Texas Instruments and its subsidiaries (TI) reserve the right to make changes to their products
or to discontinue any product or service without notice, and advise customers to obtain the latest
version of relevant information to verify , before placing orders, that information being relied on
is current and complete. All products are sold subject to the terms and conditions of sale supplied
at the time of order acknowledgment, including those pertaining to warranty , patent infringement,
and limitation of liability .
TI warrants performance of its semiconductor products to the specifications applicable at the
time of sale in accordance with TI’s standard warranty. Testing and other quality control
techniques are utilized to the extent TI deems necessary to support this warranty . Specific testing
of all parameters of each device is not necessarily performed, except those mandated by
government requirements.
Customers are responsible for their applications using TI components.
In order to minimize risks associated with the customer’s applications, adequate design and
operating safeguards must be provided by the customer to minimize inherent or procedural
hazards.
TI assumes no liability for applications assistance or customer product design. TI does not
warrant or represent that any license, either express or implied, is granted under any patent right,
copyright, mask work right, or other intellectual property right of TI covering or relating to any
combination, machine, or process in which such semiconductor products or services might be
or are used. TI’s publication of information regarding any third party’s products or services does
not constitute TI’s approval, warranty or endorsement thereof.
The TSB12LV32 (GP2Lynx) is a high-performance general-purpose IEEE P1394a link-layer controller
(LLC) with the capability of transferring data between a host controller, the 1394 Phy-link interface, and
external devices connected to the data mover port (local bus interface). The 1394 Phy-link interface
provides the connection to the 1394 physical layer device and is supported by the LLC. The LLC provides
the control for transmitting and receiving 1394 packet data between the microcontroller interface and the
Phy-link interface via internal 2K byte FIFOs at rates up to 400 Mbit/s. The TSB12LV32 transmits and
receives correctly formatted 1394 packets, generates and detects the 1394 cycle start packets,
communicates transaction layer transmit requests to the Phy , and generates and inspects the 32-bit cyclic
redundancy check (CRC). The TSB12L V32 is capable of being cycle master (CM), isochronous resource
manager (IRM), bus manager, and supports reception of isochronous data on two channels.
The TSB12LV32 supports a direct interface to many microprocessors/microcontrollers including
programmable endian swapping. TSB12LV32 has a generic 16/8-bit host bus interface which includes
support for the ColdFire microcontroller mode at rates up to 60 MHz. The microinterface may operate in
byte or word (16 bit) accesses. The data mover block in GP2Lynx is meant to handle an external memory
interface of large data blocks. The port can be configured to either transmit or receive data packets. The
packets can be either asynchronous, isochronous, or streaming data packets. Asynchronous or
isochronous receive packets will be routed to the DM port or the GRF via the receiver routing control logic.
The internal FIFO is separated into a transmit FIFO and a receive FIFO each of 517 quadlets (2 Kbytes).
Asynchronous packets may be transmitted from the DM port or the internal FIFO. If there is contention the
FIFO has priority and will be transmitted first.
The LLC also provides the capability to receive status information from the physical layer device and to
access the physical layer control and status registers by the application software.
1.2TSB12LV32 Features
•Compliant With IEEE 1394-1995 Standards and P1394a Supplement for High Performance
Serial Bus
•Supports Transfer Rates of 400, 200, or 100 Mbit/s
•Compatible With Texas Instruments Physical Layer Controllers (Phys)
•Supports the Texas Instruments Bus Holder Galvanic Isolation Barrier
•Glueless Interface to 68000 and ColdFire Microcontrollers/Microprocessors
•Supports ColdFire Burst Transfers
•2K-Byte General Receive FIFO (GRF) Accessed Through Microcontroller Interface Supports
•Programmable Microcontroller Interface With 8-Bit or 16-Bit Data Bus, Multiple Modes of
Operation Including Burst Mode, and Clock Frequency to 60 MHz.
•8-Bit or 16-Bit Data Mover Port (DM Port) Supports Isochronous, Asynchronous, and Streaming
Transmit/Receive From an Unbuffered Port at a Clock Frequency of 25 MHz.
•Backward Compatible With All TSB12LV31(GPLynx) Microcontroller and Data Mover
Functionality in Hardware.
•Four-Channel Support for Isochronous Transmit From Unbufferred 8/16 Bit Data Mover Port.
•Single 3.3-V Supply Operation With 5-V Tolerance Using 5-V Bias Terminals.
•High Performance 100-Pin PZ Package
†
Implements technology covered by one or more patents of Apple Computer, Incorporated and SGS Thomson, Limited.
The terminal functions are described in Table 1–1.
T able 1–1. Terminal Functions
TERMINAL
NAMENO.
Microcontroller/Microprocessor Interface
BCLK6IMicrointerface clock. Maximum frequency is 60 MHz. In the ColdFire mode,
COLDFIRE12IColdFire mode. To operate in this mode, COLDFIRE must be asserted high.
LENDIAN75ILittle-endian mode for the microinterface. When this terminal is pulled up, the
MA0 – MA624 – 21
19 – 17
M8BIT/SIZ013IConfiguration bit for microinterface. If the microinterface is 8 bits wide, this
MCMODE/SIZ114IMode bit for microinterface. If the microinterface wants to communicate in a
MCA4OMicrointerface cycle acknowledge. When asserted low, MCA signals an
MCS7IMicrointerface cycle start. When asserted low, MCS signals the beginning of a
MDINV11IMicrointerface data invariant mode. This terminal is meaningful only when
MD0 – MD1599 – 96
94 – 91
89 – 86
84 – 81
MWR8IMicrocontroller read/write indicator. When asserted high, MWR indicates a read
TEA3OTransfer error acknowledge. This active-low signal is asserted low for one BCLK
BCLK is the same as CLK, which is the clock-input signal to the ColdFire.
data on MD0–MD15 will be byte-swapped to little endian byte format before it is
written to the CFR or FIFO and after it is read from the CFR or FIFO.
IMicrocontroller address bus. MA0 is the most significant bit (MSB) of these 7 bits.
terminal must be pulled up to the supply voltage. In ColdFire mode, this terminal
represents burst SIZ0.
handshake manner this terminal must be pulled up to the supply voltage. When
the ColdFire mode terminal (12) is high, this terminal represents burst SIZ1.
acknowledge of the microcontroller cycle from the TSB12L V32.
microcontroller operation to the TSB12L V32.
LENDIAN (75) is high. When asserted high, the microinterface operates in the
data invariant mode. When low, the microinterface operates in address invariant
mode.
I/OMicrointerface bidirectional data bus. MD0 is the most significant bit. However,
byte significance is dependent on the state of the LENDIAN and MDINV
terminals.
access from the TSB12LV32. When asserted low , MWR
to the TSB12LV32.
cycle whenever there is an illegal transfer request by the microcontroller (i.e.,
requested data transfer size is unsupported or MCS
one BCLK cycle in ColdFire mode).
indicates a write access
is asserted low for more than
1–4
Page 13
T able 1–1. Terminal Functions (Continued)
I/O
DESCRIPTION
TERMINAL
NAMENO.
Data-Mover Port Interface
DMD0–DMD1526 – 29
31 – 34
36 – 39
41 – 44
DMCLK46OData mover clock at (SCLK/2) MHz
DMDONE50OData mover done. For transmit, this will be activated when the packet per block
DMERROR52OData mover error. DMERROR is asserted high when there is an error in the
DMPRE48OData mover predata indicator. In transmit mode, DMPRE pulses for one DMCLK
DMREADY77IData mover ready. Must be asserted high by the external logic controlling the DM
DMRW49OData mover read/write indicator. When data is being moved from 1394 to the DM
PKTFLAG51OPacket flag. When set, PKTFLAG is asserted high to indicate the first (header) or
CTL0, CTL170, 69I/OPhy-link interface control lines.
D0–D767, 66,
63–58
LINKON64ILink-on from the Phy is a 4 MHz – 8 MHz clock. This signal will be activated when
LPS53OLink power status. LPS is used to drive the LPS input to the Phy. It indicates to the
LREQ74OLink request to Phy. LREQ makes bus requests and register access requests to
SCLK72ISystem clock. SCLK is a 49.152 MHz clock supplied by the Phy. DMCLK is
I/OData mover (DM) bidirectional data port. DMD0 is the MSB of these 16 bits.
counter in the CFR counts down to zero. For receive, this terminal will pulse for
one DMCLK prior to the first byte/word available to the DM interface.
received packet or an illegal transmit speed was attempted.
prior to sending the first quadlet. In isochronous receive mode, DMPRE will pulse
for one DMCLK when the sync bit in the header matches a bit set in the
isochronous register. DMPRE is not used in asynchronous receive mode.
interface when it is ready to supply data for transmit. DMREADY must be set low
when the data mover is in receive mode.
port (receive) this signal will go active high to indicate data is available on
DMD[0:15]. When data is being moved from DM to 1394 bus (transmit) this signal
will go active high to indicate that data must be supplied to the DMD[0:15] port for
transmission.
last (trailer) quadlet of a received packet on the DM interface. PKTFLAG is not
valid in transmit mode.
Phy/Link Interface
I/OPhy-link interface data lines. Data is only expected on D0 and D1 at 100 Mbit/s,
D0–D3 at 200 Mbit/s, and D0–D7 at 400 Mbit/s. D0 is the MSB bit.
the link is inactive and the Phy has detected a link-on packet or a Phy interrupt.
This clock will persist for no more than 500 ns. When the link detects this terminal
as active, it will turn on and drive LPS.
Phy that the link is powered up and active. LPS toggles at a rate = 1/16 of BCLK.
the Phy.
generated from SCLK.
1–5
Page 14
T able 1–1. Terminal Functions (Continued)
I/O
DESCRIPTION
TERMINAL
NAMENO.
Miscellaneous Functions
CONTNDR65I/OContender. When asserted high, this terminal tells the link that this node is a
CYCLEIN76ICycle in. This input is an optional external 8-kHz clock used as the isochronous
CYSTART2OIsochronous cycle start indicator. CYSTAR T signals the beginning an isochronous
DIRECT79IIsolation terminal. When this terminal is asserted high, no isolation is present
GND5, 25, 30,
INT1OInterrupt. NOR of all internal interrupts.
RESET
STAT0–STAT254 – 56OGeneral status outputs. STATn is the output signal selected with the CFR at
TESTMODE16IThis terminal is used to place the TSB12LV32 in the test mode. In normal
VDD5V10, 35, 855 V (± 0.5V) supply voltage for 5-V tolerant inputs. Only the Phy/link interface side
V
DD
45, 57, 73,
78, 90,
100
9I
15, 20, 40,
47, 68, 71,
80, 95
contender for isochronous resource manager (IRM) or bus manager functions.
The state of the CONTNDR must match the state of the Phy contender terminal for
1394-1995 compliant Phys, and the Phy register bit for 1394.A compliant Phys.
This terminal defaults to being an input on power up. After power up, the value of
this terminal may be driven internally by the CTNDRSTAT bit (bit#12 at 08h)
cycle clock. It should only be used if attached to the cycle-master node. It is
enabled by the cycle source bit and should be tied high when not used.
cycle by pulsing for one DMCLK period.
between the TSB12LV32 and the Phy. When low, bus holder isolation becomes
active.
of the TSB12LV32 is not 5-V tolerant. T ie this terminal to the 3.3-V supply voltage if
the TSB12LV32 is not connected to any devices driving 5-V signals. Tie this
terminal to the 5-V supply voltage if the TSB12L V32 is connected to any devices
driving 5-V signals. This terminal is only used to make inputs 5-V tolerant, it is not
used for any outputs.
3.3 V (± 0.3 V) supply voltage
1–6
Page 15
1.5.1STAT0, STAT1, and STAT2 Programming
ST AT0, ST AT1 and ST AT2 terminals can be independently programmed to show one of fourteen possible
internal hardware status. The controls for the ST AT terminals are in the
of the CFR register. STAT0 is controlled by STATSEL0(bits 16–19), STAT1 is controlled by bits
STATSEL1(bits 20–23), and STAT2 is controlled by STATSEL2 (bits 24–27). Refer to Table 1–2 for
programming the STAT terminals.
T able 1–2. STAT Terminals Programming
STATSEL0,
STATSEL1, or
STATSEL2
0000ReservedReserved
0001ATFFULLATF is full. Bit 12 in CFR at 30h.
0010Bus Reset1394 Bus reset. Bit 3 in CFR at 0Ch
0011Arbitration reset gapBit 26 in CFR at 0Ch
0100CYCLEOUT
0101RXDMPKT
0110RXGRFPKT
0111BX_BUSYByte busy. This represents the OR of bits 0 – 3 of CFR at 20h
1000SUBGP
1001CYCLE_DONE
1010
1011DMACKERRDM acknowledge was not Complete. Bit 17 in CFR at 0Ch
1100DMENDM enable. Bit 26 in CFR at 04h
1101
1110ReservedReserved
1111ReservedReserved
STAT0/STAT1/STAT2DESCRIPTION
Cycle out. This is the link’s cycle clock. It is based on the timer
controls and the received cycle-start messages.
Packet received to DM interrupt. Activated at the end of a received
packet. Bit 9 in CFR at 0Ch
Packet received to GRF interrupt. Activated at the end of a received
packet. . Bit 6 of CFR at 0Ch
Subaction gap. Activated upon detection of a subaction gap. Bit 27
in CFR at 0Ch
Cycle done. Indicates the end of the isochronous period. This
happens when a subaction gap has been detected.
ATSTARTED
(default setting for STAT1)
GRFEMPTY
(default setting for STAT2)
Activated when an asynchronous packet transfer has started from
the ATF. Bit 5 in CFR at 0Ch
NOTES: A. All dark gray areas (bits) are reserved bits.
B. All light gray areas are read-only bits. All remaining are read/write bits.
2–2
STATSEL0STATSEL1STATSEL2
PHYRXADPHYRXDATA
CD
ATACKGRFUSED
GRFEMPTY
ACKCODESPD
Diagnostic
Phy Access
Reserved
FIFO Status
Bus Reset
Header0
Header1
Header2
Header3
Trailer
LPS_OFF
LPS_RESET
Asynchronous
Retry
Page 19
DIRECTION
y
Asynchronous
OF DM DATA
TRANSFER
TRANSMIT
(to 1394 Bus)
RECEIVE
(from 1394
Bus)
PACKET TYPE
Isochronous
Asynchronous/
asynchronous
streaming
Isochronous/
asynchronous
streaming
T able 2–2. Header Usage for CFRs 38h–44h
AUTO HEADER
INSERT/
EXTRACT
YES
NO
YES
NO
YES
NO
YES
NO
Header 0 CFR formatted for isochronous transmission.
Header1 – Header3 are used for additional channels.
Isochronous header supplied by DM interface. Header0 CFR
is automatically written with extracted (from transmitted
packet) isochronous header.
Header0–Header3 CFRs formatted for asynchronous transmission.
Asynchronous header supplied by DM interface. Header0 –
Header3 CFRs are automatically written with extracted (from
transmitted packet) header.
Header0 – Header3 are always automatically updated. The
isochronous header is streamed through the DM port. The
trailer quadlet is always appended to the data stream.
Header0 – Header3 are always automatically updated. The
isochronous header is streamed through the DM port along
with the payload data. The trailer quadlet is always appended
to the data stream.
Header0 – Header3 are always automatically updated.
Asynchronous headers are not streamed through the DM
port. The trailer quadlet is always appended to the data
stream.
Header0 – Header3 are always automatically updated.
Asynchronous headers are streamed through the DM port
along with data. The trailer quadlet is always appended to the
data stream.
This register controls the Data Mover port and must be set up before using the port. The power-up reset
value of this register = 0000_0000’h
BIT
NUMBER
0–11PACKET PER-
12ENDSWAPEndian SwapR/WSwap endian. When this bit is set, the quadlet formed by
13BYTEMODEByte ModeR/WByte mode. When this bit is set the DM port will only look
14HANDSHKHandshake
15AUTOUPAutomatic
16–20DMACKDM
21RESERVEDRESERVED
22–23SPEEDDM Speed
BIT NAMEFUNCTIONDIRDESCRIPTION
BLOCK
Packets per
Block
R/WNumber of packets per block. A packet is the size of the
data payload and is specified as part of the header. The
data mover logic uses this value to deactivate DMDONE.
This field is only used in transmit mode.
stacking the DM data will be byte reversed, (i.e. the
quadlet formed by fetching doublet AB01 then ‘CD02’ will
be 02CD–01AB instead of AB01CD02). In byte mode the
quadlet formed by fetching AB, 01, CD, 0 will be
02CD01AB instead of AB01CD02.
at DM0–DM7. DM8–DM15 will be ignored for transmit
and will not be driven on receive. In this mode, the
maximum speed allowed is 200 Mbps.
R/WHandshake. When this bit is 1 DMREADY and DMDONE
Mode
(CPLynx
Mode)
are in strict handshake mode (i.e., TSB12LV31
compatible mode). DMREADY must not be deactivated
until DMDONE activates. When this bit is set to 0,
DMREADY may be deactivated before DMDONE
activates.
