TF536 + 68000 relocator and ROM board project.

[stephen_usher]

It's almost as if the decoupling capacitors aren't doing their job.

[stephen_usher]

I'm completely mystified now.

One CPU (68000 socket) runs fine now. Although it does produce the ground spike it's nowhere as much as the second (Accelerator socket).

Scoping each of the CPU lines in parallel I can see no real noticable difference other than the voltage on the accelerator CPU is a teeny-tiny bit less. The only decernable difference is on A8 where the non-data "shark's fin" rise is slightly faster on the 'Accelerator' CPU but the signal in every other aspect is identical.

Yet, one CPU can run YAART all day and the other gets large numbers of errors (usually single bit and in batches where all the memory locations have the same bad bit).

None of the CPU ground connections are now via the board traces.

Oh, and I found the problem with the second CPU crashing if there was disk access... my SMT soldering struck again. /BGACK input to the 74HCT157D wasn't soldered properly so /BGACK was floating on the second processor.

[stephen_usher]

I must be doing something right. I decided to try the TF536 in the "accelerator" socket rather than a 68000 and everything's fine.

Even the floppy works, unlike with the Mk.I board.

[stephen_usher]

Well, well. It does seem to be stable running on the TF536.

YAART-TT ran on the ST RAM several times with no errors, as did Gembench and Frontier.

Even Spectre GCR works!

IMG_2084.jpg (244.62 KiB) Viewed 3221 times

I just need to find some low profile, turned pin header strips now so as to lower the whole stack by about 1cm in total. The only turned pin headers I can find have massively tall stand-offs on the pins. The alternative would be to find ones with longer mating pins so that I could push them through the board and solder underneath so that the board was flush with the top of the socket, but I can only find one reference to those on-line and no-one looks as if they stock them. (I tried with the standard ones but they're too short to make contact properly. The 10mm rather than 7mm pins should work.)

[stephen_usher]

And the Turbo LED now works. I had the transistor the wrong way around as I thought the flat side was the opposite to what it is. So, reducing the height is the next major hurdle.

If I swapped the 68000 socket on the motherboard for a spring-connector one then I could use a long square-pin header through the hole. That would reduce the height by ~3mm. I could do the same for the TF536. The down side is that the spring connector sockets wouldn't hold the cards in place so well.

[PaulJ]

@stephen_usher , look here for some really low profile pins. Don't know if you want to deal with US but they must have something comparable in the EU.

https://www.peconnectors.com/machined-h ... p-sockets/


Shows hight difference between Exxos pins and the low profile pins. Should give you the highth you need.

short.jpg (45.08 KiB) Viewed 3203 times

[stephen_usher]

That looks interesting, thanks!

[stephen_usher]

Oh, @derkom and @PhilC you may like to know that the biggest decrease in noise on the ground I got yesterday was when I fitted a second decopling capacitor on the second Vcc input on the processor. The 68000 needs two, one for each power input.

@derkom your relocator board design only has one, as does my Mk.I board, and this could introduce instability.

[derkom]

Oh, @derkom and @PhilC you may like to know that the biggest decrease in noise on the ground I got yesterday was when I fitted a second decopling capacitor on the second Vcc input on the processor. The 68000 needs two, one for each power input.

@derkom your relocator board design only has one, as does my Mk.I board, and this could introduce instability.

I was curious about this when building it, not knowing a whole lot about best practises. I decided not to concern myself too much with it, because in my particular build, one side has the TF (with two caps) and the other side has the stack beginning at the bottom with the Lightning ST (also with two caps). Since both 68K sockets were going to have two decoupling caps right on top of them, I figured even one is probably superfluous in my design.

If I had been building a board for a more general application, as you are, I probably would have been more careful about this aspect of the design.

[stephen_usher]

Yeah, decoupling caps are tricky. They need to be as close to the chip pin as it is possible for them to be. The further away they are the less effective they are. They shouldn't be more than a couple of mm away in any case, so the caps next to the socket on the main board are too far away to have any effect on your board.