP83816 MIB are less than the bit widths called for in the above specifications. It is assumed that management agent software will maintain a set of fully compliant statistic values ("software" counters), utilizing the hardware counters to reduce the frequency at which these
Table 4-3 MIB Registers
DP83816
"software" counters must be updated. Sizes for specific hardware statistic counters were chosen such that the count values will not roll over in less than 15 ms if incremented at the theoretical maximum rates described in the above specifications. However, given that the theoretical maximum counter rates do not represent realistic network traffic and events, the actual rollover rates for the hardware counters are more likely to be on the order of several seconds. The hardware counters are updated automatically by the MAC on the occurrence of each event.
Offset Tag Size
0060h RXErroredPkts 16 8 Packets received with errors. This counter is incremented for each
0064h RXFCSErrors 8 4 Packets received with frame check sequence errors. This counter is
0068h RXMsdPktErrors 8 4 Packets missed due to FIFO overruns. This counter is incremented
006Ch RXFAErrors 8 4 Packets received with frame alignment errors. This counter is
0070h RXSymbolErrors 8 4 Packets received with one or more symbol errors. This counter is
0074h RXFrameTooLong 4 2 Packets received with length greater than 1518 bytes (too long
0078h TXSQEErrors 4 2 Loss of collision heartbeat during transmission. This counter is
warning
(MS bits)
Description
packet received with errors. This count includes packets which are automatically rejected from the FIFO due to both wire errors and FIFO overruns.
incremented for each packet received with a Frame Check Sequence error (bad CRC).
Note: For the MII interface, an FCS error is defined as a resulting invalid CRC after CRS goes invalid and an even number of bytes have been received.
for each receive aborted due to data or status FIFO overruns (insufficient buffer space).
incremented for each packet received with a Frame Check Sequence error (bad CRC).
Note: For the MII interface, an FAE error is defined as a resulting invalid CRC on the last full octet, and an odd number of nibbles have been received (Dribble nibble condition with a bad CRC).
incremented for each packet received with one or more symbol errors detected.
Note: For the MII interface, a symbol error is indicated by the RXER signal becoming active for one or more clocks while the RXDV signal is active (during valid data reception).
packets). This counter is incremented for each packet received with greater than the 802.3 standard maximum length of 1518 bytes.
incremented when the collision heartbeat pulse is not detected by the PMD after a transmission.
65 www.national.com
4.0 Register Set (Continued)
4.3 Internal PHY Registers
The Internal Phy Registers are only 16 bits wide. Bits [31:16] are not used. In the following register definitions under the ‘Default’ heading, the following definitions hold true:
—RW=Read Write access —RO=Read Only access — LL=Latched Low and held until read, based upon the occurrence of the corresponding event —LH=Latched High and held until read, based upon the occurrence of the corresponding event — SC=Register sets on event occurrence and Self-Clears when event ends — P=Register bit is Permanently set to a default value —COR=Clear On Read
4.3.1 Basic Mode Control Register
Tag: BMCR Size: 16 bits Hard Reset: XX00h
Offset: 0080h Access: Read Write
Bit Bit Name Description
15 Reset Reset: Default: 0, RW/SC
1 = Initiate software Reset / Reset in Process 0 = Normal operation This self-clearing bit returns a value of one until the reset process is complete. A reset causes all PHY
registers to return to their default values (in some cases registers defaults are defined by related bits in the CFG register, offset 04h).
14 Loopback Loopback: Default: 0
1 = Loopback enabled 0 = Normal operation The loopback function enables MII transmit data to be routed to the MII receive data path. Setting this bit may cause the de-scrambler to lose synchronization and produce a 500 µs “dead time”
before any valid data will appear at the MII receive outputs.
13 Speed
12 Auto-
11 Power Down Power Down: Default: 0
10 Isolate Isolate: Default: 0
9 Restart Auto-
8 Duplex Mode Duplex Mode: Default: dependent on the setting of the ANEG_SEL bits in the CFG register
Selection
Negotiation
Enable
Negotiation
Speed Select: Default: dependent on the setting of the ANEG_SEL bits in the CFG register When auto-negotiation is disabled writing to this bit allows the port speed to be selected. 1 = 100 Mb/s 0 = 10 Mb/s
Auto-Negotiation Enable: Default: dependent on the setting of the ANEG_SEL bits in the CFG register 1 = Auto-Negotiation Enabled - bits 8 and 13 of this register are ignored when this bit is set. 0 = Auto-Negotiation Disabled - bits 8 and 13 determine the port speed and duplex mode.
1 = Power down 0 = Normal operation Setting this bit powers down the port.
1 = Isolates the port from the MII with the exception of the serial management. 0 = Normal operation
Restart Auto-Negotiation: Default: 0, RW/SC 1 = Restart Auto-Negotiation 0 = Normal operation When this bit is set, it re-initiates the Auto-Negotiation process. If Auto-Negotiation is disabled (bit 12 =
0), this bit is ignored. This bit is self-clearing and will remain a value of 1 until Auto-Negotiation is initiated, whereupon it will self-clear. Operation of the Auto-Negotiation process is not affected by the management entity clearing this bit.
When auto-negotiation is disabled writing to this bit allows the port Duplex capability to be selected. 1 = Full Duplex operation 0 = Half Duplex operation
DP83816
66 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
7 Collision Test Collision Test: Default: 0
1 = Collision test enabled 0 = Normal operation When set, this bit will cause the COL signal to be asserted in response to the assertion of TXEN within
512-bit times. The COL signal will be de-asserted within 4-bit times in response to the de-assertion of TXEN.
6:0 Reserved Reserved: Default: 0, RO
4.3.2 Basic Mode Status Register
Tag: BMSR Size: 16 bits Hard Reset: 7849h
Offset: 0084h Access: Read Only
Bit Bit Name Description
15 100BASE-T4 100BASE-T4 Capable: Default: 0
0 = Device not able to perform 100BASE-T4 mode.
14 100BASE-TX
Full Duplex
13 100BASE-TX
Half Duplex
12 10BASE-T
Full Duplex
11 10BASE-T
Half Duplex
10:7 Reserved Reserved: Write as 0, read as 0
6 Preamble
Suppression
5 Auto-
4 Remote Fault Remote Fault: Default: 0/L(H)
3 Auto-
2 Link Status Link Status: Default: 0/L(L)
1 Jabber Detect Jabber Detect: Default: 0/LH
0 Extended
Negotiation
Complete
Negotiation
Ability
Capability
100BASE-TX Full Duplex Capable: Default: 1 1 = Device able to perform 100BASE-TX in full duplex mode
100BASE-TX Half Duplex Capable: Default: 1 1 = Device able to perform 100BASE-TX in half duplex mode.
10BASE-T Full Duplex Capable: Default: 1 1 = Device able to perform 10BASE-T in full duplex mode
10BASE-T Half Duplex Capable: Default: 1 1 = Device able to perform 10BASE-T in half duplex mode
Preamble suppression Capable: Default: 1 1 = Device able to perform management transaction with preamble suppressed, 32-bits of preamble
needed only once after reset, invalid opcode or invalid turnaround. 0 = Normal management operation
Auto-Negotiation Complete: Default: 0 1 = Auto-Negotiation process complete 0 = Auto-Negotiation process not complete
1 = Remote Fault condition detected (cleared on read or by reset). Fault criteria: Far End Fault Indication or notification from Link Partner of Remote Fault.
0 = No remote fault condition detected
Auto Configuration Ability: Default: 1 1 = Device is able to perform Auto-Negotiation 0 = Device is not able to perform Auto-Negotiation
1 = Valid link established (for either 10 or 100 Mb/s operation) 0 = Link not established The criteria for link validity is implementation specific. The occurrence of a link failure condition will cause
the Link Status bit to clear. Once cleared, this bit may only be set by establishing a good link condition and a read via the management interface.
1 = Jabber condition detected 0 = No Jabber This bit is implemented with a latching function, such that the occurrence of a jabber condition causes it
to set until it is cleared by a read to this register by the management interface or by a reset. This bit only has meaning in 10 Mb/s mode.
Extended Capability: Default: 1 1 = Extended register capabilities 0 = Basic register set capabilities only
DP83816
67 www.national.com
4.0 Register Set (Continued)
4.3.3 PHY Identifier Register #1
The PHY Identifier Registers #1 and #2 together form a unique identifier for the PHY section of this device. The Identifier consists of a concatenation of the Organizationally Unique Identifier (OUI), the vendor's model number and the model revision number. A PHY may return a value of zero in each of the 32 bits of the PHY Identifier if desired. The PHY Identifier is intended to support network management. National Semiconductor's IEEE assigned OUI is 080017h.
Tag: PHYIDR1 Size: 16 bits Hard Reset: 2000h
Offset: 0088h Access: Read Only
Bit Bit Name Description
15:0 OUI_MSB OUI Most Significant Bits: Default: <0010 0000 0000 0000>
Bits 3 to 18 of the OUI (080017h) are stored in bits 15 to 0 of this register. The most significant two bits of the OUI are ignored (the IEEE standard refers to these as bits 1 and 2).
4.3.4 PHY Identifier Register #2
Tag: PHYIDR2 Size: 16 bits Hard Reset: 5C21h
Offset: 008Ch Access: Read Only
DP83816
Bit Bit Name Description
15:10 OUI_LSB OUI Least Significant Bits: Default: <01 0111>
Bits 19 to 24 of the OUI (080017h) are mapped to bits 15 to 10 of this register respectively.