R/WAutomatic update offset address. Valid only for
Address Up-
date
asynchronous transmit using header insert mode (bit 27
DMHDR set to 1). For write request asynchronous
packets, header quadlet 2 contains the destination offset
low address for the write. When this bit is set, header
quadlet 2 will be updated by the value of the payload size
(rounded up to the nearest quadlet boundary).
RDM acknowledge. This is the ack received from the
Acknowledge
receiving node. This is updated only when the transfer is
from the DM port.
R/WSpeed code. This is valid for isochronous transmit and
Code
asynchronous transmit through the DM port. The DM
logic uses this field to specify to the Phy the speed of the
isochronous transfer.
DMRX
2–4
Page 21
BIT
NUMBER
24–25CHNLCNTChannel
Count
26DMENDM EnableR/WDMEN controls the transmission of packets from the DM
27DMHDRDM Header
Insert Control
28–29AR0, AR1Receive
Control
Routing
30DMASYNCDM
Asynchronous
31DMRXDM ReceiveR/WIf this bit is set to 1 the DM port is configured to receive.
R/WChannel count. This field is valid only in isochronous
transmit. This field allows the node to transmit multiple
packets during a single isochronous period. Each packet
must have a different channel number, however,
hardware does not check this. When the isochronous
transmit header is supplied by the DM interface or
automatically inserted by the hardware, a maximum of
four different channels may be accessed in one
isochronous period. In isochronous transmit with
automatic header insert, Header0–Header3 CFRs are
used as the isochronous header registers.
port. If this bit is 0, transmission through from the DM port
is inhibited. This is used for asynchronous flow control. In
normal operation, if an asynchronous packet transmitted from the DM port receives an acknowledge from
the receiving node other than
be set to 0 and DMERROR is asserted high. Software will
need to set this bit to allow further transmission of
asynchronous packets from the DM port. The default and
power-up value is 0.
R/WDM header insert bit. When set to 0, the hardware will
automatically insert the header(s) into the DM transmit
data. In receive, setting this bit to 0 will strip off the
header(s) before routing packet to the DM. Header(s) are
always written to the CFR header registers regardless of
the value of DMHDR.
R/WReceive packet routing control encoded bits. These bits
in conjunction with DMASYNC and DMRX bits in the DM
control register controls the routing of the received
packet to either the data mover port or to the GRF . Refer
to Table 4–1.
R/WIf this bit is set to 1 the DM port is configured for
asynchronous traffic only. The DM port can not accept
both asynchronous and isochronous traffic. It must be
configured for asynchronous (DMASYNC = 1) or
isochronous (DMASYNC = 0).
The DM port cannot both transmit and receive data at the
same time, it must be configured for either transmit or
receive.
The control register dictates the basic operation of the TSB12L V32. The power-up reset value of this register
equals E004_0200’h
BIT
NUMBER
0FLSHERRFlush GRF
1RXSIDReceived
2FULLSIDSave full
3PHY_PKT_ENAPhy Packets
4BSYCTRLBusy ControlR/W BSYCTRL controls which busy status the chip returns to
5TXENTransmit
6RXENReceive
7ENA_ACCELAcceleration
8ENA_CONCATConcatenation
BIT NAMEFUNCTIONDIRDESCRIPTION
R/W This bit controls the flushing of the GRF when a packet with
on error
a data CRC error is detected. The power–up value is 1,
which means flush the GRF when a data CRC error is
detected.
R/W If set, the self-identification (SID) packets generated by Phy
Self-ID
packets
devices during the bus initialization are received and placed
into the GRF as a single packet. The default setting of this
bit is 1. When set to 0, the SIDs are not placed into the GRF .
R/W Save the full self-ID packets.When this bit is 1 the self-ID
Self-ID Packet
in GRF
data quadlet and its inverse quadlet are saved in the GRF.
When this bit is 0 only the self-ID data quadlet is saved in the
GRF .
R/W Phy packet enable allows reception of all Phy packets. If this
Receive
Enable
bit is reset to 0, all Phy packets, except for self-IDs, will be
rejected and interrupt HDERR (if not masked) will be
generated. One HDERR interrupt will be generated for
every Phy packet received.
incoming packets. When this bit is 0 the chip follows normal
busy/retry protocol, only send busy when necessary. When
this bit is 1 the chip sends a busy acknowledge to all
incoming packets following the normal busy/retry protocol.
R/W When TXEN is cleared, the transmitter does not arbitrate or
Enable
send packets. TXEN bit is cleared following a bus reset, and
all traffic through the DM port will be interrupted. TXEN must
be set before packet transmit can resume. Power-on reset
value of TXEN is 0
R/W When RXEN is cleared, the receiver does not receive any
Enable
packets. This bit is not affected by a bus reset and is set to 0
after a power-on reset.
R/W Enable acceleration. When this bit is set, fly-by acceleration
Enable
and accelerated arbitration are enabled. This bit cannot be
set while TXEN and RXEN are set. This bit must only be
used with a 1394a capable Phy.
R/W Enable concatenation. When this bit is set it allows the link
Enable
to concatenate multiple isochronous or asynchronous
packets. This bit must only be used with a 1394a capable
Phy.
2–6
Page 23
BIT
NUMBER
9ENA_
INSERT_IDLE
Insert Idle
Enable
DESCRIPTIONDIRFUNCTIONBIT NAME
R/W Per P1394a, the link is required to insert an idle state on the
control lines after the Phy grants the link control of the
Phy/link interface. If using a P1394a Phy, this bit should be
set to 1 in order for the link to drive an idle state following the
grant state from the Phy. For 1394-1995 Phys this bit must
remain low.
10RSTTXTransmitter
Reset
11RSTRXReceiver
Reset
12CTNDRSTATContenter
status
13CTNDRISINContender
Driver Enable
14RESERVEDReserved
15BUSNRSTBus number
reset enable
16–17BDIV0, BDIV1BCLK divisor
encode bits
R/W When RSTTX is set, the entire transmitter resets
synchronously. This bit clears itself.
R/W When RSTRX is set, the entire receiver resets
synchronously. This bit clears itself.
R/W Contender status. On power up, this bit reflects the status of
the CONTNDR pin. When bit 13, CTNDRISIN, is 0 this bit
will be driven out to the CONTNDR pin. If CTNDRISIN is 1
this bit is not used. (Only use on 1394–1995 Phys, or
P1394a Phys when using hardware reset, otherwise, use
the 1394a Phy registers to set the nodes contender status).
R/W Driver enable for the CONTNDR pin. On power up this bit is
set to 1 which disables the driver and allows reading of the
state of the CONTNDR pin. Writing a 0 to this bit will enable
the driver and will drive bit 12, CTNDRSTAT, to the
CONTNDR pin.
R/W When this enable is set to high, the bus number field clears
to 3FFh when a local bus reset is received.
R/W
BCLK divisors encode bits. Used to divide down the BCLK
to generate the link power status (LPS) clock to the Phy.
BDIV0 BDIV1 DESCRIPTION
00Divide by 16. Default power on value.
01Divide by 2. Recommended for BCLK
10Divide by 4. Recommended for BCLK
11Divide by 32. Recommended for BCLK
Recommended for BCLK frequencies in
the range of 8 – 88 MHz.
frequencies in the range of 1 – 1 1 MHz.
frequencies in the range of 2 – 22 MHz.
frequencies in the range of 16 – 176
MHz
18DMACKCOMPData Mover
Acknowledge
Complete
19FIFOACKCOMPFIFO
Acknowledge
Complete
R/W Data mover acknowledge complete. This bit controls the
acknowledge response to an asynchronous packet
received and routed to the DM port. The default and power
on value is 0 which means to respond with ack pending. A 1
means to respond with an ack complete for write request
packets.
R/W FIFO acknowledge complete. This bit controls the
acknowledge response to an asynchronous packet
received and routed to the GRF. The default and power on
value is 0 which means to respond with ack pending. A 1
means to respond with ack complete.
2–7
Page 24
BIT
NUMBER
20CYMASCycle MasterR/W When CYMAS is set and the TSB12LV32 is attached to the
21CYSRCCycle SourceR/W When CYSRC is set, the cycle_count field increments and
22CYTENCycle timer
enable
23CLRSIDERSelf-ID
error-code
clear
24–27SIDERCODESelf-ID error
code
28CMAUTOAuto set cycle
master
29IRP1ENIR port 1
enable
30IRP2ENIR port 2
enable
31RESERVEDReserved
root Phy, the cyclemaster function is enabled. When the
cycle_count field of the cycle timer register increments, the
transmitter sends a cycle-start packet.
the cycle_offset field resets for each positive transition of
CYCLEIN. When CYSRC is cleared, the cycle_count field
increments when the cycle_offset field rolls over .
R/W When CYTEN is set, the cycle_offset field increments.
WWhen CLRSIDER is set, the SIDERCODE field (bits 24–27)
is cleared.This bit clears itself.
R
SIDERCODE contains the error code of the first Self-ID
Error. The error code is as follows:
0000No error
0001Last self-ID received was not all child ports
0010Received Phy ID in self-ID not as expected
0011Quadlet not inverted (phase error)
0100Phy ID sequence error (two or more gaps in IDs)
0101Phy ID sequence error (large gap in IDs)
0110Phy ID error within packet
0111Quadlet not the inversion of the prior quadlet
1000Reserved
R/W When CMAUTO is high, the TSB12LV32 automatically
enables CYMAS when the this node becomes the root
following a bus reset.
R/W When IRP1EN is set, the receiver accepts isochronous
packets when the channel number matches the value in the
IR port1 field at18h
R/W When IRP2EN is set, the receiver accepts isochronous
packets when the channel number matches the value in the
IR Port2 field at18h
DESCRIPTIONDIRFUNCTIONBIT NAME
2–8
Page 25
2.2.4Interrupt/Interrupt Mask Register at 0Ch and 10h
The interrupt and interrupt mask register work in tandem to inform the host bus interface when the state of
the TSB12L V32 changes. The interrupt register is at 0Ch, the interrupt mask register is at 10h. The interrupt
register powers up all 0s, however, the interrupt mask register powers up with the INT and the MCERROR
bit set, i.e. 8000_1000h. The mask bits allows individual control for each interrupt. A 1 in the mask bit field
allows the corresponding interrupt in the interrupt register to be generated. Once an interrupt is generated
it must be cleared by writing a 1 to the bit in the interrupt register. For testing, each interrupt bit can be set
manually . This is done by first setting the REGRW bit at20h and then setting the individual interrupt bit. This
is also true for bit 0 at0Ch. In this test mode, the interrupt mask register is not used and has no effect.
BIT
NUMBER
0INTInterruptR/WINT contains the value of all interrupt and interrupt mask bits
1PHINTPhy chip
2PHRRXPhy register
3PHRSTPhy reset
4SELFIDENDSelf-ID
5ATSTARTEDAsynchronous
6RXGRFPKTGRF packet
7CMDRSTCSR register
8DMERRORData Mover
9RXDMPKTData Mover
10SELFIDERSelf-ID packet
BIT NAMEFUNCTIONDIRDESCRIPTION
ORed together
R/WWhen PHINT is set, the Phy has signalled an interrupt
interrupt
through the Phy interface
R/WWhen PHRRX is set, a register value has been transferred to
information
the Phy access register (offset 24h) from the Phy interface
received
R/WWhen PHRST is set, a Phy-LLC reconfiguration has started
started
(1394 bus reset)
R/WSelf-ID end. This bit is set at the end of the self-ID reporting
validated
process. When this bit is set, the contentF of the bus reset
CFR at34h is valid.
R/WAsynchronous transfer started. Activated when the bus has
transfer started
been granted and the first quadlet from the FIFO is about to
be popped from the ATF.
R/WReceive packet to GRF . This bit is set whenever a complete
received
packet has been confirmed into the GRF (asynchronous or
isochronous).
R/WIf CMDRST is set, the receiver has been sent a quadlet write
reset request
request to the Reset_Start CSR register(target address is
FFFF_F000_000Ch)
R/WDM error. This bit will be set whenever there is an error in the
error
DM stream. For transmit, if the DM port is configured for byte
access and the speed code in the DM control register or the
asynchronous header register is set for 400 Mbps then this
bit will be set. Under this condition DMEN will be reset to 0
preventing further transmit. For receive this bit will be set if
there is a header or data CRC error or if the DM port is configured for byte access and the data is received at 400 Mbps.
R/WReceive packet to DM. This bit is set whenever a packet is
packet receive
received to the DM port.
R/WSet if an error in the self-ID quadlet/packet has been de-
error
tected.
IARBFL
2–9
Page 26
BIT
NUMBER
11LINKONLink-ON detectR/WSet if a link-on pulse is detected on the LINKON input termi-
12ATSTKTransmitter is
stuck (AT)
13ATFEMPTYATF empty
interrupt
14SNTRJBusy
acknowledge
sent by receiver
15HDRERRHeader errorR/WWhen HDRERR is set, the receiver detected a header CRC
16TCERRTransaction
code error
17DMACKERRData Mover
acknowledge
error
18FIFOACKFIFO
acknowledge
interrupt
19MCERRORMicro-interface
error
20CYSECCycle second
incremented
21CYSTCycle startedR/WWhen CYST is set, the transmitter has sent or the receiver
22CYDNECycle doneR/WWhen CYDNE is set, an arbitration gap has been detected
23RESERVEDRESERVED
24CYLSTCycle lostR/WWhen CYLST is set, the cycle timer has rolled over twice
25CARBFLCycle
arbitration failed
26ARBGPArbitration gapR/WWhen ARBGP is set, the serial bus has been idle for an ar-
nal. This bit should be used by software to reactivate the LPS
output to the Phy .
R/WWhen ATSTK is set, the transmitter has detected invalid data
at the asynchronous transmit-FIFO interface. If the first
quadlet of a packet is not written to the ATF_First or
ATF_First&Update, the underflow of the ATF also causes an
ATStuck interrupt. When this state is entered, no asynchronous packets can be sent until the ATF is cleared by way of
the CLR ATF control bit. Isochronous packets can be sent
while in this state.
R/WATFEMPTY. This bit is set to 1 when the ATF is empty.
R/WWhen SNTRJ is set, the receiver is forced to send a busy ac-
knowledge to a packet addressed to this node because the
GRF overflowed.
error on an incoming packet that may have been addressed
to this node.
R/WWhen TCERR is set, the transmitter detected an invalid
transaction code in the data at the transmit-FIFO interface.
R/WDM acknowledge error. Set to 1 when the acknowledge re-
ceived is not
of the DM Control CFR at04h will be reset to 0 and no more
asynchronous transmit from the DM port will be allowed to
take place until DMEN is set to 1.
R/WFIFO ack interrupt. This bit will be set when an acknowledge
from a previous ATF transmit has been received.
R/WMicro-interface error. Set whenever the microcontroller write
protocol is violated.
R/WWhen CYSEC is set, the cycle-second field in the cycle timer
register has incremented. This occurs about every second
when the cycle timer is enabled.
has received a cycle-start packet.
on the bus after the transmission or reception of a cycle-start
packet. This indicates that the isochronous cycle is over.
without the reception of a cycle-start packet. This occurs
only when this node is not the cycle master. All isochronous
traffic stop once CYLST is set. However , asynchronous and
asynchronous streaming traffic will not be affected.
R/WWhen CARBFL is set, the arbitration to send a cycle-start
packet has failed.
bitration reset gap.
ack complete
DESCRIPTIONDIRFUNCTIONBIT NAME
. When this occurs, DMEN(bit 26)
2–10
Page 27
BIT
NUMBER
27SUBGPSubaction gapR/WWhen SUBGP is set, the serial bus has been idle for a subac-
28–30RESERVEDRESERVED
31IARBFLIsochronous
arbitration failed
tion gap time (fair-gap). This bit can be set only when the
REGRW bit has been set in the diagnostics register at 20h.
R/WWhen IARFL is set, the arbitration to send an isochronous
The power-up reset value of this register = 0000_0000h
BIT
NUMBER
0–1TAG1Tag Field 1R/W The TAG1 field can further qualify the isochronous reception
2–7IRPORT1Isochronous
8–9TAG2Tag Field 2R/W The TAG2 field can further qualify the isochronous reception
10–15IRPORT2Isochronous
16–23RESERVEDReserved
24–27ISYNCRCVNSynchronous
28IRCVALLReceive all
29–30RESERVEDReserved
31MONTAGMatch on tagR/W MONTAG is set when the user wants to only accept
BIT NAMEFUNCTIONDIRDESCRIPTION
for isochronous Receive PORT1 when the MONT AG bit is set.
R/W IR port1 contains the channel number of the isochronous
receive port
1 channel
number
packets that the receiver accepts. The receiver accepts
isochronous packets with this channel number when the
IRP1EN is set.
for isochronous Receive PORT2 when the MONT AG bit is set.
R/W IR port2 contains the channel number of the isochronous
receive port
2 channel
number
packets that the receiver accepts. The receiver accepts
isochronous packets with this channel number when the
IRP2EN is set.
