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
DP83816 is a single-chip 10/100 Mb/s Ethernet Controller
for the PCI bus. It is targeted at low-cost, high volume PC
motherboards, adapter cards, and embedded systems.
The DP83816 fully implements the V2.2 33 MHz PCI bus
interface for host communications with power management
support. Packet descriptors and data are transferred via
bus-mastering, reducing the burden on the host CPU. The
DP83816 can support full duplex 10/100 Mb/s transmission
and reception, with minimum interframe gap.
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,
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.
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,
PCICLK60IClock: This PCI Bus clock provides timing for all bus phases. The rising edge
DEVSELN95I/ODevice Select: As a bus master, the DP83816 samples this signal to insure that the
FRAMEN91I/OFrame: As a bus master, this signal is asserted low to indicate the beginning and
LQFP Pin
No(s)DirDescription
I/OAddress 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/OBus 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.
GNTN63IGrant: This signal is asserted low to indicate to the DP83816 that it has been
IDSEL76IInitialization Device Select: This pin is sampled by the DP83816 to identify when
INTAN61OInterrupt A: This signal is asserted low when an interrupt condition occurs as
IRDYN92I/OInitiator Ready: As a bus master, this signal will be asserted low when the
PAR99I/OParity: This signal indicates even parity across AD[31-0] and CBEN[3-0] including
PERRN97I/OParity Error: The DP83816 as a master or target will assert this signal low to
REQN64ORequest: The DP83816 will assert this signal low to request ownership of the bus
RSTN62IReset: 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
SERRN98I/OSystem Error: This signal is asserted low by DP83816 during address parity errors
STOPN96I/OStop: This signal is asserted low by the target device to request the master device
TRDYN93I/OTarget Ready: As a master, this signal indicates that the target is ready for the data
PMEN/
CLKRUNN
3VAUX122IPCI Auxiliary Voltage Sense: This pin is used to sense the presence of a 3.3V
LQFP Pin
No(s)DirDescription
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.
59I/OPower 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.
PWRGOOD123IPCI 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
COL28ICollision Detect: The COL signal is asserted high asynchronously by the external
CRS29ICarrier Sense: This signal is asserted high asynchronously by the external PMD
MDC5OManagement Data Clock: Clock signal with a maximum rate of 2.5 MHz used to
MDIO4I/OManagement Data I/O: Bidirectional signal used to transfer management
RXCLK6IReceive Clock: A continuous clock, sourced by an external PMD device, that is
RXD3/MA9,
RXD2/MA8,
RXD1/MA7,
RXD0/MA6
RXDV/MA1115IOReceive Data Valid: Indicates that the external PMD is presenting recovered and
LQFP Pin
No(s)DirDescription
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/MA1014IOReceive 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.
RXOE13OReceive Output Enable: Used to disable an external PMD while the BIOS ROM is
being accessed.
TXCLK31ITransmit 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
TXEN30OTransmit 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, TPTDM54, 53A-OTransmit Data: Differential common output driver. This differential common output
TPRDP, TPRDM46, 45A-IReceive Data: Differential common input buffer. This differential common input can
LQFP Pin
No(s)DirDescription
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
MCSN129OBIOS ROM/Flash Chip Select: During a BIOS ROM/Flash access, this signal is
MWRN131OBIOS ROM/Flash Write: During a BIOS ROM/Flash access, this signal is used to
MRDN130OBIOS 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/OBIOS 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.
OBIOS 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
X117ICrystal/Oscillator Input: This pin is the primary clock reference input for the
X218OCrystal Output: This pin is used in conjunction with the X1 pin to connect to an
LQFP Pin
No(s)DirDescription
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/MA0142OTX/RX Activity: This pin is an output indicating transmit/receive activity. This pin is
LED100N/MA2144O100 Mb/s Link: This pin is an output indicating the 100 Mb/s Link status. This pin is
LQFP Pin
No(s)DirDescription
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/MA1143O10 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
EESEL128OEEPROM Chip Select: This signal is used to enable an external EEPROM device.
EECLK/MA42OEEPROM Clock: During an EEPROM access (EESEL asserted), this pin is an
EEDI/MA31OEEPROM Data In: During an EEPROM access (EESEL asserted), this pin is an
EEDO/MD4138IEEPROM Data Out: During an EEPROM access (EESEL asserted), this pin is an
MD1/CFGDISN133I/OConfiguration Disable: When pulled low at power-on time, disables load of
Note: DP83816 supports NMC93C46 for the EEPROM device.
LQFP Pin
No(s)DirDescription
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 KΩ to ground to disable configuration
load.
9 www.national.com
2.0 Pin Description (Continued)
External Reference Interface
DP83816
Symbol
VREF40IBandgap Reference: External current reference resistor for internal Phy bandgap
LQFP Pin
No(s)DirDescription
circuitry. The value of this resistor is 10KΩ 1% metal film (100 ppm/
be connected from the VREF pin to analog ground.
o
C) which must
No Connects and Reserved
Symbol
NC34, 42, 43, 36,
RESERVED41, 50, 127These pins are reserved and cannot be connected to any external logic or net.
REGEN48Reserved and cannot be connected to any external logic or net..
LQFP Pin
No(s)DirDescription
No Connect
37, 84, 85,
124, 125, 126
This pin has an internal weak pull down.