9:4 VNDR_MDL Vendor Model Number: Default: <00 0010>
The six bits of vendor model number are mapped to bits 9 to 4 (most significant bit to bit 9).
3:0 MDL_REV Model Revision Number: Default: <0001>
Four bits of the vendor model revision number are mapped to bits 3 to 0 (most significant bit to bit 3). This field will be incremented for all major device changes.
4.3.5 Auto-Negotiation Advertisement Register
This register contains the advertised abilities of this device as they will be transmitted to its link partner during Auto­Negotiation.
Tag: ANAR Size: 16 bits Hard Reset: 05E1h
Offset: 0090h Access: Read Write
Bit Bit Name Description
15 NP Next Page Indication: Default: 0
0 = Next Page Transfer not desired 1 = Next Page Transfer desired
14 Reserved Reserved by IEEE: Writes ignored, Read as 0
13 RF Remote Fault: Default: 0
1 = Advertises that this device has detected a Remote Fault 0 = No Remote Fault detected
12:11 Reserved Reserved for Future IEEE use: Write as 0, Read as 0
10 PAUSE PAUSE: Default: dependent on the setting of the PAUSE_ADV in the CFG register
1 = Advertise that the DTE (MAC) has implemented both the optional MAC control sublayer and the pause function as specified in clause 31 and annex 31B of 802.3u.
0 = No MAC based full duplex flow control
68 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
9 T4 100BASE-T4 Support: Default: 0/ RO
1= 100BASE-T4 is supported by the local device 0 = 100BASE-T4 not supported
8 TX_FD 100BASE-TX Full Duplex Support: Default: dependent on setting of the ANEG_SEL in the CFG register
1 = 100BASE-TX Full Duplex is supported by the local device 0 = 100BASE-TX Full Duplex not supported
7 TX 100BASE-TX Support: Default: dependent on the setting of the ANEG_SEL bits in the CFG register
1 = 100BASE-TX is supported by the local device 0 = 100BASE-TX not supported
6 10_FD 10BASE-T Full Duplex Support: Default: dependent on setting of the ANEG_SEL in the CFG register
1 = 10BASE-T Full Duplex is supported by the local device 0 = 10BASE-T Full Duplex not supported
5 10 10BASE-T Support: Default: dependent on the setting of the ANEG_SEL bits in the CFG register
1 = 10BASE-T is supported by the local device 0 = 10BASE-T not supported
4:0 Selector Protocol Selection Bits: Default: <00001>
These bits contain the binary encoded protocol selector supported by this port. <00001> indicates that this device supports IEEE 802.3u.
DP83816
4.3.6 Auto-Negotiation Link Partner Ability Register
This register contains the advertised abilities of the Link Partner as received during Auto-Negotiation. The content changes after the successful auto-negotiation if Next-pages are supported.
Tag: ANLPAR Size: 16 bits Hard Reset: 0000h
Offset: 0094h Access: Read Only
Bit Bit Name Description
15 NP Next Page Indication:
0 = Link Partner does not desire Next Page Transfer 1 = Link Partner desires Next Page Transfer
14 ACK Acknowledge:
1 = Link Partner acknowledges reception of the ability data word 0 = Not acknowledged The Device's Auto-Negotiation state machine will automatically control this bit based on the incoming FLP
bursts.
13 RF Remote Fault:
1 = Remote Fault indicated by Link Partner 0 = No Remote Fault indicated by Link Partner
12:10 Reserved Reserved for Future IEEE use: Write as 0, read as 0
9 T4 100BASE-T4 Support:
1 = 100BASE-T4 is supported by the Link Partner 0 = 100BASE-T4 not supported by the Link Partner
8 TX_FD 100BASE-TX Full Duplex Support:
1 = 100BASE-TX Full Duplex is supported by the Link Partner 0 = 100BASE-TX Full Duplex not supported by the Link Partner
7 TX 100BASE-TX Support:
1 = 100BASE-TX is supported by the Link Partner 0 = 100BASE-TX not supported by the Link Partner
6 10_FD 10BASE-T Full Duplex Support:
1 = 10BASE-T Full Duplex is supported by the Link Partner 0 = 10BASE-T Full Duplex not supported by the Link Partner
69 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
5 10 10BASE-T Support:
1 = 10BASE-T is supported by the Link Partner 0 = 10BASE-T not supported by the Link Partner
4:0 Selector Protocol Selection Bits:
Link Partners’s binary encoded protocol selector.
4.3.7 Auto-Negotiate Expansion Register
This register contains additional Local Device and Link Partner status information.
Tag: ANER Size: 16 bits Hard Reset: 0004h
Offset: 0098h Access: Read Only
Bit Bit Name Description
15:5 Reserved Reserved: Writes ignored, Read as 0.
4 PDF Parallel Detection Fault:
1 = A fault has been detected via the Parallel Detection function 0 = A fault has not been detected
3 LP_NP_ABLE Link Partner Next Page Able:
1 = Link Partner does support Next Page 0 = Link Partner does not support Next Page
2 NP_ABLE Next Page Able:
1 = Indicates local device is able to send additional “Next Pages”
1 PAGE_RX Link Code Word Page Received: RO/COR
1 = Link Code Word has been received, cleared on a read 0 = Link Code Word has not been received
0 LP_AN_ABLE Link Partner Auto-Negotiation Able:
1 = Indicates that the Link Partner supports Auto-Negotiation 0 = Indicates that the Link Partner does not support Auto-Negotiation
DP83816
4.3.8 Auto-Negotiation Next Page Transmit Register
This register contains the next page information sent by this device to its Link Partner during Auto-Negotiation.
Tag: ANNPTR Size: 16 bits Hard Reset: 2001h
Offset: 009Ch Access: Read Write
Bit Bit Name Description
15 NP Next Page Indication: Default: 0
0 = No other Next Page Transfer desired 1 = Another Next Page desired
14 Reserved Reserved: Writes ignored, read as 0
13 MP Message Page: Default: 1
1 = Message Page 0 = Unformatted Page
70 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
12 ACK2 Acknowledge2: Default: 0
1 = Will comply with message 0 = Cannot comply with message Acknowledge2 is used by the next page function to indicate that Local Device has the ability to comply
with the message received.
11 TOG_TX Tog gle: Default: 0, RO
1 = Value of toggle bit in previously transmitted Link Code Word was 0 0 = Value of toggle bit in previously transmitted Link Code Word was 1 Toggle is used by the Arbitration function within Auto-Negotiation to ensure synchronization with the Link
Partner during Next Page exchange. This bit shall always take the opposite value of the Toggle bit in the previously exchanged Link Code Word.
10:0 CODE Code Field: Default: <000 0000 0001>
This field represents the code field of the next page transmission. If the MP bit is set (bit 13 of this register), then the code shall be interpreted as a "Message Page”, as defined in annex 28C of IEEE
802.3u. Otherwise, the code shall be interpreted as an "Un-formatted Page”, and the interpretation is application specific.
The default value of the CODE represents a Null Page as defined in Annex 28C of IEEE 802.3u.
4.3.9 PHY Status Register
This register provides a single location within the register set for quick access to commonly accessed information.
DP83816
Tag: PHYSTS Size: 16 bits Hard Reset: 0000h
Offset: 00C0h Access: Read Only
Bit Bit Name Description
15:14 Reserved Reserved: Write ignored, read as 0.
13 Receive Error
12 Polarity
11 False Carrier
10 Signal Detect Signal Detect: Default: 0, RO/LL
9 De-scrambler
8 Page
7 MII Interrupt MII Interrupt Pending: Default: 0, RO/LH
Latch
Status
Sense Latch
Lock
Received
Receive Error Latch:
This bit will be cleared upon a read of the RECR register. 1 = Receive error event has occurred since last read of RXERCNT (address 0xD4) 0 = No receive error event has occurred
Polarity Status:
This bit is a duplication of bit 4 in the TBTSCR register. This bit will be cleared upon a read of the TBTSCR register, but not upon a read of the PHYSTS register.
1 = Inverted Polarity detected 0 = Correct Polarity detected
False Carrier Sense Latch: Default: 0, RO/LH This bit will be cleared upon a read of the FCSCR register. 1 = False Carrier event has occurred since last read of FCSCR (address 0xD0) 0 = No False Carrier event has occurred
100BASE-TX unconditional Signal Detect from PMD.
De-scrambler Lock: Default: 0, RO/LL 100BASE-TX De-scrambler Lock from PMD.
Link Code Word Page Received:
This is a duplicate of the Page Received bit in the ANER register, but this bit will not be cleared upon a read of the PHYSTS register.
1 = A new Link Code Word Page has been received. Cleared on read of the ANER (address 0x98, bit 1) 0 = Link Code Word Page has not been received
1 = Indicates that an internal interrupt is pending, cleared by the current read 0 = No interrupt pending
71 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
6 Remote Fault Remote Fault:
1 = Remote Fault condition detected (cleared on read of BMSR (address 0x84) register or by reset). Fault criteria: notification from Link Partner of Remote Fault via Auto-Negotiation
0 = No remote fault condition detected
5 Jabber Detect Jabber Detect: This bit only has meaning in 10 Mb/s mode
This bit is a duplicate of the Jabber Detect bit in the BMSR register, except that it is not cleared upon a read of the PHYSTS register.