R/W In isochronous receive mode to the DM port, when the
Enable
ISYNCRCVN enable bits are high, the DMPRE terminal pulses
when an isochronous packet is received whose SYNC bit field
in its header matches the bit pattern in this field. The default is
0000b.
R/W When the IRCVALL bit is set high, the TSB12L V32 receives all
isochronous
packets
isochronous packets regardless of the channel number or tag
number. The default is off.
isochronous packets that match both the tag field and the
channel number field. When set, MONTAG indicates that
isochronous receive data is accepted. The default is off.
This register is used to generate test conditions. The control bits in this register allow errors to be inserted
into various places in the packets generated by this node. After the completion of error insertion, enabled
error-insertion controls are disabled. The power-up reset value of this register = 0000_0000’h
BIT
NUMBER
0E_HCRCHeader CRC
1E_DCRCData CRC
2NO_PKTNo PacketR/WIf NO_PKT is set, the next primary packet to be generated by
3F_ACKAck FieldR/W If F_ACK is set, the ack field shall be used within the next ac-
4NO_ACKR/W If NO_ACK is set, the next acknowledge packet (that would
5–7RESERVEDReserved
8–15ACKR/W The 8-bit ACK field contains the 8-bit acknowledge packet
16–23RESERVEDReserved
24–31PINGVALUEPing timer
BIT NAMEFUNCTIONDIRDESCRIPTION
R/W If E_HCRC is set, the packet header CRC component of the
Error
next primary packet generated by this node shall be in error or
shall be invalid; otherwise, this bit has no effect. After the next
packet for this node is generated, this bit will be cleared.
R/W If E_DCRC is set, the packet data CRC component of the next
Error
primary packet generated by this node shall be in error or shall
be invalid; otherwise, this bit has no effect. After the next packet
for this node is generated, this bit will be cleared to zero immediately upon transmission of the erroneous CRC.
this node shall be discarded. This bit will be cleared to zero immediately after the next packet for this node is discarded.
knowledge packet generated by this node. This bit will be
cleared to zero immediately after the next acknowledge packet
for this node is generated.
normally have been generated by this node) is not sent. This bit
will be immediately cleared to zero when the next acknowledge
packet for this node is discarded.
(ack_code and ack_parity) to be supplied when the F_ACK bit
indicates a modified acknowledge packet is to be generated.
R/W Ping timer value. This value reflects the time it takes a node
value
to respond to a ping packet. The granularity of this timer is
40 ns.
The power-up reset value of this register = 0000_4AD0’h
BIT
NUMBER
0B0_BUSYByte 0 busyRByte 0 busy. When this bit is set, no microinterface write to byte 0 of
1B1_BUSYByte 1 busyRByte 1 busy. When this bit is set, no microinterface write to byte 1 of
2B2_BUSYByte 2 busyRByte 2 busy. When this bit is set, no microinterface write to byte 2 of
3B3_BUSYByte 3 busyRByte 3 busy. When this bit is set, no microinterface write to byte 3 of
4B0_PNDByte 0
5B1_PNDByte 1
6B2_PNDByte 2
7B3_PNDByte 3
8RAM_TESTR/W This bit can be set only when TESTMODE is high. When this bit is
9REGRWRegister
10–15RESERVEDReserved
16–19STATSEL0State0
20–23STATSEL1State1
24–27STATSEL2State2
28–31RESERVEDReserved
BIT NAMEFUNCTIONDIRDESCRIPTION
any CFRs is allowed. The microinterface must first poll this bit
before writing to byte 0.
any CFRs is allowed. The microinterface must first poll this bit
before writing to byte 1.
any CFRs is allowed. The microinterface must first poll this bit
before writing to byte 2.
any CFRs is allowed. The microinterface must first poll this bit
before writing to byte 3.
RByte 0 pending. When this bit is set, it indicates that byte 0 of a word
pending
or quadlet write has been accepted and the hardware is waiting for
the remaining bytes to be written. When the full write is complete,
this bit will be cleared.
RByte 1 pending. When this bit is set, it indicates that byte 1 of a word
pending
or quadlet write has been accepted and the hardware is waiting for
the remaining bytes to be written. When the full write is complete,
this bit will be cleared.
RByte 2 pending. When this bit is set, it indicates that byte 2 of a word
pending
or quadlet write has been accepted and the hardware is waiting for
the remaining bytes to be written. When the full write is complete,
this bit will be cleared.
RByte 3 pending. When this bit is set, it indicates that byte 3 of a word
pending
or quadlet write has been accepted and the hardware is waiting for
the remaining bytes to be written. When the full write is complete
this bit will be cleared.
set, the built in self test(BIST) for the FIFOs (transmit and receive)
will be run. On completion of the test hardware will reset this bit to 0
and simultaneously set bit 30 and 31.
R/W When REGRW is set, write-protected bits in various registers can
read/write
be written to.
access
R/W Status output select bits. Used to program the output of STAT0
select
terminal. See table in
Operation
section.
R/W Status output select bits. Used to program the output of STAT1
select
terminal. See table in
Operation
section.
R/W Status output select bits. Used to program the output of STAT2
The Phy access register allows access to the registers in the attached Phy . The most significant 16 bits send
read and write requests to the Phy registers. The least significant 16 bits are for the Phy to respond to a read
request sent by the TSB12L V32. The Phy access register also allows the Phy interface to send information
back to the TSB12LV32. When the Phy interface sends new information to the TSB12LV32, the Phy
register-information-receive (PhyRRx) interrupt is set. The Phy acess register is at address 24h and is a
read/write register. The power-up reset value of this register = 0000_0000’h.
BIT
NUMBER
0RDPHYRead Phy
1WRPHYWrite Phy
2–3RESERVEDRESERVED
4–7PHYRGADPhy-register
8–15PHYRGDATAPhy-register
16–19RESERVEDRESERVED
20–23PHYRXADPhy-register
24–31PHYRXDATAPhy-register
BIT NAMEFUNCTIONDIRDESCRIPTION
register
register
address
data
received ad-
dress
received
data
R/W When RDPHY is set, the TSB12LV32 sends a read register
request with the address equal to the PHYRGAD field to the Phy
interface. This bit is cleared when the request is sent.
R/W When WRPHY is set, the TSB12LV32 sends a write register
request with the address equal to the PHYRGAD field to the Phy
interface. This bit is cleared when the request is sent.
R/W PHYRGAD is the address of the Phy register that is to be
accessed.
R/W PHYRGDATA is the data to be written to the Phy register
indicated in PHYRGAD.
R/W PHYRXAD is the address of the register from which
PHYRXDAT A came. For testing, these bits can be set only when
the REGRW bit has been set in the diagnostics register at20h.
R/W PHYRXDAT A contains the data from the register addressed by
PHYRXAD. For testing, these bits can be set only when the
REGRW bit has been set in the diagnostics register at20h.
The power-up reset value of this register = 81BF_FFC0’h.
NOTE:
The power-up reset value shown above assumes one node on the bus only. A
P1394a compliant Phy is assumed to be attached to the TSB12LV32. If a
1394-1995 Phy is attached to the TSB12LV32 link, the NODECNT filed will be 0.
This is due to the fact that a 1394-1995 compliant Phy does not report its self-ID
packet back to the local link.
BIT
NUMBER
0NRIDVALValidRWhen set, NRIDV AL indicates that the Node ID, IRM Node ID,
1RESERVEDReserved
2–7NODECNTNode countRNODECNT contains the number of nodes detected in the
8ROOTRootRRoot is set when the current node is the root node. This bit is
9CONTENDERContenderRContender contains the status of the TSB12LV32 CONTNDR
10–15IRMNODEIDIRM node
16–25BUSNUMBERBus numberR/W BUSNUMBER is the 10-bit IEEE-1212 bus number. These
26–31NODENUMBERNode
BIT NAMEFUNCTIONDIRDESCRIPTION
Node Count, and Root information are valid. This bit is read
only.
system. This field is loaded with 1 following a power-on reset.
The NODECNT field is read only.
read only.
terminal. This bit is read only.
RIRMNODEID is the isochronous resource manager node
identification
identification. If there is no IRM node present on the bus,
these bits will be equal to 3Fh. These bits are read only.
bits are set to 3FFh when RBUSNUM is set and there is a bus
reset.
R/W NODENUMBER is the node number of the current node.
number
These bits are automatically updated following a bus reset. T o
change the node number of this node (spoofing), the
TESTMODE terminal must be set high.
2–17
Page 34
2.2.13Header0 Register at 38h
Header0 register must contain the isochronous header or the first quadlet of an asynchronous header if in
header insert mode. If not in header insert mode or if in receive mode, this register will be updated with the
received header. This register is write protected such that it cannot be written to unless automatic header
insert mode is enabled and DM is in transmit mode(i.e.,DMHDR=0 and DMRX=0). The power-up reset value
of this register = 0000_0000’h
Header1 register must contain the isochronous header or the second quadlet of an asynchronous header
if in header insert mode. If not in header insert mode or if in receive mode, this register will be updated with
the received header. This register powers up with all bits reset to 0. For multiple isochronous packets (within
the same isochronous cycle), this register would contain the isochronous header of the second isochronous
packet in the same format as the header0 register, if in header insert mode. This register is write protected
such that it cannot be written to unless automatic header insert mode is enabled and DM is in transmit mode
(i.e., DMHDR=0 and DMRX=0).
BIT
NUMBER
0–15DestinationIDDestination IDR/WFor asynchronous packets this field contains the destination
Header2 register must contain the isochronous header or the third quadlet of a asynchronous header if in
header insert mode. If not in header insert mode or if in receive mode, this register will be updated with the
received header. This register powers up with all bits reset to 0. For multiple isochronous packets (within
the same isochronous cycle), this register would contain the isochronous header of the third isochronous
packet in the same format as the header0 register, if in header insert mode. This register is write protected
such that it cannot be written to unless automatic header insert mode is enabled and in transmit mode (i.e.,
DMHDR=0 and DMRX=0).
BIT
NUMBER
0–31Header2R/WHeader quadlet for asynchronous or isochronous packet.
Header3 register must contain the isochronous header or the fourth quadlet of an asynchronous header if
in header insert mode. If not in header insert mode or if in receive mode, this register will be updated with
the received header. This register powers up with all bits reset to 0. For multiple isochronous packets (within
the same isochronous cycle), this register would contain the isochronous header of the fourth isochronous
packet in the same format as the header0 register, if in header insert mode. This register is write protected
such that it cannot be written to unless automatic header insert mode is enabled and DM is in transmit mode
(i.e., DMHDR=0 and DMRX=0).
BIT
NUMBER
0–31Header3R/WHeader quadlet for asynchronous or isochronous packet.
01111Reserved
10000No ack received
10001Ack too long
10010Ack too short
1001 1 – 11111 Reserved
was sent.
00 => 100 Mbit/s, 10 = > 400 Mbits/s,
01 => 200 Mbit/s, 11 is undefined.
These codes are added by the link layer
p
IEEE 1394–1995
specification.
LPS_OFF
LPS_RESET
NOTE:
The acknowledge code specified by the IEEE 1394-1995 specification is a 4-bit
field. The AckCode field in this register is a 5-bit field. The TSB12L V32 logic core
is able to provide (specifiy) three additional ack codes, which are not part of the
original specification. The ack codes are 10000, 10001, and 10010.
2–20
Page 37
BIT
NUMBER
30LPS_RESETLPS ResetR/W Link power status reset. This bit is
31LPS_OFFLPS OffR/W Link power status off. If set to 1, this
BIT NAME
FUNCTION
NAME
DIRDESCRIPTION
set by software and is reset by
hardware. When this bit is set,
hardware will deactivate LPS for a
fixed period to ensure that the Phy
has reset the interface. It will then
reactivate LPS. When this bit is
cleared by hardware, a PHRST
interrupt will also be generated.
bit will turn off the LPS pulsed
output to the Phy . This bit can also
be turned off from the Phy. Upon
detection of the LINKON pulsed
input signal, this bit will be turned off
allowing LPS to be driven to the
Phy which will, in turn, activate
SCLK and power up the link.
The power-up reset value of this register = 0000_0000’h.
BIT
NUMBER
0–7ASYNC RETRY
8–15RETRY
16–31RESERVEDRESERVED
BIT NAME
COUNT
INTERV AL
FUNCTION
NAME
Retry countR/WAsynchronous retry count, specifies the number of times to
Retry
interval
DIRDESCRIPTION
automatically retry sending asynchronous packets from the
ATF before giving up. After the retry count is exhausted
FIFOACK interrupt will be generated and the ATACK field in
CFR 30h will be updated to reflect the timeout.
R/W Asynchronous retry interval, is the time in increments of
isochronous cycles, between asynchronous retries.
2–21
Page 38
2–22
Page 39
3 Microcontroller Interface
MODE OF OPERATION
The microcontroller interface allows the local microcontroller/microprocessor to communicate with the
internal control and configuration registers (CFR), asynchronous transfer FIFO (A TF) and general receive
FIFO (GRF). All microcontroller reads/writes are initiated by the microcontroller. The micro interface
supports read transactions from the CFR or GRF, and write transactions to the CFR or ATF.
The micro interface can operate in byte (8 bits) or word (16 bits) accesses. Each CFR, with the exception
of the cycle timer register at 14h and the Phy access register at 24h, can be addressed on byte or word
boundaries. The possible configurations for the interface are shown in T able 3–1. The TSB12L V32 can also
be directly connected to the Motorola 68000 and ColdFire line of MC/MP . Table 3–2 defines the mapping
of the micro interface pins between the TSB12L V32, the Motorola 68000 and the ColdFire microprocessor.
T able 3–1. Microcontroller Interface Modes of Operation
T able 3–2. TSB12LV32 MP/MC Interface Terminal Function Matrix
TSB12LV32
TERMINAL NAMEUSAGETERMINAL NAMEUSAGE
MA0 – MA6InputA[6:0]Output
MD0 – MD15I/OD[31:16]I/O
MCAOutputTAZInput
MCSInputTSZOutput
MWRInputR/WZOutput
MCMODE/SIZ1, M8BIT/SIZ0InputSIZ1, SIZ0Output
TEAOutputTEAZInput
BCLKInputSCLK / CLKInput
Motorola 68000/ColdFire MICROCONTROLLER
The Byte stacker allows the TSB12LV32 to be easily connected to most processors. The byte stacker
consists of a programmable 8-/16-bit data bus and a 7-bit address bus. The TSB12LV32 uses cycle-start
and cycle-acknowledge handshake signals to allow the local bus clock and the 1394 clock to be
asynchronous to one another. The TSB12L V32 is an interrupt driver to reduce cycling. All bus signal labeling
on the TSB12L V32 microcontroller interface use bit #0 to denote the most significant bit (MSB).
ColdFire is a trademark of Motorola, Inc.
3–1
Page 40
3.1Microcontroller Byte Stack (Write) Operation
The microcontroller byte stack (write) protocol is shown in Figure 3–1.
The micro interface can be configured to operate in one of the following modes: handshake, fixed-timing,
or ColdFire mode. Burst transfers are only supported in the latter two modes.
3.3.1Microcontroller Handshake Mode
Byte handshake read and word handshake read are shown in Figure 3–3 and Figure 3–4, respectively .
The MCS
Another read or write transaction can begin after the next rising edge of BCLK. Note that data size is
determined by the MCMODE/SIZ1 and M8BIT/SIZ0 signals. The ColdFire signal is only asserted high when
the micro interface is operating in ColdFire mode.
, MCA handshake timing sequence for a read transaction can be summarized as follows:
1.The host takes MCS
MCS
low and MWR high, the MD[0:15] lines are enabled and driven with the read value. For an
low to signal the start of access. When the rising edge of BCLK samples
8-bit data bus, MD[0:7] lines are not used.
2.Following the next rising edge of BCLK, the TSB12LV32 takes MCA
low to signal that the
requested operation is complete. This is ensured to take place after two BCLK cycles. MCA
remains low with the MD lines containing valid read data until the micro interface releases MCS
(high state)
3.The host takes MCS
4.The TSB12LV32 takes MCA
high to signal the end of the process.
high to acknowledge the end of the access. This 3-states the MD
lines.
BCLK
COLDFIRE
M8BIT/SIZ0
MCMODE/SIZ1
MWR
MCS
MCA
MCADR[0:6]
MD[0:7]
MD[8:15]
A1
D1
A2
D2
Figure 3–3. Byte Handshake Read
Figure 3–4 shows a word handshake read transaction. In this case, all 16 bits of the MD lines are used. Note
that MD[0] contains the MSB and MD[15] contains the LSB. As in the byte read case, another read or write
transaction can begin after the next rising edge of BCLK.
3–4
Page 43
BCLK
COLDFIRE
M8BIT/SIZ0
MCMODE/SIZ1
MWR
MCS
MCA
MA[0:6]
MD[0:7]
MD[8:15]
A1A2
D1D3
D2D4
Figure 3–4. Word Handshake Read
Byte handshake write and word handshake write are shown in Figure 3–5 and Figure 3–6. In this case, the
micro interface asserts MCA
low immediately after MCS is sampled low. The micro interface keeps MCA
low until it samples MCS high. For 8-bit accesses, the MD[0:7] lines are not used. If a transfer error condition
occurs, TEA
will be asserted low for one BCLK cycle. An error condition can occur if the MCMODE/SIZ1
or M8BIT/SIZ0 line changes state during the access cycle.
