A500 trapdoor memory: free design

[PeteAU]

I built my revC, it's working. Gary from '87.

[katarakt]

I think one of the problems why not so many people built rev C of gary adapter is because there's no combined gerber file together with the memory expansion. Therefore it would be great if someone can update the gerber files so everyone can choose

About the ram problem, i haven't built my expansion yet but have you done a ram-test with the EDO rams where you picked the rams from? I've done so with a few EDO 72pin and found one EDO which has bad adresses (checked with memtest86).

[Mq]

Ok, I redraw schematic and pcb of Mathesar's revision C from KiCad into Eagle. I had to do that because I was ordering some commercial bigger project prepared in Eagle, and then I added Gary adapter pcb to my another project for my personal use. I hope Mathesar is not offended, but I attach Eagle redrawed project. I tried to redraw everything identically. If someone want to do that then there is now possible to import main board and this Gary adapter into one Eagle project.
Finally maybe someone can find an error in my redrawed project?

[Mathesar]

Originally Posted by Mq666

Ok, I redraw schematic and pcb of Mathesar's revision C from KiCad into Eagle. I had to do that because I was ordering some commercial bigger project prepared in Eagle, and then I added Gary adapter pcb to my another project for my personal use. I hope Mathesar is not offended, but I attach Eagle redrawed project.

Eagle??? Grrrrr...

About the gary problem; I'm back from vacation and I found another Gary lying around which I can try.
It's a CSG 5719 / 2890 / 44
Closer to the "problem" Gary's than my original 1987 Gary. Which software exactly are you using for the RAM test? I can then try to replicate your problem.

PS: My motherboard is a rev 5

[Mathesar]

Originally Posted by Mq666

Hahaha:-) It seems that the adapter in revision C was built only by Mathesar, Dragmar and Mq. Nobody else built it around the world:-)

I'm going to do another test in the meantime.
I'm going to solder Gary adapter in previous revision B for this purpose.
I can test it simply because i still have Gary adapter PCB in revision B.
This test should show me if all chip RAMs which I have are working properly as 1,5MB slow, and main extensions boards have no any other mistakes.
I want to have three or four working mainboards with three RAM ICs for future tests to run last fourth IC.

If you set jumper 1&2 to 0.5MB FAST the rev C adapter should behave identically to the rev B as the 74F74 latches are then bypassed. Maybe something to try as well.

[Mq]

I abstained from testing the old adapter rev B. I have a feeling that it makes no sense.

For RAM testing I am using same software as everyone else in this thread.

Ok, I make some pictures of my results. Look at them:
First: on second Gary there is less errors than on first Gary.
And look at the test results of smaller ranges: the larger the range = more errors.

I remind you: errors are only in fourth chip of slow RAM range from d00000.
And ICs are 100% good, because I test few of them and I have same results.

Someone ask me if I test EDO RAM before I desolder them from SIMM: Yes, I tested them first on old PC. And second I tested all my ICs as first and only one on simplest configuration in this project (ramboard with one IC only, and two decoupling capacitors =512kB slow RAM working perfectly).

PS. My motherboard is rev 6A. I have two motherboards in same revisions, and I haven't another motherboards.

[PeteAU]

Are you sure your /XRAM (enable) isn't cut on the motherboard JP7A? People often did that in the old days for 1MB CHIP mod.

If I remove my enable jumper, it ends up mirroring C0 with D0 and causing errors.

[Mq]

My both motherboards 6A were never moded. But it was good point to check, so I checked all circuit for JP7A and JP7B. Everything is as it was done in manufacture and as it is in A500 rev 6A schematic.

Today I built next Gary adapter (same as previous, revision C from my pcb). I did it because testing in two Amigas was annoying when I must move adapter between Amigas.
Both Gary adapters works identically. Same results, same errors.
And on both Amigas there are same issues. Only difference is on one Amiga I have some less errors and they are usualy on other data lines.

[Mathesar]

Tonight I ran the RAM test on my 1987 Gary.
No problems found, even after more than 90 runs.
I will try the other Gary tomorrow. (No energy left today to open up the machine).

In the meantime, could it be a rev5 / rev6a problem?

[Mq]

I suppose it might be a problem on rev 6A only. I have no another motherboard than 6A. I have two 6A only. Previously dragmar told me about problems on 6A too with that 4 Garys (2 works) I told. And now my another friend is building extension, but he has 6A only too, then we will see what he will see:-). By the way, I think 6A is most popular in Poland.

In the meantime, can someone compare my Eagle redrawed schematic and pcb with Mathesar's KiCad original project? I checked it few times, but maybe someone else will see what I can't see?

By the way Mathesar: there is one mistake on Your pcb, and I redraw that mistake too:-) It's not important, but mistake prevents disabling extension by the JP4. There is wire on schematic to set default position to "more RAM". But You probably add this wire before pcb projecting, and You forgot it. Look at pcb and schematic, and You will see there is no way to proper switch from "more" to "512" by the jumper. It will not work because pins 3,9,10 of U1 are connected together and don't be wired properly to the JP4. So our extension must be always turned on and I'm very happy of that if it finally works for me:-)

[motosega]

i only have 6A boards too...

could i use 1mb chip ram on the motherboard and 512k slow ram in the trapdoor? do i need the gary adapter for that?