1 = Jabber condition detected 0 = No Jabber
4 Auto-Neg.
3 Loopback
2 Duplex Status Duplex:
1 Speed Status Speed10:
0 Link Status Link Status:
Complete
Status
Auto-Negotiation Complete:
1 = Auto-Negotiation complete 0 = Auto-Negotiation not complete
Loopback:
1 = Loopback enabled 0 = Normal operation
This bit indicates duplex status and is determined from Auto-Negotiation or Forced Modes. 1 = Full duplex mode 0 = Half duplex mode Note: This bit is only valid if Auto-Negotiation is enabled and complete and there is a valid link or if Auto-
Negotiation is disabled and there is a valid link.
This bit indicates the status of the speed and is determined from Auto-Negotiation or Forced Modes. 1 = 10 Mb/s mode 0 = 100 Mb/s mode Note: This bit is only valid if Auto-Negotiation is enabled and complete and there is a valid link or if Auto-
Negotiation is disabled and there is a valid link.
This bit is a duplicate of the Link Status bit in the BMSR register, except that it will not be cleared upon a read of the PHYSTS register.
1 = Valid link established (for either 10 or 100 Mb/s operation) 0 = Link not established
DP83816
72 www.national.com
4.0 Register Set (Continued)
4.3.10 MII Interrupt Control Register
This register implements the MII Interrupt PHY Specific Control register. Sources for interrupt generation include: Link State Change, Jabber Event, Remote Fault, Auto-Negotiation Complete or any of the counters becoming half-full. Note that the TINT bit operates independently of the INTEN bit. In other words, INTEN does not need to be active to generate the test interrupt.
Tag: MICR Size: 16 bits Hard Reset: 0000h
Offset: 00C4h Access: Read Write
Bit Bit Name Description
15:2 Reserved Reserved: Writes ignored, Read as 0
1 INTEN Interrupt Enable:
1 = Enable event based interrupts 0 = Disable event based interrupts
0 TINT Test Interrupt:
Forces the PHY to generate an interrupt at the end of each management read to facilitate interrupt testing. 1 = Generate an interrupt 0 = Do not generate interrupt
DP83816
4.3.11 MII Interrupt Status and Misc. Control Register
This register implements the MII Interrupt PHY Control and Status information. These Interrupts are PHY based events. When any of these events occur and its respective bit is not masked, and MICR:INTEN is enabled, the interrupt will be signalled in ISR:PHY.
Tag: MISR Size: 16 bits Hard Reset: 0000h
Offset: 00C8h Access: Read Write
Bit Bit Name Description
15 MINT MII Interrupt Pending: Default: 0, RO/COR
1 = Indicates that an interrupt is pending and is cleared by the current read. 0 = no interrupt pending
14 MSK_LINK Mask Link: When this bit is 0, the change of link status event will cause the interrupt to be seen by the ISR.
13 MSK_JAB Mask Jabber: When this bit is 0, the Jabber event will cause the interrupt to be seen by the ISR.
12 MSK_RF Mask Remote Fault: When this bit is 0, the Remote Fault event will cause the interrupt to be seen by the
11 MSK_ANC Mask Auto-Neg. Complete: When this bit is 0, the Auto-negotiation complete event will cause the inter-
10 MSK_FHF Mask False Carrier Half Full: When this bit is 0, the False Carrier Counter Register half-full event will
9 MSK_RHF Mask Rx Error Half Full: When this bit is 0, the Receive Error Counter Register half-full event will cause
8:0 Reserved Reserved: Writes ignored, Read as 0
ISR.
rupt to be seen by the ISR.
cause the interrupt to be seen by the ISR.
the interrupt to be seen by the ISR.
73 www.national.com
4.0 Register Set (Continued)
4.3.12 False Carrier Sense Counter Register
This counter provides information required to implement the “FalseCarriers” attribute within the MAU managed object class of Clause 30 of the IEEE 802.3u specification.
Tag: FCSCR Size: 16 bits Hard Reset: 0000h
Offset: 00D0h Access: Read Write
Bit Bit Name Description
15:8 Reserved Reserved: Writes ignored, Read as 0
7:0 FCSCNT[7:0] False Carrier Event Counter: Default: 0, RW/COR
This 8-bit counter increments on every false carrier event. This counter sticks when it reaches its max count (FFh).
4.3.13 Receiver Error Counter Register
This counter provides information required to implement the “SymbolErrorDuringCarrier” attribute within the PHY managed object class of Clause 30 of the IEEE 802.3u specification.
Tag: RECR Size: 16 bits Hard Reset: 0000h
Offset: 00D4h Access: Read Write
DP83816
Bit Bit Name Description
15:8 Reserved Reserved: Writes ignored, Read as 0
7:0 RXERCNT[7:0] RXER Counter: Default: 0, RW / COR
This 8-bit counter increments for each receive error detected. when a valid carrier is present and there is at least one occurrence of an invalid data symbol. This event can increment only once per valid carrier event. If a collision is present, the attribute will not increment. The counter sticks when it reaches its max count.
4.3.14 100 Mb/s PCS Configuration and Status Register
Tag: PCSR Size: 16 bits Hard Reset: 0100h
Offset: 00D8h Access: Read Write
Bit Bit Name Description
15:13 Reserved Reserved: Writes ignored, Read as 0
12 BYP_4B5B Bypass 4B/5B Encoding:
1 = 4B5B encoder functions bypassed 0 = Normal 4B5B operation
11 FREE_CLK Receive Clock:
1 = RX_CK is free-running 0 = RX_CK phase adjusted based on alignment
10 TQ_EN 100 Mb/s True Quiet Mode Enable:
1 = Transmit True Quiet Mode 0 = Normal Transmit Mode
9 SD_FORCE_B Signal Detect Force:
1 = Forces Signal Detection 0 = Normal SD operation
74 www.national.com
4.0 Register Set (Continued)
Bit Bit Name Description
8 SD_OPTION Signal Detect Option:
1 = Enhanced signal detect algorithm 0 = Reduced signal detect algorithm
7:6 Reserved Reserved: Read as 0
5 FORCE_100_OK Force 100 Mb/s Good Link:
1 = Forces 100 Mb/s Good Link 0 = Normal 100 Mb/s operation
4:3 Reserved Reserved: Read as 0
2 NRZI_BYPASS NRZI Bypass Enable:
1 = NRZI Bypass Enabled 0 = NRZI Bypass Disabled
1:0 Reserved Reserved: Read as 0
4.3.15 PHY Control Register
Tag: PHYCR Size: 16 bits Hard Reset: 003Fh
Offset: 00E4h Access: Read Write
DP83816
Bit Bit Name Description
15:12 Reserved Reserved
11 PSR_15 BIST Sequence select: Selects length of LFSR used in BIST
1 = PSR15 selected 0 = PSR9 selected
10 BIST_STATUS BIST Test Status: Default: 0, LL/RO
1 = BIST pass 0 = BIST fail. Latched, cleared by write to BIST start bit.
9 BIST_START BIST Start: BIST runs continuously until stopped. Minimum time to run should be 1 ms.
1 = BIST start 0 = BIST stop
8 BP_STRETCH Bypass LED Stretching:
This will bypass the LED stretching and the LEDs will reflect the internal value. 1 = Bypass LED stretching 0 = Normal operation
7 PAUSE_STS Pause Compare Status: Default: 0, RO
0 = Local Device and the Link Partner are not Pause capable 1 = Local Device and the Link Partner are both Pause capable
6:5 Reserved Reserved
4:0 PHYADDR[4:0] PHY Address: Default: <11111b>, RW
PHY address for the port.
75 www.national.com
4.0 Register Set (Continued)
4.3.16 10BASE-T Status/Control Register
Tag: TBTSCR Size: 16 bits Hard Reset: 0804h
Offset: 00E8h Access: Read Write
Bit Bit Name Description
15:9 Unused
8 LOOPBACK_10_DIS 10BASE-T Loopback Disable:
This bit is OR’ed with bit 14 (Loopback) in the BMCR. 1 = 10 Mb/s Loopback is enabled 0 = 10 Mb/s Loopback is disabled
7 LP_DIS Normal Link Pulse Disable:
1 = Transmission of NLPs is disabled 0 = Transmission of NLPs is enabled
6 FORCE_LINK_10 Force 10 Mb/s Good Link:
1 = Forced Good 10 Mb/s Link 0 = Normal Link Status
5 FORCE_POL_COR Force 10 Mb/s Polarity Correction:
1 = Force inverted polarity 0 = Normal polarity
4 POLARITY 10 Mb/s Polarity Status: RO/LH
This bit is a duplication of bit 12 in the PHYSTS register. Both bits will be cleared upon a read of either register.
1 = Inverted Polarity detected 0 = Correct Polarity detected
3 AUTOPOL_DIS Auto Polarity Detection & Correction Disable:
1 = Polarity Sense & Correction disabled 0 = Polarity Sense & Correction enabled
2 Reserved Reserved
This bit must be written as a one.
1 HEARTBEAT_DIS Heartbeat Disable: This bit only has influence in half-duplex 10 Mb/s mode.
1 = Heartbeat function disabled 0 = Heartbeat function enabled When the device is operating at 100 Mb/s or configured for full duplex, this bit will be ignored - the
heartbeat function is disabled.
