Acorn Cambridge Workstation Troubleshooting Guide

Fault finding on the Acorn Cambridge Workstation (ACW) splits along its three halves: the BBC Model B+ derived I/O processor board in the base, the 32016 processor card on the right-hand tray, and the Winchester system on the left-hand tray. The Tube switch on the back of the keyboard lets the I/O processor run on its own, without the 32016 card. The checks, pin numbers and limits come from Acorn's ACW 443 Service Manual (0420,001, issue 1, January 1987); printed page numbers are cited.[1] Read the safety section of the Acorn Cambridge Workstation Maintenance Guide before opening the case: the monitor holds high voltages for hours after switch-off.
What a working machine shows
[edit | edit source]| Condition | Expected |
|---|---|
| Tube off, CTRL and BREAK | "Cambridge Workstation", "Acorn ADFS", then the > prompt
|
| Tube off, CTRL, D and BREAK | "Cambridge Workstation", "ACW DFS", then >
|
| Tube on, after reset | "PANDORA 32016 version X.XX" and "Memory size 4096 kilobytes", or "PANDORA 32000 2.00 4096 KB", then "Acorn ADFS" and the * prompt
|
| Winchester at power-up | Spins up in about 10 seconds with a rising hum, distinct from the steady fan noise |
Sources: I/O board situation (c), p. 61; 32016 Tube signals, p. 102; Winchester fault finding, p. 117.[1]
Acorn's I/O board fault finding starts with the Tube switched off at the keyboard, so that the 32016 card takes no part; the 32016 card has its own set of checks.[2][3]
Power checks
[edit | edit source]| Measurement | Expected | Page |
|---|---|---|
| +5 V (red to black) | 4.9 to 5.2 V, measured with all connectors in place | 6, 117 |
| +12 V (orange to black) | 11.4 to 12.6 V | 6, 117 |
| +5 V at the I/O board | 5 V ±0.1 V; ICs around the board within 5 % | 36, 62 |
| +5 V at the 32016 board | 5.00 V or more at both pairs of Fastons | 100 |
| I/O board +5 V current | 1.5 to 2.2 A, depending on upgrades | 62 |
| Winchester drive | +5 V about 1 to 1.5 A; +12 V up to 4.5 A, falling to about 2 A at speed | 118 |
| Adaptec controller board | +5 V about 1.5 A; +12 V about 250 mA | 118 |
| Host adapter board | +5 V about 500 mA; +12 V not used | 118 |
Sources: ACW 443 Service Manual, pages as given.[1]
Measuring the I/O board current
[edit | edit source]Acorn's method finds a short or an open feed on the I/O board:[2]
- With the supply off, unplug the three red +5 V leads and the purple −5 V lead from the I/O board.
- Set a meter to its 10 A DC range and connect it between one VCC (+5 V) post on the board and the three red wires joined together.
- Switch on only long enough to read the current, because a fault can overheat the tracks. Expect 1.5 to 2.2 A.
- Repeat for each of the three +5 V posts.
- Zero or very low current at one post points to the loom, a connector or a broken track. Zero at all of them means either a short that has tripped the supply (likely) or a completely open feed (unlikely); measure +5 V across the board to tell which.
- Refit the leads: red to VCC, black to 0 V, purple to −5 V. Every Faston post looks the same, so check polarity before switching on.
The same method, with the meter in series with the red or orange wire of each four-way plug, checks the Winchester drive, controller and host adapter against the currents in the table. Take the readings after power-up has settled, because the drive's +12 V current falls once it is up to speed.[4]
I/O processor board
[edit | edit source]Switch the Tube off for all of these checks. Acorn's equipment list for them is its PORT tester (DRAM and every port), a 10 A multimeter, a logic probe, a 20 MHz dual-trace oscilloscope, colour and composite monitors, a cassette player, a disc drive and a frequency counter. The checks below need only the meter, probe and oscilloscope.[2]
| Symptom | Probable cause | Action |
|---|---|---|
| Noise on the screen, no start-up beep, keyboard LEDs dark or wrong | Supply dead, or a fault in the core of the computer | No sound and no LEDs at all: check the supply. Otherwise work through the CPU checks below |
| Screen synchronised but only a flashing cursor at top left | Usually the keyboard | Check the keyboard connection and cable |
| Banner wrong or incomplete | CPU runs and reads the OS ROM | Fragmented banner: check the CRTC address lines for shorts. "Language?" prompt: check link S13 and the paged ROM latch |
| Start-up beep and Caps Lock LED, but no picture | Video circuitry | See video checks |
| Starts normally but a port fails | I/O circuit | Test the port concerned |
Source: section 4.7.2, pp. 60-62.[2]
CPU, clocks and ROM
[edit | edit source]- Check HS and VS at IC78 (6845 CRTC) pins 39 and 40: clean TTL levels, HS every 64 µs, VS every 20 ms. If they run, the CPU has programmed the 6845 and so can read the OS ROM. If they run at the wrong rate, look for a data line fault to the CRTC.