[alenppc]

I am also using rev 6a. Remember that you have to cut the default JP2 position when using this expansion.

[Mathesar]

Originally Posted by Mq666

By the way Mathesar: there is one mistake on Your pcb, and I redraw that mistake too:-) It's not important, but mistake prevents disabling extension by the JP4. There is wire on schematic to set default position to "more RAM". But You probably add this wire before pcb projecting, and You forgot it. Look at pcb and schematic, and You will see there is no way to proper switch from "more" to "512" by the jumper. It will not work because pins 3,9,10 of U1 are connected together and don't be wired properly to the JP4. So our extension must be always turned on and I'm very happy of that if it finally works for me:-)

It is not a bug, it is a feature

I did not forgot to remove the wire but I put it there on purpose; Pete's rev B adapter has it too (although it is not drawn in the schematics) so I copied it over. This way you don't have to place JP4 and if you do, you will have to cut the trace between pin 1&2.
Like you said, if everything works you will always have JP4 enabled anyway.

[PeteAU]

Originally Posted by Mathesar

It is not a bug, it is a feature

I did not forgot to remove the wire but I put it there on purpose; Pete's rev B adapter has it too (although it is not drawn in the schematics) so I copied it over. This way you don't have to place JP4 and if you do, you will have to cut the trace between pin 1&2.
Like you said, if everything works you will always have JP4 enabled anyway.

Actually it is a small bug on your board, you're missing a trace to U1 from the middle pad. It's sometimes handy for testing plain old 512K boards without having to remove the gary adapter.

If you ever decide to re-spin, I suggest you print the "74F00" & "74F74" silk screen upside down to match what's written on the chip. It had me stumped for a few minutes, because my chips don't have a notch or circle.
---

So Mq666, did you fabricate your eagle board or Mathesar's from kicad? Perhaps there is a small mistake somewhere.

[Mathesar]

Originally Posted by PeteAU

Actually it is a small bug on your board, you're missing a trace to U1 from the middle pad.

That's a mistake indeed. I should have wired that net from the middle pad! Now the "cut the trace" trick doesn't work...

Originally Posted by PeteAU

If you ever decide to re-spin, I suggest you print the "74F00" & "74F74" silk screen upside down to match what's written on the chip.

Good one.

[Mq]

PeteAU: In my every tests in this thread I use my Eagle redrawed pcb only. I don't have any other Gary adapters. I have few rev B pcb, but I don't solder them at all. And You are right about upside down prints on 74's. If someone will merge final version of Gary adapter with main pcb in one project, then it worth to correct it.

alenppc: I have cutted JP2 on my motherboards.

[Mq]

Ok guys. My problems are partially solved.
First, in two issues where 2nd chip not works there are bad soldering on one or two data lines. Solved.
Second, my friend built this extension with my "eagle" gary adapter and everything works for him.
Third, I built another one complete extension with memory chips from another manufacturer, and it works perfectly for me on my both Amigas and every Garys I have.

Conclusion is there is some problem with that memory chips I have in my first extension. Each of them separately works perfectly, but in some combinations between them one interfere with others. And that first extension works perfectly with three chips, but I can't find fourth chip which works with that another three chips. I try three or four different chips and always I have issues.

I think I will leave this one extension as it is with three chips only.

And finally here is a picture of my working complete extension with 4 chips:

[Mathesar]

Congratulations!

Good to hear that you solved it. So we can rule out the Gary then? (than I don't have to open up my 500 tonight).

I wonder why the ram chips won't cooperate with each other. Could it be a power / decoupling problem (always a worry on a 2 layer board)?

[katarakt]

Originally Posted by PeteAU

Actually it is a small bug on your board, you're missing a trace to U1 from the middle pad. It's sometimes handy for testing plain old 512K boards without having to remove the gary adapter.

If you ever decide to re-spin, I suggest you print the "74F00" & "74F74" silk screen upside down to match what's written on the chip. It had me stumped for a few minutes, because my chips don't have a notch or circle.
---

So Mq666, did you fabricate your eagle board or Mathesar's from kicad? Perhaps there is a small mistake somewhere.

Originally Posted by Mathesar

That's a mistake indeed. I should have wired that net from the middle pad! Now the "cut the trace" trick doesn't work...


Good one.

Hey great to hear that it's all good with rev.C of gary adapter. So, do you correct the mentioned issues with this missing trace and the silk screen? Would be great, if you do so please name it rev.D of gary adapter ....and finally with merged main expansion in one eagle and/or kicad file

[Mq]

I think we can definitely rule out the Gary. At least in my case.

I don't know why the RAM chips won't cooperate with each other.
It's strange, especially that i have my problems with that first extension where I have four identical chips desoldered from one SIMM. Next I tryed to change last fourth chip with another few chips, but still same issues occured.

I have no idea why is it like that. I have some another chips to tests, so I would try to mix them and maybe they starts to work with others. Now it is evidently problem of cooperation of some chips as I think. I will check it yet, but in a few days, because I fought with this for about week, and for now I have to catch up on my professional work:-)