Supply Pins
Symbol
LQFP Pin
No(s)DirDescription
C1 19SConnect to GND through 10uF and 0.1uF external capacitors in parallel.
IAUXVDD39, 47SConnect to isolated Aux 3.3V supply VDD
AUXVDD9, 21, 27, 33,
56, 58, 137
PCIVDD69, 80, 94,
107, 117
VSS8, 16, 20, 26,
32, 35, 38, 44,
49, 51, 52, 55,
57, 65, 77, 90,
103, 114, 136
SConnect to Aux 3.3V supply VDD
SPCI VDD - connect to PCI bus 3.3V VDD
SVSS
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
60b4b6B2B46B-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 Ω 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/S10 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/S10 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.
— (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 Ω
LED100N
453 Ω
LED10N
453 Ω
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
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 100Ω 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:
SDUDN⊕()=
UDSDN⊕()=
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.
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 100Ω termination
network to accommodate UTP cable. The internal
impedance of RD± (typically 1.1KΩ) is in parallel with two
54.9Ω resistors to approximate the 100Ω 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 1KΩ 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 KΩ pull-down resistor on
the MDIO signal. If the PHY will always be connected then
there should be a 1.5 kΩ pull-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.
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
00011110000 000
IdleStart
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
IdleStart
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 000000000000
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
OffsetTagDescriptionAccess
00hCFGIDConfiguration Identification RegisterRO
04hCFGCSConfiguration Command and Status RegisterR/W
08hCFGRIDConfiguration Revision ID RegisterRO
0ChCFGLATConfiguration Latency Timer RegisterRO
10hCFGIOAConfiguration IO Base Address RegisterR/W
44hPMCSRPower Management Control and Status RegisterR/W
48-FFhReserved (reads return zero)
DP83816
4.1.1 Configuration Identification Register
This register identifies the DP83816 Controller to PCI system software.
Tag: CFGIDSize: 32 bitsHard Reset: 0020100Bh
Offset: 00hAccess: Read OnlySoft Reset: Unchanged
BitBit NameDescription
31-16DEVIDDevice ID
This field is read-only and is set to the device ID assigned by National Semiconductor to the DP83816,
which is 0020h.
15-0VENIDVendor 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.
Refer to the description in the PCI V2.2 specification.
30SSERRSignaled SERR
Refer to the description in the PCI V2.2 specification.
29RMABTReceived Master Abort
Refer to the description in the PCI V2.2 specification.
28RTABTReceived Target Abort
Refer to the description in the PCI V2.2 specification.
27STABTSent Target Abort
Refer to the description in the PCI V2.2 specification.
26-25DSTIMDEVSELN Timing
This field will always be set to 01 indicating that DP83816 supports “medium” DEVSELN timing.
24DPDData Parity Detected
Refer to the description in the PCI V2.2 specification.
23FBBFast Back-to-Back Capable
DP83816 will set this bit to 1.
22-21unused
(reads return 0)
20NCPENNew 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.
19-16Unused
(reads return 0)
15-10Unused
(reads return 0)
9FBBENFast 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).
8SERRENSERRN 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.
7Unused
(reads return 0)
DP83816
31 www.national.com
4.0 Register Set (Continued)
BitBit NameDescription
6PERRSPParity 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.
5-3Unused
(reads return 0)
2BMENBus 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.
1MSENMemory Space Address
When set, DP83816 responds to memory space accesses. When reset, DP83816 ignores memory
space accesses.
0I/OSENI/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: CFGRIDSize: 32 bitsHard Reset: 02000000h
Offset: 08hAccess: Read OnlySoft Reset: Unchanged
BitBit NameDescription
31-24BASECLBase Class
Returns 02h which specifies a network controller.
23-16SUBCLSub Class
Returns 00h which specifies an Ethernet controller.
15-8PROGIFProgramming IF
Returns 00h which specifies the first release of the DP83816 Software Interface Specification.
7-0REVIDSilicon 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.
Set by software to the number of PCI clocks that DP83816 may hold the PCI bus.
7-0CLSCache 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 Writefor 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.
This is set by software to the base I/O address for the Operational Register Map.
7-2IOSIZESize indication
Read back as 0. This allows the PCI bridge to determine that the DP83816 requires 256 bytes of I/O
space.
1Unused
(reads return 0).
0 IOINDI/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.
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.
Set to 0 indicating a requirement for 64K bytes of Boot ROM space. Read only.
10-1unused
(reads return 0)
0ROMENROM 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: CAPPTRSize: 32 bitsHard Reset: 00000040h
Offset: 34hAccess: Read OnlySoft Reset: unchanged
BitBit NameDescription
31-8unused
(reads return 0)
7-0CLOFSCapabilities 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.
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.
23-16MNGNTMinimum 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.
15-8IPINInterrupt Pin
Read Only, always return 0000 0001 (INTA).
7-0ILINEInterrupt 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: PMCAPSize: 32 bitsHard Reset: FF820001
Offset: 40hAccess: Read OnlySoft Reset: unchanged
BitBit NameDescription
31-27PMESPME 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.
26D2SD2 Support
This bit is set to a 1 when the DP83816 supports the D2 state.
25D1SD1 Support
This bit is set to a 1 when the DP83816 supports the D1 state.
36 www.national.com
4.0 Register Set (Continued)
BitBit NameDescription
24-22AUX_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)
21DSIDevice 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.
