-NMRA packets from the computer to the PIC should consist of 8-bit bytes.
+PROTOCOL BETWEEN HOST AND MASTER PIC
+====================================
-bits 6-0 contain the data to be sent to the track
-bit 7 is 1 if the byte is the last in the packet and 0 otherwise
-
-The first 14 data bits in the NMRA packet should be 1s.
-(i.e. the first two complete bytes should be 01111111 01111111)
-(packets beginning with some other sequence are reserved for non-NMRA messages)
+9600 8N1 over the serial port. The PIC must obey the host's flow
+control line, so that if the host gets backed up none of the PICs
+messages can get lost. (If this is too hard, then the PIC should
+attempt to buffer some data while the host is busy but if the PIC's
+buffer gets too full it should panic.)
-Maximum NMRA message length = 15 bytes (i.e. 105 bits).
-Messages longer than this will cause things to get confused at the moment.
+Each message consists of a number of 8-bit bytes. The top bit of each
+byte is 1 iff there is another byte in the message.
-baud rate 9600
-most significant bit first
-8n1
+ First Second ASCII Message Brief
+ Byte byte etc. or hex name description
+
+From host to PIC:
+
+ > 1 0100 TTT 0 TTTTTTT (a0) POINT Point T fire
+ > 1 1111 111 .... (ff) NMRADATA NMRA data
+ > 1 0001 XXX 0 XXXXXXX (88+) PING Ping `X' (please Pong `X')
+ > 1 0010 RRR E RRR... (90+) POLARITY Set polarity
+ > 1 0011 000 0 MMMMMMM (98+) WATCHDOG W'dog reset, t/o <M*16>ms from now
+ > 0 0010 001 (11) ON Power on
+ > 0 0010 000 (10) OFF Power off
+
+;> 00000000 CRASHED Acknowledge panic, go to readout mode
+;> 00001010 (0a) TELLMODE Confirm mode - say HELLO or CRASHED
+; if crashed, undoes the effect of ack
+
+; In crash readout mode:
+;
+; 00000000 MS Select crash readout mode if not already
+; Reset crash readout pointer to 0
+;
+; 00001000 Acknowledge RS232 framing error or overrun
+; 00001001 Reboot
+;
+; 1vvvvvvv M Prepare byte 0vvvvvvv for transmission to the slave
+;
+; 00000nnn M (n>0) prepare to receive nnnn bytes from slave
+; 0001nnnn M (n>0) transmit nnnn bytes of our own from the
+; readout pointer
+;
+; 001sssss M Select slave S^0x10. Then:
+; After 1vvvvvvv
+; Transmit just 0vvvvvvv to slave
+; and then send some some unspecified byte to host
+; After 0000nnnn
+; Receive nnnn bytes, forwarding each one
+; to the host.
+; After the transaction is complete, 1vvvvvvv
+; or 0000nnnn must be specified again before
+; 001sssss is repeated.
+;
+; 01bbbbbb MS Supply 6 bits for crash readout pointer
+; (crash readout mode only)
+; Effect is FSR << 6; FSR |= bbbbbb
+
+From PIC to host:
+
+ < 1 001Y SSS 0 SSSSSSS (9?) DETECT Train is (Y=1) or is not (Y=0) at S
+ < 1 0001 XXX 0 XXXXXXX (88+) PONG Pong `X' (reply to Ping `X')
+ < 1 100 0001 0 NNNNNNN (ENQ) SPURIOUS Number of spurious fault interrupts
+ < 0 000 1001 (HT) HELLO I am booted
+ < 0 000 1011 (VT) AAARGH Followed by debug chars (only)
+ < 0 000 1101 (CR) WTIMEOUT Watchdog timeout happened
+ < 0 000 0111 (BEL) FAULT Fault exists
+ < 0 000 0110 (ACK) FIXED Fault now fixed
+ < 0 0100 PPP (20+) POINTED Point change done using capacitor P
+ < 0 0101 PPP (28+) CHARGED Point capacitor P is now charged
+ < 0 00000 FF NMRADONE Have processed F NMRADATA message(s)
+
+ < 0000 1010 (LF) } debugging output 0x0a (newline) and
+ < 001C CCCC } (works with terminal 0x20-0x7e
+ < 01CC CCCC except 0111 1111 } emulator, or host logs) (printing ASCII)
+
+(These are all shown big-endian, and all of the numerical
+representations are big-endian too. Where a number is split across
+two or more bytes, the relevant bits are to be concatenated, in the
+order shown, ie bits from the MS byte first, into a larger number.)
+
+
+HELLO, AAARGH and debugging output
+----------------------------------
+
+When the master PIC starts up and has confirmed that all is well (all
+of the other PICs are there, etc), it should send HELLO once.
