16K RAM Pack Schematic / Chip Layout

[TMAOne]

Success! I think. But I don't understand it.

When I explain, you guys are going to think I'm bullshiting you, but I'm not.

To review, I had a pack with 1 bad bit. Its measured voltages were low, but it seemed to run fine except for that one stuck bit. Then I started replacing the electrolytic caps, and it no longer booted at all.

I now know why THAT was. After replacing all the electrolytics, I next swapped the transistor with a donor unit, and prepared to take the voltage measurements again. When I bent the ribbon bridge between the two halves to get better access to the RAM pins, it fell apart.

Makes sense. The darn bridge is stiff steel wire and I'd been stressing it. 20-20 hindsight--of course it broke eventually. No wonder the pack stopped working altogether.

BIG exercise in soldering,... I cut a suitable piece of ribbon from an old IDE controller cable, and de-soldered and re-soldered 44 connections, going slightly cross-eyed in the process.

It boots again. Yay!

Measured voltages. Still bad. Pin 1 is -0.27 instead of -5. Not even close. Pin 8 is +9.40 intead of 12. Pin 0 is 4.8--a good enough 5.

Then I did a silly thing. I ran my test program that hammers (POKEs) the offending byte, alternating between 0 and 255, and PEEKs the values back.

The stuck bit is working now. Huh?

I'm running the RAMTEST program from the "The Explorer's Guide to the ZX81" (Mike Lord) even as we speak. It is a minimalist program so as to test as much memory as possible without committing suicide. So far, no errors.

What's the deal here? When I tested it yesterday, after putting the original RAM chip back in, it functioned with only the one bad bit as before, so I can rule out that heating up the chip somehow "cured" it.

Is it possible that I had a bad connection in the ribbon bridge that only affected ONE bit of ONE byte out of 16K bytes? That hardly seems likely.

I'm watching for black cats now because I think someone is screwing with The Matrix,...

Ian

[1024MAK]

I would say that the power supply circuit that provides the -5V and +12V is not working. The +5V line comes from the 7805 voltage regulator in the Zeddy.

16k RAM pack ZX81 & Timex1016_Schematics (Power supply only).png (23.28 KiB) Viewed 6862 times

You are most likely getting +9.4V on the +12V rail due to the 1N4002 diode (top most diode in the schematic) and or the BA221 diode. If you test (using a meter) the voltage on the +9V rail (anode of the 1N4002 diode) it is likely to be about 10.1V to 10.5V.

The following assumes (1) that the version you have is the same as the schematic and (2) that all the components you have changed are new or known good (I recommend new) and are correctly fitted (polarity) and the soldering is good (no "dry" joints or broken tracks/traces).

If you have a diode test function on your meter, check each of the diodes in the circuit (both ways round). For most of them you will need to desolder and lift one lead out wire so that the rest of the components do not affect the results. Or if you have spares, renew them anyway. It is still worthwhile testing the removed/replaced components.

Test the resistors as well (although none of these are likely to have failed).

If none of that helps, then it is a case of putting a new ZTX750 (or a ZTX751) transistor in and maybe new 5.1V and 12V zener diodes.

If you know how to use your 'scope, when the circuit is working, you should get an ac waveform (spikes) at the anode of the BA221 diode. The frequency is likely to in the kHz range, it depends on actual component values and the actual load current.

Mark

[TMAOne]

Thank-you Mark. That's just the kind of insightful guidance I need, and I appreciate the time and thought that goes into laying that out for me. I will follow your advice next time I work on this (soon).

Strangely, the problem came back after I boxed the RAM pack back up in its case and put it back on the Zeddy. Immediately, as if heat is NOT a factor. So the sticky bit is not "working" after all. I DO get 255's now, whereas I never did before. But my newly enhanced BYTETEST program is getting between roughly 20 and 70% success, with the boards bare, and close to 0% with it in the case.

No matter. It suggests the chip is healthy enough to work if it was properly fed, so the power supply problem is where I'll concentrate.

In my travels I did socket the "bad" chip, and use the location to test my stock of 4116's, although the testing is tainted by the power issue. All but 1 of 16 failed miserably. The one trooper would boot, but the ZX91 would set RAMTOP to a random value, once about 8K, and once close to 64K! Dreamer. But it shows me that if I get the power issue sorted out I can use this pack as a "4116 tester".

I really appreciate you guys helping me on this. It's great to work on my ZX81 again.

Ian

[TMAOne]

I've put in a ton of soldering, pulling parts and replacing them with what I have on hand, or parts from other non-working or semi-working RAM packs, in a effort to determine what components are giving me the trouble. Feeling pretty frustrated right now.