0 JABBER_DIS Jabber Disable:
Applicable only in 10BASE-T Full Duplex. 1 = Jabber function disabled 0 = Jabber function enabled
DP83816
76 www.national.com
5.0 Buffer Management
DP83816
The buffer management scheme used on the DP83816 allows quick, simple and efficient use of the frame buffer memory. Frames are saved in similar formats for both transmit and receive. The buffer management scheme also uses separate buffers and descriptors for packet information. This allows effective transfers of data from the receive buffer to the transmit buffer by simply transferring the descriptor from the receive queue to the transmit queue.
The format of the descriptors allows the packets to be saved in a number of configurations. A packet can be stored in memory with a single descriptor and a single packet fragment, or multiple descriptors each with a single fragment. This flexibility allows the user to configure the DP83816 to maximize efficiency. Architecture of the specific system’s buffer memory, as well as the nature of network traffic, will determine the most suitable configuration of packet descriptors and fragments.
Table 5-1 DP83816 Descriptor Format
Offset Tag Description
0000h link 32-bit "link" field to the next descriptor in the linked list. Bits 1-0 must be 0, as
descriptors must be aligned on 32-bit boundaries.
0004h cmdsts 32-bit Command/Status Field (bit-encoded).
0008h bufptr 32-bit pointer to the first fragment or buffer. In transmit descriptors, the buffer can
begin on any byte boundary. In receive descriptors, the buffer must be aligned on a 32-bit boundary.
5.1 Overview
The buffer management design has the following goals: — simplicity, — efficient use of the PCI bus (the overhead of the buffer
management technique is minimal), — low CPU utilization, — flexibility. Descriptors may be either per-packet or per-packet-
fragment. Each descriptor may describe one packet fragment. Receive and transmit descriptors are symmetrical.
5.1.1 Descriptor Format
DP83816 uses a symmetrical format for transmit and receive descriptors. In bridging and switching applications this symmetry allows software to forward packets by simply moving the list of descriptors that describe a single received packet from the receive list of one MAC to the transmit list of another. Descriptors must be aligned on an even long word (32-bit) boundary.
The original DP83810A Descriptor format supported multiple fragments per descriptor. DP83816 only supports a single fragment per descriptor. By default, DP83816 will use the descriptor format shown above. By setting CFG:EUPHCOMP, software may force compatibility with the previous DP83810A Descriptor format (although still
Table 5-2 cmdsts Common Bit Definitions
Bit Tag Description Usage
31 OWN Descriptor Ownership Set to 1 by the data producer of the descriptor to transfer
30 MORE More descriptors Set to 1 to indicate that this is NOT the last descriptor in a packet
29 INTR Interrupt Set to 1 by software to request a “descriptor interrupt" when
28 SUPCRC
INCCRC
Suppress CRC / Include CRC
only single fragment descriptors are supported). When CFG:EUPHCOMP is set, then bufptr is at offset 0Ch, and the 32-bit bufcnt field at offset 08h is ignored.
Some of the bit definitions in the cmdsts field are common to both receive and transmit descriptors:
ownership to the data consumer of the descriptor. Set to 0 by the
data consumer of the descriptor to return ownership to the data producer of the descriptor. For transmit descriptors, the driver is
the data producer, and the DP83816 is the data consumer. For receive descriptors, the DP83816 is the data producer, and the driver is the data consumer.
(there are MORE to follow). When 0, this descriptor is the last descriptor in a packet. Completion status bits are only valid when this bit is zero.
DP83816 transfers the ownership of this descriptor back to software.
In transmit descriptors, this indicates that CRC should not be appended by the MAC. On receives, this bit is always set, as the CRC is always copied to the end of the buffer by the hardware.
77 www.national.com
5.0 Buffer Management (Continued)
DP83816
27 OK Packet OK In the last descriptor in a packet, this bit indicates that the packet
26-16 --- The usage of these bits differ in receive and transmit descriptors.
15-12 (reserved)
11-0 SIZE Descriptor Byte Count Set to the size in bytes of the data.
Table 5-3 Transmit Status Bit Definitions
Bit Tag Description Usage
26 TXA Transmit Abort Transmission of this packet was aborted.
25 TFU Transmit FIFO
Underrun
24 CRS Carrier Sense Lost Carrier was lost during the transmission of this packet. This
23 TD Transmit Deferred Transmission of this packet was deferred.
22 ED Excessive Deferral The length of deferral during the transmission of this packet was
21 OWC Out of Window
Collision
20 EC Excessive Collisions The number of collisions during the transmission of this packet
19-16 CCNT Collision Count If TXCFG register ECRETRY=0, this field indicates the number of
was either sent or received successfully.
See below for details.
Transmit FIFO was exhausted during the transmission of this packet.
condition is not reported if TXCFG:CSI is set.
excessive (> 3.2 ms), indicating transmission failure.
The MAC encountered an "out of window" collision during the transmission of this packet.
was excessive, indicating transmission failure. If TXCFG register ECRETRY=0, this bit is set after 16 collisions. If TXCFG register ECRETRY=1, this bit is set after 4 Excessive
Collision events (64 collisions).
collisions encountered during the transmission of this packet. If TXCFG register ECRETRY=1, CCNT[3:2] = Excessive Collisions (0-3) CCNT[1] = Multiple Collisions CCNT[0] = Single Collision Note that Excessive Collisions indicate 16 attempts failed, while
multiple and single collisions indicate collisions in addition to any excessive collisions. For example a collision count of 33 includes 2 Excessive Collisions and will also set the Single Collision bit.
78 www.national.com
5.0 Buffer Management (Continued)
Table 5-4 Receive Status Bit Definitions
Bit Tag Description Usage
26 RXA Receive Aborted Set to 1 by DP83816 when the receive was aborted, the value of
this bit always equals RXO. Exists for backward compatibility.
25 RXO Receive Overrun Set to 1 by DP83816 to indicate that a receive overrun condition
occurred. RXA will also be set.
24-23 DEST Destination Class When the receive filter is enabled, these bits will indicate the
destination address class as follows:
00 - Packet was rejected 01 - Destination is a Unicast address 10 - Destination is a Multicast address 11 - Destination is a Broadcast address
If the Receive Filter is enabled, 00 indicates that the packet was rejected. Normally packets that are rejected do not cause any bus activity, nor do they consume receive descriptors. However, this condition could occur if the packet is rejected by the Receive Filter later in the packet than the receive drain threshold (RXCFG:DRTH).
Note: The DEST bits may not represent a correct DA class for runt packets received with less than 6 bytes.
22 LONG Too Long Packet
Received
21 RUNT Runt Packet Received The size of the receive packet was less than 64 bytes (inc. CRC).
20 ISE Invalid Symbol Error (100 Mb/s only) An invalid symbol was encountered during the
19 CRCE CRC Error The CRC appended to the end of this packet was invalid.
18 FAE Frame Alignment Error The packet did not contain an integral number of octets.
17 LBP Loopback Packet The packet is the result of a loopback transmission.
16 COL Collision Activity The receive packet had a collision during reception.
If RXCFG:ALP=0, this flag indicates that the size of the receive packet exceeded 1518 bytes.
If RXCFG:ALP=1, this flag indicates that the size of the receive packet exceeded 2046 bytes.
reception of this packet.
DP83816
5.1.2 Single Descriptor Packets
To represent a packet in a single descriptor, the MORE bit in the cmdsts field is set to 0.
single descriptor / single fragment
link
64
0
ptr
MAC hdr
netwk hdr
data
Figure 5-1 Single Descriptor Packets
79 www.national.com
5.0 Buffer Management (Continued)
5.1.3 Multiple Descriptor Packets
A single packet may also cross descriptor boundaries. This is indicated by setting the MORE bit in all descriptors except the last one in the packet. Ethernet applications (bridges, switches, routers, etc.) can optimize memory utilization by using a single small buffer per receive descriptor, and allowing the DP83816 hardware to use the minimum number of buffers necessary to store an incoming packet.
multiple descriptor / single fragment
DP83816
5.1.4 Descriptor Lists
Descriptors are organized in linked lists using the link field. The system designer may also choose to implement a "ring" of descriptors by linking the last descriptor in the list back to the first. A list of descriptors may represent any number of packets or packet fragments.
1
MAC hdr
addr 10100
10140
addr 10100
10140
link
ptr
14
link
1
ptr
netwk hdr
20
Figure 5-2 Multiple Descriptor Packets
Descriptors Organized in a Ring
addr 10140
10180
addr 10180
101C0
Descriptors Organized in a Linked List
addr 10140
10180
addr 10180
101C0
0
link
30
ptr
data
addr 101C0
10100
addr 101C0
00000
Figure 5-3 List and Ring Descriptor Organization
80 www.national.com
5.0 Buffer Management (Continued)
5.2 Transmit Architecture
The following figure illustrates the transmit architecture of the DP83816 10/100 Ethernet Controller.
Software/Memory Hardware
Transmit Descriptor
link cmdsts ptr
Packet
TxHead
link cmdsts ptr
Tx DMA
Figure 5-4 Transmit Architecture
When the CR:TXE bit is set to 1 (regardless of the current state), and the DP83816 transmitter is idle, then DP83816 will read the contents of the current transmit descriptor into the TxDescCache. The DP83816’s TxDescCache can hold a single fragment pointer/count combination.