20Reserved
(reads return 0)
19PMECPME Clock
Returns 0 to indicate PCI clock not needed for PMEN.
18-16PMVPower Management Version
This bit field indicates compliance to a specific PM specification rev level. Currently set to 010b.
15-8NLIPTRNext 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.
7-0CAPIDCapability 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.
Sticky bit which represents the state of the PME logic, regardless of the state of the PMEEN bit.
14-9Reserved
(reads return 0)
8PMEENPME 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.
7-2Unused
(reads return 0)
1-0PSTATEPower State
This 2 bit field is used to determine the current power state of DP83816, and to set a new power state.
00 - D010 - D2
01 - D111 - 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
OffsetTagDescriptionAccess
MAC/BIU Registers
00hCRCommand RegisterR/W
04hCFGConfiguration RegisterR/W
08hMEAREEPROM Access RegisterR/W
0ChPTSCRPCI Test Control RegisterR/W
10hISRInterrupt Status RegisterRO
14hIMRInterrupt Mask RegisterR/W
18hIERInterrupt Enable RegisterR/W
1ChIHRInterrupt Holdoff RegisterR/W
20hTXDPTransmit Descriptor Pointer RegisterR/W
24hTXCFGTransmit Configuration RegisterR/W
28-2ChReservedReserved
30hRXDPReceive Descriptor Pointer RegisterR/W
34hRXCFGReceive Configuration RegisterR/W
38ReservedReserved
3ChCCSRCLKRUN Control/Status RegisterR/W
40hWCSRWake on LAN Control/Status RegisterR/W
44hPCRPause Control/Status RegisterR/W
48hRFCRReceive Filter/Match Control RegisterR/W
4ChRFDRReceive Filter/Match Data RegisterR/W
50hBRARBoot ROM AddressR/W
54hBRDRBoot ROM DataR/W
58h SRRSilicon Revision RegisterRO
5ChMIBCManagement Information Base Control RegisterR/W
60-78hMIBManagement Information Base Data RegistersRO
7ChReservedReserved
Internal Phy Registers
80hBMCRBasic Mode Control RegisterR/W
84hBMSRBasic Mode Status RegisterRO
88hPHYIDR1PHY Identifier Register #1RO
8ChPHYIDR2PHY Identifier Register #2RO
90hANARAuto-Negotiation Advertisement RegisterR/W
94hANLPARAuto-Negotiation Link Partner Ability RegisterR/W
98hANERAuto-Negotiation Expansion RegisterR/W
9ChANNPTRAuto-Negotiation Next Page TXR/W
A0-BChReservedReserved
C0hPHYSTSPHY Status RegisterRO
C4hMICRMII Interrupt Control RegisterR/W
C8hMISRMII Interrupt Status RegisterR/W
CChReservedReserved
D0hFCSCRFalse Carrier Sense Counter RegisterR/W
D4hRECRReceive Error Counter RegisterR/W
D8hPCSR100 Mb/s PCS Configuration and Status RegisterR/W
DCh-E0hReservedReserved
E4hPHYCRPHY Control RegisterR/W
E8hTBTSCR10Base-T Status/Control RegisterR/W
ECh-FChReservedReserved
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.
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.
7SWISoftware Interrupt
Setting this bit to a 1 forces the DP83816 to generate a hardware interrupt. This interrupt is mask-able
via the IMR.
6unused
5RXRReceiver 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.
4TXRTransmit 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.
3RXDReceiver 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.
2RXEReceiver 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).
1TXDTransmit 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.
0TXETransmit 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.
Link status of the internal phy. Asserted when link is good. RO
30SPEED100Speed 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
29FDUPFull 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
28POL10 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
27ANEG_DNAuto-negotiation Done
Auto-negotiation done indicator from internal phy. Asserted when auto-negotiation process has
completed or is not active. RO
26-24unused
23-18PHY_CFGPhy Configuration
Miscellaneous internal phy Power-On-Reset configuration control bits.
17PINT_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
16PAUSE_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
15-13ANEG_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
101 Auto-negotiation enabled, advertise 100 Mb/s half & full duplex
111 Auto-negotiation enabled, advertise 10/100 Mb/s half & full duplex
12EXT_PHYExternal 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
11Reserved
DP83816
40 www.national.com
4.0 Register Set (Continued)
BitBit NameDescription
10PHY_RSTReset 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
9PHY_DISDisable internal Phy
When set to a 1, this bit forces the internal phy to its low-power state. R/W
8EUPHCOMP 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
7REQALGPCI 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
6SBSingle 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
5POWProgram 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
4EXDExcessive 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
3PESELParity 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
2BROM_DIS Disable Boot ROM interface
When set to 1, this bit inhibits the operation of the Boot ROM interface logic. R/W
1Reserved
(reads return 0)
0BEMBig 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.
