EAW P8000 WDC Emulator

Home

QMail HOWTO

- html [de]

- rtf [de]

EAW P8000

Zilog S8000

history

imprint

links







Valid HTML 4.01 Transitional

  

2.x

v2 Prototype v2 Prototype v2.0 v2.1 v2.1
v2 Prototype (2015) v2 Prototype (2016) v2.0 (2016) v2.1 (2016) v2.1.1 (2016)

1.x

v1 Prototype v1.0 v1.2
v1 Prototype (2012) v1.0 (2012) v1.2 (2013)


Features

  • ATMega 1284P running at 18.4320MHz
  • Data storage on
    • SD-Card
    • SDHC-Card
    • PATA (IDE) hard disk
  • compact dimensions
    • fits in a 5.25" Slot in the P8000 Compact
    • can be mounted easily in a 5.25" case for the P8000
  • standard power connector (5.25" Molex plug)

Hardware Downloads

Release Date Version Notes PCB Layout Circuit Diagram
2016-06-22 2.1.1
  • Add missing 4.7k resistor to keep RST2 on high when the P8000 Compact is connected
PDF PDF
2016-04-24 2.1
  • Replace Resistor Networks and one 74245 in the P8000 Interface by a 74573 and 74243
  • Replace the 7403 by a 7400
  • Replace the LM317 which was not running in its specs by a AM1117 3.3V
PDF PDF
2016-03-15 2.0
  • major rework of the PATA interface which increases AVR<->PATA performance by 100%
    • 74573 latches removed from layout, PATA und P8000-interface now sharing the same ports
    • /CS1 is now generated through a negated /CS0
    • remove 7404, 7432, 74138 from layout
  • better RESET-generation with extra cap and diode
  • attach 2L1 (PATA-LED) and 1L1 (Power-LED) to the AVR to indicate different system status (harddisk found, uninitialized harddisk, no harddisk found)
  • completly new PCB layout
  • checked polarisation indicators for all the changed components on the PCB
  • replace MAX3232 by MAX232 which is sufficient
  • changed the tantal e-cap footprints to radial alu e-cap footprints for the MAX232
  • migrate from 2012 KiCad to 4.0.1
PDF PDF
2013-09-04 1.2
  • RST is now generated via the 7403
  • mini DIN connector removed
  • footprints of the 5.25" molex plug and the LED where fixed
  • new arangement of the components and larger board to make the overall construction smaller to fit into the P8000 Compact
  • SD-Card Slot was moved to the back of the board
  • a second 5.25" molex plug was added
  • resistor namings on the PCB (R6, 3R4) where corrected
  • TO92 graphic was fixed to correctly indicate how to place the transistor
  • polarisation indicators for several components where added on the PCB to ease the placement
  • use two 2x13 Pin header instead of the DB25 + EFS26 connector
PDF PDF
2012-06-27 1.0
  • Initial Release
PDF -

Software Downloads

Release Date Version Notes matching Board Software
2016-07-02 2.10
  • When no drive or an unformatted drive is attached, specific errors are reported to the P8000. In some circumstances the errors the Emulator reported where not correct. This is now fixed.
  • always wait some time before an error gets reported back to the P8000 to make sure the P8000 is ready to receive it
  • enhance P8000-Reset detection during Emulator-operation
  • fix LED<->port assignment
  • when a P8000-Reset occured during Emulator-operation, set the STATUS lines back to 0
  • fix the logic which checks if the specified block from the P8000 exceeds the maximum available block. Reserved blocks where considered twice before that fix
2.x Github Release Page
2016-03-15 2.00
  • spend more sleeping time on bootup to fix initialization problems because of ground loops
  • retry SD-Card initialisation up to 10 times because of cards getting up slowly after /RST
  • indicate system status by LED
  • correctly determine if a SD-Card wants block or byte addressing
  • speed up P8000-communication by reducing nano-delays in communication
  • implement CHS support for the PATA driver to support old drives
  • fix the PATA code for one specific drive which refused to work with my emulator by issuing a "Initialize Drive" ATA-Command once at the bootup
  • the emulator does not hangup when PATA is selected but no drive is connected
  • implement parameter validation for the block number, cylinder and head specified in a command by the P8000
  • correct implementation of command code 0x01 (read sector)
  • implement command code 0x02 (write sector) which is used by sa.diags
tar.gz
elf
0.92
  • implement interpreting the drive argument in several command codes (only access to drive 0 allowed)
  • reimplement UART function to get the code BSD licensed
  • little more output on the terminal about found disks on bootup
  • code cleanup, minor tweaks after checking with Cppcheck
1.x never released
2013-04-28 0.91
  • Initial Release
tar.gz
elf

Performance

This are the transfer rates I measured in WEGA. So the times are: P8000 16Bit accesses the Harddisk attached to the original WDC or the Emulator. I executed 10 or more runs of each method. The methods where:

  • dd reading /dev/z with a blocksize of 4K and 3840 blocks (RAM-Disk only 768 blocks)
  • dd reading /dev/rz with a blocksize of 4K and 3840 blocks (RAM-Disk only 768 blocks)
  • dd reading a file in /z with a blocksize of 4K and 3840 blocks (RAM-Disk only 512 blocks)
  • Bonnie which was compiled with the unsegmented WEGA system compiler cc with -O optimisation switch working with a 15MiB file (RAM-Disk only 2MiB file)
  P8000 WDC Emu PATA Emu SD-Card
MR-535 HH-1050 ST251-1 3MiB RAM-Disk
bonnie write intelligently 59.32 KiB/sec 52.08 KiB/sec 59.58 KiB/sec 61.13 KiB/sec 66.84 KiB/sec 58.20 KiB/sec
bonnie read intelligently 51.59 KiB/sec 40.23 KiB/sec 52.22 KiB/sec 100.89 KiB/sec 76.12 KiB/sec 77.52 KiB/sec
dd /dev/z read 58.34 KiB/sec 59.22 KiB/sec 59.65 KiB/sec 128.00 KiB/sec 99.48 KiB/sec 100.91 KiB/sec
dd /dev/rz read 113.69 KiB/sec 82.58 KiB/sec 91.98 KiB/sec 132.35 KiB/sec 224.11 KiB/sec 179.89 KiB/sec
dd /z/testfile read 47.16 KiB/sec 39.07 KiB/sec 47.78 KiB/sec 93.09 KiB/sec 73.25 KiB/sec 74.53 KiB/sec

The following transfer rates are direct accesses from the ATMega without any P8000 communication. The rates strongly depend on the PATA drive and more important on the SD-Card used! With some SD-Cards I experienced less then half of the transfer rates. While the Emulator runs at 18.432MHz for propper baud rate generation, I also tested the speed at 20 MHz.

  Multiblock (4096B) Singleblock (512B)
18.432 MHz PATA SD-Card PATA SD-Card
write 1586 KiB/Sec 546 KiB/Sec 1413 KiB/Sec 211 KiB/Sec
read 1594 KiB/Sec 537 KiB/Sec 1490 KiB/Sec 648 KiB/Sec
20.000 MHz PATA SD-Card PATA SD-Card
write 1683 KiB/Sec 582 KiB/Sec 1502 KiB/Sec 216 KiB/Sec
read 1722 KiB/Sec 581 KiB/Sec 1605 KiB/Sec 673 KiB/Sec

Story

In May 2012 I finally decided to build a replacement for the old P8000 WDC Controller. The P8000 WDC Controller is a Z80 system interfacing with up to three ST506 hard disks. The WDC Controller communicates with the P8000 Computer via a Z80-PIO which is located on the 16Bit-Board of the P8000 Computer. This PIO is dedicated to the communication with the WDC Controller. As usual, the Z80-PIO is divided into two parts - the PIO-A and the PIO-B each one being able to serve 8 Bits of data.

In this implementation here, PIO-A serves as the data-input/output interface transporting the data blocks from the hard disk as well as the so called "command codes". The PIO-A is programmed in mode 0 (output) and mode 1 (input) and is set to each mode during run-time by the firmware of the Z8001. The PIO-B is used in mode 3 (control) and serves as a status information monitor.



Communication between Computer and Controller in detail

The signals

Signal Port I/O
at WDC
Meaning
D0-D7 PIO-A I/O 8Bit data received from and transmitted to the Z8001
/RST PIO-B I When active, the WDC will sleep and accept no commands
/ASTB PIO-A O acknowledges the receipt of data from Port A, or tells the PIO to load data into Port A
/TR PIO-B O Indicates towards the Z8001, that a transfer of data will happen
STATUS0-2 PIO-B O Status Codes to indicate, that an error happened on processing the command or that everything went fine
/TE PIO-B I Indicates, that a transfer of data towards the WDC is about to start
/ARDY PIO-A but inverted I Indicates, that data can be read from the PIO-A, or data can be sent to the PIO-A

Initialisation
Before any operation starts, the Z8001 will set the /Reset Signal to high which literally wakes up the WDC. As long as /Reset remains high, the WDC will listen on the Z80-PIO Port A for receiving Command Codes telling the WDC what to do next.

The Command Code
Every communication from the Z8001 to our WDC starts with a Command Code. Such a Command Code is a string length of nine bytes. The first byte always indicates the action the WDC has to do. The following five bytes contain data that has to be interpreted depending on the command code. Such data could be a block number to write on, a head/cylinder/sector to format or even a RAM address to store or read data from. The seventh and eighth byte may also contain valid data for some commands. If they are used, they contain the length of the data to receive from or transmit to the Z8001 after the Command Code arrival. The last, ninth byte is always 0xFF.