Pack #1:

The pack I've mainly been working on, ('ll call this Pack 1), has had almost everything in the power supply area changed out. Nothing bad was ever identified, except for a questionable power transistor. So I put one in place from another pack (#2, that would not boot but who's voltages were GOOD at the time), thinking I had it solved.

This pack 1, which previously ran close to perfect (one bad bit even with crap voltages), has now stopped working altogether. So I put the original transistor back in. Still no boot. I've fried something.

I have tested the one RAM chip that is socketed (for my bad bit experiments) in another pack. The chip is bad. The other 7 probably are too. They are still soldered in.

Voltages are: pin 1 is +.66 (instead of -5.0), pin 8 is +4.35 (instead of +12), and pin 9 is +3.84 (instead of +5.0)

I will only undertake to replace RAM if I get those sorted out.


Pack #2;

Pack 2 is a very old one I'd forgotten I had. I must have pulled all the RAM chips and socketed them at some point in the past trying ignorantly to fix it. There are currently no RAM chips in the sockets, although I have chips available.

Testing with no RAM, the voltages at the pins are wrong on this one too.

Voltages are: pin 1 is -.72 (instead of -5.0), pin 8 is +8.27 (instead of +12), and pin 9 is +8.34 (instead of +5.0)

I can quickly populate it with proven 4116's if I get the voltages sorted out.


Pack #3:

Pack number 3 represents my sole ability to run programs. The voltaages are good and the RAM works, but the video to be distorted. Not sure why a RAM pack does this.

I will include a picture, but it's hard to see. It's like horizontal streaks that take the text for a ride to the right or left a bit. Not pretty, but legible. I should mention that this ZX81 has the one transistor video monitor mod, and the modulator is not being used.

At the risk of breaking this one semi-good pack, I de-soldered one RAM chip and socketed it, making this pack into a "4116 tester". To my surprise the 16 or so loose chips I have all work! (8 of those belong in pack #2.)


Status:

I have ordered some parts from eBay, since I do not have a good electronics supplier in my city (which really needs one), and from my experience they will take about a month to arrive from China.

I bought 2SB892 transistors which can replace ZTX750 (although the pinouts are different and I'll have to watch that), assorted diodes including 1N4148 and 1N4002, and assorted Zeners including 12V, as well as 6.8V and 7.5V which I hope are close enough to 7.1V

I'm sick of soldering in parts I rescued from some scrap-outed carcus of dead electronics, not knowing if they're good or a close enough match to what I need. Moving parts about from pack to pack is wearing things out and getting me nowhere. I'm afraid I've lost track of which parts came from the power supply that used to be good.

So I'm taking a month or so "off" from this project while my supply boat comes in.

If any of you electronics engineers can tell something from those voltages for the first two packs, by all means make suggestions and I'll start my "rebuilds" there.

Similarly, if anyone has any idea what could be causing video interference from a RAM pack (noisy transistor?) by all means share.

Cheers,
Burnt Fingers Ian.

[1024MAK]

I bought 2SB892 transistors which can replace ZTX750 (although the pinouts are different and I'll have to watch that), assorted diodes including 1N4148 and 1N4002, and assorted Zeners including 12V, as well as 6.8V and 7.5V which I hope are close enough to 7.1V

Err, for the -5V line you need 5.1V zener diodes...
Not sure where you want to use 7.1V zener diodes

Mark

PS short of time at the moment, will have a proper read through and reply later

[TMAOne]

Oops. Dunno where I came up with 7.1--I was getting tired. But 5.1V Zeners are in the "assortment" I bought so I'm good.

[RetroTechie]

Pack 2 is a very old one I'd forgotten I had. I must have pulled all the RAM chips and socketed them at some point in the past trying ignorantly to fix it. There are currently no RAM chips in the sockets, although I have chips available.

Testing with no RAM, the voltages at the pins are wrong on this one too.

Voltages are: pin 1 is -.72 (instead of -5.0), pin 8 is +8.27 (instead of +12), and pin 9 is +8.34 (instead of +5.0)

Either you're measuring wrong, or there's a short circuit on that board. Regardless of what components are okay or not.

4116 DRAM pins 9 connect to the same +5V that the ZX81 uses internally. The only way to force that to +8.34V is to pump current into it (using a higher supply voltage), with some power behind it. That requires a circuit which connects to +12V and/or unregulated DC (+9V nominal).

A failed 4116 can't do that when you only have empty sockets in those locations.
A failed logic IC can't do that since they only have +5V and ground as supply, input signals are also in that range & don't have power behind them.
Power supply decoupling capacitors can't do it as they only sit between a supply and ground. Failure as open circuit would have no visible effect, failure as short circuit would pull that supply towards ground.
The remaining circuitry (voltage conversion) connects to +9V input supply, +12V and -5V output supplies for the DRAMs, but NOT to +5V supply. So it can't lift +5V above it's usual range either.