+
+If the host makes a mistake (eg, sends an unknown command, or does
+something else wrong) or something goes horribly wrong, the master PIC
+should send AAARGH.
+
+The PIC may always send printing ASCII characters and spaces and
+newlines (ie, bytes 0x0a, 0x20-0x7e). These will print out nicely in
+a terminal emulator, if that's what's running on the host. If the
+host is running the real software, that software will put the
+characters sent in its log or somewhere else nicely accessible.
+
+Apart from debugging output, the PIC should send nothing before HELLO
+and nothing after AAARGH.
+
+
+POWER AND FAULT
+---------------
+
+The host can send ON and OFF to turn the track (and various other
+stuff) on and off. After ON, the track power should be enabled and
+transmitting NMRA idle, and the CDU should be enabled.
+
+If the power is ON, and a track power short circuit is detected, the
+PIC should send FAULT. When the short circuit is removed, the PIC
+should send FIXED but not fully reenable track power; track power
+should be reenabled when the host transmits ON.
+
+
+ Track and CDU Track and CDU
+ disabled -------ON-------> enabled
+ .
+ /|\ |
+ | |Short circuit detected
+ \ |
+ \FIXED FAULT
+ \ |
+ \__________________ V
+ operator `
+ fixes the short Short circuit
+ (User Fault indicator lit)
+
+
+If OFF is issued in the first 500ms following a short circuit, it will
+be ignored; if the short circuit persists, OFF during the following
+period will go to the `Track and CDU disabled' state without sending
+FIXED.
+
+
+POINTS and CDU
+--------------
+
+The ON command should cause the CDU to be enabled (and of course all
+point motor outputs should be disabled first - see README.circuitry).
+
+Following ON the host must wait until it receives CHARGED before
+attempting to change a point. After CHARGED it may send POINT, to
+activate the point and direction specified by T. The PICs will report
+POINTED when the point has stopped moving, and CHARGED when the CDU is
+ready to change another point (the host may not send POINT for a point
+on the same CDU until then).
+
+Currently there is only one CDU so P is always 0 (but the PICs need
+not check that the received P value is 0; they may simply assume it).
+
+
+ ----ON-----> CDU is ------CHARGED---> CDU is charged
+ charging _. and ready
+ /|
+ ,----CHARGED--' |
+ / |POINT
+ / |
+ Point has ' V
+ changed; CDU
+ is recharging <----POINTED---- Point is changing
+
+
+Note that OFF will turn the CDU off, and a short circuit (FAULT) will
+turn it off 500ms later (this delay avoids losing races where the host
+sends a point change instruction just before a short happens).
+
+
+PING and PONG
+-------------
+
+The host may send PING at any time; the PIC should reply with PONG
+with the same X as was in the PING message. The host may not send
+another PING until the first one's PONG has come back.
+
+
+POLARITY and POLARISED
+----------------------
+
+The POLARITY command may be sent whether the track power is enabled or
+disabled. The polarity of each segment is `unreversed' after ON; it
+remains constant until from then on except as modified by POLARITY.
+
+The command is of variable length (but at least two bytes):
+
+ > 1 0010 RRR E RRR... POLARITY Set polarity
+
+Each byte after the first contains 7 more R bits. The first R bit
+(most significant R bit in the first byte) corresponds to track
+reversal segment 1; The next bit (2nd most significant bit in the
+first byte) corresponds to track reversal segment 2; and so on.
+
+Bits which do not correspond to defined reversal segments will be
+ignored by the PICs. The host must send exactly as many bytes as are
+necessary to include all of the reversal segments for each reversers
+board (for every potential reversal segment, regardless of whether
+that segment is a defined segment corresponding to some actual track;
+however a board with _no_ reversers segments used does not count).
+
+For example, if there are 14 reversible segments (numbered 1 to 14)
+then the following message
+ 1 0010 000 1 000 1000 0 111 1010 Actual message
+ (E RRR) (E RRR RRRR) (E RRR R---) } helpful annotations
+ 1 111 1111 } and commentary
+ 123 456 7890 123 4567 }
+specifies to reverse segments 7 and 11 to 14. The trailing bits are
+for segments 15 to 17 and are ignored. (Note that the assignment of
+physical segments to segment numbers is complex due to bit-twiddling.
+see detpic/reverse.asm and layout/data2safety.)
+
+The PIC will reply to POLARITY with POLARISED when the polarity change
+is complete. The host must not send another POLARITY until then.
+
+
+NMRADATA and NMRAFULL
+---------------------
+
+The data bits in all of the bytes of the NMRADATA command (including
+the first) are simply transmitted as NMRA data to the track (most
+significant bit first). The top `end of packet' bit is not
+transmitted, though.
