RAM keeps failing

[mcarlson_sb]

EDIT: Originally I thought I had a display problem.
It turns out my RAM keeps failing.
No display Is just how it shows itself.

ORIGINAL POST:
I just got some new toys to play with - and I can't because my display says "No Signal" - again.

Symptoms:
1) The signal looks terrible (see attached pic)
2) Sometimes I get no signal at all on ULA pin 16 - it just stays flat when I plug in.

Things I've tried:
I've checked that the Z80 is responding
I've checked that the clock signal looks good at Z80 pin 6
I've checked that the clock signal looks good at ULA pin 14

Swapped out ULA for known good.
Swapped out Z80 for known good.
Swapped out ROM for another - not known good, but had the exact same issue so unlikely to be the problem

I just swapped out RAM a few weeks ago.
So, while that might be it, if it is then there is probably some deeper problem.

Even so, I used a known good 16k external RAM pack and it didn't help

It worked less than a week ago. I tried to turn it on today and no dice.
It's been hot here in the bay area (over 100 F) and there has been ash in the air from all the fires, so I took it out of the case and inspected but didn't see anything out of the ordinary.

From the scope it looks like I should at *least* be getting a black/gray/static screen - but instead I get "No signal" on two different displays.
The large screen has always been very tolerant of the zx81 signals - but I'm getting no signal there also.

Any ideas on what to try next?

Image is signal from ULA pin 16

[mcarlson_sb]

Crap - it's memory again
I removed my 16k internal upgrade and put back in the stock 2k
Booted fine

What the fresh hell is blowing my memory?

[mcarlson_sb]

For reference I'm now on my 3rd 62256
I still have the two previous to test (some day - when I figure out how) with the hope of gaining some insight into what about them failed.
And one more from the batch I bought

I guess it's possible they were just a bad batch.
But I'm afraid there is something else going on.

[mcarlson_sb]

So, I replaced the RAM a few days ago and I'm now having the same problem again.

Clearly there is an underlying issue.
But I'm stumped as to how to find it.

Possibility -
I've been doing a lot with edge connector hardware builds.
RAM failing seems to correlate with me plugging in homemade boards. But that might just be because I do a lot of it.

I spent a day just playing after I changed the RAM four days ago and it was fine.

I tried to plug in and check my Talk-Back recreation this evening and I'm back to failure mode. (I *should* have tested the computer without the board first - but I didn't )

Any ideas would be greatly appreciated.
I've only got one 62256 left

Cheers,
Matthew

[1024MAK]

So here are my thoughts...

First, these SRAM chips are sensitive to static electricity.

So when handling any uncased expansion, try not to touch any conductors or components. With cased items, don’t touch the edge-connectors.

The PSUs used in the U.K. and Europe don’t have any connection to earth/ground. And most TVs also don’t have any connection to earth/ground. Is this the same where you live?

This may then be a problem if you are connecting expansions, as static can build up between different parts of the system. The expansion may be at a different static voltage to that of the computer and TV...

A second problem arrises with not everything being earthed/grounded, or if some items that are not normally connected are earthed/grounded but other items are not earthed/grounded.

The interference suppression capacitors and the stray capacitance in switched mode PSUs (as used in modern TVs and other consumer gear) means that a weak mains voltage can appear on the low voltage parts. In the U.K. this voltage is often around half mains voltage (about 100V to 140V). To a human, it can “wake you up”, be a slight sensation or not even be noticed. But to a CMOS chip on a 5V digital input pin, it can be rather unhealthy...

As I have mentioned before, in a normal production ZX81 or TS1000, due to partial address decoding, even though there is only 8k bytes of ROM and only 1k bytes or 2k bytes of RAM, the Z80 sees the whole 64k byte address range filled with memory.

So if an expansion is added, that does not include circuitry to disable the internal ROM chip, and / or the internal RAM chip(s) in the machine, when the Z80 reads from any address provided by the expansion, the relevant chip in the expansion will attempt to put data on the data bus, but either the internal ROM, or the internal RAM chip will also try to do the same. This is known as a bus clash or bus collision. Although very undesirable, most 1980s chips survive, as the time that their outputs are engaged fighting the other chip/device is short (although the processor may not be able to receive the wanted data). But it is possible that some chips may be damaged by this.