Receiving a Command Code
The transfer of the Command Code is only initiated by the Z8001. But the Z8001 will only send a command code when the WDC sets the correct status from its 3 status lines (0x01 = get command). Every transfer will be then set up by the Z8001 by setting the /TE signal to High. Basically, after /Reset went to High, the Emulator sets status 0x01 and then sits there and waits until /TE goes High. When /TE goes high, this tells the Emulator that data will now arrive. After /TE has gone high, the Emulator then waits until the /ARDY signal goes high as well.
Attention! The /ARDY signal of the WDC is the inverted ARDY signal of the Z80-PIO. So - when the /ARDY signal is high on the WDC, it is low (not active) at the Z80-PIO.
After the /ARDY signal is high at the WDC (=Now data is written into the Z80-PIO by the Z8001) a loop is entered. The loop now waits until the /ARDY signal goes to low (high at Z80-PIO) which indicates that data can now be read from the Z80-PIO. At the same moment, the Emulator sets the /ASTB signal to low (=active) telling the Z80-PIO, that now the data transmission happens. As soon as the data has been read by the WDC, the /ASTB signal is set to high to indicate the Z80-PIO, that the data transfer has been finished. Now the WDC waits until /ARDY goes to high again which indicates, that the Z80-PIO realised, that the data was transferred. The loop will now continue with waiting until /ARDY gets low again and so on. This happens nine times to receive all nine bytes of the Command code.

Analysing of the Command Code
TODO description

Receiving Data
TODO description

Sending Data
TODO description

List of all Command Codes

All known codes are listed. Only the Command Codes used by any available Software where implemented. Some commands have a "Drive" parameter. It is checked, by the emulator that the specified "Drive" by the host does not exceed the number of attached drives. Since only 1 drive is attachable, the "Drive" number must always be 0.

Code Meaning Byte 1 Byte 2 Byte 3 Byte 4 Byte 5 Byte 6 Byte 7
0x01 read data beginning at given sector Drive Cylinder Head Sector No. of bytes to be transfered
0x02 write sector Drive Cylinder Head Sector No. of bytes to be transfered
0x04 format a track not implemented
0x08 read data from WDC RAM the RAM address No. of bytes to be transfered
0x11 read sector taking the Bad Track Table (BTT) into account not implemented
0x12 write sector taking the BTT into account not implemented
0x14 format and reread a track Drive Cylinder Head Sector
0x18 write data to WDC RAM the RAM address No. of bytes to be transfered
0x21 read data beginning at given block Drive the disk block number No. of bytes to be transfered
0x22 write data beginning at given block Drive the disk block number No. of bytes to be transfered
0x24 format a track and provide BTT info Drive Cylinder Head Sector
0x28 read parameter block from WDC RAM No. of bytes to be transfered
0x38 read error statistics No. of bytes to be transfered
0x44 verify track Drive Cylinder Head Sector
0x48 delete BTT from WDC RAM no additional parameters
0x58 read BTT from WDC RAM No. of bytes to be transfered
0x68 write BTT to WDC RAM No. of bytes to be transfered
0x78 write parameter block to WDC RAM No. of bytes to be transfered
0x81 verify sector not implemented
0x82 write and re-read sector not implemented
0x84 delete a track not implemented
0x92 write and re-read sector taking BTT into account not implemented
0xa1 copy block from drive 1 to drive 0 not implemented
0xa2 write and re-read block not implemented
0xc2 store parameter block and BTT on the harddisk no additional parameters

Command Code Usage:


Source Code 0x01 0x02 0x04 0x08 0x11 0x12 0x14 0x18 0x21 0x22 0x24 0x28 0x38 0x44 0x48 0x58 0x68 0x78 0x81 0x82 0x84 0x92 0xa1 0xa2 0xc2
firmware/MON16/p.disk.s X X X X
WEGA/src/cmd/standalone/sa.diags/sa.test.s X X X X
WEGA/src/cmd/standalone/sa.format.c X X X X X X X X X
WEGA/src/cmd/standalone/sa.shipdisk.c X X
WEGA/src/cmd/standalone/sa.verify.c X X X X
WEGA/src/uts/dev/disk.s X X

Open topics

  • Check if BTT handling could or should be implemented. Right now, the BTT is ignored by all accesses
  • Check how the error report in sa.format (key "t" at the last question) looks with the original WDC and how it looks with the emulator

    Original WDC - formating CHS 81/0/0 - 81/8/17

    Total Errors since RESET: 42106---> 55330 * 'A'/30076 * 'B'/13175 * 'C'/48051 * '10'/41851 * '11'/50231 * '12'

    Emulator - formatting CHS 145/0/0 - 145/4/16

    Total Errors since RESET: 14202---> 32968 * 'A'/23583 * 'B'/6972 * 'C'/64069 * '10'/16400 * '11'/1282 * '12'
  • Check how the output of the timing constants behaves in sa.format

    Original WDC

    PAR --- z40: 203  z41: 209  zmin40: 251  zmax40: 253  zmin41: 241  zmax41: 243

    Emulator

    PAR --- z40: 0  z41: 0  zmin40: 0  zmax40: 0  zmin41: 0  zmax41: 0

Other Stuff