+
+The first 14 data bits in the NMRA packet should be 1s. (i.e. the
+first two complete bytes should be 11111111 11111111). Packets
+beginning with a different first byte are reserved for other commands
+to the PIC and the 14 idle bits are a requirement of the NMRA
+specification.
+
+The maximum NMRA message length is 15 bytes each carrying 7 bits of
+actual NMRA data (i.e. 105 bits).
+
+Up to three NMRADATA commands may be supplied by the host to the
+master PIC, and their will be transmitted in sequence. After each
+NMRADATA is completed, the PIC will send an NMRAFULL message to the
+host. In the NMRAFULL message, F is the number of completely-received
+NMRADATA commands awaiting transmission to the track.
+
+If the PIC runs out of NMRA data, it will transmit an NMRA idle
+stream. It is an error for the host to try to have more than three
+outstanding NMRADATA commands.
+
+
+DETECT
+------
+
+The DETECT command indicates to the host whether there is currently a
+train being detected at a specific location. The PIC must send a
+DETECT with Y=1 when a train is detected in a location where there was
+previously none, and with Y=0 when a train ceases to be detectable for
+more than a small amount of time.
+
+At HELLO, the host will assume that no trains are being detected.
+
+
+RAM (data) memory map
+=====================
+
+The data memory map (for PIC18F458) looks like this:
+
+ 0x000-0x05f Access bank RAM - RAM locations accessible via
+ access bank instructions; also form part of
+ RAM page 0
+ 0x060-0x0ff Remainder of RAM page 0, accessible only via correct
+ BSR setting (ie, BSR==0), INDF, etc.
+
+ 0x100-0x1ff RAM page 1, accessible only via bank switching etc.
+ 0x200-0x2ff RAM page 2, accessible only via bank switching etc.
+ 0x300-0x3ff RAM page 3, accessible only via bank switching etc.
+ 0x400-0x4ff RAM page 4, accessible only via bank switching etc.
+ 0x500-0x5ff RAM page 5, accessible only via bank switching etc.
+
+ 0x600-0xeff Nothing here, don't try to access.
+
+ 0xf00-0xf5f SFR's (memory-mapped peripherals etc.) accessible
+ only via correct BSR, INDF, etc - but these are only
+ the CAN SFR's and we do not use the CAN controller.
+ 0xf60-0xfff SFR's accessible via access bank (also form part
+ of RAM page 15).
+
+
+See common.inc for actual uses of the RAM areas.
+
+
+Program (flash etc.) memory map
+===============================
+
+Program memory map (for PIC18F458) looks like this:
+
+ 0x00 0000- Program memory
+ 0x00 7fff Contains actual program instructions and can also
+ contain preprogrammed data provided via special .asm
+ files. Notable contents and addresses:
+ 0x00 0000 reset vector
+ 0x00 0008 high-priority interrupt vector
+ 0x00 0018 low-priority interrupt vector
+ See common.inc for some special tables in here, for
+ morse messages, pin/hardware-object definitions, etc.
+
+ 0x20 0000- ID locations
+ 0x20 0007 Programming which varies per PIC. Programmed by
+ idlocs*.asm which are made by make-idlocs and
+ included in perpic*.hex. Contents:
+
+ 0x20 0000
+ bits 7-5 = 000
+ bits 4-0 = PIC number (guaranteed to be
+ in the range 0..31 inclusive)
+ 0x20 0001
+ bit 7 = 1 for the main PIC (#0)
+ 0 otherwise
+ bit 6 = 1 for Reversers board, 0 for Detectors
+ bits 0-5 = currently unused, set to 0
+
+ 0x20 0002- } not currently used,
+ 0x20 0007 } may contain anything
+
+ 0x30 0000- Hardware configuration
+ 0x30 000f Defines (clock source, WDT operation, etc.)
+ Probably best not to touch. `config.asm' provides
+ correct contents, which is included in *-withcfg.hex
+ and perpic*.hex.
+
+ 0x3f fffe- Hardware device ID
+ 0x3f ffff Fixed at manufacturing time; can be read to discover
+ hardware type and version (probably not very useful)
+
+ 0xf0 0000- EEPROM data area
+ 0xf0 00ff Not currently used by us
+
+ 0x01 0000- } These locations, not listed above,
+ 0x1f ffff } do not correspond to anything - there
+ 0x20 0008- } is no hardware or memory in the chip
+ 0x2f ffff } at these locations.
+ 0x30 0010- }
+ 0x3f fffd } Accessing them isn't useful
+ 0x40 0000- } and should probably be avoided.
+ 0xef ffff }
+
+
+(Buffer page 50 0000h reserved for NMRA) XXXX these look wrong
+(Buffer page 40 0000h reserved for i2c) XXXX -iwj
+
+
+
+I2C
+===
+(slave addresses will be 10xxxxx where xxxxx=PIC number above)