Controls the value of the MDC pin. When set, the MDC pin is 1; when clear the MDC pin is 0. R/W
5MDDIRMII 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
4MDIOMII Management Data
Software access to the MDIO pin (see MDDIR above). R/W
3EESELEEPROM 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
2EECLKEEPROM 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
1EEDOEEPROM 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
0EEDIEEPROM Data In
Controls the value of the EEDI pin. R/W
DP83816
4.2.4 EEPROM Map
EEPROM
Address
0000hCFGSID[0:15]D008h
0001hCFGSID[16:31]0400h
0002hCFGINT[24:31],CFGINT[16:23]2CD0h
0003hCFGCS[20],PMCAP[31],PMCAP[21],PMCSR[8],
0004hSOPAS[1:16]0000h
0005hSOPAS[17:32]0000h
0006hSOPAS[33:47],PMATCH[0]000Nh
0007hPMATCH[1:16]NNNNh
0008hPMATCH[17:32]NNNNh
0009hPMATCH[33:47],WCSR[0]NNNNh
000AhWCSR[1:4],WCSR[9:10],RFCR[20],RFCR[22],
000Bhchecksum valueXX55
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.
Setting this bit to 1 allows the SRAM BIST engine to be reset. R/W
9-8Reserved for NSC internal use only.
Must be written as a 00 otherwise. R/W
7RBIST_ENSRAM BIST Enable
Setting this bit to 1 starts the SRAM BIST engine. R/W
6RBIST_DONESRAM 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
5RBIST_RXFAILRX FIFO BIST Fail
This bit is set to 1 if the SRAM BIST detects a failure in the RX FIFO SRAM. RO
4RBIST_TXFAILTX FIFO Fail
This bit is set to 1 if the SRAM BIST detects a failure in the TX FIFO SRAM. RO
3RBIST_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
2EELOAD_ENEnable 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
1EEBIST_ENEnable 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
0EEBIST_FAILEE 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.
Indicates that a requested transmit reset operation is complete.
24RXRCMPReceive Reset Complete
Indicates that a requested receive reset operation is complete.
23DPERRDetected Parity Error
This bit is set whenever CFGCS:DPERR is set, but cleared (like all other ISR bits) when the ISR register
is read.
22SSERRSignaled System Error
The DP83816 signaled a system error on the PCI bus.
21RMABTReceived Master Abort
The DP83816 received a master abort generated as a result of target not responding.
20RTABTReceived Target Abort
The DP83816 received a target abort on the PCI bus.
19-17unused
16RXSOVRRx Status FIFO Overrun
Set when an overrun condition occurs on the Rx Status FIFO.
15HIBERRHigh Bits Error Set
A logical OR of bits 25-16.
14PHYPhy interrupt
Set to 1 when internal phy generates an interrupt.
13PMEPower Management Event
Set when WOL conditioned detected.
12SWISoftware Interrupt
Set whenever the SWI bit in the CR register is set.
11MIBMIB Service
Set when one of the enabled management statistics has reached its interrupt threshold. (See
Section 4.2.24)
10TXURNTx Underrun
Set when a transmit data FIFO underrun condition occurs.
9TXIDLETx 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).
8TXERRTx 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)
BitBit NameDescription
7TXDESCTx Descriptor
This event is signaled after a transmit descriptor when the INTR bit in the CMDSTS field has been
updated.
6TXOKTx Packet OK
This event is signaled after the last transmit descriptor in a successful transmission attempt has been
updated with valid status.
5RXORNRx Overrun
Set when a receive data FIFO overrun condition occurs.
4RXIDLERx 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).
3RXEARLYRx 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.
2RXERRRx Packet Error
This event is signaled after the last receive descriptor in a failed packet reception has been updated with
valid status.
1RXDESCRx Descriptor
This event is signaled after a receive descriptor with the INTR bit set in the CMDSTS field has been
updated.
0RXOKRx 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.
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.
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.
7-0IHInterrupt 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.
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).
1-0unused
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.
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.
30HBIHeartBeat 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
29MLBMAC 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
28ATPAutomatic 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)
BitBit NameDescription
27-26IFGInterframe 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
25-24Reserved for NSC internal use only.
Must be written as a 00 otherwise. R/W
23ECRETRYExcessive 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.
22-20MXDMAMax 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.
19unused
18Reserved 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.
17-14unused
13-8FLTHTx 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.
7-6unused
5-0DRTHTx 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.
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.
1-0unused
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.
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.
30ARPAccept 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.
29unused
28ATXAccept 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.
27ALPAccept 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.
26unused
25-23unused
Writes are ignored, reads return 000b.
22-20MXDMAMax 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)
19-6unused
DP83816
51 www.national.com
4.0 Register Set (Continued)
BitBit NameDescription
5-1DRTHRx 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.
0Reserved
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.
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.
14-9reserved
(reads return 0)
8PMEENPME 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.
7-1unused
(reads return 0)
0CLKRUN_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.
Set to 1 if a Magic Packet has been detected and the WKMAG bit is set. RO, cleared on read.
30PATM 3Pattern 3 match
Associated bit set to 1 if a pattern 3 match is detected and the WKPAT3 bit is set. RO, cleared on read.
29PATM 2Pattern 2 match
Associated bit set to 1 if a pattern 2 match is detected and the WKPAT2 bit is set. RO, cleared on read.
28PATM 1Pattern 1 match
Associated bit set to 1 if a pattern 1 match is detected and the WKPAT1 bit is set. RO, cleared on read.
27PATM 0Pattern 0 match
Associated bit set to 1 if a pattern 0 match is detected and the WKPAT0 bit is set. RO, cleared on read.
26ARPRARP Received
Set to 1 if an ARP packet has been detected and the WKARP bit is set. RO, cleared on read.
