A3660 CPLD's: to socket or not to socket?

[Mathesar]

Any a3660 builders on this forum?
I am finally building mine.

I had collected all the parts months ago but I never had the time until now.
Tomorrow I am going to program the CPLD's at work and I like to whether other builders have used sockets. I prefer to solder the CPLD's directly onto the board for a cleaner look but I am also a bit scared that I might need to replace them if things don't work.


What is your experience?

[ajk]

Sockets are good for projects that are still evolving, but in this case it's probably not likely that the A3660 would be receiving significant updates. It is a pretty straightforward accelerator, after all.

If you are careful to program the CPLDs correctly, and don't mix them up when installing (like I did...), it's fine to leave out the sockets. And even then it's not that hard to unsolder them with a bit of hot air if you later need to.

[grelbfarlk]

I would say don't put sockets down. If you run into problems, then you aren't sure whether it's the socket or the PLDs you burnt. And before you laugh and say YOU ARE WEAK AND CANT SOLDER SOCKETS, they can be finicky unless you have the proper tools. Even a continuity test can't really tell you whether it's soldered or not because they sit so flat that when you test for continuity it will work. But when you put a chip in it it can have just a slightly bad connection-then when it doesn't boot you're back to, hmm is it the socket or the burning.

But then again if you're not sure you're going to burn them correctly definitely sockets. So I'll say put sockets down.

But on the third hand if you have a scope to check it and some notion of what chips are supposed to do and when then definitely don't use sockets.

On the fourth hand if it's easy for you to remove chips and reburn them correctly next time, just solder them directly.

But if you're not sure then sockets might be easier to insert probes for troubleshooting, so in that case go with sockets.

So in summary sockets are a land of contrasts.

[Chucky]

most tgimes people contacted me with issues soldering the 3660s.. when they skipped the sockets the board booted. (not always but VERY VERY often)

I have a fully socketed 3660 (the first one, actually) that I test all my GALs on first to be sure I have programmed them correctly.. it is not like there will be updates.

[stachu100]

+1 to all above.
*AVOID* sockets for SMD components.

Sockets are ok for PGA chips like 030/040/060, or DIL chips like IC's found on A1000/A2000/A500 or kickstart.

[Chucky]

I have a test 4000 and 1200 board fully socketd.. those boards now more or less only boots diagrom.. sigh. need to change those sockets (but there it is understandable as I test chips so I know the chips I got is working in my small stock)

so.. nah.. sockets will just introduce alot of issues

[grelbfarlk]

Originally Posted by Chucky

I have a test 4000 and 1200 board fully socketd.. those boards now more or less only boots diagrom.. sigh. need to change those sockets (but there it is understandable as I test chips so I know the chips I got is working in my small stock)

so.. nah.. sockets will just introduce alot of issues

That's funny I have an A4000D with several of the GALs around Buster socketed (it needed major fixing) and now that it's all working every other day I have to poke at them to get the thing to boot.

[Mathesar]

Oke, no sockets then .
I have dug up this old programmer at work: http://www.artbv.eu/support/programmers/up1/. It seems to work but of course things are never easy .


First of all, the jedec files I downloaded from chucky's site are not recognized as jedec files? However, the ones from the amiga-wiki site do work. The programmer does ask whether the default fuses should be zero or one? I chose one.
Second, I am using ATF16V8C and ATF22V8C devices. However, when selecting these devices in the programmer I get (only a few) verify errors. When I select the LV variants (so ATF16LV8C) it verifies ok.


Problems, problems

[grelbfarlk]

Originally Posted by Mathesar

Oke, no sockets then .
I have dug up this old programmer at work: http://www.artbv.eu/support/programmers/up1/. It seems to work but of course things are never easy .


First of all, the jedec files I downloaded from chucky's site are not recognized as jedec files? However, the ones from the amiga-wiki site do work. The programmer does ask whether the default fuses should be zero or one? I chose one.
Second, I am using ATF16V8C and ATF22V8C devices. However, when selecting these devices in the programmer I get (only a few) verify errors. When I select the LV variants (so ATF16LV8C) it verifies ok.


Problems, problems

Yeah, that's the reason I try sockets when I'm not sure about burning a particular file. My suggestion, ship your chips to somebody who has burnt them and send them back. Most kind hearted Amiga folks would probably do it for free.

[Mathesar]

Originally Posted by grelbfarlk

Yeah, that's the reason I try sockets when I'm not sure about burning a particular file. My suggestion, ship your chips to somebody who has burnt them and send them back. Most kind hearted Amiga folks would probably do it for free.

It worries me a bit too... but I have been playing a bit more with the programmer and I am now able to program an ATF16V8C correctly.
The thing is, my programmer lists the Atmel ATF16V8C in both DIP and PLCC version. According to the description (in the device selection screen of the programmer) the DIP version has 2195 fuses and the PLCC version has 2194 fuses. The voltages and programming times are also different according to the programmer. I am using a simple ebay PLCC20 to DIP20 adapter btw.

Now comes the weird thing, if I select the DIP version the chip programs and verifies 100%! If I use the PLCC version I get errors at certain addresses...
It all sounds a bit strange to me. The datasheet doesn't mention any electrical difference between the DIP and PLCC version as far as I can tell. One would think the chips are exactly the same except for the package.

