REV 3 - REV 5 - The beginning (ST536)

[PhilC]

Have you tried a different CPU to rule that out?

I Just swapped the CPU just to doubly make sure.. (was swapping CPU yesterday in fact) . Made no odds :(

I think this problem might be backing onto a similar thing when I thought 32bit ROM copy wasn't working.

My rev 5.50 seems to have fits of random oddness.. I've not seen the problem yet on the 5.62 board, but haven't done much testing with that yet.

There must be some odd timing issue somewhere, or the cache malfunctions without good enough caps.. no idea.

At least that rules out the CPU and its cache. What revision was the board you'd sent to me?

[exxos]

At least that rules out the CPU and its cache. What revision was the board you'd sent to me?

5.50 I think.. I've got 2 of those boards, One randomly acts odd.

I think is why I get so confused about it all because the board will not work for hours, and I will change something in the code, and then it will start working fine.. So I think that code change fixed it.. When in reality.. Because the problem is intermittent it all becomes red herrings.. And the problem will probably come back a couple of days later again anyway :roll:

The cache problem is just odd.. I mean why would it fail on ROM copy, then later fail on TTram test.. You would think at some point the cache would get flushed or overwritten or something.. I suppose it depends what the cache ends up with , assuming some sort of corruption issue. But it would have to be corrupting something that destabilises the whole system in various routines in TOS.. It just does not seem likely somehow.. But here we are.. :lol: :roll:

[exxos]

I've been plastering ground wires and fiddling with the code and don't seem to make any progress..

Most significant change was just changing the DEBUG output from CLKCPU to 1'b1. It then completes ROM verification but then locks up solid.

I think there must be some sort of ground bounce thing going on somewhere because I been going round in circles for ages with all this already :roll: I think the code changes are just changing the internal layout of the PLD to the point where it just alters the ground currents with tiny little bit which can improve or make things worse.

Take for example, if I run ROM at 8mhz, its massively unstable. It's better running it faster, which is odd. But if I do a similar thing of running it at 50mhz, and adding waitstates, slowing it down that way, it still "works fine". Granted theres more clock switching going on in one method which could be why not clock switching works better because there's less logic changing internally..

I think it must be just similar that disabling the cache with CDIS just changes GND bounce just enough to work. I think it's just borderline somehow. But this is why I re-spun the board and slapped more copper on the power rails and more capacitors.. Maybe I should just forget these earlier boards under start with the rev 5.62+ boards and build up some of those to see if a small batch of boards all works fine.. Rather time consuming experiment but obviously doing what I'm doing now just isn't going to work.

The rev5 board is most sulky. The 5.50 "mostly works". REV 5.62 I've been swapping about randomly and is not tripped up once yet . That one basically has doubled up power planes and more capacitors.

[exxos]

Built up a second rev 5.62 board...

Booted right up as well.

IMG_4624.JPG

IMG_4625.JPG

IMG_4626.JPG

The main difference with rev 5.62 is I doubled up on power rails where I could. It had a LOT more decoupling capacitors. And they also moved to a different voltage regulator and used two of them separately. One to power the PLD and the other to power the SDRAM.

On one of the earlier boards I did plaster a lot of capacitors on the bottom everywhere and while it did change sustainability certainly was not a "cure all". The doubling up of the power rails should not really have been a huge factor either. Mostly there is just a slab of extra copper under the PLD and CPU. So it's not exactly "significant".

The only avenue I have not explored yet, is trying to find a different regulator in the same package, so I can do a simple swap on the earlier rev5 boards to see if it makes any difference or not.

[coonsgm]

Well, you're making me feel a bit better that I really wasn't crazy.....although I'm bummed I didn't catch the ROM socket needing resoldering.

and yes, I didn't clean the board before sending....lots of flux....so sorry.

[exxos]

Well, you're making me feel a bit better that I really wasn't crazy.....although I'm bummed I didn't catch the ROM socket needing resoldering.

That's the odd thing, it looked fine and I looked at it several times. Eventually I just went around resoldering everything and after the ROM socket resolder, it sprung to life ! I'm starting to wonder about looking again for a through-hole socket.. I think I did not use them last time as they were slightly too wide and too tall.. I don't think I could find anything small enough other than the SMT sockets.

I don't know if you've been following all the posts but it seems the earlier boards (which is most of them I think) are starting to have weird issues again. Going to look for some better regulators because the faults make no sense whatsoever.

and yes, I didn't clean the board before sending....lots of flux....so sorry.

ahh it was me who gunked it up mostly :)

[coonsgm]

My guess is that my soldering of the socket was a bit borderline and with all the back and forth with ROM IC's I broke a connection.

Greatly appreciate all your work on this!!

[exxos]

The samples I tried to order don't seem to be making any progress :roll: The status has just been stuck on "open" for the past few days. As to what that actually means, anyone's guess.. But not waiting around on the off chance they might get sent out any more.

Ordering from farnell has been a nightmare. Pages constantly failed to load and checking out was literally impossible. I would enter the credit card information, many would say enter CCV.. So I would do that, and then you would say enter credit card number, so I would do that, and it was a enter CCV again.. so I would fill both out.. Then it would say enter credit card number again .. I even tried adding in my credit card on a totally separate page and that was secretly as broken :pullhair: in the end I switched over to Firefox and things seem to actually work better..

Generally I gave up ordering from Farnell some years ago. Prices were just too high, stuff was always out of stock, stuff which said was in stock then mysteriously went out of stock as soon as I the order and will be put on back order for months. The website hasn't worked properly for the past decade or more now. I've complained so much to their sales rep they don't ring me up any more :lol: it was the only place I could find that actually listed them in the UK and had them in stock..

I managed ordered the regulators from Farnell in the end and they should be here tomorrow..

[Badwolf]

Generally I gave up ordering from Farnell some years ago. Prices were just too high

I used to order a lot from Farnell when their shipping was cheap. Haven't for a couple of years, but I never had issues with the website.

Are you running an adblocker that could be interfering?

BW

[exxos]

Are you running an adblocker that could be interfering?

Good point that had slipped my mind.. I don't think Firefox has the advert blockers,so certainly a possibility. Never had any troubles on any other sites though.

I do run a filtered IP address aswell, but as it worked in Firefox, that is unlikely to be the problem.