- Check the reset pin of the 6512A, IC42 pin 40: high when running, pulsing low at power-up and when BREAK is pressed. Stuck low: look for shorts or damaged parts around the 555 (IC43), the keyboard and BREAK key. Stuck high: check IC43, the keyboard ribbon and connectors, and the BREAK key.
- Check for activity on SYNC (IC42 pin 7) and R/NW (pin 34). A stuck SYNC means the CPU has stalled; look for a short or open address or data line, or an OS ROM that is not being selected.
- Check the 2 MHz phi1 and phi2 clocks on IC42 pins 3 and 37 (non-overlapping). If they are wrong, check 2M from the video processor IC53, IC33 pin 3, and SYNC 1M at IC41 pin 8; if SYNC 1M is stuck high, find which input of IC41 is stuck low and check the 1 MHz device on that input.
- Check 8, 4, 2 and 1 MHz on IC53 pins 7, 6, 5 and 4. If they are missing but 16 MHz is present on IC53 pin 8, replace IC53; if 16 MHz is missing, check the oscillator formed by half of IC26 and crystal X1. Check 1E on IC25 pin 6.
- If address activity starts after BREAK and then stops, the CPU cannot read the OS ROM. Check that the ROM is seated with every pin in its socket, that NOE pulses low at 2 MHz after BREAK, and that NCS goes low and stays low for a time. Try a known good OS/BASIC ROM and check that link S19 is made East.
- After BREAK, check that the CRTC chip select, IC78 pin 25, pulses low.[2]
Memory, video and ports
[edit | edit source]- DRAM: check that RAS and CAS (pins 4 and 15 of the eight DRAMs) are active. RAS stalled low can destroy the DRAMs; check the RAS and CAS generation from IC53 (8M, 4M, 2M), half of IC31 and the gating. Check the DRAM data lines and the data bus buffer enable at IC49 pin 19.[2]
- Video: if no monitor works, the video processor IC53 is the likely part; check that it is being selected (IC53 pin 3 active while a program writes to &FE20). A distorted RGB picture means a wrongly fitted DIN plug or CSYNC of the wrong polarity, set by link S27.[2]
- Sideways ROMs: write ROM numbers 0 to 15 to the ROM select latch at &FE30 and check that they appear on IC45 pins 11 to 14 (pin 11 most significant). If not, check NPGLD on IC36 pin 13; if that is active, replace IC45. If the latch is right but ROMs still fail, check decoder IC46.[2]
- Cassette and RS423 share the 6850 ACIA (IC82) and the serial processor (IC85). IC85 pin 25 must have a stable 16 MHz ÷ 13 clock (812 ns period) from IC18. For RS423, pin 2 of driver IC95 swings 0 to 5 V and pin 15 swings −5 V to +5 V; on the receive side, IC94 pin 4 is at RS423 levels and pin 7 at TTL. Linking RTS to CTS and Dout to Din makes a loopback.[2]
- Floppy interface: the 1770 needs links S3 South, S4 South, S5 broken, S7 East and S8 made.