Does anyone has any experience with this?
I have used the programmer before to burn new Lattice GAL's for my A3000 and 27C400 eprom's. These all worked.

[Mathesar]

Oh, another quick question:
Are the files from the Amiga-Wiki site identical to the ones from Chucky's site?
In other words, do they work?

Update:
Found the answer here: http://eab.abime.net/showpost.php?p=...6&postcount=48

[Mathesar]

Out of curiosity, what do you guys have your delayline tap set to?
Most pictures I find of populated boards have it set to 5ns.

(see here: https://www.ebay.com/itm/New-Pro-Pro...-/173752433859
here: http://eab.abime.net/showpost.php?p=...4&postcount=13
also here: http://www.amibay.com/showthread.php...rd-newly-built
and here: https://acillclassics.wordpress.com/...3660-with-zif/)

Afaik, for a non-overclocked CPU that is wrong. For a 25Mhz bus clock the period is 40ns. To get the required 90degree phaseshift the delayline should thus be set to 10ns. 5ns might be useful when overclocking the CPU.

Am I missing something here? Mine is set to 10ns (like in the original A3640)

[hooverphonique]

My A4000 w/ A3640 works up to 35MHz with a CLK90 delay of 10ns. At 37MHz, it needs 5ns.
AFAIK, the delay does not specifically need to be 90 degrees, just "short enough", so if the machine can handle running @5ns, it shouldn't be a problem (at any clock speed).

[Chucky]

Originally Posted by hooverphonique

My A4000 w/ A3640 works up to 35MHz with a CLK90 delay of 10ns. At 37MHz, it needs 5ns.
AFAIK, the delay does not specifically need to be 90 degrees, just "short enough", so if the machine can handle running @5ns, it shouldn't be a problem (at any clock speed).

well also a thing with A3000 it seems that it runs much more stable at 10 than 5..

[Mathesar]

Originally Posted by Mathesar

Oh, another quick question:
Are the files from the Amiga-Wiki site identical to the ones from Chucky's site?

I am still waiting for the PLCC28/DIP24 to arrive and in the meantime I have been going through the GAL files.
I can confirm that the files for U209, U211 and U401 are the same on John's site and Amiga wiki. For U203,U204,U205, U207 and U208 I will be using SpeedGeek's waitstate modded files.


However, I am confused about U213 and U401. The files from John and Amiga wiki are not the same. In fact, John's file for U213 has a different fuse size for an ATF22V10C and won't even load. U213 gives different checksums for both files.


Can anyone shed a light on this?

[Mathesar]

Originally Posted by Mathesar

However, I am confused about U213 and U401. The files from John and Amiga wiki are not the same. In fact, John's file for U213 has a different fuse size for an ATF22V10C and won't even load. U213 gives different checksums for both files.
Can anyone shed a light on this?

Ok, all the GAL's have been programmed. I have recompiled the PLD files (taken from the Dave Haynie archives) for U213 and U401. There is still a difference between John's U213 and the recompiled U213 file but that is due to an extra config fuse that is ignored by the ATF22V10C. So, the files are actually the same. U401 remains a mystery to me though, the recompiled file, John's file and the file from Amiga Wiki are all different.

Very strange as U401 is just a simple GAL with some combinatorial functions in it. I will take my chance with the recompiled file. What can go wrong?

Here are the programmed chips, now I can finish the card:

[Chucky]

Yes. U213 is different due to this simple reason: my TL866 REFUSED to progra that chip. it verified it all nicluy but it never worked. I always had to re-read it from a U213 and copyy it to a new.. but wen I read one U213 and wroite that file. THEN it worked.

so blame TL866 for being a crappy and sucky programmer for GALs.
I now never EVER program GALs with my TL! I cannot trust that crap.. I use my superpro 610S instead. then it works.

so . reason is: crappy programmer

[dalek]

Originally Posted by Mathesar

...
However, I am confused about U213 and U401. The files from John and Amiga wiki are not the same. In fact, John's file for U213 has a different fuse size for an ATF22V10C and won't even load. U213 gives different checksums for both files.
...

You can use something like jedrevenge https://sourceforge.net/p/jedrevenge/wiki/Home/ to decompile each of the 3 similar jed files in to their corresponding eqn files.


Then just diff the eqn files to see what's changed.

[Mathesar]

Originally Posted by Chucky

Yes. U213 is different due to this simple reason: my TL866 REFUSED to progra that chip.
...
so . reason is: crappy programmer

Understood, I read more stories about that programmer.
The Art UP-1 programmer I found at work is actually not that bad. It supports all the Atmel chips (A,B,C, low power types, etc). It only acted up on the ATF16V8C PLCC type but I think that it is because of an error in the database. DIP works fine.

Anyway, your file and the recompiled file are the same. I think yours has an extra fusebit in it because you read it from a 'B' type GAL.

[Mathesar]

Originally Posted by dalek

You can use something like jedrevenge https://sourceforge.net/p/jedrevenge/wiki/Home/ to decompile each of the 3 similar jed files in to their corresponding eqn files.


Then just diff the eqn files to see what's changed.

Good idea! I used jed2eqn.exe in the OpalJr package. It wouldn't read the file from Amiga Wiki but it did read john's file and my recompiled file.

The equation are globally the same but John's file has an enormous amount of extra terms summed to it. The extra terms might be doing nothing (eg, they don't hurt) but it does explain why the files are different.
See here: u401_equations.zip