Repairing ZX Spectrum issue two

User avatar
1024MAK
Posts: 5151
Joined: Mon Sep 26, 2011 10:56 am
Location: Looking forward to summer in Somerset, UK...

Re: Repairing ZX Spectrum issue two

Post by 1024MAK »

Testing the DRAM chips is a good idea. Don’t forget, DRAM needs either a refresh cycle, or a write or read at regular intervals or it forgets. If your tester can do 4164 chips, you can test the 32K bit ‘upper’ DRAM. I recommend building your tester to test this first. That way you can get your tester working before having to deal with the +12V and -5V supplies and the 4116 chips.
Pygmalion wrote: Thu Sep 21, 2023 1:46 am I did not quite understand the part of the plan with multiplexers. I simply remove them or replace them? What should I see after removing them? I do have a simple desoldering gun, so this is not completely far fetched.
If you have or can get new chips, remove the existing ones, fit sockets (if there is room under that heatsink/cooler) and fit the new chips.
Pygmalion wrote: Thu Sep 21, 2023 1:46 amSince I am already ordering from AliExpress, should I also order some solid electrolytic capacitors like these

https://vi.aliexpress.com/item/1005002973596027.html

I assume that solid means polymer, therefore solving aging problem. They are suspiciously cheap, but with excellent reviews.

Do you have any indication that caps could be be culprit for at least some of my problems?
Okay, a few, okay, a lot of words on capacitors and ZX Spectrum computers and the recapping nonsense that has spread like a plague across the echo sounding interweb.

First off, all the ZX Spectrum (and for that matter, ZX80, ZX81, TS1000, QL etc.) boards were built long before the so called capacitor plague (the capacitor plague was due to capacitors being manufactured incorrectly). And despite the bull and propaganda from some people and organisations, Sinclair did not use substandard components. The original electrolytic capacitors were normal quality parts used throughout the electronics industry at the time. These capacitors do not normally leak. The ZX Spectrum issue one, two, 3 and 3B (plus some later) boards were built in the U.K. using parts available in the U.K.

Within the ZX Spectrum 16K, 48K and plus (+) machines, I class some of the electrolytic capacitors as critical components. This is because either they are subjected to heat (those under or near components that get hot, such as under the heatsink fitted to the 7805) or because some of them (those in the DC/DC converter part of the power supply circuitry), can cause other parts to fail, if one of these capacitors degrades or fails.

Replacement of all the other electrolytic capacitors is optional, they can be left until there are symptoms that indicate that they are life expired. By life expired, I mean that the capacitance has fallen or the ESR value has increased. Typically this is because the electrolyte has ‘dried up’.

For ZX Spectrum 16K/48K/+ models, capacitors that I consider critical are:

For Issue 2, 3 and 3B boards:
Capacitors C34, C44, C45, C46, C47 (where fitted), C50, C74 and C79 (where fitted)

It’s also worthwhile thinking about carrying out the power circuitry (DC/DC converter) modification (see here). Although this is more involved on issue two boards.

It is up to you what type of capacitor you fit, but normal quality electrolytic capacitors are fine. They should last typically twenty years or more (assuming you buy normal or good quality parts from a reputable supplier). Especially if the 7805 is replaced with a DC/DC converter / switching regulator to reduce the heat in the case.

With the fault that you have, assuming the +5V, +12V and -5V supply rails are reasonably free of ripple and noise (best tested using an oscilloscope or an expensive multimeter that has a wide bandwidth on the AC mV ranges, but if your voltages are within specification, this is a good indicator that the power supply circuitry is okay) it’s unlikely to be an electrolytic capacitor problem.

It’s fairly common for people on finding that their ZX Spectrum does not work, to replace all the electrolytic capacitors, only to find it did not make any difference.

Having said all that (sorry if it’s seen as a rant), as long as good capacitors are used, replacing all the electrolytic capacitors, if done correctly, doesn’t do any harm. Some say that it also improves the quality of the video output. But back when the Speccy was designed, people used 12” and 14” CRT TVs, so the quality of the video was not so critical compared to today when viewing on large high definition telly… And there are other things that can be done to significantly improve the video quality.