*ROMSshould list DFS. With SHIFT, D and BREAK held, drive 0 should start; if not, check MO (IC16 pin 20) high and NMOTOR (IC8 pin 10) low, then IC17 and IC7. No catalogue: check S7 East. Read data on IC16 pin 19 should be negative pulses every 4 or 8 µs.[2] - Printer port: write &00 and then &FF (better &AA and &55) to the VIA at &FE61 with &FF in its direction register at &FE63, and check the data pins on PL9 and on IC10 pins 2 to 9; if the VIA pins are right and PL9 is not, the buffer IC5 is faulty. Strobe (IC10 pin 39) should give 4 µs low pulses alternating with ACK (IC10 pin 40).[2]
- Analogue port: VREF at IC84 pin 8 should be about 1.8 V; the end-of-conversion line, IC84 pin 28, pulses low about every 10 ms after CTRL and BREAK.[2]
- 1 MHz bus: FRED and JIM on IC28 pins 4 and 5 should pulse low for 1 or 1.5 µs when &FC00 and &FD00 are accessed. The bus must be terminated: the terminators RP2, RP3 and RP4 sit on the Winchester host adapter and are fitted when the Winchester is the only device on the bus. Remove them, and keep them, if external 1 MHz bus devices are added.[2]
- Tube: the Tube buffer enable, IC14 pin 19, should be active while a program accesses &FEE0; check the data lines through IC14 to PL12.[2]
- Econet: Acorn left proper Econet service to Econet service centres. Check that every Econet part is fitted, that NNMI (IC42 pin 6) is not held low, and use the shunts S23 to find a broken or shorted station ID track.[2]
- Keyboard: one bad key is a worn switch or a broken track; a group of dead keys points to the connectors, the keyboard cable or the internal cable from the rear panel to the board.[2]
For BBC B+ community repair experience on the same board, see BBC Micro B+ Troubleshooting Guide.
32016 processor board
[edit | edit source]Warning: the 32016 is in a ceramic package that cracks easily when it is levered out of its socket, and Acorn notes that the part is expensive.[3]
- Supply: at least 5.00 V at both pairs of Faston connectors, before and after any work.
- Clocks: XCTL1, PHI1 and PHI2 on TCU pins 13, 11 and 10, and the TTL copy of PHI1 on pin 16. PHI1 and PHI2 rise on alternate rising edges of FCLK, and must reach the CPU and FPU.
- Reset: while pressing and releasing BREAK, HRST (IC1 pin 20), PRST (IC1 pin 37), RSTI (IC32 pin 7), RSTO (IC32 pin 8, IC31 pin 34 and IC2 pin 15) should all go low and rise again.
- Bus activity: NADS on TCU pin 6 pulses at the start of each bus cycle, roughly every four clocks during reads; NRD and NWR appear on IC4 pins 3 and 4.
- Instruction flow: a stream of pulses on NPFS (CPU pin 39) shows the processor executing. If reset and clocks are good but NPFS is quiet, the processor is probably at fault.
- Boot: the BOOTING signal on IC15 pin 6 goes low for an instruction or two after reset, then stays high; if it does, the processor is fetching from the EPROMs.
- Tube: while the Pandora start-up message is passed across, the Tube chip select, IC1 pin 21, should be active.[3]
Floating point unit
[edit | edit source]A faulty 32081 either stops responding, which can make the whole card look dead because Pandora tests for an FPU when the links say one is fitted, or more rarely gives wrong answers. Acorn's quick test is to change the configuration link so that Pandora no longer expects an FPU, then press BREAK; if the card then works, the FPU is at fault. Acorn words this as removing the FPU configuration link, while its link table says link S1 fitted means "no FPU", so check which way your board is set. Wrong answers are best found by substituting a known good FPU.[3][5]
Memory
[edit | edit source]Memory faults are the most common on this board, often from upgrade kits fitted wrongly. Symptoms are a wrong memory size in the Pandora banner, small programs that run while larger ones do not, a program that will not run with an editor loaded, and random corruption of loaded files.[3]
- Upgrades are 512 KB kits of sixteen 256 Kbit DRAMs of 150 ns, and memory must be fitted contiguously from the first bank of sockets. Acorn advises against mixing DRAM makers within one 1 MB bank, because the noise immunity of some makes can be impaired.[5]