The only other possible explanation for above reading on the +5V would be a failed +5V regulator inside the ZX81. Which then likely wouldn't work at all regardless if/what RAM pack is inserted.

This RAM pack seems like the best candidate to fix I'd say. Figure out the above, and you have a RAM pack with good voltages in which you can test up to 8 pieces 4116 at a time. I suggest you use some resistors to put a minimum load on -5V and +12V (say, a mA or so each). When you're measuring the correct voltages, remove resistors, put in 4116 chips, then check voltages again.

Finally, IMPORTANT: Stop running software tests etc on a RAM pack where the voltages are off, even if it seems to be the best working RAM pack you have. If there's 4116 chips in the RAM pack & voltages aren't OK, power off at once. Only power on for a few seconds to to measure voltages after you've replaced suspect parts. Run the RAM pack longer with voltages not OK, and chances are you'll kill more 4116 chips that were OK before (and you wouldn't even know it, if the RAM pack doesn't work). Which would be stupid since it's well known that 4116's are killed that way.

Save software testing for when you've got 4116's powered with the correct voltages.

[gozzo]

Get the voltages sorted before replacing any ramchips! - I've read somewhere that the -5v line must be applied first or at the same time as the others and removed last - otherwise the chip can be instantly killed!!! So this one is the important one!! Or I think it may be possible to use 4164's - maybe easier to get and more reliable? - if you change all the chips in one go and remove the troublesome voltage generator circuitry - you will need to change a couple of power lines round to get the +5v on the correct pins of the 4164, and short the unrequired a7 line to ground, apart from this they appear to be mostly pin compatible..??? Never tried it myself, yet, but I don't see why it shouldn't work..?? Other solution is to remove all existing chips and rig up a static ram replacement on a bit of stripboard stuck on top of the original one - not original, I admit, but will probably give less trouble!!

[TMAOne]

Thank-you RetroTechie and gozzo.

No more running Pack #1 that I thought was almost good--got it! Can't anyway, the RAM has fried. I feel like it deserves a medal for working so long, with perfectly fine video and only 1 bit failing to perform write/reads. It didn't even occur to me to check supply voltages until you guys suggested it. I'll pick up the project on pack #1 again when I have more sockets and there's a good movie on TV to listen to while I unsolder and labouriously pry out the other 7 valiant chips, then work on voltage again. For posterity and anyone reading this forum, the AMD AM9016EPC version of 4116 is incredibly tolerant of bad voltages, and when it dies it should be buried with full military honours. 21 gun salute at dealer's choice.

When voltages good, repopulate with RAM.

Pack #2 is shorted somewhere--got it again. Insert brick, hit head. Analysis makes perfect sense. The +9 is getting in there somewhere. This I can look at while I'm waiting for parts.

Again, when voltages good, repopulate with RAM. I actually do have 16 good chips.

Update on Pack #3. It works fine (no video distortion) when connected to a different ZX81. (I have one a ZX81 with an internal ram expansion; I fitted the RAM pack to it anyway to test something else entirely.) To my surprise--good video. The first ZX81--the one I've been testing the RAM packs with, has always had bad video. I originally suspected the modulator, and subsequently substituted a video out modification. Passable, but low contrast with a slow wavy ripple drifting up the page. I the suspected voltage on the Zeddy so changed both the wall-wart to a known good 1A one, and swapped out the 7805. No difference. I now think it is a faulty ULA. Putting that ULA in another mother produces bad video too. As long as I use the one-transistor video out circuit, it is passable, but somehow fitting a RAM pack (with a working power section!), exasperates the problem.

Since ULAs are hard to come by I will just live with the video. Somehow, pack #1 must have gotten away with decent video because of its internal voltage issues failing to aggravate the ULA to the extent that Pack #3 does. Bit of a bummer. Always on the lookout for Zeddy/Timex and working ULAs though.

gozzo, I know there are better RAM solutions than the original packs. This for me is an exercise in learning, romanticism, and downright stubbornness. Twice a pack failed and I didn't know how to fix it. Twice I searched for and bought another one, and then it failed too. Murphy shall not prevail! I am determined to have 3 working packs, if somewhat burnt by prolific soldering and having non-original IC sockets in two of them.

Thanks again for the guidance,
Ian

[gozzo]

All these problems you are having with rampacks, and I have about eight or so of the things that I have got hold of through ebay over the past few years, in a box in my barn waiting for when I get the time to test them - I wonder how many will actually WORK....!!