A3640 dead, help me

[cpiac64]

Amiga DiagROM V1.2.1 - 16-Jul-20 - By John (Chucky/The Gang) Hertell


Testing if serial loopbackadapter is installed: <> NOT DETECTED
- Parallel Code $ff - Start of ROM, CPU Seems somewhat alive
Checking status of mousebuttons for different startups:

Set all Interrupt enablebits (INTENA $dff09a) to Disabled: Done
Set all Interrupt requestbits (INTREQ $dff09c) to Disabled: Done
Set all DMA enablebits (DMACON $dff096) to Disabled: Done

Testing if OVL is working: OK
- Parallel Code $fe - Test UDS/LDS line
- Test of writing word $FFFF to $400 OK
- Test of writing word $00FF to $400 OK
- Test of writing word $FF00 to $400 OK
- Test of writing word $0000 to $400 OK
- Test of writing byte (even) $ff to $400 OK
- Test of writing byte (odd) $ff to $401 OK
- Parallel Code $fd - Start of chipmemdetection

Addr $00000400
Write: $0F0FF0F0 00001111000011111111000011110000
Read: $0F0FF000 00001111000011111111000000000000 FAILED
Addr $00008400
Write: $0F0FF0F0 00001111000011111111000011110000
Read: $0F0FF000 00001111000011111111000000000000 FAILED
Addr $00010400
Write: $0F0FF0F0 00001111000011111111000011110000
Read: $0F0FF000 00001111000011111111000000000000 FAILED
......
NO Chipmem detected
- Parallel Code $fc - Trying to find some fastmem (as no chipmem found)

[alexh]

I don't know what you're asking. The error msg does not seem to indicate a DEAD A3640, just the opposite, the fact that it ran DIAG ROM and produced output at all means the A3640 is working fine.

The error msg seems to imply you have a problem with CHIP RAM. A common fault on A4000s which have had battery leakage.

If you are an accomplished electronics person who can solder you might attempt to diagnose and fix yourself. If not then try to find a company near you who will repair it for you. There are a couple here in the UK but there must be someone nearer, in the EU, perhaps Germany or Netherlands.

[cpiac64]

i am a technician, the 3640 tested on 3 working a4000, partially work and only with diag rom

[alexh]

The A3640 is regularly fitted with capacitors the wrong way round. The silk screen is wrong too. These capacitors are prone to breaking and leaking. Worth checking out and looking for any damage?

http://amiga.serveftp.net/A3640_capacitor.html

Is your A3640 a Rev3.1 or later? The A3640 Rev3.0 can be updated to 3.2 with a small modification and a GAL upgrade.

If you need to upgrade it might be advisable to go with a newer FW from SpeedGeek.

http://eab.abime.net/showthread.php?p=778105

[cpiac64]

rev 3.1

i have changed all 74hct543, but same problem

[alexh]

Any signs of capacitor leaks? (C107, 105, and 106 are usually soldered the wrong way around and prone to leaking).

Any sign the board has been modified for overclocking? i.e. a socket for the xtal? delay line changes?

[cpiac64]

leak moderate, cap remowed, overclock removed

xtal and delay lane working fine

u204 without label, need verify

[alexh]

If the board was previously Overclocked, that usually requires board modifications.

If you revert the overclock but don't revert the board modifications then it might not work reliably at 25MHz anymore. (e.g. delay line 10ns)

Perhaps look through the two or three A3640 modification instructions and see if you can determine which overclock and the boards current state?

This is the newer definitive A3640 overclock thread :
http://eab.abime.net/showthread.php?p=777332#post777332

This is an older overclock.
http://members.iinet.net.au/~davem2/...ock/a3640.html

[cpiac64]

overclock removed time ago

[alexh]

I see you've had this issue a long time (at least 8 years)

https://eab.abime.net/showpost.php?p=937716&postcount=8

Have you tried a different CPU? A different 68040?

Have you tried a different PSU? The 68040 puts a high load on the +5v line.

[cpiac64]

i have 7 a3640, one ok, one with problem and five dead
i use atx or original psu
i have 10 68040 tested

[alexh]

Where are you getting the DIAGROM output from? The screen? The serial port?

[cpiac64]

serial port

the screen start black, after few seconds blink green, next go black and stop the serial

[alexh]

A3640 Jumper settings?

J400 pins 3-4?
J100 all pins open?

A4000 jumper settings?

J100/4 pins 2-3?
J213 pins 1-2?
J151 pins 1-2?

• You know the A4000 motherboard & Chip RAM are good because it works with a different accelerator?
• You know the 68040 chip is good because you've tried it in another accelerator or you've tried several chips in this A3640 and they are all the same?
• You know the A3640 can read the ROM (which it must be able to because it is running DIAGROM? And it can talk to UART.) This means most of the PINS on the CPU slot are working.

That only leaves timing, something on the A3640 board is not back to factory settings after a modification?

[cpiac64]

tested on 2 a4000d and one a4000t

mobo tested with some a3640, and 2 a3630

soon verify jumper

[Kin Hell]

Talk to Chucky Hertell. - Maybe he has a Rev 3.3 PCB in his drawer?

[Kin Hell]

See here:

https://eab.abime.net/showpost.php?p...postcount=1907

[cpiac64]

i have one empty pcb, but i prefer repair this board and others five

[alexh]

Let's examine the DiagROM failure more closely.

Addr $00000400
                    BYTE0    BYTE1     BYTE2     BYTE3
                 0000 0000 1111 1100 2222 1111 3322 2222
                 7654 3210 5432 1098 3210 9876 1098 7654
Write: $0F0FF0F0 0000 1111 0000 1111 1111 0000 1111 0000
Read : $0F0FF000 0000 1111 0000 1111 1111 0000 0000 0000 FAILED

Addr $00008400
                    BYTE0    BYTE1     BYTE2     BYTE3
                 0000 0000 1111 1100 2222 1111 3322 2222
                 7654 3210 5432 1098 3210 9876 1098 7654
Write: $0F0FF0F0 0000 1111 0000 1111 1111 0000 1111 0000
Read : $0F0FF000 0000 1111 0000 1111 1111 0000 0000 0000 FAILED

Addr $00010400
                    BYTE0    BYTE1     BYTE2     BYTE3
                 0000 0000 1111 1100 2222 1111 3322 2222
                 7654 3210 5432 1098 3210 9876 1098 7654
Write: $0F0FF0F0 0000 1111 0000 1111 1111 0000 1111 0000
Read : $0F0FF000 0000 1111 0000 1111 1111 0000 0000 0000 FAILED

It looks like one byte is not written.

The data bus on the CPU slot is bidirectional and the fact that the CPU is able to read the ROM contents seems to negate the possibility of a failure of the data bus pins/wiring?

But it is strange that it is always the same 4-bits (maybe 8-bits).

It would be interesting to see the code for the RAM test in DiagROM. Is it available anywhere?

I'd like to see the point in the test with these failures, see what the test is doing, the address range it uses is and the data word size.

Is there more to this log? More failures? Or just these 3 addresses?

[cpiac64]

confirm, 11110000 is not written

but i have a doubt, on top of 11110000 read 3, why is 0,1,2,3 ?