Mark
ZX81 Variations
ZX81 Chip Pin-outs
ZX81 Video Transistor Buffer Amp

:!: Standby alert :!:
There are four lights!
Step up to red alert. Sir, are you absolutely sure? It does mean changing the bulb :!:
Looking forward to summer later in the year.
User avatar
Pygmalion
Posts: 47
Joined: Mon Sep 18, 2023 6:03 pm
Location: European Union

Re: Repairing ZX Spectrum issue two

Post by Pygmalion »

OK, good news. Using my ad-hoc DRAM tester and the program that I found here https://forum.defence-force.org/viewtopic.php?t=1699, I successfully tested all DRAM chips.

Among lower memory chips, one is dead (started failing at $1), the rest tested OK.

Among higher memory chips, four were faulty (started failing at $0 / lots of failures starting at $AC__ / lots of failures starting at $9CCE / three faulty addresses $AFAF, $E7A7, $E7E7). The first one is obviously dead, the last one probably OK for use because ZX Spectrum only uses lower half of the memory, the middle two probably OK too? Please comment on that.

Now to your suggestions. First of all thanks for all the help you provided so far.

I will try to repair one and only one ZX Spectrum in my life. So buying or making complex testers does not make sense. Especially if they are sold outside of European Union and AliExpress, making shipping and customs super complicated and expensive. I used to buy a lot of model railroading and electronic stuff from UK (wider choice and prices than Germany), but since Brexit just not worth the effort. On the other hand, AliExpress solved problems of both shipping and customs pretty neatly.

Therefore, after testing DRAMs, I think the best course is to order 4164 DRAMs and multiplexers (is there any difference between choosing 74LS157 or 74LS257?) from AliExpress - I am sure there will be enough working to complete the job. Then replace all faulty DRAMs and if this does not help, replace multiplexers. If at this point we can positively know that buying one expensive chip (ULA or some other) is probably the final solution, I might try that too.

Only if Spectrum gets repaired, I will think about buying switching regulator, do recapping with solid (polymer) capacitors (they are supposed to work "forever" like their cheremic brethren). And maybe just maybe also buying ULA to remove the last heating problem within the box.

Does that sound reasonable?
User avatar
1024MAK
Posts: 5151
Joined: Mon Sep 26, 2011 10:56 am
Location: Looking forward to summer in Somerset, UK...

Re: Repairing ZX Spectrum issue two

Post by 1024MAK »

Pygmalion wrote: Thu Sep 21, 2023 6:46 pm Among higher memory chips, four were faulty (started failing at $0 / lots of failures starting at $AC__ / lots of failures starting at $9CCE / three faulty addresses $AFAF, $E7A7, $E7E7). The first one is obviously dead, the last one probably OK for use because ZX Spectrum only uses lower half of the memory, the middle two probably OK too? Please comment on that.
There are were two different types of 32K DRAM chips from Texas Instruments, each with a slightly different part number. One type had one half (let’s say higher addresses) of the chip as not fully functional. The other type had the other half (let’s say lower addresses) of the chip as not fully functional.

So, off the top of my head, I can’t really say which is which… But it is determined by the state of pin 9 (A7 / AR) and if the multiplexed address is a row or a column.

Other DRAM manufacturers that also offered 32K DRAM chips had similar arrangements. But laid out the addressing differently. Hence why issue two boards needed a modification if 32K DRAM from someone other than Texas Instruments was used.

This led to issue 3 and later boards having more link options as standard for the DRAM selection.