- Acorn's 32016 memory test runs from disc (part 2201,257; SHIFT, D and BREAK) or from a pair of EPROMs fitted in place of the two Pandora ROMs with the power off (parts 2201,229 low and 2201,230 high). Keys 0 to 7 test 0.5 MB to 4 MB, 8 runs a refresh test and 9 a simple FPU test. A full 4 MB test takes about 20 minutes.[3][6]
- Reading the reports: errors at a few addresses with one stuck bit point to one DRAM (appendix G maps bit and address to IC); one bit wrong at every address points to a DRAM fitted backwards, a faulty DRAM or a short on the board; a correct pattern shifted up or down in memory points to an address line; random data read back points to the refresh circuit.[6]
Winchester and floppy drives
[edit | edit source]Only the host adapter board is dealer-serviceable; the drive and the Adaptec controller go back for repair or replacement.[7]
| Symptom | Probable cause | Action |
|---|---|---|
| Winchester does not spin up | Power to the drive | Check the four-way supply plug and measure +5 V and +12 V with everything connected; compare currents with the table above |
| ACW will not start after a mouse adaptor is fitted | Badly made IDC connector on the new 1 MHz bus cable | Recheck every IDC fitting on the 1 MHz bus cable |
| Winchester not found, or all 1 MHz bus devices fail | 1 MHz bus buffer enabled or shorted | With the bus idle, D0 to D7 and A0 to A7 on the 1 MHz bus should read about 2.5 V and SCSI DB0 to DB7 about 3 V. Lines at 0 V, or a mix of 0 V and 2.5 V, mean a buffer is enabled or a line is shorted |
| Disc errors after a reformat or drive change | Defect list or drive parameters wrong | Re-enter the defects from the drive label and the Service Card; check the drive parameters (NEC 5126: 4 heads, 615 cylinders, step rate code 2, reduced write current cylinder 128, landing zone 49) |
| Floppy faults | Drive or interface | Unplug the drive's data and power cables, test the interface with a known good 80-track drive, and replace the drive if the interface is good |
Sources: pp. 117-121 for the Winchester, appendix B for formatting, appendix K for the mouse adaptor, section 6.2 and 4.7.11 for the floppy drive.[4][8][9]
Formatting the Winchester erases it completely. The formatter (WFORM, on the ACW ADFS utilities disc) takes about two minutes to format and four to verify, and adds new defects found during verification to the list. Error numbers it reports are offset by &80 when the address is valid, so errors 19 and 99 are the same.[8]
Monitor
[edit | edit source]Acorn refers monitor repairs to Microvitec's service manual. Do not touch HT preset VR4, which is factory-set to 124 V as a safety adjustment. A missing earth lead from the CRT to the CRT PCB can damage the monitor.[10] Discharge the tube first: see CRT Discharge Procedure.
Related pages
[edit | edit source]- Acorn Cambridge Workstation
- Acorn Cambridge Workstation Maintenance Guide
- Acorn Cambridge Workstation Capacitor Guide
- Acorn Cambridge Workstation Documentation
- BBC Micro B+ Troubleshooting Guide
- Hard Drive Maintenance and Repair
- Floppy Drive Repair
References
[edit | edit source]- ↑ 1.0 1.1 1.2 Acorn Computers, ACW 443 Service Manual, reference 0420,001, issue 1, 7 January 1987. Hosted on this wiki: File:Acorn ACW 443 Service Manual.pdf; described on Acorn Cambridge Workstation Documentation.
- ↑ 2.00 2.01 2.02 2.03 2.04 2.05 2.06 2.07 2.08 2.09 2.10 2.11 2.12 2.13 2.14 2.15 ACW 443 Service Manual, section 4.7, pp. 59-74: I/O processor fault finding.
- ↑ 3.0 3.1 3.2 3.3 3.4 3.5 ACW 443 Service Manual, section 5.7, pp. 100-104: 32016 processor PCB fault finding.
- ↑ 4.0 4.1 ACW 443 Service Manual, section 6.9, pp. 117-121: Winchester power checks, address decoding, handshake and bus-free voltages.
- ↑ 5.0 5.1 ACW 443 Service Manual, section 5.5, p. 97: configuration links S1 (no FPU) and S2 (no MMU), other links, memory expansion.
- ↑ 6.0 6.1 ACW 443 Service Manual, appendices G to J, pp. 146-151: memory chip and address map, oscilloscope sync aid, test software part numbers, memory test fault reports.
- ↑ ACW 443 Service Manual, sections 6.2 to 6.4, pp. 105-108.
- ↑ 8.0 8.1 ACW 443 Service Manual, appendix B, pp. 127-130: formatting floppy and Winchester discs; the WFORM formatter on the ADFS utilities disc; defect lists and the Service Card.
- ↑ ACW 443 Service Manual, appendix K, pp. 152-153: fitting and testing the mouse adaptor.
- ↑ ACW 443 Service Manual, chapter 3, pp. 25-31, and section 2.4.2, p. 18.