5.2.1 Transmit State Machine
The transmit state machine has the following states:
Current Tx Desc Ptr
Tx Desc Cache
Tx Data FIFO
DP83816
txIdle The transmit state machine is idle.
txDescRefr Waiting for the "refresh" transfer of the link field of a completed descriptor from the PCI bus.
txDescRead Waiting for the transfer of a complete descriptor from the PCI bus into the
TxDescriptorCache.
txFifoBlock Waiting for free space in the TxDataFIFO to reach TxFillThreshold.
txFragRead Waiting for the transfer of a fragment (or portion of a fragment) from the PCI bus to the
TxDataFIFO.
txDescWrite Waiting for the completion of the write of the cmdsts field of an intermediate transmit
descriptor (cmdsts.MORE == 1) to host memory.
txAdvance (transitory state) Examine the link field of the current descriptor and advance to the next
descriptor if link is not NULL.
The transmit state machine manipulates the following internal data spaces:
TXDP A 32-bit register that points to the current transmit descriptor.
CTDD An internal bit flag that is set when the current transmit descriptor has been completed, and
ownership has been returned to the driver. It is cleared whenever TXDP is loaded with a new value (either by the state machine, or the driver).
TxDescCache An internal data space equal to the size of the maximum transmit descriptor supported.
descCnt Count of bytes remaining in the current descriptor.
fragPtr Pointer to the next unread byte in the current fragment.
txFifoCnt Current amount of data in the txDataFifo in bytes.
txFifoAvail Current amount of free space in the txDataFifo in bytes (size of the txDataFifo - txFifoCnt).
Inputs to the transmit state machine include the following events:
CR:TXE Driver asserts the TXE bit in the command register (similar to SONIC).
XferDone Completion of a PCI bus transfer request.
FifoAvail TxFifoAvail is greater than TxFillThreshold.
81 www.national.com
5.0 Buffer Management (Continued)
Table 5-5 Transmit State Tables
State Event Next State Actions
txIdle CR:TXE && !CTDD txDescRead Start a burst transfer at address TXDP and a length
derived from TXCFG.
CR:TXE && CTDD txDescRefr Start a burst transfer to refresh the link field of the current
txDescRefr XferDone txAdvance
txDescRead XferDone && OWN txFIFOblock
XferDone && !OWN txIdle Set ISR:TXIDLE.
txFIFOblock FifoAvail txFragRead Start a burst transfer into the TxDataFIFO from fragPtr.
(descCnt == 0) &&
txDescWrite Start a burst transfer to write the status back to the
MORE
(descCnt == 0) &&
txAdvance Write the value of TXDP to the txDataFIFO as a handle.
!MORE
txFragRead XferDone txFIFOblock
txDescWrite XferDone txAdvance
txAdvance link != NULL txDescRead TXDP <- txDescCache.link. Clear CTDD. Start a burst
link == NULL txIdle Set CTDD. Set ISR:TXIDLE. Clear CR:TXE.
descriptor.
The length will be the minimum of txFifoAvail and descCnt.
Decrement descCnt accordingly.
descriptor, clearing the OWN bit.
transfer at address TXDP with a length derived from TXCFG.
DP83816
txDescRefr
XferDone
txAdvance
XferDone
CR:TXE && CTDD
link = NULL
txIdle
link != NULL
descCnt == 0 && !(cmdsts & MORE)
txFifoBlocktxDescWrite
CR:TXE && !CTDD
XferDone && !OWN
txDescRead
XferDone && OWN
FifoAvail
txFragRead
descCnt == 0 && (cmdsts & MORE)
Figure 5-5 Transmit State Diagram
82 www.national.com
XferDone
5.0 Buffer Management (Continued)
5.2.2 Transmit Data Flow
In the DP83816 transmit architecture, packet transmission involves the following steps:
1. The device driver receives packets from an upper layer.
2. An available DP83816 transmit descriptor is allocated. The fragment information is copied from the NOS specific data structure(s) to the DP83816 transmit descriptor.
3. The driver adds this descriptor to it’s internal list of transmit descriptors awaiting transmission and sets the OWN bit.
4. If the internal list was empty (this descriptor represents the only outstanding transmit packet), then the driver must set the TXDP register to the address of this descriptor, else the driver will append this descriptor to the end of the list.
5. The driver sets the TXE bit in the CR register to insure that the transmit state machine is active.
6. If idle, the transmit state machine reads the descriptor into the TxDescriptorCache.
7. The state machine then moves through the fragment described within the descriptor, filling the TxDataFifo with data. The hardware handles all aspects of byte alignment; no alignment is assumed. Fragments may start and/or end on any byte address. The transmit state machine uses the fragment pointer and the SIZE field from the cmdsts field of the current descriptor to keep the TxDataFifo full. It also uses the
DP83816
MORE bit and the SIZE field from the cmdsts field of the current descriptor to know when packet boundaries occur.
8. When a packet has completed transmission (successful or unsuccessful), the state machine updates the upper half of the cmdsts field of the current descriptor in main memory, relinquishing ownership, and indicating the packet completion status. This update is done by a bus master transaction that transfers only the upper 2 bytes to the descriptor being updated. If more than one descriptor was used to describe the packet, then completion status is updated only in the last descriptor. Intermediate descriptors only have the OWN bits modified.
9. If the link field of the descriptor is non-zero, the state machine advances to the next descriptor and continues.
10. If the link field is NULL, the transmit state machine suspends, waiting for the TXE bit in the CR register to be set. If the TXDP register is written to, the CTDD flag will be cleared. When the TXE bit is set, the state machine will examine CTDD. If CTDD is set, the state machine will "refresh" the link field of the current descriptor. It will then follow the link field to any new descriptors that have been added to the end of the list. If CTDD is clear (implying that TXDP has been written to), the state machine will start by reading in the descriptor pointed to by TXDP.
83 www.national.com
5.0 Buffer Management (Continued)
5.3 Receive Architecture
The receive architecture is as "symmetrical" to the transmit architecture as possible. The receive buffer manager prefetches receive descriptors to prepare for incoming
Software/Memory Hardware
Receive Descriptor List
DP83816
packets. When the amount of receive data in the RxDataFIFO is more than the RxDrainThreshold, or the RxDataFIFO contains a complete packet, then the state machine begins filling received buffers in host memory.
RxHead
link cmdsts ptr
link cmdsts ptr
link cmdsts ptr
link cmdsts ptr
Rx DMA
Rx Descriptor Cache
Rx Data FIFO
Figure 5-6 Receive Architecture
When the RXE bit is set to 1 in the CR register (regardless of the current state), and the DP83816 receive state machine is idle, then DP83816 will read the contents of the descriptor referenced by RXDP into the Rx Descriptor
read an entire descriptor in a single burst, and reduces the number of bus accesses required for fragment information to 1. The DP83816 Rx Descriptor Cache holds a single buffer pointer/count combination.
Cache. The Rx Descriptor Cache allows the DP83816 to
5.3.1 Receive State Machine
The receive state machine has the following states:
rxIdle The receive state machine is idle.
rxDescRefr Waiting for the "refresh" transfer of the link field of a completed descriptor from the PCI bus.
rxDescRead Waiting for the transfer of a descriptor from the PCI bus into the RxDescCache.
rxFifoBlock Waiting for the amount of data in the RxDataFifo to reach the RxDrainThreshold or to represent a
complete packet.
rxFragWrite Waiting for the transfer of data from the RxDataFIFO via the PCI bus to host memory.
rxDescWrite Waiting for the completion of the write of the cmdsts field of a receive descriptor.
The receive state machine manipulates the following internal data spaces:
RXDP A 32-bit register that points to the current receive descriptor.
CRDD An internal bit flag that is set when the current receive descriptor has been completed, and ownership
has been returned to the driver. It is cleared whenever RXDP is loaded with a new value (either by the state machine, or the driver).
RxDescCache An internal data space equal to the size of the maximum receive descriptor supported.
descCnt Count of bytes available for storing receive data in all fragments described by the current descriptor.
fragPtr Pointer to the next unwritten byte in the current fragment.
rxPktCnt Number of packets in the rxDataFifo. Incremented by the MAC (the fill side of the FIFO). Decremented
by the receive state machine as packets are processed.
rxPktBytes Number of bytes in the current packet being drained from the rxDataFifo, that are in fact currently in the
rxDataFifo (Note: packets larger than FIFO size, this number will never be greater than the FIFO size).
Inputs to the receive state machine include the following events:
CR:RXE The RXE bit in the Command Register has been set.
XferDone completion of a PCI bus transfer request.
FifoReady (rxPktCnt > 0) or (rxPktBytes > rxDrainThreshold)... in other words, if we have a complete packet in the
FIFO (regardless of size), or the number of bytes that we do have is greater than the rxDrainThreshold, then we are ready to begin draining the rxDataFifo.
84 www.national.com
5.0 Buffer Management (Continued)
Table 5-6 Receive State Tables
State Event Next State Actions
rxIdle CR:RXE && !CRDD rxDescRead Start a burst transfer at address RXDP and a length
derived from RXCFG.