A ZX Spectrum should have all the ‘upper’ DRAM using the exact same part numbers if using 32K DRAM chips. You can however replace any or all with 4164 64k chips.
Pygmalion wrote: Thu Sep 21, 2023 6:46 pm So buying or making complex testers does not make sense. Especially if they are sold outside of European Union and AliExpress, making shipping and customs super complicated and expensive. I used to buy a lot of model railroading and electronic stuff from UK (wider choice and prices than Germany), but since Brexit just not worth the effort.
I fully understand. That’s why I try to think of as many tests and checks that can be done without expensive test gear/equipment. And yes, brexit is an absolute pain the rear, both for people outside of the U.K. and us within it. Handling fees are unregulated and excessive. I’m not in favour of VAT or customs charges anyway, but that’s off topic.
Pygmalion wrote: Thu Sep 21, 2023 6:46 pmTherefore, after testing DRAMs, I think the best course is to order 4164 DRAMs and multiplexers (is there any difference between choosing 74LS157 or 74LS257?)
Most definitely get some ‘new’ DRAM. I put ‘new’ in quotes because (1) none of it is very new, having been out of production for decades, and (2) some sellers are selling used parts as ‘new’.

There are other part numbers for equivalent DRAM. But, check the access speed, the DRAM used must be fast enough, especially for the ‘lower’ DRAM (150ns or faster). Use a generic search rather than just 4116 or 4164. Then check against the datasheet or against an equivalent list. The ‘upper’ DRAM can be either 128 or 256 cycle refresh type.

There is a difference between 74LS157 or 74LS257, but both work in a ZX Spectrum, the output enable/tri-state pin is tied low, so the differences are irrelevant to the Speccy.
Pygmalion wrote: Thu Sep 21, 2023 6:46 pmOnly if Spectrum gets repaired, I will think about buying switching regulator, do recapping with solid (polymer) capacitors (they are supposed to work "forever" like their cheremic brethren). And maybe just maybe also buying ULA to remove the last heating problem within the box.

Does that sound reasonable?
Yes. I only offer advice and suggestions (unless I get on my soapbox…!)

Mark
ZX81 Variations
ZX81 Chip Pin-outs
ZX81 Video Transistor Buffer Amp

:!: Standby alert :!:
There are four lights!
Step up to red alert. Sir, are you absolutely sure? It does mean changing the bulb :!:
Looking forward to summer later in the year.
User avatar
Pygmalion
Posts: 47
Joined: Mon Sep 18, 2023 6:03 pm
Location: European Union

Re: Repairing ZX Spectrum issue two

Post by Pygmalion »

1024MAK wrote: Fri Sep 22, 2023 5:06 pm
Pygmalion wrote: Thu Sep 21, 2023 6:46 pm Among higher memory chips, four were faulty (started failing at $0 / lots of failures starting at $AC__ / lots of failures starting at $9CCE / three faulty addresses $AFAF, $E7A7, $E7E7). The first one is obviously dead, the last one probably OK for use because ZX Spectrum only uses lower half of the memory, the middle two probably OK too? Please comment on that.
There are were two different types of 32K DRAM chips from Texas Instruments, each with a slightly different part number. One type had one half (let’s say higher addresses) of the chip as not fully functional. The other type had the other half (let’s say lower addresses) of the chip as not fully functional.

So, off the top of my head, I can’t really say which is which… But it is determined by the state of pin 9 (A7 / AR) and if the multiplexed address is a row or a column.

Other DRAM manufacturers that also offered 32K DRAM chips had similar arrangements. But laid out the addressing differently. Hence why issue two boards needed a modification if 32K DRAM from someone other than Texas Instruments was used.

This led to issue 3 and later boards having more link options as standard for the DRAM selection.

A ZX Spectrum should have all the ‘upper’ DRAM using the exact same part numbers if using 32K DRAM chips. You can however replace any or all with 4164 64k chips.
Aaaah, after reading this http://www.breakintoprogram.co.uk/hardw ... rdware/ram, I finally get it. All my upper RAM chips are either TMS4532-15NL3 or TMS4532-20NL3, hence the lower half of the chip is useable. And since three of the chips have errors in the upper half, they are actually fine. So I only have to replace one chip in lower and one chip in upper DRAM!
1024MAK wrote: Fri Sep 22, 2023 5:06 pm There are other part numbers for equivalent DRAM. But, check the access speed, the DRAM used must be fast enough, especially for the ‘lower’ DRAM (150ns or faster). Use a generic search rather than just 4116 or 4164. Then check against the datasheet or against an equivalent list. The ‘upper’ DRAM can be either 128 or 256 cycle refresh type.
Well, I am not sure if I am able to make a good generic search. So I simply ordered 10 ‘new’ 4164s from AE. The faulty 4116 will be replaced by 4164 with pin 1 cut and pin 8 connected to pin 9. 4116s are just too expensive.
User avatar
Pygmalion
Posts: 47
Joined: Mon Sep 18, 2023 6:03 pm
Location: European Union