25BCASTRBroadcast Received
Set to 1 if a broadcast packet has been detected and the WKBCP bit is set. RO, cleared on read.
24MCASTRMulticast Received
Set to 1 if a multicast packet has been detected and the WKMCP bit is set. RO, cleared on read.
23UCASTRUnicast Received
Set to 1 if a unicast packet has been detected the WKUCP bit is set. RO, cleared on read.
22PHYINTPhy Interrupt
Set to 1 if a Phy interrupt was detected and the WKPHY bit is set. RO, cleared on read.
21ReservedReserved
RO, cleared on read.
20SOHACKSecureOn 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.
19-11unused
returns 0
10MPSOEMagic Packet SecureOn Enable
Enable Magic Packet SecureOn feature. Only applicable when bit 9 is set. R/W
9WKMAGWake on Magic Packet
Enable wake on Magic Packet detection. R/W
8WKPAT 3Wake on Pattern 3 match
Enable wake on match of pattern 3. R/W
7WKPAT 2Wake on Pattern 2 match
Enable wake on match of pattern 2. R/W
6WKPAT 1Wake on Pattern 1 match
Enable wake on match of pattern 1. R/W
DP83816
54 www.national.com
4.0 Register Set (Continued)
BitBit NameDescription
5WKPAT 0Wake on Pattern 0 match
Enable wake on match of pattern 0. R/W
4WKARPWake on ARP
Enable wake on ARP packet detection. R/W
3WKBCPWake on Broadcast
Enable wake on broadcast packet detection. R/W
2WKMCPWake on Multicast
Enable wake on multicast packet detection. R/W
1WKUCPWake on Unicast
Enable wake on unicast packet detection. R/W
0WKPHYWake 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.
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
30PS_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
29PS_DAPause 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
28-24unused
returns 0
23PS_ACTPause Active
This bit is set to a 1 when the TX MAC logic is actively timing a pause interval. RO
22PS_RCVDPause 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
21PSNEGPause 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
20-17unused
returns 0
16MLD_ENManual 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
15-0PAUSE_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.
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.
30AABAccept 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.
29AAMAccept 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.
28AAUAccept 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.
27APMAccept 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.
26-23APATAccept 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.
22AARPAccept 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.
21MHENMulticast 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.
20UHENUnicast 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.
19ULMU/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.
18-10Unused
returns 0
DP83816
57 www.national.com
4.0 Register Set (Continued)
BitBit NameDescription
9-0RFADDRReceive Filter Extended Register Address
Selects which internal receive filter register is accessible via RFDR:
Used as byte mask values for pattern match template data.
15-0RFDATAReceive 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
# 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 Xbyte63byte6223E
X Xbyte61byte6023C
X Xbyte5byte4204
X Xbyte3byte2202
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 Xbyte1byte0200
Bit# 17 16 15
870
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.
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.
4.2.23 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.
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.
2ACLRClear 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.
1FRZFreeze 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.
0WRNWarning 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.
OffsetTagSize
0060hRXErroredPkts168Packets received with errors. This counter is incremented for each
0064hRXFCSErrors84Packets received with frame check sequence errors. This counter is
0068hRXMsdPktErrors84Packets missed due to FIFO overruns. This counter is incremented
006ChRXFAErrors84Packets received with frame alignment errors. This counter is
0070hRXSymbolErrors84Packets received with one or more symbol errors. This counter is
0074hRXFrameTooLong42Packets received with length greater than 1518 bytes (too long
0078hTXSQEErrors42Loss 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: BMCRSize: 16 bitsHard Reset: XX00h
Offset: 0080hAccess: Read Write
BitBit NameDescription
15ResetReset: 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).
14LoopbackLoopback: 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.
13Speed
12Auto-
11Power Down Power Down: Default: 0
10IsolateIsolate: Default: 0
9Restart Auto-
8Duplex 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)
BitBit NameDescription
7Collision 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.
6:0ReservedReserved: Default: 0, RO
4.3.2 Basic Mode Status Register
Tag: BMSRSize: 16 bitsHard Reset: 7849h
Offset: 0084hAccess: Read Only
BitBit NameDescription
15100BASE-T4 100BASE-T4 Capable: Default: 0
0 = Device not able to perform 100BASE-T4 mode.
14100BASE-TX
Full Duplex
13100BASE-TX
Half Duplex
1210BASE-T
Full Duplex
1110BASE-T
Half Duplex
10:7ReservedReserved: Write as 0, read as 0
6Preamble
Suppression
5Auto-
4Remote Fault Remote Fault: Default: 0/L(H)
3Auto-
2Link StatusLink Status: Default: 0/L(L)
1Jabber Detect Jabber Detect: Default: 0/LH
0Extended
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: PHYIDR1Size: 16 bitsHard Reset: 2000h
Offset: 0088hAccess: Read Only
BitBit NameDescription
15:0OUI_MSBOUI 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: PHYIDR2Size: 16 bitsHard Reset: 5C21h
Offset: 008ChAccess: Read Only
DP83816
BitBit NameDescription
15:10OUI_LSBOUI 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.
9:4VNDR_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).