CR:RXE && CRDD rxDescRefr Start a burst transfer to refresh the link field of the current
rxDescRefr XferDone rxAdvance
rxDescRead XferDone && !OWN rxFIFOblock
XferDone && OWN rxIdle Set ISR:RXIDLE.
rxFIFOblock FifoReady rxFragWrite Start a burst transfer from the RxDataFIFO to host
(descCnt == 0) &&
(rxPktBytes > 0)
rxPktBytes == 0 rxDescWrite Start a transfer to write the cmdsts back to the descriptor,
rxFragWrite XferDone rxFIFOblock
rxDescWrite XferDone rxAdvance
rxAdvance link!= NULL rxDescRead RXDP <- rxDescCache.link. Clear CRDD. Start a burst
link == NULL rxIdle Set CRDD. Set ISR:RXIDLE.
rxDescWrite Start a burst transfer to write the status back to the
descriptor.
memory at fragPtr. The length will be the minimum of rxPktBytes and descCnt. Decrement descCnt accordingly.
descriptor, setting the OWN bit, and setting the MORE bit. We'll continue the packet in the next descriptor.
setting the OWN bit and clearing the MORE bit, and filling in the final receive status (CRC, FAE, SIZE, etc.).
transfer at address RXDP with a length derived from RXCFG:MXDMA.
DP83816
85 www.national.com
5.0 Buffer Management (Continued)
CR:RXE && CRDD
rxDescRefr
DP83816
CR:RXE && !CRDD
XferDone
link = NULL
rxIdle
link != NULL
rxAdvance
XferDone
(descCnt == 0) && (rxPktBytes > 0)
rxPktBytes == 0
Figure 5-7 Receive State Diagram
5.3.2 Receive Data Flow
With a bus mastering architecture, some number of buffers and descriptors for received packets must be pre-allocated when the DP83816 is initialized. The number allocated will directly affect the system's tolerance to interrupt latency. The more buffers that you pre-allocate, the longer the system will survive an incoming burst without losing receive packets, if receive descriptor processing is delayed or preempted. Buffers sizes should be allocated in 32 byte multiples.
1. Prior to packet reception, receive buffers must be described in a receive descriptor list (or ring, if preferred). In each descriptor, the driver assigns ownership to the hardware by clearing the OWN bit. Receive descriptors may describe a single buffer.
2. The address of the first descriptor in this list is then written to the RXDP register. As packets arrive, they are placed in available buffers. A single packet may occupy one or more receive descriptors, as required by the application.The device reads in the first descriptor into the RxDescCache.
XferDone && OWN
rxDescRead
XferDone && !OWN
FifoReady
rxFifoBlockrxDescWrite
XferDone
rxFragWrite
3. As data arrives in the RxDataFIFO, the receive buffer management state machine places the data in the receive buffer described by the descriptor. This continues until either the end of packet is reached, or the descriptor byte count for this descriptor is reached.
4. If end of packet was reached, the status in the descriptor (in main memory) is updated by setting the OWN bit and clearing the MORE bit, by updating the receive status bits as indicated by the MAC, and by updating the SIZE field. The status bits in cmdsts are only valid in the last descriptor of a packet (with the MORE bit clear). Also for the last descriptor of a packet, the SIZE field will be updated to reflect the actual amount of data written to the buffer (which may be less the full buffer size allocated by the descriptor).
If the receive buffer management state machine runs out of descriptors while receiving a packet, data will buffer in the receive FIFO. If the FIFO overflows, the driver will be interrupted with an RxOVR error.
86 www.national.com
6.0 Power Management and Wake-On-LAN
DP83816
6.1 Introduction
The DP83816 supports Wake-On-LAN (WOL) and the PCI Power Management Specification version 1.1. These features allow the device to enter a power saving mode, and to signal the system to return to a normal operating state when a wake event occurs. This section describes the power management operation on the DP83816.
6.2 Definitions (for this document only)
• Power Management - a PCI specification that defines power-saving states of PCI devices and systems. A spec-compliant device implements two PCI Configu­ration registers to control and report status for its Power Management function.
• Wake event - An event that causes a PCI device in Power Management mode to signal the system.
• PME Enable (PMEEN) - bit 8 of the Power Manage­ment Control/Status Register (PMCSR - offset 44h in the PCI configuration space). Setting this bit to 1 al­lows the device to assert the PMEN pin when it de­tects a wake event.
• Sleep mode - A device is in sleep mode if it is pro­grammed to a Power Management state other than the fully operational state and is not allowed to signal a wake event to the system. In this mode, the PME Enable bit is 0.
• Wake-On-LAN mode - A device is in Wake-On-LAN (WOL) mode if it is programmed to a Power Manage­ment state other than the fully operational state and is allowed to signal a wake event to the system. In this mode, the PME Enable bit is 1.
• PMEN (pin59) - this pin is similar in function to a sys­tem interrupt (INTAN pin). When asserted, it signals the system that a wake event has occurred.
• PME Status - bit 15 of PMCSR. When 1, indicates the device detected a wake event. If PME Enable is also set to 1, the device will assert PMEN whenever PME Status is 1. Software writes a 1 to this bit to clear it.
• Magic Packet: “A specific packet of information sent to remotely wake up a sleeping or powered off PC on a network, it is handled in the LAN controller. The Magic Packet must contain a specific data se­quence which can be located anywhere within the packet but must be preceded by a synchronization stream. The packet must also meet the basic require­ments for the LAN technology chosen (e.g. ethernet frame). The specific data sequence consists of 16 du­plications of the MAC address of the machine to be awakened. The synchronization stream is defined as 6 bytes of FFh.”
• ACPI-compatible operating system - An operating system that takes advantage of the PCI Power Man­agement interface. These include Windows 98 (when installed with ACPI), Windows 2000, and Windows ME (when installed with ACPI).
6.3 Packet Filtering
When the PME Enable bit is set to 1, incoming packets are filtered based on settings in the Receive Filter Control Register (RFCR - offset 48h in operational registers) and the Wake Command/Status Register (WCSR - offset 40h in operational registers). In other words, a packet must pass both filters to be accepted. This is a desirable feature in WOL mode since it prevents non-wake packets from filling the receive FIFO. However, it is not desirable in normal operating mode since it will not allow non-wake packets from being received. Therefore, the driver should ensure that the PME Enable bit is set to 0 for normal operation.
6.4 Power Management
The Power Management Specification presents a low-level hardware interface to PCI devices for the purpose of saving power. The DP83816 supports power states D0, D1, D2, D3hot, and D3cold as defined in the PCI Power Management Specification. These states provide increasing power reduction in the order they are listed. Table 6-1 lists the different Power Management modes and the methods of power reduction in DP83816 devices.
Table 6-1 Power Management Modes
Power State
D0 (SW sets to 0) Unconfigured Normal On On
D1 Don’t Care Don’t Care WOL On On
D2 Don’t Care Don’t Care WOL May be Off On
D3hot Off Don’t Care Sleep May be Off Off
D3hot Don’t Care Unconfigured Sleep May be Off Off
D3hot On Configured WOL May be Off On
D3cold Off Don’t Care Sleep Off Off
D3cold Don’t Care Unconfigured Sleep Off Off
D3cold On Configured WOL Off On
PME Enable
(PMEEN)
Wake
Conditions
Power Management
Mode
87 www.national.com
PCICLK
Physical Layer
Cell
6.0 Power Management and Wake-On-LAN (Continued)
6.4.1 D0 State
The D0 state is the normal operational state of the device. The PME Enable bit should be set to 0 to prevent packet filtering based on the settings in the Wake Control/Status Register (WCSR). It is also advisable to turn off all WOL conditions in WCSR to prevent unnecessary PME interrupts.
6.4.2 D1 State
The D1 state is the least power-saving Power Management state, and might not be used by the operating system. The device will only respond to PCI configuration transactions and therefore will not transmit data. The only bus activity the device can initiate is the assertion of the PMEN pin (assuming the PME Enable bit is set to 1); no DMA activity or interrupts will occur. The device will continue to receive packets up to the limit of the receive FIFO size. Upon returning to the D0 state, the system must re-enable I/O and memory space in the device and turn on bus master capability.
6.4.3 D2 State
The D2 state has the same features as the D1 state, and the system may turn off the PCI clock, further reducing power. The device will continue to receive packets up to the limit of the receive FIFO size. Like the D1 state, the D2 state might not be used by the operating system.
6.4.4 D3hot State
The D3hot state is often known as the Standby state. If the PME Enable bit is 0, or WOL is unconfigured, the device saves power by turning off the Physical Layer Cell (PHY). The system may turn off the PCI clock. In order to receive packets in the D3hot state, both WOL mode and PME Enable must be turned on. Like the D2 and D1 states, the device will respond to PCI configuration transactions as long as the PCI clock is running.
When the device exits the D3hot state, all PCI configuration registers except for the PME Enable and PME Status bits are reset to their default values. This means the operating system must reinitialize the device’s PCI configuration registers with valid base addresses, etc. If PME Enable or WOL mode were not turned on, the device must be fully reinitialized.
6.4.5 D3cold State
The D3cold state is the highest power-saving state; it is often known as the Hibernate state. The PCI bus is turned off, as is the PCI clock. If the PME Enable bit or WOL is turned off, the PHY is turned off. This allows the device to consume the least amount of power. The device must be fully reinitialized after exiting this mode.
6.5 Wake-On-LAN (WOL) Mode
Wake-On-LAN Mode is a system-level function that allows a network device to alert the system that a wake event has occurred. It works in conjunction with the PCI Power Management states detailed in the previous section. The DP83816 supports several wake events including, but not limited to, Wake on PHY Interrupt (i.e. link change), Wake on Magic Packet, and Wake on Pattern Match. The supported wake events appear in the device’s Wake Command/Status Register (WCSR).
