Freshly built TF536 no post, dodgy ebay cpu?

[Evilged]

Hi,
I'm getting a dark red screen like you would not having a CPU in.
I've double checked all the soldering (under microscope), given it another IPA wash down, re-flashed the CPLD again from the Pi.
I suspect it may be the ebay chip, a ceramic gold 68030RC50, it was sold as an 'untested' pull and looks like it have a rough life as its was scuffed and I had to straighten most of the pins. I went for this rather than the ones from china as I thought there was more chance of it being original and not a re-badge.
Is this the most likely cause ( excluding me being an idiot :D ), do they usually just fail dead or go unstable?
I don't suppose there any way of checking it is the CPU without another one?
I presume a lesser 33mhz one would post at least if not run at 50 stable?
Its been fun building this anyway!
More of a moral support post to check i haven't missed anything stupid before I go cpu shopping again, built my hopes up after finishing at 1am last night, thought I would be playing with wondrous power of the TF tonight.
My Amiga obsession is getting expensive, just don't tell the Mrs :D
Thanks,
Ged

[terriblefire]

Hi,
I'm getting a dark red screen like you would not having a CPU in.
I've double checked all the soldering (under microscope), given it another IPA wash down, re-flashed the CPLD again from the Pi.
I suspect it may be the ebay chip, a ceramic gold 68030RC50, it was sold as an 'untested' pull and looks like it have a rough life as its was scuffed and I had to straighten most of the pins. I went for this rather than the ones from china as I thought there was more chance of it being original and not a re-badge.
Is this the most likely cause ( excluding me being an idiot :D ), do they usually just fail dead or go unstable?
I don't suppose there any way of checking it is the CPU without another one?
I presume a lesser 33mhz one would post at least if not run at 50 stable?
Its been fun building this anyway!
More of a moral support post to check i haven't missed anything stupid before I go cpu shopping again, built my hopes up after finishing at 1am last night, thought I would be playing with wondrous power of the TF tonight.
My Amiga obsession is getting expensive, just don't tell the Mrs :D
Thanks,
Ged

I would expect it would post even with the worst CPU in the world. So long as the CPU isnt dead. This is because the TF536 runs at 7Mhz when accessing ROM and CHIP ram on the Amiga.

Have you tried DiagROM? Get a serial cable on there and check its working with a normal 68000 in the machine fist. Probably would start there. It will only run the CPU at 7mhz so would avoid all issues related to CPU speed.

Next thing to test is the CPUCLK signal at the 030.. See if its there. Next check reset goes high and look for wiggling on AS20/AS

Check that DTACK/DSACK1/DSACK0 are not stuck low.

Some hi res images would be useful.

EDIT: if you havent got the OBV files already try https://github.com/terriblefire/terribl ... ster/tf536

[Evilged]

I would expect it would post even with the worst CPU in the world. So long as the CPU isnt dead. This is because ROM and CHIP ram run at 7Mhz on the Amiga.

Have you tried DiagROM? Get a serial cable on there and check its working with a normal 68000 in the machine fist. Probably would start there. It will only run the CPU at 7mhz so would avoid all issues related to CPU speed.

Next thing to test is the CPUCLK signal at the 030.. See if its there. Next check reset goes high and look for wiggling on AS20/AS

Check that DTACK/DSACK1/DSACK0 are not stuck low.

Some hi res images would be useful.

Brilliant thank you!
I will pop a DiagRom in and give that a go, then check for CPUCLK and go from there.
I'll try to take some good pictures, have to see what I can get done in between meetings tomorrow.
Thanks,
Ged

[Evilged]

Had to reflash Diagrom on a new chip as the other one seems to have died (pin detect error when trying to read it) so that is fried, probably the A500+ I got today I tried it in, I didn't notice some of its 'issues' before i turned it on, impatience always gets you.
Anyway I get no output on serial when the tf536 is in as apposed to with the 68000 which is fine.
Tested the CPUCLK on the 030 I think there is a signal (there is voltage) but my meter only measure up to 10Mhz and my scope 35Mhz (old Hameg) so can't vouch for what it is exactly.
I will probe the other pins with my cheap logic analyser and try a reflow to make sure there is nothing i'm missing under the scope, sleep first though else I'll just break something :D
Thanks again,
Ged

[Evilged]

Progress.
I found some solder bridges on pins 105-107 on the Xylinx (I am working honest ;) )
Its amazing the difference not as tired eyes and the right light angle makes to seeing things.
Sorted those out but still the same result.

So another reflow along 37-72 and SUCCESS it now posts.
I also think my cpu socket was a a bit cruddy and it doesn't like the relocator I have without holding it down which was complicated things.
Looks like my dodgy soldering to blame not the cpu.
But...
Next problem, no fast memory detected in diagrom, well it detects it in autoconfigure-detailed but doesn't seem to assign.

Diagrom-autodetection-detailed.jpg

Unfortunately I had to wipe my 3.1 rom to put Diagrom on it so cant test under that, unless I wipe and reflash that again.
Another reflow?
Thanks

[arkadiusz.makarenko]

Can you run memory test manually(diagrom menu 2) in range 0x40...to0x44... (can't remeber how many zeros is there).

[terriblefire]

So i never believe that there are no shorts.. even when i tell it to myself.

Ok so try a few things..

goto Diagrom memory menu (2) and try complete memory detection(6). It should find the memory between 40000000 and 44000000.

If that doesnt work go to memory menu (2) and manual memory edit (8) and press G, enter 40000000 .. start typing... see if anything is retained or lost / mirrored. Particularly interested in how the first 4 byte behave as that indicates the bits of the data bus and ram chips.

[Evilged]

Yep its always a shorts :D, like with programming I find its always the data I am usually more careful just over tired at the moment, too much work too little time to play.

Complete memory detection no joy.
Manual memory edit

memory-write-40000000.jpg

The first two bytes are un-writable as you can see they stay as FF 3rd and 4th will update as shown with 47/G.
I'm guessing there is still a dry joint somewhere then?
Thanks

[terriblefire]

I would suspect the connections to the ram chip nearest the CPLD.

[Evilged]

Brilliant I will get on that as soon I get the chance to ignore work some more :D
Thanks