3–5
Page 44
BCLK
COLDFIRE
M8BIT/SIZ0
MCMODE/SIZ1
MWR
MCS
MCA
MA[0:6]
MD[0:7]
MD[8:15]
BCLK
COLDFIRE
M8BIT/SIZ0
MCMODE/SIZ1
MWR
MCS
MCA
A1A2
D1D2
Figure 3–5. Byte Handshake Write
3–6
MA[0:6]
MD[0:7]
MD[8:15]
A1A2
D1D3
D2D4
Figure 3–6. Word Handshake Write
Page 45
3.3.2Microcontroller Fixed-Timing Mode
Byte and word fixed-timing reads shown in Figure 3–7 and Figure 3–8. Fixed-timing mode supports burst
transfers. If MCS is asserted low for more than one BCLK cycle, burst mode is enabled. The fixed-timing
burst mode does not have a limit on the maximum burst size allowed.
The timing sequence in the fixed-timing a read transaction can be summarized as follows:
1.The host pulses MCS
cycle will enable burst mode. The number of BCLK cycles during which MCS
low to signal the start of access. Pulsing MCS low for more than once clock
is asserted low
determines the burst size.
2.When the rising edge of BCLK samples MCS
low and MWR high, the register value or GRF data
pointed to by MA is latched onto the MD lines. The MD lines will latch on every rising edge of BCLK
if MCS
is asserted low.
3.After 2 BCLK cycles, the TSB12LV32 pulses MCA
of the requested operation. If MCS
for
n
cycles. Note that MA needs only contain valid data during the first cycle in which MCS is low.
is pulsed low for n BCLK cycles, MCA will also be pulsed low
low for one clock cycle to signal the completion
Except for the first one, every data transfer takes only one BCLK cycle. If a read transaction is
accessing the CFR, it may not cross any register boundary.
Another read or write transaction can begin after the next rising edge of BCLK. Note that data size is
determined by the MCMODE/SIZ1 and M8BIT/SIZ0 signals. The ColdFire signal is only asserted high when
the micro interface is operating in ColdFire mode.
BCLK
COLDFIRE
M8BIT/SIZ0
MCMODE/SIZ1
MWR
MCS
MCA
MA[0:6]
MD[0:7]
MD[8:15]
A1A2
D1D2D3D4D5
Figure 3–7. Byte Fixed-Timing Read
3–7
Page 46
BCLK
COLDFIRE
M8BIT/SIZ0
MCMODE/SIZ1
MWR
MCS
MCA
MA[0:6]
MD[0:7]
MD[8:15]
A1A2
D1D3D5D7D9
D2D4D6D8D10
Figure 3–8. Word Fixed-Timing Read
Byte fixed timing write and word fixed timing write are shown in Figure 3–9 and Figure 3–10.
The first data transfer takes one extra wait cycle. All subsequent data transfers take only one BCLK cycle.
For an 8-bit data bus, MD[0:7] is not used (don’t care) and is driven with zeros. If the write transaction is
accessing the CFR register, it may not cross any register boundary. First write data for each ATF quadlet
must start at byte0. Write accesses to the A TF must be quadlet aligned. The micro interface will wait for all
bytes of each quadlet to be available before creating a write request to the A TF. If a transfer error condition
occurs, TEA
will be asserted low for one BCLK cycle. An error condition can occur if the MCMODE/SIZ1
or M8BIT/SIZ0 lines transition during the access cycle.
3–8
Page 47
BCLK
COLDFIRE
M8BIT/SIZ0
MCMODE/SIZ1
MWR
MCS
MCA
MA[0:6]
MD[0:7]
MD[8:15]
BCLK
COLDFIRE
M8BIT/SIZ0
MCMODE/SIZ1
MWR
MCS
MCA
MA[0:6]
A1A2
D2D2D3D4D5
Figure 3–9. Byte Fixed-Timing Write
A1A2
MCD[0:7]
MD[8:15]
D1
D2
D3D5D7D9
D4D6D8D10
Figure 3–10. Word Fixed-Timing Write
3–9
Page 48
3.3.2.1GRF READ
The timing requirements when performing a GRF read access in fixed–timing mode are different from other
access modes. In fixed–timing mode, the GRF must be accessed only on a quadlet boundary. In other
words, only quadlet fetches are legal. If the microinterface is configured for a byte access, this means that
MCSZ must be asserted low for 4 BCLK cycles, as shown in Figure 3–1 1. If configured for word access, then
MCSZ must only be asserted for 2 BCLK cycles, as shown in Figure 3–12.
The TSB12LV32 supports a glueless interface to the ColdFire family of microcontrollers. To enable this
mode, the ColdFire pin must be asserted and kept high for the entire access cycle. The timing diagram for
a ColdFire read operation is shown in Figure 3–13.
The timing sequence for a ColdFire read access can be summarized as follows:
1.The ColdFire pulses MCS
be asserted for one clock cycle.
2.When the rising edge of BCLK samples MCS
not yet contain valid data. The MA lines should contain the address information at this point. MA
is only required to be available for one BCLK cycle. The data transfer size is determined by the
state of the MCMODE/SIZ1 and M8BIT/SIZ0 lines.
3.The TSB12LV32 pulses MCA
The number n depends on the data transfer size specified by the MCMODE/SIZ1 and
M8BIT/SIZ0 lines. The CFR register value or GRF memory data pointed to by the MA lines is
latched onto the MD lines. MCA
The microinterface does burst transfers if the MCMODE/SIZ1 and M8BIT/SIZ0 lines indicate more than
2-bytes (1 word) of data. The TSB12LV32 does not support 1-byte transfers in the ColdFire mode. If a
transfer error condition occurs, TEA
if the MCMODE/SIZ1 and M8BIT/SIZ0 lines specify a transfer size of 1-byte or if their state changes during
the access cycle. Note that all 16-bits of the MD lines are always used in this mode.
BCLK
COLDFIRE
low for one BCLK cycle to signal the start of access.MCS must only
low and MWR high, MD lines are enabled, but do
low for n clock cycles to signal the requested operation is complete.
will pulse for one clock cycle on every word (2-Bytes) transfer.
will be asserted low for one BCLK cycle. An error condition can occur
MCMODE/SIZ1
M8BIT/SIZ0
MWR
MCS
MCA
MA[0:6]
MD[0:7]
MD[8:15]
A1A2
D1D3
D2D6
D4
A3
D5
Figure 3–13. ColdFire Read
The ColdFire write transaction is shown in Figure 3–14. Unlike the handshake and fixed-timing write modes,
the ColdFire write operation requires the data on the MD lines be available one BCLK cycle after the address
on the MA lines is sampled. Violating this timing requirement may result in a transfer error , causing TEA
to
be asserted low for one BCLK cycle.
3–12
Page 51
BCLK
Setup time to BCLK
ns
Hold time from BCLK
ns
COLDFIRE
MCMODE/SIZ1
M8BIT/SIZ0
MWR
MCS
MCA
MA[0:6]
MD[0:7]
MD[8:15]
TEA
A1A2
D1D3
D2D6
D4
Figure 3–14. ColdFire Write
3.3.4Microcontroller Critical TIming
PARAMETERTERMINAL NAMEACCESS TYPEMINMAXUNIT
t
d0
t
t
t
t
t
t
t
t
t
t
t
t
t
t
†
Delay time (BCLK to Q)
d1
d2
su0
su1
su2
su3
su4
su5
h0
h1
h2
h3
h4
h5
All parameters are referenced to the rising edge of BCLK.
The term
consider a 32-bit processor; any 32-word consists of four bytes which may be stored in memory in one of
two ways. Of the four bytes, either byte 3 will be considered the most significant byte and byte 0 the least
significant byte, or vice versa (see Figures 3–15 and 3–16). A little endian type memory considers byte 0
the least significant byte, whereas a big endian type memory considers byte 3 to be the least significant byte.
endianness
refers to the way data is referenced and stored in a processor’s memory. For example,
Byte #0
(Most Significant Byte)
Byte #1Byte #2
Figure 3–15. Big Endian Format
Byte #3
(Least Significant Byte)
Byte #3
(Most Significant Byte)
Byte #2Byte #1
(Least Significant Byte)
Byte #0
Figure 3–16. Little Endian Format
The TSB12L V32 configuration register space (CFR) and FIFO memory , both of which are 32-bits wide, use
a big endian architecture. The TSB12L V32 uses the same endianness as the internal P1394 link core. This
means that the most significant byte is the left-most byte (byte 0) and the least significant byte is the right
most byte (byte 3).
3.3.5.1Data and Address Invariance for Little Endian Processors
For little-endian processors, there are two modes of byte swapping, address invariant and data invariant.
Address invariance preserves byte ordering between the internal system (GP2Lynx registers and FIFO) and
external system (microcontroller/processor). Data invariance preserves the bit significance of the data, but
changes the byte significance between the internal and external systems. The MDINV pin controls how the
write/read data is swapped at the data bus (i.e., determines how the received bytes from the microcontroller
are mapped into the TSB12L V32 internal registers and memory space). Note that when the COLDFIRE pin
is high, the MDINV pin has no affect and data is always interpreted in as big endian. Refer to Literature
3–14
Page 53
Number
INTERFACE
SLLA021.pdf ENDIANNESS AND THE TSB12LV41 (MPEG2LYNX) MICROPROCESSOR
for a detailed description of endianness.
The pin settings for all the swapping operation are shown in Table 3–3. Note that in performing the byte
swapping operation in the little-endian mode, only the two least significant bits of the 32-bit address inside
are involved. This is because there is a total of four bytes associated with the swapping operation.
Table 3–3. Endian Swapping Operation
LENDIANM8BIT/SIZ0MDINVDESCRIPTION
0XXBig endian mode, no manipulation on byte address and data bytes
11
(8-bits wide)
101
(16-bits wide)
11016-bit little endian address invariance mode, swap data between MD[0:7] and
1108-bit little endian address invariance mode, no manipulation on byte address
1Little endian data invariance mode, swap the low order 2 bit address:
External low order 2-bit address Internal low order 2-bit address
1Little endian data invariance mode, swap the low order 2 bit address:
External low order 2-bit address Internal low order 2-bit address
Since the TSB12L V32’s microprocessor interface is either 8 bits or 16 bits wide, but the internal configuration
registers are 32 bits wide, a byte stacking (for writes) and a byte unstacking (for reads) operation must be
performed on the data bus. For little endian processors, the TSB12L V32 can perform the swapping of bytes
on the data bus required to allow both the processor and the TSB12LV42 to interpret the data the same.
There are two methods of swapping the data bytes, address invariant and data invariant. Both of these
methods are described below.
NOTE:
For the host processor to work correctly with the TSB12L V32, users
must
correctly
connect the address and data busses of their microprocessor to the TSB12L V32’s
microprocessor port. Users must connect the MSB (most significant bit) of their
address/data bus to the address/data MSB of the TSB12L V32. This must be done
regardless of bit number labeling or which type of endianness their microprocessor
uses.
3.3.5.2Data Invariant System Design
Figure 3–17 shows a little endian data invariant system design example. In this system, the actual value of
the data as it was stored in the processor’s memory is preserved. Data invariant designs do not preserve
the addresses when mapping between endian domains. If the data represents an integer, it is interpreted
the same by both systems. If the data represents a string, an array, or some other type of byte indexed
structure, it is interpreted differently by both systems.
3–15
Page 54
Byte 3
(MSByte)
Byte 2
Byte 1
Byte 0
(LSByte)
Little Endian Processor
Memory
TSB12LV32 CFR/Memory
Figure 3–17. Little-Endian Data Invariance Illustration Chart
Figure 3–18 shows a little-endian address invariant system design example. In this case, the byte ordering
between both systems is preserved (i.e., byte address is preserved). For example, byte 3 in the little endian
processor memory is also byte 3 in the TSB12LV32 CFR space. As Figure 3–18 shows, the byte ordering
is automatically maintained by the TSB12L V32 when in the address-invariance mode by swapping the order
in which the incoming bytes on the microprocessor are written to the CFRs.
3–16
Page 55
4 Link Core
This section describes the link core components and operations. Figure 4–1 shows the link core
components.
Transmitter
Cycle Timer
CRC
Cycle Monitor
Receiver
Figure 4–1. Link Core Components
Physical Interface
4.1Physical Interface
The physical (Phy) interface provides Phy-level services to the transmitter and receiver. This includes
gaining access to the serial bus, sending packets, receiving packets, and sending and receiving
acknowledge packets. The Phy interface module also interfaces to the Phy chip and implements Texas
Instruments patent-pending bus-holder galvanic isolation.
4.2Transmitter
The transmitter retrieves data from either the asynchronous transmit FIFO (ATF) or the data mover (DM)
port and creates correctly formatted serial-bus packets to be transmitted through the Phy interface. When
data is present at the A TF interface to the transmitter , the TSB12L V32 Phy interface arbitrates for the serial
bus and sends a packet. When data is present at the DM Port, the TSB12L V32 arbitrates for the serial bus
during the next isochronous cycle. The transmitter autonomously sends the cycle-start packets when the
chip is a cycle master.
4.3Receiver
The receiver takes incoming data from the Phy interface and determines if the incoming data is addressed
to this node. When the incoming packet is addressed to this node, the CRC of the packet is checked. If the
header CRC is good, the header is confirmed in the general-receive FIFO (GRF). For block and isochronous
packets, the remainder of the packet is confirmed one quadlet at a time. The receiver places a status quadlet
in the GRF after the last quadlet of the packet is confirmed in the GRF . The status quadlet contains the error
code for the packet. In the case of asynchronous packets, the error code is the acknowledge code that is
sent (returned) for that packet. For isochronous and broadcast packets that do not need acknowledge
packets, the error code is the acknowledge code that would have been sent. This acknowledge code tells
the transaction layer whether or not the data CRC is good or bad. If the header CRC is bad, the header is
flushed and the rest of the packet is ignored. When a cycle-start packet is received, it is detected and the
cycle-start packet data is sent to the cycle timer. Cycle-start packets are not placed in the GRF like other
quadlet packets.
4–1
Page 56
4.4Cycle Timer
The cycle timer is only used by nodes that support isochronous data transfer. The cycle timer is a 32-bit
cycle-timer register. Each node with isochronous data-transfer capability has a cycle-timer register as
defined by the IEEE 1394–1995 specification. In the TSB12LV32, the cycle-timer register is implemented
in the cycle timer located in the IEEE-1212 initial register space at location 200h and can also be accessed
through the local bus at TSB12L V32 CFR address 14h. The low-order 12 bits of the timer are a modulo 3072
counter, which increments once every 24.576-MHz clock periods (or 40.69 ns). The next 13 higher-order
bits are a count of 8,000-Hz (or 125-µs) cycles, and the highest 7 bits count seconds. The cycle timer
contains the cycle-timer register. The cycle-timer register consists of three fields: cycle of fset, cycle count,
and seconds count. The cycle timer has two possible sources. First, when the cycle source (CYSRC) bit
in the configuration register (bit 21 at 08h) is set, then the CYCLEIN input causes the cycle count field to
increment for each positive transition of the CYCLEIN input (8 kHz) and the cycle offset resets to all zeros.
CYCLEIN should only be the source when the node is the cycle master. The timer can also be disabled using
the cycle-timer-enable bit (CYTEN) in the control register. The second cycle-source option is when the
CYSRC bit is cleared. In this state, the cycle-offset field of the cycle-timer register is incremented by the
internal 24.576-MHz clock. The cycle timer is updated by the reception of the cycle-start packet for the
non-cycle master nodes. The cycle-offset field in the cycle-start packet is used by the cycle-master node
to keep all nodes in phase and running with a nominal isochronous cycle of 125 µs. The cycle-start bit is
set when the cycle-start packet is sent from the cycle master node or received by a noncycle master node.
4.5Cycle Monitor
The cycle monitor is only used by nodes that support isochronous data transfer. The cycle monitor observes
chip activity and handles scheduling of isochronous activity . When a cycle-start message is received or sent,
the cycle monitor sets the cycle-started interrupt bit. It also detects missing cycle-start packets and sets the
cycle-lost interrupt bit when this occurs. When the isochronous cycle is complete, the cycle monitor sets the
cycle-done-interrupt bit. The cycle monitor instructs the transmitter to send a cycle-start message when the
cyclemaster bit (CYMAS) is set in the control register.
4.6Cyclic Redundancy Check (CRC)
The CRC module generates a 32-bit CRC for error detection. This is done for both the header and the data.
The CRC module generates the header and data CRC for transmitting packets and checks the header and
data CRC for received packets (see the IEEE-1394–1995 standard for details on the generation of the CRC).
4.7Packet Routing Control Logic
Asynchronous and Isochronous receive packets will be routed to the DM port or the GRF via the receiver
routing control logic. Any asynchronous packet addressed in the range of 0 to 0000_FFFF_FFFF and is of
write request
the
set, all isochronous data will be routed according to Table 4–1. Note that self-ID packets and Phy packets
are always received by the GRF regardless of the routing control settings.