3:0MDL_REVModel 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: ANARSize: 16 bitsHard Reset: 05E1h
Offset: 0090hAccess: Read Write
BitBit NameDescription
15NPNext Page Indication: Default: 0
0 = Next Page Transfer not desired
1 = Next Page Transfer desired
14ReservedReserved by IEEE: Writes ignored, Read as 0
13RFRemote Fault: Default: 0
1 = Advertises that this device has detected a Remote Fault
0 = No Remote Fault detected
12:11ReservedReserved for Future IEEE use: Write as 0, Read as 0
10PAUSEPAUSE: 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)
BitBit NameDescription
9T4100BASE-T4 Support: Default: 0/ RO
1= 100BASE-T4 is supported by the local device
0 = 100BASE-T4 not supported
8TX_FD100BASE-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
7TX100BASE-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
610_FD10BASE-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
51010BASE-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
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: ANLPARSize: 16 bitsHard Reset: 0000h
Offset: 0094hAccess: Read Only
BitBit NameDescription
15NPNext Page Indication:
0 = Link Partner does not desire Next Page Transfer
1 = Link Partner desires Next Page Transfer
14ACKAcknowledge:
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.
13RFRemote Fault:
1 = Remote Fault indicated by Link Partner
0 = No Remote Fault indicated by Link Partner
12:10ReservedReserved for Future IEEE use: Write as 0, read as 0
9T4100BASE-T4 Support:
1 = 100BASE-T4 is supported by the Link Partner
0 = 100BASE-T4 not supported by the Link Partner
8TX_FD100BASE-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
7TX100BASE-TX Support:
1 = 100BASE-TX is supported by the Link Partner
0 = 100BASE-TX not supported by the Link Partner
610_FD10BASE-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)
BitBit NameDescription
51010BASE-T Support:
1 = 10BASE-T is supported by the Link Partner
0 = 10BASE-T not supported by the Link Partner
4:0SelectorProtocol 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: ANERSize: 16 bitsHard Reset: 0004h
Offset: 0098hAccess: Read Only
BitBit NameDescription
15:5ReservedReserved: Writes ignored, Read as 0.
4PDFParallel Detection Fault:
1 = A fault has been detected via the Parallel Detection function
0 = A fault has not been detected
3LP_NP_ABLE Link Partner Next Page Able:
1 = Link Partner does support Next Page
0 = Link Partner does not support Next Page
2NP_ABLENext Page Able:
1 = Indicates local device is able to send additional “Next Pages”
1PAGE_RXLink 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
0LP_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: ANNPTRSize: 16 bitsHard Reset: 2001h
Offset: 009ChAccess: Read Write
BitBit NameDescription
15NPNext Page Indication: Default: 0
0 = No other Next Page Transfer desired
1 = Another Next Page desired
14ReservedReserved: Writes ignored, read as 0
13MPMessage Page: Default: 1
1 = Message Page
0 = Unformatted Page
70 www.national.com
4.0 Register Set (Continued)
BitBit NameDescription
12ACK2Acknowledge2: 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.
11TOG_TXTog 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.
10:0CODECode 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: PHYSTSSize: 16 bitsHard Reset: 0000h
Offset: 00C0hAccess: Read Only
BitBit NameDescription
15:14ReservedReserved: Write ignored, read as 0.
13Receive Error
12Polarity
11False Carrier
10Signal Detect Signal Detect: Default: 0, RO/LL
9De-scrambler
8Page
7MII 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.
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)
BitBit NameDescription
6Remote 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
5Jabber 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
4Auto-Neg.
3Loopback
2Duplex Status Duplex:
1Speed Status Speed10:
0Link StatusLink 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: MICRSize: 16 bitsHard Reset: 0000h
Offset: 00C4hAccess: Read Write
BitBit NameDescription
15:2ReservedReserved: Writes ignored, Read as 0
1INTENInterrupt Enable:
1 = Enable event based interrupts
0 = Disable event based interrupts
0TINTTest 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: MISRSize: 16 bitsHard Reset: 0000h
Offset: 00C8hAccess: Read Write
BitBit NameDescription
15MINTMII Interrupt Pending: Default: 0, RO/COR
1 = Indicates that an interrupt is pending and is cleared by the current read.
0 = no interrupt pending
14MSK_LINK Mask Link: When this bit is 0, the change of link status event will cause the interrupt to be seen by the ISR.
13MSK_JABMask Jabber: When this bit is 0, the Jabber event will cause the interrupt to be seen by the ISR.
12MSK_RFMask Remote Fault: When this bit is 0, the Remote Fault event will cause the interrupt to be seen by the
11MSK_ANCMask Auto-Neg. Complete: When this bit is 0, the Auto-negotiation complete event will cause the inter-
10MSK_FHFMask False Carrier Half Full: When this bit is 0, the False Carrier Counter Register half-full event will
9MSK_RHFMask Rx Error Half Full: When this bit is 0, the Receive Error Counter Register half-full event will cause
8:0ReservedReserved: 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.
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.
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
1 = Polarity Sense & Correction disabled
0 = Polarity Sense & Correction enabled
2ReservedReserved
This bit must be written as a one.
1HEARTBEAT_DISHeartbeat 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.
0JABBER_DISJabber 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
OffsetTagDescription
0000hlink32-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.
0004hcmdsts32-bit Command/Status Field (bit-encoded).
