TF536 + 68000 relocator and ROM board project.

[stephen_usher]

I removed the 68030 and checked the /AS line on the socket. That's pulled up to 5V as expected. However, the /AS on the DIP connector was hard 3.3V so that looks like the output from the Xilinx device unfortunately.

Would you mind having a look at it, if you have time? I'd, of course, pay any expenses for potentially fixing it.

[PhilC]

@stephen_usher yes no problem, send it over.

[stephen_usher]

OK... I think it's my board. I've tried it in the other socket and it's trying to boot. (The worn socket on the motherboard isn't helping.) So it *may* be OK.

[stephen_usher]

Got the socket working properly again with a bit of switch cleaner!

I think tomorrow I'm going to pull the trigger on ordering the 4 layer boards... a tad expensive but should actually work properly.

In the meantime I may try some extra pull-ups on the data and address lines to see if that helps with stability.

[stephen_usher]

After a "quick" replacement of the CPU socket on the motherboard I'm back in business.

The 68000 and TF536 will both boot into TOS 2.06 and it says that I have 70MB of memory. *phew"

However, the TF won't boot into EmuTOS 1.0. It constantly reboots. I'm guessing that the TF needs a firmware update, and I'll have to await the replacement "Xilinx" programmer clone to see if that works. If not then I'll have to take up Phil's offer for him to do the update.

[stephen_usher]

What fun I'm having diagnosing the board. This second board is not as good as the first for some reason.

Weird thing is that I've currently got two 68000s in the two slots. One fails the memory tests in the diag cart (and is stable under TOS) and the other doesn't, and it's the one furthest from the connector (in terms f the traces) which works.

On the CPU which fails some of the failing memory locations are always the same and always the same bit, which makes me suspect that I'm just on the limit of something or other and some of the RAM chips are just failing to recognise something. e.g. 0A62C0 always has bit 1 bad.

I've tried adding 2.2K pull-ups to the address bus. If anything it's made it slightly worse rather than better, so it doesn't look to be a data bus issue, or /AS given that I've pulled that up as well. I may investigate the address bus but how could that be affecting individual bits in memory?

Seeing as the problem moves with the CPU, it looks like it's some sort of tolerance issue on the CPU. Goodness knows what.

Anyway, I think I will definitely have to go 4 layer for the final board. This is just not working well enough.

P.S. The CPU which errors on the board works perfectly in the motherboard socket.

[exxos]

Anyway, I think I will definitely have to go 4 layer for the final board. This is just not working well enough.

Your pushing a lot of signals though a small ground wire. it will just act as a multiple screwy oscillator and cause all sorts of odd faults.. I said it right back at the start I couldn't even run a stock 68000 just by running a ground track between CPU pins.. it has to be solid copper. I have kept that philosophy on every board since and everything works as expected. Just look at the grounding on the original ST's and the problems it causes.. Its why the H4 / H5 has solid ground layers as I wouldn't even dream of a board with a "routed ground" system.

With good enough pullups 4.7K - 2.2K you push a fraction more current though the ground which compounds the problem you have. Its not that 2.2K is "worse" its showing you more clearly you have issues with something. Slap a few 10uF ceramic caps across all the main power routes and all the PCB related faults should be no more. If it still doesn't work, at least you solved a lot of problems which should make solving the remaining ones easier.

[stephen_usher]

I've replaced all the power/ground to the CPUs with pretty thick wires, but I take your point. I wanted to experiment with this second board as it costs nothing effectively to do so and I wanted to isolate the individual fixes as much as possible, which with the original board I could no-longer do.

Now I know I can more confidently move forward. I need to check that I have enough leeway on my credit card before I can order the 4 layer boards as I've also just ordered a replacement gaming PC and that's a tad expensive.

[exxos]

I've replaced all the power/ground to the CPUs with pretty thick wires, but I take your point. I wanted to experiment with this second board as it costs nothing effectively to do so and I wanted to isolate the individual fixes as much as possible, which with the original board I could no-longer do.

Take it from someone whos been down that road.. more wires doesn't do anything. Thicker wires will likely compound the problems.. your dealing with "skin effect" you need the surface area...

[stephen_usher]

I've replaced all the power/ground to the CPUs with pretty thick wires, but I take your point. I wanted to experiment with this second board as it costs nothing effectively to do so and I wanted to isolate the individual fixes as much as possible, which with the original board I could no-longer do.

Take it from someone whos been down that road.. more wires doesn't do anything. Thicker wires will likely compound the problems.. your dealing with "skin effect" you need the surface area...

Well, it helps a small amount as the circumferance of the wires I'm using is still far greater than the width of a track and the routing is far more direct too. The resistance certainly is far lower.

I've put the order into PCBWay for the new boards. Let's see what they make of them. This is the first time trying to do 4 layers and hopefully KiCad + Freerouting has done the right thing. It looks OK.