Re: Repairing ZX Spectrum issue two

Post by Pygmalion »

Obviously I cannot be idle for 2-3 weeks while my chips are traveling from China. So what I did is I used one 32kB chip from upper memory and put it into lower memory to replace a 4116 that was proven faulty:
20230925_092346a.jpg
Most of my upper memory chips are TMS4532-20NL3, which means they are too slow, but two are TMS4532-15NL3. NL3 means that the lower part of DRAM is OK and if you connect pins 8 and 9 then you use upper part of DRAM... but this particular chip tested OK in both upper and lower part. So I thought, why not give it a try...

Unfortunately, ZX Spectrum is still not working, but there is an improvement. There are thick black-white stripes at startup and finally I get stable picture with white lines:

https://youtu.be/LtHGNAw3cDM

As far as my DRAM tester is concerned, all lower memory chips should be OK now, so this is pretty disheartening. Also, pretty worryingly, reset button on C27 is no longer working.

Anyway, I did another test. I connected my HANTEK USB oscilloscope to all power lines. I did not see any ripple at -5V and 5V lines, but there is 2V pp ripple at 12V line (measured on C44):
12Vline_3.png
Is it possible that this is the culprit? Should I replace a few caps and which ones?
User avatar
Pygmalion
Posts: 47
Joined: Mon Sep 18, 2023 6:03 pm
Location: European Union

Re: Repairing ZX Spectrum issue two

Post by Pygmalion »

OK after I rearranged the DRAMs, the reset works again, but the black square is no longer neat and different every single time, while the image is no longer stable because it is interrupted by white stripes...

https://youtu.be/zTLHSmwAOlA

So when I rotate lower memory chips, the result constantly changes, but there is one and only thing that is consistently present, and that is this pattern, which is on the following picture easily visible by white boxes:
20230926_080848.jpg
Can someone please just tell me honestly if it's still worth it or if the computer is already beyond repair given my extremely limited expertise and the very poor condition of the computer.
User avatar
Pygmalion
Posts: 47
Joined: Mon Sep 18, 2023 6:03 pm
Location: European Union

Re: Repairing ZX Spectrum issue two

Post by Pygmalion »

Can someone at least comment on 50Hz 1.7V pp ripple in 12V line reported above? This should be a no-brainer for experts.

As an amateur, I think this is too much and that replacing C44 and C45 would probably be a fix. Anyone?
redgatemodels
Posts: 172
Joined: Tue Sep 05, 2023 11:56 am

Re: Repairing ZX Spectrum issue two

Post by redgatemodels »

Think folks are busy elsewhere, I've posted a couple of questions on my topic with no response yet. Time is given free etc, so if real life is pressing then our little issues have to take a back seat for a bit I guess ;)
User avatar
Pygmalion
Posts: 47
Joined: Mon Sep 18, 2023 6:03 pm
Location: European Union

Re: Repairing ZX Spectrum issue two

Post by Pygmalion »

redgatemodels wrote: Fri Sep 29, 2023 11:48 am Think folks are busy elsewhere, I've posted a couple of questions on my topic with no response yet. Time is given free etc, so if real life is pressing then our little issues have to take a back seat for a bit I guess ;)
I understand. We should be grateful for what we get. I just thought nobody has any idea.

You are using plural? Are there many? I see only 1024MAK responding.

I gave up on the other forum, after all, only 1024MAK responded there too. But maybe there are other people there.
User avatar
Paul
Posts: 1536
Joined: Thu May 27, 2010 8:15 am
Location: Germanys west end

Re: Repairing ZX Spectrum issue two

Post by Paul »

Many on here are mainly interested in ZX81 and it's relatives ;)
In theory, there is no difference between theory and practice. But, in practice, there is.
Post Reply