0008hbufptr32-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
BitTagDescriptionUsage
31OWNDescriptor OwnershipSet to 1 by the dataproducer of the descriptor to transfer
30MOREMore descriptors Set to 1 to indicate that this is NOT the last descriptor in a packet
29INTRInterruptSet to 1 by software to request a “descriptor interrupt" when
28SUPCRC
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 dataconsumer 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 dataproducer, and the DP83816 is the dataconsumer. 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
27OKPacket OKIn 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)
11-0SIZEDescriptor Byte Count Set to the size in bytes of the data.
Table 5-3 Transmit Status Bit Definitions
BitTagDescriptionUsage
26TXATransmit AbortTransmission of this packet was aborted.
25TFUTransmit FIFO
Underrun
24CRSCarrier Sense LostCarrier was lost during the transmission of this packet. This
23TDTransmit DeferredTransmission of this packet was deferred.
22EDExcessive DeferralThe length of deferral during the transmission of this packet was
21OWCOut of Window
Collision
20ECExcessive CollisionsThe number of collisions during the transmission of this packet
19-16CCNTCollision CountIf 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.
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
BitTagDescriptionUsage
26RXAReceive AbortedSet to 1 by DP83816 when the receive was aborted, the value of
this bit always equals RXO. Exists for backward compatibility.
25RXOReceive OverrunSet to 1 by DP83816 to indicate that a receive overrun condition
occurred. RXA will also be set.
24-23DESTDestination ClassWhen 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.
22LONGToo Long Packet
Received
21RUNTRunt Packet Received The size of the receive packet was less than 64 bytes (inc. CRC).
20ISEInvalid Symbol Error(100 Mb/s only) An invalid symbol was encountered during the
19CRCECRC Error The CRC appended to the end of this packet was invalid.
18FAEFrame Alignment Error The packet did not contain an integral number of octets.
17LBPLoopback PacketThe packet is the result of a loopback transmission.
16COLCollision ActivityThe 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
txIdleThe transmit state machine is idle.
txDescRefrWaiting for the "refresh" transfer of the link field of a completed descriptor from the PCI bus.
txDescReadWaiting for the transfer of a complete descriptor from the PCI bus into the
TxDescriptorCache.
txFifoBlockWaiting for free space in the TxDataFIFO to reach TxFillThreshold.
txFragReadWaiting for the transfer of a fragment (or portion of a fragment) from the PCI bus to the
TxDataFIFO.
txDescWriteWaiting 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.
CTDDAn 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.
descCntCount of bytes remaining in the current descriptor.
fragPtrPointer to the next unread byte in the current fragment.
txFifoCntCurrent amount of data in the txDataFifo in bytes.
txFifoAvailCurrent 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:TXEDriver asserts the TXE bit in the command register (similar to SONIC).
XferDoneCompletion of a PCI bus transfer request.
FifoAvailTxFifoAvail is greater than TxFillThreshold.
81 www.national.com
5.0 Buffer Management (Continued)
Table 5-5 Transmit State Tables
StateEventNext StateActions
txIdleCR:TXE && !CTDDtxDescReadStart a burst transfer at address TXDP and a length
derived from TXCFG.
CR:TXE && CTDDtxDescRefrStart a burst transfer to refresh the link field of the current
txDescRefrXferDonetxAdvance
txDescReadXferDone && OWNtxFIFOblock
XferDone && !OWNtxIdleSet ISR:TXIDLE.
txFIFOblockFifoAvailtxFragReadStart a burst transfer into the TxDataFIFO from fragPtr.
(descCnt == 0) &&
txDescWriteStart a burst transfer to write the status back to the
MORE
(descCnt == 0) &&
txAdvanceWrite the value of TXDP to the txDataFIFO as a handle.
!MORE
txFragReadXferDonetxFIFOblock
txDescWriteXferDonetxAdvance
txAdvancelink != NULLtxDescReadTXDP <- txDescCache.link. Clear CTDD. Start a burst
link == NULLtxIdleSet 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:
rxIdleThe receive state machine is idle.
rxDescRefrWaiting for the "refresh" transfer of the link field of a completed descriptor from the PCI bus.
rxDescReadWaiting for the transfer of a descriptor from the PCI bus into the RxDescCache.
rxFifoBlockWaiting for the amount of data in the RxDataFifo to reach the RxDrainThreshold or to represent a
complete packet.
rxFragWriteWaiting for the transfer of data from the RxDataFIFO via the PCI bus to host memory.
rxDescWriteWaiting for the completion of the write of the cmdsts field of a receive descriptor.
The receive state machine manipulates the following internal data spaces:
RXDPA 32-bit register that points to the current receive descriptor.
CRDDAn 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.
descCntCount of bytes available for storing receive data in all fragments described by the current descriptor.
fragPtrPointer to the next unwritten byte in the current fragment.
rxPktCntNumber 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.
rxPktBytesNumber 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:RXEThe RXE bit in the Command Register has been set.
XferDonecompletion 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
StateEventNext StateActions
rxIdleCR:RXE && !CRDDrxDescReadStart a burst transfer at address RXDP and a length
derived from RXCFG.