4–2
type will be routed to the DM port according to Table 4–1. When bit IRCV ALL (at 18h) is
01Receive no dataReceives all data
11Receives addressed write
request asynchronous packets in
the address range of
0000_0000_0000 0000_FFFF_FFFFh
X0Receives no dataReceives all data
X1Receive all data (asynchronous
and isochronous)
Receives all data (asynchronous and
isochronous); power-on default
Receives all other asynchronous
packets and isochronous packets
Receives isochronous/asynchronous
streaming data only
Receives asynchronous data
Receives addressed asynchronous
packets in the address range of
0001_0000_0000-FFFF_FFFF_FFFFh.
Also receives any asynchronous
packets not going to the DM port
regardless of the address. Receives
isochronous packets too.
Receives no data
4–3
Page 58
4–4
Page 59
5 Data Mover Port Interface
The data mover (DM) port in the TSB12LV32 is the physical medium by which autonomous streams of
different types are piped to/from an application that uses the TSB12L V32. The DM port is meant to handle
an external memory interface of large data packets.
The DM port can support three types of packets:
•Asynchronous
•Isochronous
•Asynchronous Streaming (1394a supported format)
The port can be configured to either transmit or receive data packets at one time (half duplex). A typical
system diagram is shown in Figure 5–1:
Application
Stream Process
(External Memory Interface)
µProcessor/µController
(ColdFire)
Data Mover I/F
Microcontroller Interface
TSB12LV32
Phy/Link I/F
Phy
Figure 5–1. A Typical System Diagram
The DM port will perform all operations synchronously , utilizing a 24.576 MHz output clock called DMCLK.
DMCLK is essentially SCLK/2. There is no asynchronous logic within the DM block. All data transfers are
synchronized to the DMCLK output. The DM operates by setting the DM control register at 04h and the
control register at 08h of the CFR. The DM interfaces internally with the configuration register (CFR) block
and the link core (Link) block and interfaces externally with the data mover external interface.
The advantages of the DM port can be summarized as follows:
•Transmits or receives large blocks of data at speeds up to 400 Mbits/s.
•Allows for a large external FIFO specific to an individual application.
•Handles asynchronous, isochronous, or asynchronous streaming packets.
5–1
Page 60
No
Isochronous DM Idle
(DMDONE is high)
DMEN is 1, DMASYNC is 0
DMREADY is High?
Isochronous DM Go
No
(DMDONE is low)
New Isochronous Cycle Started?
Yes
Isochronous Arbitrate/Xmit
Arbitrate for Isochronous Transmit
and Send One Isochronous Packet
No
No
(DMDONE is low)
All Channels Done
(DMDONE is low)
End of All Channels?
Data Block Done
(DMDONE is low)
End of All Packets
For This Data Block?
Handshake
Mode?
Handshake
No
(DMDONE is high)
DMREADY is Low?
Yes
Yes
Yes
Yes
No
Yes
Yes
NOTE A: DMEN and DMASYNC are configuration register bits, DMREADY is an input terminal, and DMDONE is an
output terminal.
Figure 5–2. Isochronous DM Flow Control (TSB12LV32 Transmit)
5–2
Page 61
MSBLSB
Data Mover Data
DMD0 – DMD15
1394 Packet Data
Bit 0
Bit 1 Bit 2 Bit 3 Bit 4 Bit 5 Bit 6 Bit 7 Bit 8 Bit 9 Bit 10 Bit 11 Bit 12 Bit 13 Bit 14 Bit 15
Bit 0 Bit 1 Bit 2 Bit 3 Bit 4 Bit 5 Bit 6 Bit 7 Bit 8 Bit 9 Bit 10 Bit 11 Bit 12 Bit 13 Bit 14 Bit 15
. . .Bit 31
Figure 5–3. Transmit Data Path
The DM isochronous transmit reads data from the DM interface (DMD[0–15] lines) and passes it to the 1394
isochronous transmit interface in accordance with Figure 5–2. The data path is shown in Figure 5–3. The
asynchronous header registers will contain the latest extracted header from the asynchronous stream when
the header is supplied via the DM port. For automatic header insertion, the
datalength
field in the header
register will be automatically updated by the payload size of the previous asynchronous transmit packet.
This option can be turned off by setting the appropriate bits in the DM control registers. The DM
asynchronous transmit reads data from the DM interface (DMD[0:15] lines) and passes it to the 1394
asynchronous interface in accordance with Figure 5–4
Asynchronous DM Idle
(DMDONE is high)
DMEN is 1, DMASYNC is 0
DMREADY is High?
Yes
No
Asynchronous Arbitrate/Xmit
Arbitrate for Asynchronous Transmit
No
NOTE: DMEN and DMASYNC are configuration register bits, DMREADY is an input terminal, and DMDONE is an output
terminal.
(DMDONE is low)
and Send Asynchronous Packet
Yes
Wait for Acknowledge
Ack Complete received?
Yes
Data Block Done
(DMDONE is low)
End of All Packets
For This Data Block?
Yes
No
DM Disabled
Re-enable via
Software
Figure 5–4. Asychronous DM Flow Control (TSB12LV32 Transmit)
The DM Interfaces to the configuration register (CFR), the link core (Link), and the external data mover
interface (DMI).
5–3
Page 62
5.1Data Mover Data Flow Diagram
The data mover has eight modes of operation. There are four modes for transmit and four modes for receive.
Definitions
•Data mover port configured to operate in transmit mode means that the packet data is received
through the data mover port and forwarded (unbuffered) to the link core transmit logic to be sent
to the physical layer device (Phy), which will, in turn,
transmit
the data onto the 1394 bus.
•Data mover port configured to operate in receive mode means that the packet data is
received
by the link core receive logic from the 1394 bus through the Phy . The data is then routed by the
link core to the data mover port without any internal buffering.
5.1.1Isochronous Receive
In all the isochronous receive modes, the packet header information is always loaded into the header
registers. The packet header quadlet is loaded into the Header0 register at 38h and the packet trailer quadlet
is loaded into the trailer register at 48h.
5.1.1.1Isochronous Packet Receive Without Header and Trailer
Step 1: Isochronous packet is received through the receiver logic of the link core
Step 2:The packet header is stripped off from the packet and loaded into the header0 register at
38h.
Step 3:Packet data (payload only) is routed directly to the DM port without any buffering.
Step 4:Trailer quadlet is loaded into the trailer register at 48h
CFR REGISTER
Header0 Register at 38h
Step 2
Trailer Register at 48h
LINK CORE
Transmitter
Data
Mover
Port
Step 3 (Packet Data)
Step 4
Receiver
Step 1
Packet received from
1394 bus through the
Phy
Figure 5–5. Isochronous Receive Without Header and Trailer
5.1.1.2Isochronous Packet Receive With Header and Trailer
Step 1:Isochronous packet is received through the receiver block of the link core.
Step 2:The header quadlet is both loaded into the header0 register at 38h and routed to the DM
port without any buffering.
Step 3:Packet data (payload only) is sent directly through the DM port only.
Step 4:Trailer quadlet is loaded into the trailer register at 48h. It is also forwarded to the DM port.
5–4
Page 63
CFR REGISTER
Step 2
Step 4
LINK CORE
Transmitter
Receiver
Step 1
Packet received from
1394 bus through the
Phy
Data
Mover
Port
Header0 Register at 38h
Trailer Register at 48h
Step 4 (trailer quadlet)
Step 3 (packet data)
Step 2 (header quadlet)
Figure 5–6. Isochronous Receive With Header and Trailer
5.1.2Isochronous Transmit
There are two ways (modes of operation) to transmit isochronous data through the data mover:
•Isochronous packet transmit with automatic header insertion.
•Isochronous packet transmit without automatic header insertion.
The difference between the two modes lies in the mechanism in which the header information is inserted
into the data stream. However, in both cases the header information is always loaded into the link core
transmitter from the header register.
5.1.2.1Isochronous Packet Transmit With Automatic Header Insertion
In this mode, the header information is first loaded into the header0 register through the microcontroller
interface. The header will subsequently be automatically inserted into the data once the data mover starts
streaming it through to the link core transmitter logic. The following steps further illustrate the process:
Step 1
Write
Header
Information
CFR REGISTER
Micro-
controller
Interface
Step 1
Header
loaded
Step 3 (Packet Data)
Header0 Register at 38h
Step 3
Data
Mover
Port
Step 2
(packet data)
Figure 5–7. Isochronous Transmit With Auto Header Insertion
LINK CORE
Transmitter
Receiver
Step 4
Packet sent
to 1394 bus
through the
Phy
5–5
Page 64
Step 1:Isochronous header quadlet is loaded into header0 register at 38h through a write
operation from the microcontroller interface.
Step 2:Header quadlet is forwarded to the transmitter of the link core.
Step 3:Packet data (payload only) is transmitted through the data mover directly to the transmitter
of the link core.
Step 4:Isochronous packet is sent to the 1394 bus through the Phy.
NOTE:
The data coming through the data mover port is typically supplied by an external
fast memory block (i.e., FIFO, DRAM). This external memory logic may begin
transmitting data through to the data mover port exactly one DMCLK cycle after the
DMPRE output pin on the GP2Lynx is asserted high.
5.1.2.2Isochronous Packet Transmit Without Automatic Header Insertion
In this mode, the packet header and data information is loaded through the data mover port. This mode is
sometimes called isochronous packet transmit with manual header insertion. This is because the header
quadlet is not preloaded into the header0 register via the microcontroller interface. Instead, it is inserted
manually
the process:
into the data stream at the same time as the rest of the packet. The following steps further illustrate
Step 1:Isochronous header information (only one header quadlet in this case) is fetched into the
header0 register at 38h through the data mover port.
Step 2:Header quadlet is forwarded to the transmitter of the link core.
Step 3:Packet data (payload only) is transmitted through the data mover directly to the transmitter
of the link core.
Step 4:Isochronous packet is sent to the 1394 bus through the Phy.
CFR REGISTER
Step 1 (header fetched)
Step 1
(header supplied)
Step 3
(packet data)
Data
Mover
Port
Header0 Register at 38h
Step 3 (Packet Data)
Figure 5–8. Isochronous Transmit Without Auto Header Insertion
5–6
Step 2
LINK CORE
Transmitter
Receiver
Step 4
Packet sent
to 1394 bus
through the
Phy
Page 65
5.1.3Asynchronous Receive
In all the asynchronous receive modes, the packet header information is always loaded into the header
registers. In quadlet receive mode, the first three header quadlets are loaded into the header0at 38h,
header1at 3Ch, and header2at 40h registers, respectively. The trailer quadlet is loaded into the trailer
register at 48h. In block receive mode, the only additional step performed is loading the fourth header
quadlet received into the header3 register at 44h.
5.1.3.1Asynchronous Packet Receive Without Headers and Trailer
Step 1:Asynchronous packet is received through the receiver logic of the link core
Step 2:The packet headers are stripped off from the packet and loaded into the header registers:
a) If in quadlet receive mode, the three header quadlets are loaded into the header0–
header2 registers.
b) If in block receive mode, the four header quadlets are loaded into the header0–
header3 registers.
Step 3:Packet data (payload only) is routed directly to the DM port without any buffering.
Step 4:Trailer quadlet is loaded into the trailer register at 48h
CFR REGISTER
Loaded only in
Block Receive
Data
Mover
Port
Header0 Register at 38h
Header1 Register at 3Ch
Header2 Register at 40h
Header3 Register at 42h
Trailer Register at 48h
Step 3 (packet data)
Quadlet#0
Quadlet#1
Quadlet#2
Quadlet#3
Step 2
Step 4
LINK CORE
Transmitter
Receiver
Figure 5–9. Asynchronous Receive Without Headers and Trailer
Step 1
Packet received
from 1394 bus
through the Phy
5–7
Page 66
5.1.3.2Asynchronous Packet Receive With Headers and Trailer
Step 1:Asynchronous packet is received through the receiver logic of the link core.
Step 2:The header quadlets are loaded into their respective header registers AND routed to the
DM port without any buffering.
Step 3:Packet data is routed directly to the DM port (no buffering performed).
Step 4:Trailer quadlet is loaded into the trailer register at 48h.
CFR REGISTER
Quadlet#0
Quadlet#1
Quadlet#2
Quadlet#3
LINK CORE
Transmitter
Receiver
Step 1
Packet received
from 1394 bus
through the Phy
Loaded only in
Block Receive
Data
Mover
Port
Header0 Register at 38h
Header1 Register at 3Ch
Header2 Register at 40h
Header3 Register at 42h
Trailer Register at 48h
Step 4 (trailer quadlet)
Step 3 (packet data)
Step 2 (header quadlet)
Figure 5–10. Asynchronous Receive With Headers and Trailer
5.1.4Asynchronous Transmit
There are two ways (modes of operation) to transmit asynchronous data through the data mover:
•Asynchronous packet transmit with automatic header insertion.
•Asynchronous packet transmit without automatic header insertion.
The difference between the two modes lies in the mechanism in which the header information is inserted
into the data stream. However, in both cases, the header information is always loaded into the link core
transmitter from the header registers.
5–8
Page 67
5.1.4.1Asynchronous Packet Transmit With Automatic Header Insertion
In this mode, the header information is first loaded into the header0–header3 registers through the
microcontroller interface. The headers will subsequently be automatically inserted into the data once the
data mover starts streaming it through to the link core transmitter logic. The following steps further illustrate
the process:
Step 1:Asynchronous header quadlets (3 quadlets in quadlet receive mode and 4 quadlets in block
receive mode) are loaded into header0–header3 registers through a write operation from
the microcontroller interface. Loading one header requires a single write operation.
Step 2:Header quadlets are forwarded to the transmitter of the link core.
Step 3:Packet data (payload only) is transmitted through the data mover directly to the transmitter
of the link core.
Step 4:Asynchronous packet is sent to the 1394 bus through the Phy.
NOTE:
The data coming through the data mover port is typically supplied by an external
fast memory block (i.e., FIFO, DRAM). This external memory logic may begin
transmitting data through to the data mover port exactly one DMCLK cycle after the
DMPRE output pin on the GP2Lynx is asserted high.
CFR REGISTER
Step 1
Write
Header
Quadlets
Quadlet#0
Quadlet#1
Quadlet#2
Quadlet#3
Step 2
Step 3
Micro-
controller
Interface
Loaded only in
Block Receive
Step 1
Headers
loaded
Step 3 (packet data)
Header0 Register at 38h
Header1 Register at 3Ch
Header2 Register at 40h
Header3 Register at 42h
Data
Mover
Port
Figure 5–11. Asynchronous Transmit With Auto Header Insertion
LINK CORE
Transmitter
Receiver
Step 4
Packet sent
to 1394 bus
through the
Phy
5–9
Page 68
5.1.4.2Asynchronous Packet Transmit Without Automatic Header Insertion
In this mode, the packet headers and data information are loaded through the data mover port. This mode
is sometimes called asynchronous packet transmit with manual header insertion. This is because the
header quadlets are not preloaded into the header registers via the microcontroller interface. Instead, they
are inserted
further illustrate the process:
manually
into the data stream at the same time as the rest of the packet. The following steps
Step 1:Asynchronous header quadlets (3 quadlets in quadlet receive mode and 4 quadlets in block
receive mode) are fetched into the header registers through the data mover port.
Step 2:The header quadlets are then forwarded to the transmitter of the link core.
Step 3:Packet data (payload only) is transmitted through the data mover directly to the transmitter
of the link core.
Step 4:Asynchronous packet is sent to the 1394 bus through the Phy.
CFR REGISTER
Quadlet#0
Quadlet#1
Quadlet#2
Quadlet#3
Step 2
LINK CORE
Transmitter
Receiver
Step 4
Packet sent
to 1394 bus
through the
Phy
(headers fetched)
Step 1
(Headers supplied)
Step 3 (packet data)
Step 1
Mover
Header0 Register at 38h
Header1 Register at 3Ch
Header2 Register at 40h
Header3 Register at 42h
Step 3
(package data)
Data
Port
Figure 5–12. Asynchronous Transmit Without Auto Header Insertion
5.2Data Mover Modes of Operation
The data mover (DM) port in the GP2Lynx is meant to handle an external memory interface of large data
packets. The port can be configured to either transmit or receive data packets. The data can be either
asynchronous or isochronous packets. All traffic through the data mover is synchronous to the rising edge
of DMCLK. DMCLK is an output signal at 25 MHz.
The data mover operates by setting the DM control registers. If the DM is configured for transmit mode, it
waits for DMREADY to be asserted before it can drive DMDONE low and fetch the entire block of data one
packet at a time. Upon transmitting the last packet in the block, the DM will drive DMDONE high for a
minimum of one DMCLK cycle (~ 40 ns). The next DMCLK cycle in which DMDONE finds DMREADY high,
the process will be restarted.
The data mover has eight modes of operation which are specified by the DMASYNC, DMHDR, and DMRX
bits in the DM control register at 04h. Table 5–1 shows all the DM modes of operation.