6.5.1 Entering WOL Mode
The following steps are required to place the DP83816 into WOL mode:
1. Disable the receiver by writing a 1 to the Receiver Dis­able bit 3 (RXD) in the Command Register (CR - offset 00h in operational registers).
2. Write 0 to the Receive Descriptor Pointer Register (RXDP - offset 30h in operational registers) to reset the receive pointer.
3. Enable the receiver (now in “silent receive” mode) by writing a 1 to the Receiver Enable bit 2 in the Com­mand Register (CR:RXE).
4. Configure the Receive Filter Control Register (RFCR) to enable the receive filter (RFCR:RFEN - bit 31) and accept the desired type of wakeup packets. Note that the Receive Filter Enable bit must be set to 1 for Wake on PHY Interrupt as well.
5. If Wake on PHY Interrupt is desired, additionally con­figure registers MICR (offset C4h in operational regis­ters) and MISR (offset C8h in operational registers).
6. Configure the Wake Command/Status Register (WCSR) with the desired type of wake events. An ACPI-compatible operating system should notify the driver of these events.
7. Write a 1 to PME Enable, and set the desired Power State in PMCSR. These can be done in one operation, or PME Enable can be written first. An ACPI-compati­ble operating system should handle this step.
8. If the Power Management state is D3cold, the system will assert PCI reset, stop the PCI clock, and remove power from the PCI bus.
The following two examples show the corresponding register settings for Wake on Magic Packet mode and Wake on PHY Interrupt mode respectively:
Entering Wake on Magic Packet mode:
1. CR = 00000008h (disable the receiver)
2. RXDP = 00000000h (reset the receive pointer)
3. CR = 00000004h (enable the receiver)
4. RFCR = F0000000h (enables the receive filter and allows Broadcast, Multicast and Unicast packets to be received - a Magic Packet could be any of those.)
5. WCSR = 00000200h (sets the Wake on Magic Packet bit)
6. PMCSR = 00008103h (clears the PME status bit 15, sets the PME Enable bit 8 and sets the Power State bits [1:0] to D3hot)
Entering Wake on PHY Interrupt mode:
1. CR = 00000008h (disable the receiver)
2. RXDP = 00000000h (reset the receive pointer)
3. CR = 00000004h (enable the receiver)
4. RFCR = 80000000h (enables the receive filter)
5. MICR = 00000002h (sets the Interrupt Enable bit 1)
6. MISR = 00000000h (unmasks the change of link sta­tus event)
7. WCSR = 00000001h (sets the Wake on PHY interrupt bit)
8. PMCSR = 00008103h (clears the PME status bit 15, sets the PME Enable bit 8 and sets the Power State bits [1:0] to D3hot)
DP83816
88 www.national.com
6.0 Power Management and Wake-On-LAN (Continued)
6.5.2 Wake Events
If the device detects a wake event while in WOL mode, it will assert the PMEN pin low to signal the system that a wake event has occurred. The system should then bring the device out of WOL mode as described below.
6.5.3 Exiting WOL Mode
The following steps are required to bring the device out of WOL mode (with or without an accompanying wake event):
1. If the Power Management state is D3cold, the system will assert PCI reset, restore PCI bus power, and restart the PCI clock. This will also return the Power State to D0. The PCI configuration registers (i.e. base addresses, bus master enable, etc.) must be reinitial­ized.
2. Write a 0 to Power State bits [0:1] in the PMCSR (in case the WOL Power State was not D3hot or D3cold) and PME Enable. These can be done in one opera­tion, or Power State can be written first. Turning off PME Enable will cause the device to de-assert the PMEN pin, if it was asserted.
3. If the WOL Power State was D3hot or D3cold, reinitial­ize the PCI configuration registers (i.e. base addresses, bus master enable, etc.). An ACPI-com­patible operating system should handle this step. Note that operational registers will not be accessible until this step is completed.
4. If a wake event occurred, read the WCSR to deter­mine what the event was.
5. Write a 1 to PME Status. This will clear any wake event in the device. An ACPI-compatible operating system will perform this write to the PMCSR; a driver can perform this write using the Clockrun Control/Sta­tus Register (CCSR).
6. If the wake event was a PHY interrupt from an internal PHY, clear the event in the PHY registers. Refer to the MISR in Section 4.3.11.
7. Clear all bits in WCSR.
8. Disable the receiver by writing a 1 to the Receiver Dis­able bit in the Command Register (CR:RXD).
9. Reconfigure RFCR as appropriate for normal opera­tion.
10. Write a valid receive descriptor pointer to the Receive Descriptor Pointer Register (RXDP)
11. Enable the receiver by writing a 1 to the Receiver Enable bit in the Command Register (CR:RXE). If the wake event was a packet, this will now be emptied from the receive FIFO via DMA.
6.6 Sleep Mode
Sleep Mode is a system-level function that allows a device to be placed in a lower power mode than WOL mode. In sleep mode, the device will not be able to detect wake events or signal the system that it needs service.
6.6.1 Entering Sleep Mode
The following steps are required to enter Sleep Mode:
1. Disable the receiver by writing a 1 to the Receiver Dis­able bit in the Command Register (CR:RXD).
2. Write 0 to the Receive Descriptor Pointer Register (RXDP)
3. Force the receiver to reread the descriptor pointer by writing a 1 to the Receiver Enable bit in the Command Register (CR:RXE).
4. Do not configure any wake events in WCSR.
5. Write a 0 to PME Enable, and set the desired Power State in PMCSR. These can be done in one operation. An ACPI-compatible operating system should handle this step.
6. If the Power Management state is D3cold, the system will assert PCI reset, stop the PCI clock, and remove power from the PCI bus.
6.6.2 Exiting Sleep Mode
The following steps are required to bring the DP83816 out of Sleep Mode:
1. If the Power Management state is D3cold, the system will assert PCI reset, restore PCI bus power, and restart the PCI clock. This will also return the Power State to D0. The PCI configuration registers (i.e. base addresses, bus master enable, etc.) must be reinitial­ized.
2. Write a 0 to Power State bits [0:1] in the PMCSR (in case the sleep Power State was not D3hot or D3cold).
3. If the sleep Power State was D3hot or D3cold, reinitial­izeaddresses, bus master enable, etc.). An ACPI-com­patible operating system should handle this step. Note that operational registers will not be accessible until this step is completed.
4. Disable the receiver by writing a 1 to the Receiver Dis­able bit in the Command Register (CR:RXD).
5. Write a valid receive descriptor pointer to the Receive Descriptor Pointer Register (RXDP)
6. Enable the receiver by writing a 1 to the Receiver Enable bit in the Command Register (CR:RXE).
6.7 Pin Configuration for Power Management
Refer to Table 6-2 for proper pin connection for power management configuration:
Pin Name Pin No. Power Mgt No Power Mgt
PMEN 59 *PME# 3.3V
3VAUX 122 *3.3Vaux GND
Note: *Refer to Demo Board schematics for additional information.
DP83816
Table 6-2 PM Pin Configuration
89 www.national.com
7.0 DC and AC Specifications
DP83816
Absolute Maximum Ratings
Supply Voltage (VDD) -0.5 V to 3.6 V
DC Input Voltage (V
DC Output Voltage (V
Storage Temperature Range (T
Power Dissipation (P
Lead Temp. (T
ESD Rating
(R
ZAP
ESD for TPTD + pins only
θ
(@0 cfm, 0.5 Watt) 46.4 °C/W
ja
(@1 Watt) 9.29 °C/W
θ
jc
) -0.5 V to 5.5 V
IN
) -0.5 V to VDD + 0.5 V
OUT
) -65 °C to 150 °C
STG
) 504 mW
D
) (Soldering, 10 sec) 260 °C
L
= 1.5k, C
= 120 pF)
ZAP
2.0 KV
1.0 KV
Recommended Operating Conditions
Supply voltage (V
Normal Operating Temperature (T
Note: Absolute maximum ratings are values beyond which
operation is not recommended or guaranteed. Extended exposure beyond these limits may affect device reliability. They are not meant to imply that the device should be operated at these limits.
) 3.3 Volts + 0.3V
DD
)
A
0 to 70
7.1 DC Specifications
TA = 0oC to 70oC, VDD = 3.3 V ±0.3V, unless otherwise specified
Symbol Parameter Conditions Min Typ Max Units
V
OH
V
OL
V
IH
V
I
IN
I
OZ
I
DD
Minimum High Level Output Voltage IOH = -6 mA (Note 1) VDD - 0.5 V
Maximum Low Level Output Voltage IOL = 6 mA (Note 1) 0.4 V
Minimum High Level Input Voltage
Maximum Low Level Input Voltage
IL
(Note 1)
(Note 1)
2.0 V
Input Current VIN = VDD or GND -10 10 µA
TRI-STATE Output Leakage Current V
Operating Supply Current I
= VDD or GND -10 10 µA
OUT
= 0 mA (Note 2) 116 140 mA
OUT
WOL Standby 90 105 mA
Sleep Mode 16 20 mA
0.8 V
°C
R
INdiff
V
TPTD_100
V
TPTDsym
V
TPTD_10
C
IN
C
OUT
SD
THon
SD
THoff
V
TH1
Note 1: These values ensure 3.3V and 5V compatibility.