CR:RXE && CRDDrxDescRefrStart a burst transfer to refresh the link field of the current
rxDescRefrXferDonerxAdvance
rxDescReadXferDone && !OWNrxFIFOblock
XferDone && OWNrxIdleSet ISR:RXIDLE.
rxFIFOblockFifoReadyrxFragWriteStart a burst transfer from the RxDataFIFO to host
(descCnt == 0) &&
(rxPktBytes > 0)
rxPktBytes == 0rxDescWriteStart a transfer to write the cmdsts back to the descriptor,
rxFragWriteXferDonerxFIFOblock
rxDescWriteXferDonerxAdvance
rxAdvancelink!= NULLrxDescReadRXDP <- rxDescCache.link. Clear CRDD. Start a burst
link == NULLrxIdleSet CRDD. Set ISR:RXIDLE.
rxDescWriteStart 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)UnconfiguredNormalOnOn
D1Don’t CareDon’t CareWOLOnOn
D2Don’t CareDon’t CareWOLMay be OffOn
D3hotOffDon’t CareSleepMay be OffOff
D3hotDon’t CareUnconfiguredSleepMay be OffOff
D3hotOnConfiguredWOLMay be OffOn
D3coldOffDon’t CareSleepOffOff
D3coldDon’t CareUnconfiguredSleepOffOff
D3coldOnConfiguredWOLOffOn
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
PMEN59*PME#3.3V
3VAUX122*3.3VauxGND
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
= 1.5kΩ, 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
SymbolParameterConditionsMinTypMaxUnits
V
OH
V
OL
V
IH
V
I
IN
I
OZ
I
DD
Minimum High Level Output Voltage IOH = -6 mA (Note 1)VDD - 0.5V
Maximum Low Level Output Voltage IOL = 6 mA (Note 1)0.4V
Minimum High Level Input Voltage
Maximum Low Level Input Voltage
IL
(Note 1)
(Note 1)
2.0V
Input CurrentVIN = VDD or GND-1010µA
TRI-STATE Output Leakage Current V
Operating Supply CurrentI
= VDD or GND-1010µA
OUT
= 0 mA (Note 2)116140mA
OUT
WOL Standby90105mA
Sleep Mode1620mA
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 ResistanceRD+/−56kΩ
100 Mb/s Transmit VoltageTD+/−0.9511.05V
100 Mb/s Transmit Voltage
Symmetry
10 Mb/s Transmit VoltageTD+/−2.22.52.8V
CMOS Input Capacitance8pF
CMOS Output Capacitance8pF
100BASE-TX Signal detect turn-on
threshold
100BASE-TX Signal detect turn-off
threshold
10BASE-T Receive ThresholdRD+/−300585mV
Measurements, outputs are not loaded.
DD
TD+/−±2%
RD+/−1000mV diff
pk-pk
RD+/−200mV diff
pk-pk
90www.national.com
7.0 DC and AC Specifications (Continued)
7.2 AC Specifications
7.2.1 PCI Clock Timing
PCICLK
NumberParameterMinMaxUnits
7.2.1.1
7.2.1.2
7.2.1.3
PCICLK Low Time12ns
PCICLK High Time12ns
PCICLK Cycle Time30ns
T1
DP83816
T3T3
T2T1T2
∞
7.2.2 X1 Clock Timing
X1
NumberParameterMinMaxUnits
7.2.2.1
7.2.2.2
7.2.2.3
T3T3
T1
X1 Low Time16ns
X1 High Time16ns
X1 Cycle Time4040ns
T2T1T2
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
NumberParameterMinMaxUnits
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
NumberParameterMinMaxUnits
7.2.4.1
Note: Minimum reset complete time is a function of the PCI, transmit, and receive clock frequencies.
RSTN to Output Float40ns
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
NumberParameterMinMaxUnits
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 duration2000us
60us
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.
NumberParameterMinMaxUnits
7.2.6.1
Input Setup Time7ns
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 Time0ns
Output Valid Delay211ns
Output Float Delay (t
Output Valid Delay for REQN - point to point212ns
Input Setup Time for GNTN - point to point10ns
time)28ns
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
DataDataData
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
NumberParameterMinMaxUnits
7.2.7.1
7.2.7.2
7.2.7.3
7.2.7.4
7.2.7.5
EECLK Cycle Time4us
EECLK Delay from EESEL Valid1us
EECLK Low to EESEL Invalid2us
EECLK to EEDO Valid2us
EEDI Setup Time to EECLK2us
7.2.7.6
EEDI Hold Time from EECLK2us
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
NumberParameterMinTypUnits
7.2.8.1
7.2.8.2
7.2.8.3
T13
T17
T5
T2
T1
Data Setup Time to MRDN Invalid20ns
Address Setup Time to MRDN
Address Hold Time from MRDN Invalid0ns
T14
T3T4
T15T16
T9
T6
T8
T10
Valid
T11T12
T7
30ns
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 Valid180ns
MRDN Pulse Width180ns
Data Hold Time from MRDN Invalid0ns
Data Invalid from MWRN Invalid60ns
Data Valid to MWRN Valid30ns
Address Setup Time to MWRN Valid30ns
MRDN Invalid to MWRN Valid150ns
MWRN Pulse Width150ns
Address/MRDN Cycle Time210ns
MCSN Valid to MRDN Valid30ns
MCSN Invalid to MRDN Invalid0ns
MCSN Valid to MWRN Valid30ns
MWRN Invalid to MCSN Invalid30ns
MCSN Valid to address Valid0ns
100 www.national.com
Loading...
+ hidden pages
You need points to download manuals.
1 point = 1 manual.
You can buy points or you can get point for every manual you upload.