5–10
Page 69
T able 5–1. Modes of Operation
DMASYNCDMHDRDMRXMODE OF OPERATION
000Isochronous packet transmit with auto header insertion
001Isochronous packet receive without header and trailer
010Isochronous packet transmit without header insertion
011Isochronous packet receive with header and trailer
100Asynchronous packet transmit with auto header insertion
101Asynchronous packet receive without headers and trailer
110Asynchronous packet transmit without header insertion
111Asynchronous packet receive with headers and trailer
5.2.1Isochronous Transmit With Automatic Header Insertion
Upon receiving a high on DMREADY, the following sequence of operations is performed:
Step 1:DMDONE will be asserted low (deactivated) at the next DMCLK cycle.
Step 2:The data mover will take the header that has been loaded into the header0 register at 38h
and request the link core to transmit the data onto the 1394 bus.
Step 3:The link core will fetch the header from the header0 register.
Step 4:DMPRE will pulse for one DMCLK cycle before the first data quadlet is sent.
Step 5:The data mover will then begin to fetch the data payload by asserting DMRW high.
Step 6:When the link core has fetched the last data quadlet, the data mover checks if the number
of channels specified by the control registers have been sent. If all channels have been sent
the data mover waits for a subaction gap to occur before asserting DMDONE high to
indicate the end of the cycle. Otherwise the data mover will provide the header in the next
header register and then begin fetching the data payload until all channels are complete.
The timing diagrams in Figures 5–13 to 5–15 illustrate this mode of operation at different transmit speeds.
For simplification, these diagrams show three quadlets of data payload.
DMCLK
DMRW
DMD[0:15]
DMREADY
DMPRE
DMDONE
DMCLK
DMRW
DMD[0:15]
DMREADY
DMPRE
DMDONE
Figure 5–13. Isochronous Transmit With Auto Header Insertion at 400 Mbps
Figure 5–14. Isochronous Transmit With Auto Header Insertion at 200 Mbps
5–11
Page 70
DMCLK
DMRW
DMD[0:15]
DMREADY
DMPRE
DMDONE
Figure 5–15. Isochronous Transmit With Auto Header Insertion at 100 Mbps
5.2.2Isochronous Transmit Without Automatic Header Insertion
Upon receiving a high on DMREADY, the following sequence of operations is performed:
Step 1:DMDONE will be asserted low (deactivated) at the next DMCLK cycle.
Step 2:DMPRE will pulse for one DMCLK cycle before the first header quadlet is sent.
Step 3:The data mover will fetch the header by asserting DMRW high.
Step 4:The data mover will then load the header into the header0 register and request the data
to be transmitted out on the 1394 bus by the link core.
Step 5:The link will fetch the header.
Step 6:DMPRE will pulse for one DMCLK cycle before the first data quadlet is sent.
Step 7:The data mover will then begin to fetch the data payload by asserting DMRW high.
Step 8:When the last data quadlet has been fetched by the link, the data mover will check if the
number of channels specified by the control registers have been sent. If all channels have
been sent the data mover will wait for a subaction gap to occur before asserting DMDONE
high to indicate the end of the cycle. Otherwise the DM will fetch the next header and load
it into the next header register and then begin fetching the data payload until all channels
are complete.
Figure 5–16 shows the timing diagram for this mode at a data transmit rate of 400 Mbps. The dashed
sections indicate repetitive behaviour (when the payload is more than two quadlets long).
DMCLK
DMRW
DMD[0:15]
DMREADY
DMPRE
DMDONE
5–12
Figure 5–16. Isochronous Transmit Without Auto Header Insertion
Page 71
5.2.3Isochronous Packet Receive Without Header and Trailer
In this mode, when the link receives an isochronous packet that is addressed to it, the following sequence
of operations are performed:
Step 1:The packet router control logic will route the packet to the data mover. If the sync bit field
in the header quadlet matches a bit pattern in the ISYNCRCVN field of the isochronous port
register at 18h, DMPRE will be asserted high for one DMCLK cycle.
Step 2:After the header is sent through, DMDONE will be asserted high for one DMCLK cycle.
DMRW is then asserted high as the data payload comes through.
Step 3:After all data has been received on the DMD[0:15] lines, DMRW will be asserted low and
the trailer quadlet will then come out on the DMD[0:15] lines.
PKTFLAG is never asserted high in this mode. Figure 5–17 shows the timing diagram for this mode at
400 Mbps. Figure 5–17 shows the case where DMPRE is asserted high for one DMCLK cycle to indicate
that the sync bits of the received isochronous header matches the contents of the ISYNCRCVN field.
DMCLK
DMRW
DMD[0:15]
PKTFLAG
DMDONE
DMPRE
Header quadlet
Trailer quadlet
Figure 5–17. Isochronous Receive Without Header and Trailer
5.2.4Isochronous Packet Receive With Header and Trailer
In this mode, when the link receives an isochronous packet that is addressed to it, the following sequence
of operations are performed:
Step 1:The packet router control logic will route the packet to the data mover. If the sync bit field
in the header quadlet matches a bit pattern in the ISYNCRCVN field of the isochronous port
register at 18h, DMPRE will be asserted high for one DMCLK cycle. At the same time
DMDONE will be asserted high for one DMCLK cycle.
Step 2:This is followed by DMRW asserted high as the packet comes through. PKTFLAG is only
asserted high when the header quadlet is being received.
Step 3:After all the data payload has been received on the DMD[0:15] lines, PKTFLAG will be
asserted high again as the trailer quadlet is being received. Once the entire packet is
received, the DMRW line will be asserted low .
Figure 5–18 shows the timing diagram for this mode at 400 Mbps. Also, Figure 5–18 shows the case where
DMPRE is asserted high for one DMCLK cycle to indicate that the sync bits of the received isochronous
header matches the contents of the ISYNCRCVN field.
DMCLK
DMRW
DMD[0:15]
PKTFLAG
DMDONE
DMPRE
Trailer quadletHeader quadlet
Figure 5–18. Isochronous Receive With Header and Trailer
5–13
Page 72
Figure 5–19 shows the timing diagram at 200 Mbps when the received packet contains only one quadlet
of payload.
DMCLK
DMRW
DMD[0:15]
PKTFLAG
DMDONE
DMPRE
0000
Packet payload
0000
Trailer quadletHeader quadlet
Figure 5–19. Isochronous Receive With Header and Trailer at 200 Mbps
5.2.5Asynchronous Packet Transmit With Automatic Header Insertion
Upon receiving a high on DMREADY, the following sequence of operations are performed:
Step 1:DMDONE will be asserted low (deactivated) at the next DMCLK cycle.
Step 2:The data mover will take the headers that have been loaded into the header0–header3
registers and request the link core to transmit the data onto the 1394 bus.
Step 3:The link core will fetch the headers from the header0–header3 registers.
Step 4:DMPRE will pulse for one DMCLK cycle before the first data quadlet is sent.
Step 5:The data mover will then begin to fetch the data payload by asserting DMRW high.
Step 6:When the link core has fetched the last data quadlet, the data mover waits until the
destination node returns an
ack_complete
received, the data mover will assert DMERROR high and become disabled.
immediate response. If an
ack_complete
is not
Figure 5–20 and Figure 5–21 show the timing diagram for this mode for the quadlet transmit and the block
transmit cases, respectively. For simplicity , a data block size of three quadlets was selected in Figure 5–20.
Figure 5–22 shows the block transmit case at 400 Mbps.
DMCLK
DMRW
DMD[0:15]
DMREADY
DMPRE
DMDONE
Figure 5–20. Asynchronous Quadlet Transmit With Automatic Header Insertion
DMCLK
DMRW
DMD[0:15]
DMREADY
DMPRE
DMDONE
Figure 5–21. Asynchronous Block Transmit With Automatic Header Insertion at 200 Mbps
5–14
Page 73
DMCLK
DMRW
DMD[0:15]
DMREADY
DMPRE
DMDONE
Figure 5–22. Asynchronous Block Transmit With Automatic Header Insertion at 400 Mbps
5.2.6Asynchronous Packet Transmit Without Automatic Header Insertion
Upon receiving a high on DMREADY, the following sequence of operations are performed:
Step 1:DMDONE will be asserted low (deactivated) at the next DMCLK cycle.
Step 2:DMPRE will pulse for one DMCLK cycle before the header quadlets are sent.
Step 3:The data mover will fetch the headers by asserting DMRW high.
Step 4:The data mover will then load the headers into the header0–header3 registers and request
the data to be transmitted out on the 1394 bus by the link core.
Step 5:The link will fetch the headers.
Step 6:DMPRE will pulse for one DMCLK cycle before the first data quadlet is sent.
Step 7:The data mover will then begin to fetch the data payload by asserting DMRW high.
Step 8:When the link core has fetched the last data quadlet, the data mover waits until the
destination node returns an
ack_complete
received, the data mover will assert DMERROR high and become disabled.
immediate response. If an
ack_complete
is not
Figure 5–23 and Figure 5–24 show the timing diagram for this mode for the quadlet transmit and the block
transmit cases, respectively. For simplicity , a data block size of three quadlets was selected in Figure 5–24.
DMCLK
DMRW
DMD[0:15]
DMREADY
DMPRE
DMDONE
Header Quadlets
Data Quadlet
Figure 5–23. Asynchronous Quadlet Transmit Without Automatic Header Insertion at 400 Mbps
DMCLK
DMRW
DMD[0:15]
DMREADY
DMPRE
DMDONE
Header QuadletsData Quadlets
Figure 5–24. Asynchronous Block Transmit Without Automatic Header Insertion at 400 Mbps
5–15
Page 74
5.2.7Asynchronous Packet Receive With Headers and Trailer
In this mode, when the link receives an isochronous packet that is addressed to it, the following sequence
of operations is performed:
Step 1:The packet router control logic will route the packet to the data mover. At the same time
DMDONE will be asserted high for one DMCLK cycle.
Step 2:This is followed by DMRW asserted high as the packet comes through. PKTFLAG is only
asserted high when the header quadlets are being received.
Step 3:After all the data payload has been received on the DMD[0:15] lines, PKTFLAG will be
asserted high again as the trailer quadlet is being received. Once the entire packet is
received, the DMRW line will be asserted low .
Figure 5–25 and Figure 5–26 show the timing diagram for this mode for the quadlet receive and the block
receive cases, respectively. For simplicity, a data block size of three quadlets was selected in Figure 5–26
DMCLK
DMRW
DMD[0:15]
PKTFLAG
DMDONE
Figure 5–25. Asynchronous Quadlet Receive With Headers and Trailer at 400 Mbps
DMCLK
DMRW
DMD[0:15]
PKTFLAG
DMDONE
Figure 5–26. Asynchronous Block Receive With Headers and Trailer at 400 Mbps
5.2.8Asynchronous Packet Receive Without Headers and Trailer
In this mode, when the link receives an isochronous packet that is addressed to it, the following sequence
of operations are performed:
Step 1:The packet router control logic will route the packet to the data mover. After the headers
are sent through, DMDONE will be asserted high for one DMCLK cycle.
Step 2:DMRW is then asserted high as the data payload comes through.
Step 3:After all data has been received on the DMD[0:15] lines, DMRW will be asserted low and
the trailer quadlet will then come out on the DMD[0:15] lines.
5–16
Page 75
Figure 5–27 and Figure 5–28 show the timing diagram for this mode for the quadlet receive and the block
receive cases, respectively. For simplicity, a data block size of three quadlets was selected in Figure 5–28
DMCLK
DMRW
DMD[0:15]
PKTFLAG
DMDONE
Figure 5–27. Asynchronous Quadlet Receive Without Headers and Trailer at 400 Mbps
DMCLK
DMRW
DMD[0:15]
PKTFLAG
DMDONE
Figure 5–28. Asynchronous Block Receive Without Headers and Trailer at 400 Mbps
5.3Data Mover Byte Mode
In this mode the DMD lines are only 1 byte wide => the maximum speed is only 200 Mbit/sec. Only bits 0–7
will be used for the data bus. DMERROR will be asserted if transmission of a 400 Mbit/sec packet is
attempted.
5.4Data Mover Endian Swapping
In this mode the DMD[0:15] bytes are swapped. If the data mover is in byte mode, the least significant byte
is fetched first (Figure 5–29). If the data mover is not in byte mode, the least significant word is fetched first
and the byte order is then swapped (Figure 5–30).
DMCLK
DMRW
DMD[0:15]
NORM_LINK_DATA
SWAP_LINK_DATA
01020304A1B2C3D4
01020304
04030201D4C3B2A1
A1B2C3D4
DMCLK
DMRW
DMD[0:15]
NORM_LINK_DATA
SWAP_LINK_DATA
Figure 5–29. Endian Swapping in Byte Mode
01020304A1B2C3D4
01020304
04030201
Figure 5–30. Endian Swapping in Word Mode
A1B2C3D4
D4C3B2A1
5–17
Page 76
5.5Data Mover Handshake Mode
Delay time (DMCLK to Q)
ns
Setup time to DMCLK
ns
Hold time from DMCLK
ns
In this mode, when DMDONE is asserted high it will check for DMREADY low as an acknowledge. This is
equivalent to the mode used in the TSB12LV31 (GPLynx), as shown in Figure 5–31.
DMCLK
DMRW
DMD[0:15]
DMREADY
DMPRE
DMDONE
Figure 5–31. Data Mover Handshake Mode (GPLynx mode)
5.6Data Mover Critical Timing
Table 5–2. CLK to Output Timing With Respect to DMCLK
PARAMETERTERMINAL NAMEMINMAXUNIT
t
d0
t
d1
t
d2
t
d3
t
d4
t
d5
t
su0
t
su1
t
h0
t
h1
NOTE: All timing parameters are with respect to the rising edge of DMCLK
Figure 5–32. Clock to Output Timing With Respect to DMCLK
D0
D5
5–19
Page 78
5–20
Page 79
6 FIFO Memory Access
Access to all FIFO memories is fundamentally the same, only the addresses to where the write is made
changes. Figure 6–1 shows the FIFO-address access map. The FIFO is separated into an asynchronous
transmit FIFO (ATF) and a general receive FIFO (GRF), each of 517 quadlets (2 Kbytes). Since
asynchronous packets may also be transmitted through the data mover port and the ATF always has priority
and its data wil be transmitted first.
The suffix _First denotes a write to the FIFO location where the first quadlet of a packet should be written
when the writer wants to transmit the packet. The first quadlet is held in the FIFO until a quadlet is written
to an update location. The suffix _Continue denotes a write to the FIFO location where the second through
n-1 quadlets of a packet should be written. The second through n-1 quadlets are held in the FIFO until a
quadlet is written to an update location. The suffix_Continue & Update denotes a write to the FIFO location
where the last quadlet of a multiple quadlet packet should be written.
ATF_First
ATF_Continue
ATF_Continue & Update
ATF_Burst_Write
GRF Data
Reserved
ATF_First_Update
6.2ATF Access
The procedure for accessing the ATF for a quadlet write operation is accomplished in three successive
steps. T o ensure that an A TF underflow condition does not occur , loading of the ATF in the following manner
is highly recommended:
First Quadlet of the Packet
Successive (N–1) Quadlets of the Packet
•
•
•
Last (Nth) Quadlet of the Packet
Figure 6–2. Asynchronous Packet With N Quadlets (ATV Loading Operation)
Each quadlet can be written into the A TF register on byte (8-bit) boundary or word (16-bit) boundary. T o write
to the A TF in a byte fashion, the following steps should be followed:
Step 1:Writing the first quadlet of the packet:
a) Write the first 8-bits of the quadlet to ATF location 50h.
b) Write the second 8-bits of the quadlet to ATF location 51h.
c)Write the third 8-bits of the quadlet to ATF location 52h.
d) Write the fourth 8-bits of the quadlet to ATF location 53h.
The data is not yet confirmed for transmission.
6–1
Page 80
Step 2:Writing the next (n-1) quadlets of the packet:
a) Write the first 8-bits of each quadlet to ATF location 54h.
b) Write the second 8-bits of each quadlet to ATF location 55h.
c)Write the third 8-bits of each quadlet to ATF location 56h.
d) Write the fourth 8-bits of each quadlet to ATF location 57h.
The data is not yet confirmed for transmission.
Step 3:Last (N
To write to the A TF in a word fashion, the following steps should be followed:
Step 1:Writing the first quadlet of the packet:
Step 2:Writing the next (n-1) quadlets of the packet:
Step 3:Last (N
All writes to the A TF must be quadlet aligned (i.e., only an even number of write accesses is allowed). If the
first quadlet of a packet is not written to the A TF_First location, the transmitter enters a state denoted by an
A TStuck interrupt. An underflow of the A TF also causes an A TSTK interrupt. When this state is entered, no
asynchronous packets can be sent until the ATF is cleared by way of the ATFCLR control bit (bit 0 at CFR
30h). However isochronous packets may still be sent while the ATF is in this state.
th
) quadlet of the packet:
a) Write the first 8-bits of the quadlet to ATF location 58h.
b) Write the second 8-bits of the quadlet to ATF location 59h.
c)Write the third 8-bits of the quadlet to ATF location 5Ah.
d) Write the fourth 8-bits of the quadlet to ATF location 5Bh.