Note 2: For I
Differential Input Resistance RD+/− 56 k
100 Mb/s Transmit Voltage TD+/− 0.95 1 1.05 V
100 Mb/s Transmit Voltage Symmetry
10 Mb/s Transmit Voltage TD+/− 2.2 2.5 2.8 V
CMOS Input Capacitance 8 pF
CMOS Output Capacitance 8 pF
100BASE-TX Signal detect turn-on threshold
100BASE-TX Signal detect turn-off threshold
10BASE-T Receive Threshold RD+/− 300 585 mV
Measurements, outputs are not loaded.
DD
TD+/− ±2%
RD+/− 1000 mV diff
pk-pk
RD+/− 200 mV diff
pk-pk
90 www.national.com
7.0 DC and AC Specifications (Continued)
7.2 AC Specifications
7.2.1 PCI Clock Timing
PCICLK
Number Parameter Min Max Units
7.2.1.1
7.2.1.2
7.2.1.3
PCICLK Low Time 12 ns
PCICLK High Time 12 ns
PCICLK Cycle Time 30 ns
T1
DP83816
T3T3
T2T1 T2
7.2.2 X1 Clock Timing
X1
Number Parameter Min Max Units
7.2.2.1
7.2.2.2
7.2.2.3
T3T3
T1
X1 Low Time 16 ns
X1 High Time 16 ns
X1 Cycle Time 40 40 ns
T2T1 T2
91 www.national.com
7.0 DC and AC Specifications (Continued)
7.2.3 Power On Reset (PCI Active)
Power Stable
T1
RSTN
T2
1st PCI Cycle
PCICLK
Reset Complete
Number Parameter Min Max Units
7.2.3.1
7.2.3.2
Note: Minimum reset complete time is a function of the PCI, transmit, and receive clock frequencies.
Note: Minimum access after reset is dependent on PCI clock frequency. Accesses to DP83816 during this period will be ignored.
Note: EE is disabled for non power on reset.
RSTN Active Duration from PCICLK stable
Reset Disable to 1st PCI Cycle EE Enabled EE Disabled
1ms
1500
1
us us
DP83816
7.2.4 Non Power On Reset
Output
T1
RSTN
Number Parameter Min Max Units
7.2.4.1
Note: Minimum reset complete time is a function of the PCI, transmit, and receive clock frequencies.
RSTN to Output Float 40 ns
1st PCI Cycle
92 www.national.com
7.0 DC and AC Specifications (Continued)
7.2.5 POR PCI Inactive
VDD
T1
EESEL
DP83816
TPRD
T2
Number Parameter Min Max Units
7.2.5.1
7.2.5.2
VDD stable to EE access VDD indicates the digital supply (AUX
power plane, except PCI bus power.) Guaranteed by design.
EE Configuration load duration 2000 us
60 us
93 www.national.com
7.0 DC and AC Specifications (Continued)
7.2.6 PCI Bus Cycles
The following table parameters apply to ALL the PCI Bus Cycle Timing Diagrams contained in this section.
Number Parameter Min Max Units
7.2.6.1
Input Setup Time 7 ns
DP83816
7.2.6.2
7.2.6.3
7.2.6.4
7.2.6.5
7.2.6.6
PCI Configuration Read
PCICLK
FRAMEN
AD[31:0]
T1
T1
T1
C/BEN[3:0]
T1
IDSEL
IRDYN
TRDYN
Input Hold Time 0 ns
Output Valid Delay 2 11 ns
Output Float Delay (t
Output Valid Delay for REQN - point to point 2 12 ns
Input Setup Time for GNTN - point to point 10 ns
time) 28 ns
off
T2
T2
Addr
T3
T3
T4
Data
T2
Cmd
T1
BE
T2
T2
T2
T1
T4
T3
T3
T4
DEVSELN
PAR
PERRN
T1
T2
T3
T3
T1
94 www.national.com
T4
T1
T2
7.0 DC and AC Specifications (Continued)
PCI Configuration Write
PCICLK
T1
FRAMEN
T2
DP83816
AD[31:0]
T1
T1
C/BEN[3:0]
T1
IDSEL
IRDYN
TRDYN
DEVSELN
PAR
PERRN
PCI Bus Master Read
PCICLK
FRAMEN
AD[31:0]
C/BEN[3:0]
IRDYN
TRDYN
DEVSELN
PAR
PERRN
T2
Addr
T2
Cmd
T2
T3
T3
Addr
T3
Cmd
T1
T1
T1
T1
T3
T4
T3
Data
T2
T3
T3
T3
BE
T4
T4
BE
T1
T1
T1
T3
T3
T2
T2
T2
T3
T1
T1
T2
Data
T4
T3
T2
T2
T4
T4
T2
T4
T2
T3
T4
T4
95 www.national.com
7.0 DC and AC Specifications (Continued)
PCI Bus Master Write
PCICLK
DP83816
FRAMEN
AD[31:0]
C/BEN[3:0]
IRDYN
TRDYN
DEVSELN
PAR
PERRN
T3
T3
Addr
T3
Cmd
T3
T3
Data
T3
T3
T3
BE
T4
T3
T1
T1
T1
T4
T4
T2
T2
T3
T4
T4
T2
T4
PCI Target Read
PCICLK
FRAMEN
AD[31:0]
C/BEN[3:0]
IRDYN
TRDYN
DEVSELN
PAR
PERRN
T1
T1
T1
T2
T2
Addr
T2
Cmd
T1
T1
T1
BE
T3
T2
T3
T3
T1
T3
T3
T3
T4
Data
T2
T2
T4
T4
T4
T1
T2
T4
96 www.national.com
7.0 DC and AC Specifications (Continued)
PCI Target Write
PCICLK
T1
FRAMEN
T2
DP83816
AD[31:0]
C/BEN[3:0]
IRDYN
TRDYN
DEVSELN
PAR
PERRN
PCI Bus Master Burst Read
PCICLK
FRAMEN
AD[31:0]
C/BEN[3:0]
IRDYN
T1
Addr
T1
Cmd
T3
T3
Addr
T3
T2
T2
Cmd
T1
T1
T1
T1
T4
T3
Data
T2
T3
T3
BE
BE
T3
T1
Data Data Data
T3
T3
T3
T2
T2
T4
T4
T2
T4
T4
T2
T4
T4
T2
TRDYN
DEVSELN
PAR
PERRN
T3
T4
T1
T1
T2
T3
T2
T4
T1
97 www.national.com
7.0 DC and AC Specifications (Continued)
PCI Bus Master Burst Write
PCICLK
DP83816
FRAMEN
AD[31:0]
C/BEN[3:0]
IRDYN
TRDYN
DEVSELN
PAR
PERRN
T3
T3
Addr
T3
Cmd
T3
Data
T3
T3
T1
T1
T3
BE
Data
T3
T3
Data
T1
T4
T4
T2
T2
T3
T4
T4
T4
T2
PCI Bus Arbitration
PCICLK
REQN
GNTN
T5
T5
T6
T2
98 www.national.com
7.0 DC and AC Specifications (Continued)
7.2.7 EEPROM Auto-Load
DP83816
EECLK
EESEL
EEDO
EEDI
T1T1
T2
T3
T4
T6
T5
Refer to NM93C46 data sheet
Number Parameter Min Max Units
7.2.7.1
7.2.7.2
7.2.7.3
7.2.7.4
7.2.7.5
EECLK Cycle Time 4 us
EECLK Delay from EESEL Valid 1 us
EECLK Low to EESEL Invalid 2 us
EECLK to EEDO Valid 2 us
EEDI Setup Time to EECLK 2 us
7.2.7.6
EEDI Hold Time from EECLK 2 us
99 www.national.com
7.0 DC and AC Specifications (Continued)
7.2.8 Boot PROM/FLASH
DP83816
MCSN
MRDN
MA[15:0]
MD[7:0]
MWRN
Number Parameter Min Typ Units
7.2.8.1
7.2.8.2
7.2.8.3
T13
T17
T5
T2
T1
Data Setup Time to MRDN Invalid 20 ns
Address Setup Time to MRDN
Address Hold Time from MRDN Invalid 0 ns
T14
T3 T4
T15 T16
T9
T6
T8
T10
Valid
T11T12
T7
30 ns
7.2.8.4
7.2.8.5
7.2.8.6
7.2.8.7
7.2.8.8
7.2.8.9
7.2.8.10
7.2.8.11
7.2.8.12
7.2.8.13
7.2.8.14
7.2.8.15
7.2.8.16
7.2.8.17
Note: T10 is guaranteed by design.
Note: Timings are based on a 30ns PCI clock period.
Address Invalid from MWRN Valid 180 ns
MRDN Pulse Width 180 ns
Data Hold Time from MRDN Invalid 0 ns
Data Invalid from MWRN Invalid 60 ns
Data Valid to MWRN Valid 30 ns
Address Setup Time to MWRN Valid 30 ns
MRDN Invalid to MWRN Valid 150 ns
MWRN Pulse Width 150 ns
Address/MRDN Cycle Time 210 ns
MCSN Valid to MRDN Valid 30 ns
MCSN Invalid to MRDN Invalid 0 ns
MCSN Valid to MWRN Valid 30 ns
MWRN Invalid to MCSN Invalid 30 ns
MCSN Valid to address Valid 0 ns
100 www.national.com
Loading...