Mark

[Paul]

I know it's weird but are you shure your RAM is 5V? From the readings on it they are. But did you buy your RAM from a reliable source?
It may have been relabeled and it only allows 3.6V?
That would explain why it works ok for a while and then gives up.
Kind regards Paul

[siggi]

There might be a timing problem, which occurs using modern and fast ram chips. See there:
https://www.sinclairzxworld.com/viewtop ... 2620#p2620
"
The problem is the old design of the ZX80/81, where the internal RAM's output drivers are enabled together with the complete chip ( /CE conneted to /OE). That causes a crash at the data bus, when the Z80 wants to write data into the RAM, because the RAM is enabled immediately when the address is valid and then also drives the data bus. But the Z80 also drives the data bus! 1 clock cycle after the address is valid the Z80 enables the /WR signal to write the data into the ram. And that is the moment, where the RAM releases the data bus (disables its output drivers). But then the data from the Z80 may not have enough time to stabilize and nonsense might be written into RAM.
That design works with old and slow RAM chips, but not with modern and fast chips, which enable their output drivers much faster!

So I feeded the /RD signal to /OE of the RAM chip. And the problem was solved!
"

So try this: bend up the ram's /OE pin and connect it to Z80 /RD-signal (eg. at edge connector). Then this timing problem should be solved.

But I also think, that ESD (when connecting things to the edge connector) could be the problem!
How to avoid it: see
https://forum.tlienhard.com/phpBB3/view ... f=3&t=3361

HTH Siggi

[mcarlson_sb]

Thank you all.

ESD -
Hadn't thought of this. I remember in the 80's being very careful about it but have become complacent. I'll have to think about how to change my workspace to be safer in this regard.

CMOS -
I hadn't even realized these were CMOS ICs
I bought them when I was just getting back into electronics for a, breadboard computer (all TTL) as a way to learn. I've just been assuming they were TTL because they are TTL compatible. (here is where I need a face-palm emoji)

I replaced the 7508 with a switching regulator.
I didn't add any filter capacitors.
It's possible the switching noise is a constant hazard for this RAM

Timing -
The RAM is 55ns
What is the Z80 timing? How fast is too fast?
I need to dig into this a bit more to understand it.
Thank you for the links

Bus Clash -
This is interesting. I need to think through the ramifications (no pun intended)
I've been recreating old designs that don't take this into account.
Maybe I should build a back plane that does.
I'll have to think about how that might work.

One more thought comes to mind -
Because these are 32k RAM and I'm only using 16K
I'm hoping to try testing them and see if one might be usable if I switch out an address pin for the unused area.

I need to grab an arduino and put together a memory test rig.

Again, thank you all for the help.

Cheers,
Matthew

[1024MAK]

I replaced the 7508 with a switching regulator.
I didn't add any filter capacitors.
It's possible the switching noise is a constant hazard for this RAM

Not to kill it, no. Lack of filtering may cause unreliable operation.

Timing -
The RAM is 55ns
What is the Z80 timing? How fast is too fast?
I need to dig into this a bit more to understand it.

It’s not about clock speed as such, more about how fast the chip responds to the Z80 control signals.

Bus Clash -
This is interesting. I need to think through the ramifications (no pun intended)
I've been recreating old designs that don't take this into account.
Maybe I should build a back plane that does.
I'll have to think about how that might work.

It’s tricky doing it that way unless your backplane circuitry fully decodes the memory space. But keep in mind that the RAM MUST be echoed in high memory (enabled when A14=1, and A15=0 or A15=1) so that the screen display will continue working.

Mark

[mcarlson_sb]

Thanks Mark

With all of that, most look like they would show up as intermittent faults. This is consistent.

That points again to ESD

I guess I switch out the RAM again, add grounding to my work space, and try to be more careful.