The data is now confirmed for transmission.
a) Write the first 16-bits of the quadlet to ATF location 50h.
b) Write the second 16-bits of the quadlet to ATF location 52h.
The data is not yet confirmed for transmission.
a) Write the first 16-bits of each quadlet to ATF location 54h.
b) Write the second 16-bits of each quadlet to ATF location 56h.
The data is not yet confirmed for transmission.
th
) quadlet of the packet:
a) Write the first 16-bits of the quadlet to ATF location 58h.
b) Write the second 16-bits of the quadlet to ATF location 5Ah.
The data is now confirmed for transmission.
6.3ATF Burst Access
It is allowable to perform a burst write into location 54h (A TF_Continue), which allows multiple quadlets to
load into ATF, but the data is not confirmed for transmission. It is also allowable to perform burst write to
location 58h (ATF_Continue & Update), which allows multiple quadlets to load into ATF, and the data is
confirmed for transmission. Write accesses to address 5Ch (A TF_Burst Write) writes the whole packet into
the ATF. The first quadlet written into ATF has the control bit set to 1 to indicate this is the first quadlet of
the packet, and the rest of the quadlets have the CD bit set to 0. The last quadlet written into A TF confirms
the packet for transmission.
To do a burst write operation the host bus master must continually drive MCS
MD0–MD15 to the A TF during each rising edge of BCLK while MCS
(MCA is always one cycle behind MCS) low. The CD bit is 0 for A TF_Continue and A TF_Continue & Update.
The ATF_First_Update is a unique address location optimised for transmitting zero length isochronous
packets (including asynchronous streaming packets). A zero-length packet contains no data payload, and
only the packet header and header CRC are transmitted.
is low. At the same time it asserts MCA
low. The TSB12LV32 loads
6.4General-Receive-FIFO (GRF)
Access to the GRF is done with a read from the GRF, which requires a read from address 60h. The GRF
will accumulate self-ID packets upon bus-reset. All quadlets of a self-ID packet are saved in the GRF after
6–2
Page 81
power up. Hardware will check to insure the second quadlet is indeed the complement (logical inverse) of
the first quadlet. If there are any errors associated with the self-ID process, a self-ID interrupt will be
generated and the self-ID check register at 38h will be updated to reflect the error(s). This option can be
turned off by setting the FULLSID bit in the control register at 08h to 0.
6.5GRF Stored Data Format
Each quadlet in the GRF is internally 33-bits wide. The most significant bit (extra bit) is used to indicate
whether it is a packet token or a regular received quadlet (received header CRC and data CRC are checked
and not stored in GRF). This bit is called the CD bit, which value is reflected in bit #16 of the FIFO status
register. If CD bit is 1, the next quadlet read from the GRF is a packet token. If the CD bit is 0, the next quadlet
read from the GRF is a regular received quadlet. A packet token is stored as the first quadlet for each
received confirmed packet. The definition for packet token is shown in Table 6-1. Bit 0 is most significant
bit and bit 32 is the least significant bit.
T able 6–1. Packet Token Definition
BITSNAMEDESCRIPTION
0CDCD bit is 1 for packet token. This bit should only be read from the FIFO status
1–2ReservedReserved
3–16QUADLET_COUNTExpected quadlet count after packet token for this received packet.
17–19ReservedReserved
20–24ackCodeIf bit 20 is 0, bits[21:24] are used as the Ack code that was sent back to the
25–26ReservedReserved
27–28SPEEDThe speed code for the received packet.
29–32ReservedReserved
Register at 30h
transmitting node. If bit 20 is 1, it is an error condition and an error Ack code is sent to
the transmitting node.
00 – 100 Mb/s
01 – 200 Mb/s
10 – 400 Mb/s
6–3
Page 82
6–4
Page 83
7 TSB12LV32 Data Formats
The data formats for transmission and reception of data are shown in the following sections. The transmit
format describes the expected organization of data presented to the TSB12LV32 at the host-bus interface.
The receive formats describe the data format that the TSB12L V32 presents to the host-bus interface.
7.1Asynchronous Transmit (Host Bus to TSB12LV32)
Asynchronous transmit refers to the use of the asynchronous-transmit FIFO (ATF) interface. The
general-receive FIFO (GRF) is shared by asynchronous data and isochronous data. There are two basic
formats for data to be transmitted and received. The first is for quadlet packets, and the second is for block
packets. For transmits, the FIFO address indicates the beginning, middle, and end of a packet. For receives,
the data length, which is found in the header of the packet, determines the number of bytes in a block packet.
7.1.1Quadlet Transmit
The quadlet-transmit format is shown in Figure 7–1 and 7–2, are described in Table 7–1. The first quadlet
contains packet control information. The second and third quadlets contain the 64-bit, quadlet-aligned
address. The fourth quadlet is data used only for write requests and read responses. For read requests and
write responses, the quadlet data field is omitted.
Figure 7–2. Quadlet-Transmit Format (Read Response)
prioitytCoderttLabelspd
7–1
Page 84
T able 7–1. Quadlet-Transmit Format Functions
FIELD NAMEDESCRIPTION
spdThe spd field indicates the speed at which the current packet is to be sent. 00 = 100 Mb/s,
tLabelThe tLabel field is the transaction label, which is a unique tag for each outstanding transaction
rtThe rt field is the retry code for the current packet is: 00 = new, 01 = retry_X, 10 = retryA, and
tCodeThe tCode field is the transaction code for the current packet (see T able 6–10 of IEEE–1394
priorityThe priority field contains the priority level for the current packet. For cable implementation,
destinationIDThe destinationID field is the concatenation of the 10-bit bus number and the 6-bit node
destination OffsetHigh,
destination OffsetLow
quadlet dataFor write requests and read responses, the quadlet data field holds the data to be transferred.
rcode
01 = 200 Mb/s, and 10 = 400 Mb/s, and 11 is undefined for this implementation.
between two nodes. This field is used to pair up a response packet with its corresponding
request packet.
11 = retryB.
standard).
the value of the bits must be zero (for backplane implementation, see clause 5.4.1.3 and
5.4.2.1 of the IEEE–1394 standard).
number that forms the destination node address of the current packet.
The concatenation of these two fields addresses a quadlet in the destination node address
space. This address must be quadlet aligned (modulo 4).
For write responses and read requests, this field is not used and should not be written into the
FIFO.
Specifies the result of the read request transaction. The response codes that may be
returned to the requesting agent are defined as follows:
Response
Code
0
NameDescription
resp_complete
Node successfully completed requested operation
1–3
4
5
6
7
8-Fh
reserved
resp_conflict_error
resp_data_error
resp_type_error
resp_address_error
reserved
Resource conflict detected by responding agent.
Request may be retried.
Hardware error. Data not available.
Field within request packet header contains un
supported or invalid vallue.
Address location within specified node not accessible
7.1.2Block Transmit
The block-transmit format is shown in Figure 7–3 and is described in Table 7–2. The first quadlet contains
packet-control information. The second and third quadlets contain the 64-bit address. The first 16 bits of the
fourth quadlet contains the dataLength field. This is the number of bytes of data in the packet. The remaining
16 bits represent the extended_tCode field (see T able 6–1 1 of the IEEE–1394 standard for more information
on extended_tCodes). The block data, if any , follows the extended_tCode.
SpdThe spd field indicates the speed at which the current packet is to be sent. 00 = 100 Mb/s,
TLabelThe tLabel field is the transaction label, which is a unique tag for each outstanding transaction
RtThe rt field is the retry code for the current packet is 00 = new, 01 = retry_X, 10 = retryA, and
TCodetCode is the transaction code for the current packet (see Table 6–10 of IEEE–1394 standard).
priorityThe priority level for the current packet. For cable implementation, the value of the bits must
destinationIDThe destinationID field is the concatenation of the 10-bit bus number and the 6-bit node
destination OffsetHigh,
destination OffsetLow
dataLengthThe dataLength filed contains the number of bytes of data to be transmitted in the packet.
extended_tCodeThe block extended_tCode to be performed on the data in the current packet (see Table 6–1 1
block dataThe block data field contains the data to be sent. If dataLength is 0, no data should be written
01 = 200 Mb/s, and 10 = 400 Mb/s, and 11 is undefined for this implementation.
between two nodes. This field is used to pair up a response packet with its corresponding
request packet.
11 = retryB.
be zero. For backplane implementation, see clause 5.4.1.3 and 5.4.2.1 of the IEEE–1394
standard.
number that forms the node address to which the current packet is being sent.
The concatenation of the destination OffsetHigh and the destination OffsetLow fields
addresses a quadlet in the destination node address space. This address must be quadlet
aligned (modulo 4). The upper 4 bits of the destination OffsetHigh field are used as the
response code for lock-response packets and the remaining bits are reserved.
of the IEEE–1394 standard).
into the FIFO for this field. Regardless of the destination or source alignment of the data, the
first byte of the block must appear in byte 0 of the first quadlet.
7.1.3Quadlet Receive
The quadlet-receive format through the FIFO is shown in Figure 7–4 and is described in T able 7–3. The first
quadlet (trailer) contains the packet-reception status that is added by the TSB12LV32. The first 16 bits of
the second quadlet contain the destination node and bus ID, and the remaining 16 bits contain
packet-control information. The first 16 bits of the third quadlet contain the node and bus ID of the source,
and the remaining 16 bits of the third quadlet and the fourth quadlet contain the 48-bit, quadlet-aligned
destination offset address. The last quadlet contains data that is used by write requests and read responses.
For read requests and write responses, the quadlet data field is omitted.
quadlet data (for write request and read response)
destinationOffsetHigh
prioitytCoderttLabel
Figure 7–4. FIFO Quadlet-Receive Format
The quadlet-receive format through the DM is shown in Figure 7–5 and is described in T able 7–3. This format
is similar to the quadlet receive format for the TSB12LV31(GPLynx). The first 16 bits of the first quadlet
contain the destination node and bus ID, and the remaining 16 bits contain packet-control information. The
first 16 bits of the second quadlet contain the node and bus ID of the source, and the remaining 16 bits of
the second and third quadlets contain the 48-bit, quadlet-aligned destination offset address. The fourth
quadlet contains data that is used by write requests and read responses. For read requests and write
responses, the quadlet data field is omitted. The last quadlet (trailer) contains the packet-reception status
that is added by the TSB12L V32.
quadlet data (for write request and read response)
00000
destinationOffsetHigh
ackCode
000000
spdnumofQuadlets
prioitytCoderttLabel
7–4
Figure 7–5. Data Mover Quadlet-Receive Format
Page 87
T able 7–3. Quadlet–Receive Format Functions
FIELD NAMEDESCRIPTION
numofQuadletsTotal number of quadlets in the current packet (payload and header quadlets only).
ackCodeThis 5-bit field holds the acknowledge code sent by the receiver for the current packet (see
SpdThe spd field indicates the speed at which the current packet was sent. 00 = 100 Mbits/s,
destinationIDThe destinationID field contains the concatenation of the 10-bit bus number and the 6-bit
tLabelThe tLabel field is the transaction label, which is a unique tag for each outstanding transaction
rtThe rt field is the retry code for the current packet is 00 = new, 01 = retry_X, 10 = retryA, and
tCodeThe tCode field is the transaction code for the current packet (see Table 6-10 of the
priorityThe priority field contains the priority level for the current packet. For cable implementation,
sourceIDThe sourceID field contains the node ID of the sender of the current packet.
destination OffsetHigh,
destination OffsetLow
quadlet dataFor write requests and read responses, the quadlet data field holds the transferred data. For
Table 6-13 in the draft standard).
01 =200 Mbits/s, 10 = 400 Mbits/s, and 11 is undefined for this implementation.
node number that forms the node address to which the current packet is being sent.
between two nodes. This field is used to pair up a response packet with its corresponding
request packet.
11 = retryB.
IEEE–1394 standard).
the value of the bits must be zero (for backplane implementation, see clause 5.4.1.3 and
5.4.2.1 of the IEEE–1394 standard).
The concatenation of the destination OffsetHigh and the destination OffsetLow fields addresses a quadlet in the destination nodes address space. This address must be quadlet
aligned (modulo 4). (The upper four bits of the destination OffsetHigh field are used as the
response code for lock-response packets, and the remaining bits are reserved.)
write responses and read requests, this field is not present.
7.1.4Block Receive
The block-receive format through the FIFO is shown in Figure 7-6 and is described in Table 7-5. The first
16 bits of the first quadlet contain the node and bus ID of the destination node, and the last 16 bits contain
packet-control information. The first 16 bits of the second quadlet contain the node and bus ID of the source
node, and the last 16 bits of the second quadlet and all of the third quadlet contain the 48-bit, quadlet-aligned
destination offset address. All remaining quadlets, except for the last one, contain data that is used only for
write requests and read responses. For block read requests and block write responses, the data field is
omitted. The last quadlet contains packet-reception status.
numofQuadletsTotal number of quadlets in the current packet (payload and header quadlets only)
ackCodeThis 5-bit field holds the acknowledge code sent by the receiver for the current packet (see
destinationIDThe destinationID field is the concatenation of the 10-bit bus number and the 6-bit node
tLabelThe tLabel field is the transaction label, which is a unique tag for each outstanding transaction
rtThe rt field contains the retry code for the current packet is 00 = new, 01 = retry_X, 10 = retryA,
tCodeThe tCode field is the transaction code for the current packet (see Table 6-10 of the
priorityThe priority field contains the priority level for the current packet. For cable implementation,
sourceIDThe sourceID field contains the node ID of the sender of the current packet.
destination OffsetHigh,
destination OffsetLow
dataLengthFor write request, read responses, and locks, the dataLength field indicates the number of
extended_tCodeThe extended_tCode field contains the block extended_tCode to be performed on the data in
block dataThe block data field contains any data being transferred for the current packet. Regardless of
spdThe spd field indicates the speed at which the current packet was sent. 00 = 100 Mb/s,
Table 6-13 in the draft standard).
number that forms the node address to which the current packet is being sent.
between two nodes. This field is used to pair up a response packet with its corresponding
request packet.
and 11 = retryB.
IEEE-1394 standard).
the value of the bits must be zero (for backplane implementation, see clause 5.4.1.3 and
5.4.2.1 of the IEEE-1394 standard).
The concatenation of the destination OffsetHigh and the destination OffsetLow fields
addresses a quadlet in the destination nodes address space. This address must be quadlet
aligned (modulo 4). The upper 4 bits of the destination OffsetHigh field are used as the
response code for lock-response packets and the remaining bits are reserved.
bytes being transferred. For read requests, the dataLength field indicates the number of
bytes of data to be read. A write-response packet does not use this field. Note that the number
of bytes does not include the head, only the bytes of block data.
the current packet (see Table 6-11 of the IEEE-1394 standard).
the destination address or memory alignment, the first byte of the data appears in byte 0 of the
first quadlet of this field. The last quadlet of the field is padded with zeros out to four bytes, if
necessary.
01 = 200 Mb/s, 10 = 400 Mb/s, and 11 is undefined for this implementation.
7–6
Page 89
7.2Isochronous Transmit (Host Bus to TSB12LV32)
The format of the isochronous-transmit packet is shown in Figure 7–8 and is described in Table 7–5. The
data for each channel must be presented to the isochronous-transmit FIFO interface in this format in the
order that packets are to be sent. The transmitter sends any packets available at the isochronous-transmit
interface immediately following reception or transmission of the cycle-start message. The speed at which
the current packet is sent is determined by the
dataLengthThe dataLength field indicates the number of bytes in the current packet
TAGThe TAG field indicates the format of data carried by the isochronous packet (00 = formatted,
01 – 11 are reserved).
chanNumThe chanNum field carries the channel number with which the current data is associated.
tCodeThe transaction code for the current packet (tCode=Ah).
syThe sy field carries the transaction layer-specific synchronization bits.
isochronous dataThe isochronous data field contains the data to be sent with the current packet. The first byte of
data must appear in byte 0 of the first quadlet of this field. If the last quadlet does not contain four
bytes of data, the unused bytes should be padded with zeros.
speed
field in the DM control register (bits 22-23)
TAGdataLength
isochronous data
chanNum
tCode
01
10
sy
7.2.1Isochronous Receive (TSB12LV32 to Host Bus)
The format of the iscohronous-receive data through the DM is shown in Figure 7–8 and is described in
Table 7–6. The data length, which is found in the header of the packet, determines the number of bytes in
an isochronous packet. For iscohronous-receive through the FIFO, the last quadlet will be inserted as the
first quadlet in the receive data, as shown in Figure 7–9.
dataLengthThe dataLength field indicates the number of bytes in the current packet.
TAGThe TAG field indicates the format of data carried by isochronous packet (00 = formatted, 01 –- 1 1
chanNumThe chanNum field contains the channel number with which this data is associated.
tCodeThe tCode field carries the transaction code for the current packet (tCode = Ah).
syThe sy field carries the transaction layer-specific synchronization bits.
isochronous dataThe isochronous data field has the data to be sent with the current packet. The first byte of data
spdThe spd field indicates the speed at which the current packet was sent.
numofQuadletsT otal number of quadlets in the current packet (payload and header quadlets only).
errCodeThe errCode field indicates whether the current packet has been received correctly. The
are reserved).
must appear in byte 0 of the first quadlet of this field. The last quadlet should be padded with zeros.
possibilities are Complete, DataErr, or CRCErr , and have the same encoding as the corresponding
acknowledge codes.
7–8
Page 91
7.3Phy Configuration
The format of the Phy configuration packet is shown in Figure 7-12 and is described in Table 7-8. The Phy
configuration packet transmit contains two quadlets, which are loaded into the ATF. The first quadlet is
written to address 50h. The second quadlet is written to address 58h. The 00E0h in the first quadlet (bits
16–31) tells the TSB12L V32 that this quadlet is the Phy configuration packet. The Eh is then replaced with
0h before the packet is transmitted to the Phy interface.
Logical inverse of first 16 bits of first quadlet11111111111111
1716
tcode = ’E’
000
11
Figure 7–11. Phy Configuration Packet Format
The Phy configuration packet can perform the following functions:
•Set the gap count field of all nodes on the bus to a new value. The gap count, if set intelligently ,
can optimize bus performance.
•Force a particular node to be the bus root after the next bus reset.
It is not valid to transmit a Phy configuration packet with both the R bit and T bit set to zero. This would cause
the packet to be interpreted as an extended Phy packet.
T able 7–7. Phy Configuration Packet Functions
FIELD NAMEDESCRIPTION
00The 00 field is the Phy configuration packet identifier.
root_IDThe root_ID field is the physical_ID of the node to have its force_root bit set (only meaningful when R
†
R
TWhen T is set, the PHY_CONFIGURATION.gap_count field of all the nodes is set to the value in the
gap_cntThe gap_cnt field contains the new value for PHY_CONFIGURATION.gap_count for all nodes. This
is set).
When R is set, the force-root bit of the node identified in root_ID is set and the force_root bit of all
other nodes are cleared. When R is cleared, root_ID is ignored.
gap_cnt field.
value goes into effect immediately upon receipt and remains valid after the next bus reset. After the
second reset, gap_cnt is set to 63h unless a new Phy configuration packet is received.
The format of a received Phy-configuration packet is shown in Figure 7–12 and is described in Table 7–8.
When PHY_PKT_ENA (bit 3 of the control register @08h) is set, all Phy packets will be received in the GRF .
One HDRERR interrupt will be generated for every Phy packet received.
Figure 7–12. Received Phy Configuration Packet Format
T able 7–8. Receive Phy–Configuration Packet
FIELD NAMEDESCRIPTION
numofQuadlets Total number of quadlets in the packet. This field is equal to 2 in this case.
ackCodeThis 5–bit field holds the acknowledge code sent by the receiver for the current
spdThe spd field indicates the speed at which the current packet was sent. In this
packet. In this case, the ackCode is equal to 1 (ack_complete)
case, the spd field is equal to ‘00’ (S100)
7.3.1Extended Phy Packets
7.3.1.1Ping Packets
The reception of a Phy ping packet causes the node identified by Phy_ID to transmit Self-ID packet(s) that
reflect the current configuration and status of the Phy . The ping packet provides a method of measuring the
round-trip delay of packets between two nodes on the bus that are farthest from one another in terms of cable
hops. The format of this packet is shown in Figure 7–13 and described in Table 7–9.
transmitted first
00Phy_ID00type (0)00000000
Logical inverse of the first quadlet
Figure 7–13. Ping Packet Format
T able 7–9. Ping Packet Fields
FIELD NAMEDESCRIPTION
Phy_IDPhysical node identifier of the destination of this packet
typeExtended Phy packet type (zero identifies ping packet)
7–10
0000000000
transmitted last
Page 93
7.3.1.2Remote Access Packets
The remote access packet provides a method for a node to access the Phy registers of another node on
the bus. The reception of a remote access packet causes the node identified by the Phy_ID field to read the
selected Phy register and subsequntly return a remote reply packet that contains the current value of the
Phy register. The format of this packet is shown in Figure 7–14 and described in Table 7–10.
transmitted first
00Phy_ID00type (8)cmmd00111
Logical inverse of the first quadlet
port000000
transmitted last
Figure 7–14. Remote Access Packet Format
T able 7–10. Remote Access Packet Fields
FIELD NAMEDESCRIPTION
Phy_IDPhysical node identifier of the destination of this packet
typeExtended Phy packet type ( 8 identifies command packet)
portThis field selects one of the Phy’s ports
cmmdCommand:
0 = NOP (No operation)
1 = Transmit TX_DISABLE_NOTIFY then disable port
2 = Initiate suspend (i.e., become a suspend initiator)
4 = Clear the port’s Fault bit to zero
5 = Enable port
6 = Resume port
7.3.1.3Remote Command Packets
The remote command packet provides a method for one node to issue a number of Phy-specific commands
to the selected port within the target Phy . The reception of a remote command packet shall request the node
identified by the Phy_ID field to perform the operation specified in the
a remote confirmation packet. The format of this packet is shown in Figure 7–15 and described in T able 7–1 1.
transmitted first
cmnd
field and subsequently return
00Phy_ID00type (8)cmnd00111
Logical inverse of the first quadlet
0 0 0port0 0 0
Figure 7–15. Remote Command Packet Format
transmitted last
7–11
Page 94
T able 7–11. Remote Command Packet Fields
FIELD NAME DESCRIPTION
Phy_IDPhysical node identifier of the destination of this packet
typeExtended Phy packet type (8 identifies command packet)
portThis field selects one of the Phy’s ports
cmndCommand:
0 = NOP (No operation)
1 = Transmit TX_DISABLE_NOTIFY then disable port
2 = Initiate suspend (i.e., become a suspend initiator)
4 = Clear the port’s Fault bit to zero
5 = Enable port
6 = Resume port
7.3.1.4Resume Packets
The resume packet is a broadcast packet to all the Phys on the bus. It commands all suspended ports on
the bus to resume normal operation. The reception of the resume packet causes any node to commence
resume operations for all Phy ports that are both connected and suspended. A resume packet requires no
reply . The format of this packet is shown in Figure 7–16 and described in Table 7–12.
transmitted first
00Phy_ID00type (F16)00001
Logical inverse of the first quadlet
00001100000
Figure 7–16. Resume Packet Format
T able 7–12. Resume Packet Fields
FIELD NAMEDESCRIPTION
Phy_IDPhysical node identifier of the source of this packet
typeExtended Phy packet type (F identifies resume packet)
transmitted last
7–12
Page 95
7.4Receive Self-ID Packet
Based on the settings of the RXSID and FULLSID bits in the control register @08h, the self-ID packets can
be either ignored or received into the GRF. Refer to Table 7–13.
Table 7–13. GRF Receive Self-ID Setup Using Control Register Bits (RXSID and FULLSID)
RXSID
(bit 1)
0XSelf-ID packets are not received by the link.
10Only the data quadlet (first quadlet) of the self-ID packets are received into the GRF.
11
Figures 7–17 and 7–18 show the format of a received self-ID packet. For completeness, the figures assume
the cable Phy on the bus implements the maximum number of ports allowed by the P1394a specification.
Both figures show one received self-ID packet. The contents are described in Table 7–14.
FULLSID
(bit 2)
OPERATION
soth the data quadlet (first quadlet) and the logical inverse quadlet (second quadlet) of all
Self-ID packets are received into the GRF.
Figure 7–17. Receive Self-ID Packet Format (RXSID=1, FULLSID=1)
Figure 7–18 shows the format of the received self-ID packet when the FULLSID is cleared. In this case, only
the first quadlet of each self-ID packet is received in the GRF.
Figure 7–18. Receive Self-ID Packet Format (RXSID=1, FULLSID=0)
T able 7–14. Receive Self-ID Function
FIELD NAMEDESCRIPTION
Self-ID Data Quadlet First 32-bits of the first self-ID packet
Logical Inverse of the
Self-ID Quadlet
ACKWhen the ACK field is set (0001), the data in the Self-ID packet is correct. When ACK is ≠ 0001,
Second 32-bits of the first self-ID packet
the data in the self-ID packet is incorrect.
7–13
Page 96
The cable Phy sends one to three self-ID packets at the base rate (100 Mbits/s) during the self-ID phase
of arbitration or in response to a ping packet. The number of self-ID packets sent depends on the number
of ports. Figures 7–19, 7–20, and 7–21 show the format of the cable Phy self-ID packets. Inside the GRF,
the first received quadlet of a self-ID packet is always 0000_00E0h, and the final quadlet is always the
quadlet containing the acknowledgement code.
When there is only one node (i.e., one Phy/link pair) on the bus, following a bus reset, the GRF contains
0000_00E0h and the acknowledge quadlet only.
Example: If there are three 1394.a compliant nodes on the bus, each with a Phy containing three or less
ports, the GRF of any one of the links is shown below. The FULLSID bit is assumed to be set
in this example.
GRF CONTENTSDESCRIPTION
0000_00E0hHeader quadlet for Self-ID Phy packet
Self-ID1Self_ID quadlet for Phy #1
Self-ID1 (inverse)Logical inverse quadlet for Self_ID of Phy #1
Self-ID2Self_ID quadlet for Phy #2
Self-ID2 (inverse)Logical inverse quadlet for Self_ID of Phy #2
Self-ID3Self_ID quadlet for Phy #3
Self-ID3 (inverse)Logical inverse quadlet for Self_ID of Phy #3
0000_000_ACKTrailing acknowledgement quadlet
GRF contents (following a bus reset) with three nodes on the bus
7–14
Page 97
T able 7–15. Phy Self-ID Packet Fields
FIELD
NAME
10The 10 field is the self-ID packet identifier.
LIf set, this node has an active link and transaction layers. In discrete Phy implementations, this shall be the
gap_cntThe gap_cnt field contains the current value for the current node PHY_CONFIGURATION.gap_count
sp
cIf set and the link_active flag is set, this node is contender for the bus or isochronous resource manager as
pwr
p0 – p15
iIf set, this node initiated the current bus reset (i.e., it started sending a bus_reset signal before it received
mIf set, another self-ID packet for this node will immediately follow (i.e., if this bit is set and the next Self-ID
nExtended self-ID packet sequence number
rsvReserved and set to all zeros
†
There is no way to ensure that exactly one node has this bit set. More than one node can be requesting a bus reset at
the same time.
‡
The link is enabled by the link-on Phy packet described in clause 7.5.2 of the IEEE 1394.a spec.; this packet may also
enable application layers.
logical AND of Link_active and LPS active.
field.
The sp field contains the Phy speed capability. The code is:
0098.304 Mbits/s
0198.304 Mbits/s and 196.608 Mbits/s
1098.304 Mbits/s 196.608 Mbits/s, and 393.216 Mbits/s
11Extended speed capabilities reported in Phy register 3
described in clause 8.4.2 of IEEE Std 1394–1995.
Power consumption and source characteristics:
000Node does not need power and does not repeat power.
001Node is self-powered and provides a minimum of 15W to the bus.
010Node is self-powered and provides a minimum of 30W to the bus.
011Node is self-powered and provides a minimum of 45W to the bus.
100Node may be powered from the bus and is using up to 3W. No additional power is needed to enable
the link‡.
101Reserved for future standaraization.
110Node is powered from the bus and is using up to 3W. An additional 3W is needed to enable the link‡.
111Node is powered from the bus and is using up to 3W. An additional 7W is needed to enable the link‡.
The p0 – p15 field indicates the port connection status. The code is:
00Not present on the current Phy
01Not connected to any other Phy
10Connected to the parent node
11Connected to the child node
one)†.
packet received has a different Phy_ID, the a self-ID packet was lost)
DESCRIPTION
7–15
Page 98
7–16
Page 99
8 TSB12LV32/Phy Interface
This section provides an overview of the digital interface between a TSB12L V32 and a physical layer device
(Phy). The information that follows can be used as a guide through the process of connecting the
TSB12L V32 to a 1394 Phy. The part numbers referenced, the TSB41L V03A and the TSB12L V32, represent
the T exas Instruments implementation of the Phy (TSB41LV03A) and link (TSB12LV32) layers of the IEEE
1394-1995 and P1394a standards.
The specific details of how the TSB41L V03A device operates are not discussed in this document. Only those
parts that relate to the TSB12L V32 Phy interface are mentioned.
8.1Principles of Operation
The TSB12LV32 is designed to operate with a Texas Instruments physical-layer device. The following
paragraphs describe the operation of the Phy-LLC interface assuming a TSB41LV03A Phy. The
TSB41LV03A is an IEEE 1394a three port cable transceiver/arbiter Phy capable of 400 Mbits/s speeds.
The interface to the Phy consists of the SCLK, CTL0–CTL1, D0–D7, LREQ, LPS, LINKON, and DIRECT
terminals on the TSB12LV32, as shown in Figure 8–1. Refer to Texas Instruments
SLLA044
for a detailed description of the electrical interface between the TSB12L V32 and TSB41LV03.
Application Report
TSB12LV32
Link-Layer
Controller
CTL0–CTL1
LINKON
DIRECT
Phy/LLC Interface
SCLK
D0–D7
LREQ
LPS
ISODIRECT
TSB41LV03A
Physical-Layer
Device
SYSCLK
CTL0–CTL1
D0–D7
LREQ
LPS
C/LKON
ISO
Figure 8–1. Phy-LLC Interface
The SYSCLK from the Phy terminal provides a 49.152 MHz interface clock. All control and data signals are
synchronized to, and sampled on, the rising edge of SYSCLK.
The CTL0 and CTL1 terminals form a bidirectional control bus, which controls the flow of information and
data between the TSB41L V03A and TSB12LV32.
The D0–D7 terminals form a bidirectional data bus, which is used to transfer status information, control
information, or packet data between the devices. The TSB41L V03A supports S100, S200, and S400 data
transfers over the D0–D7 data bus. In S100 operation only the D0 and D1 terminals are used; in S200
operation only the D0–D3 terminals are used; and in S400 operation all D0–D7 terminals are used for data
8–1
Page 100
transfer. When the TSB41LV03A is in control of the D0–D7 bus, unused Dn terminals are driven low during
S100 and S200 operations. When the TSB12L V32 is in control of the D0–D7 bus, unused Dn terminals are
ignored by the TSB41LV03A.
The LREQ terminal is controlled by the TSB12LV32 to send serial service requests to the Phy in order to
request access to the serial-bus for packet transmission, read or write Phy registers, or control arbitration
acceleration.
The LPS and LINKON terminals are used for power management of the Phy and TSB12LV32. The LPS
terminal indicates the power status of the TSB12L V32, and may be used to reset the Phy-LLC interface or
to disable SYSCLK. The C/LKON terminal is used to send a wake-up notification to the TSB12LV32 and
to indicate an interrupt to the TSB12L V32 when either LPS is inactive or the Phy register LCtrl bit is zero.
The DIRECT
and ISO terminals are used to enable the output differentiation logic on the CTL0–CTL1 and
D0–D7 terminals. Output differentiation is required when an Annex J type isolation barrier is implemented
between the Phy and TSB12LV32.
The TSB41LV03A normally controls the CTL0–CTL1 and D0–D7 bidirectional buses. The TSB12LV32 is
allowed to drive these buses only after the TSB12L V32 has been granted permission to do so by the Phy.
There are four operations that may occur on the Phy-LLC interface: link service request, status transfer, data
transmit, and data receive. The TSB12LV32 issues a service request to read or write a Phy register, to
request the Phy to gain control of the serial-bus in order to transmit a packet, or to control arbitration
acceleration.
The Phy may initiate a status transfer either autonomously or in response to a register read request from
the TSB12L V32. The Phy initiates a receive operation whenever a packet is received from the serial-bus.
The Phy initiates a transmit operation after winning control of the serial-bus following a bus-request by the
TSB12LV32. The transmit operation is initiated when the Phy grants control of the interface to the
TSB12LV32.
The encoding of the CTL0–CTL1 bus is shown in Table 8–1 and Table 8–2.
Table 8–1. CTL Encoding When the Phy Has Control of the Bus
CTL0CTL1NAMEDESCRIPTION
00IdleNo activity (this is the default mode)
01StatusStatus information is being sent from the Phy to the TSB12LV32.
10ReceiveAn incoming packet is being sent from the Phy to the TSB12LV32.
11GrantThe TSB12LV32 has been given control of the bus to send an outgoing packet.
Table 8–2. CTL Encoding When the TSB12LV32 Has Control of the Bus
CTL0CTL1NAMEDESCRIPTION
00IdleThe TSB12L V32 releases the bus (transmission has been completed)
01HoldThe TSB12LV32 is holding the bus while data is being prepared for transmission, or
10TransmitAn outgoing packet is being sent from the TSB12LV32 to the Phy.
11ReservedReserved
8–2
indicating that another packet is to be transmitted (concatenated) without arbitrating
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.