Sinclair ZX80 / ZX81 / Z88 Forums

Hi everyone! I inherited a ZX 48k that was in really rough shape and was destined to be a donor board. The ZX is a 48k 6a model. Almost everything was busted on this motherboard...

Here's what I have done so far:

- Bad voltage regulator - replaced with TRACO
- 2 bad rams were bringing down the 12v and -5v lines (IC7 and IC11)
- Replaced D19 as it wasn't testing well
- Replaced TR4 and TR5 (TR4 was dead)
- Verified voltages are normal after these replacements
- Verified TR1, TR2, TR8, and TR9 are working normally (pulled from board and tested out of circuit)
- Did a full recap with axial caps
- Did the composite video mod
- Socketed and replaced LM1889 which improved the video quality
- Socketed and replaced all lower RAM
- Removed all upper RAM
- Pin 39 on the ULA had a very unstable 14Mhz. I replaced X1 and now I get proper, steady value. (pin 32 gives a good 3.5Mhz)
- Socketed and tested ROM with a known working ROM a friend lent me

Someone messed with the 22uF lower RAM decoupling capacitors on this board, and C57 was missing entirely. C6 and C7 appeared to have been desoldered at one point, but then not completely resoldered back into place. Replacing all of these is my next step. In the meantime, I soldered down the others, and dropped a radial 22uF in C57 for the time being. (but I think this cap would only affect upper memory?)

Currently, the ZX powers on, all voltages are good, and I get a white border, but the paper shows the screen garbage loop that is seen when the ZX has no memory installed at all. This is what I see with all the lower RAM populated:

https://www.youtube.com/watch?v=izcGFap ... WL&index=1

I had a loaned ByteDelight diag board for a bit... When the board is inserted, I get the same garbage on the screen, the border changes color as expected, and eventually the border displays that all the RAM is bad. I have a couple of dozen 4116 chips, but they're all untested. (I have a tester on the way, and it should be here in a couple of weeks.) However, whatever RAM I try, the diag board always says that all the RAM is bad, which I find hard to believe, and it makes me think something else is going on...

Here's a video of a run with the diag cart:

https://youtu.be/9h1h9DkiUmA

I have checked pins for address, data, etc. on the RAM, ULA, ROM and Z80 and everything *seems* fine, but I have to admit I'm a total noob with Spectrums, so there's probably a lot I'm missing here.

So... Some initial questions, for the kind souls who are still with me:

- Does anyone have any pointers on where to look next to debug this?
- I'm beginning to think the culprit might be the PCF1306P. Does anyone know how to test this chip?
- I'm really confused as to why I get garbage on the screen when I use the diag board. Don't diag boards override the ZX ROM/RAM and can still run regardless of the state of the memory?

Thanks in advance for the help!

astroturtle wrote: ↑Thu Mar 20, 2025 8:59 pm Someone messed with the 22uF lower RAM decoupling capacitors on this board, and C57 was missing entirely. C6 and C7 appeared to have been desoldered at one point, but then not completely resoldered back into place. Replacing all of these is my next step. In the meantime, I soldered down the others, and dropped a radial 22uF in C57 for the time being. (but I think this cap would only affect upper memory?)

Do you mean 22nF rather than 22µF (which are rather bigger!) for C6, C7 and C57?

C5, C6, C7 and/or C8 can be changed for larger values (e.g. 220nF) or two of them (e.g. C5 and C7 or C6 and C8) can be changed for 1µF multilayer ceramic types. Doing either of these two things may reduce or eliminate the "jail bars" vertical lines.

I know of no reason to mess with C57.

astroturtle wrote: ↑Thu Mar 20, 2025 8:59 pm - Does anyone have any pointers on where to look next to debug this?
- I'm beginning to think the culprit might be the PCF1306P. Does anyone know how to test this chip?
- I'm really confused as to why I get garbage on the screen when I use the diag board. Don't diag boards override the ZX ROM/RAM and can still run regardless of the state of the memory?

Alas, everything is pointing to IC27 (PCF1306P / ZX8401) being toast. This chip does the multiplexing for all the address lines from the Z80A to ALL the DRAM chips. With a dead IC27, the Z80A can't see any of the DRAM.

Is it possible to test it? Yes, but I have never tried and I don't know of anyone describing doing it. You would either have to build your own test circuit or reprogram a chip tester to test it.

The other possibility is either a partial fault with the ULA (IC1) or a problem with one or more connections between the the ULA (IC1) and the lower DRAM chips.

No, the diagnostic boards don't override any DRAM. They switch out the internal ROM chip yes. The Z80 CPU can run without any working RAM if you limit the program to a sub-set of the full instruction set. That's what the first part of the diagnostic board own ROM (EPROM/EEPROM/flash) chip does. It switches in it's own ROM and runs it's RAM test program. As this is written to run without any working RAM, as long as the Z80A and ULA are okay and there are no address, control or data line faults, normally it will run it's code and you will see the border colours change and / or hear the bleeper and / or see the diagnostic boards LEDs change state (depending on which diagnostic board you have).

If it doesn't find any working DRAM, that's it. If some working DRAM is found, it moves on to carry out further testing.

If IC27 is in a socket, carefully remove it. Now power up the computer and see if there is any difference.

Worse news is that the chip (PCF1306P / ZX8401) for IC27 is long obsolete and new working chips are extremely rare and rather expensive (if you actually find one). Slightly better news, is that modules may be available that can replace IC27.

Mark

Hi 1024MAK! Thanks for taking the time to reply.

Do you mean 22nF rather than 22µF (which are rather bigger!) for C6, C7 and C57?

Oops! Yes, that was a typo, I meant 22nF.

Alas, everything is pointing to IC27 (PCF1306P / ZX8401) being toast. This chip does the multiplexing for all the address lines from the Z80A to ALL the DRAM chips.

Oh! So always reporting bad memory makes sense if this chip is fried. My IC27 is soldered on to the motherboard, but pulling it and testing will be next on my list. I'm not sure what the rules are here for linking to commercial sites, but there is a retro store in the UK that sells a ZXMUX board (like the ULA replacements) that replaces the PCF1306P. They're not super expensive, (like £10) so if removing IC27 results in different errors, I'll order one and some 40pin DIP sockets and see if that works. There's not much left to replace!

Thanks again for the feedback! I'll report back (eventually) on how this goes.

So I desoldered the PCF1306P. I have never had a 40 pin chip come out so easily. This IC27 wanted *out*.

Now, instead of the no memory loop, I'm now just getting this on the screen:



So, progress, I guess? 1024MAK is in line with what you were expecting?

Thanks again!

Hi astroturtle

You should be able to upload JPEG pictures to this forum. If using a PC drag and drop an image file into the text edit box. Or regardless of what you are using, you can use the Attachments tab below the text edit box.

We have no problems with people using external image/picture/file sharing sites. As you have discovered, if the link you use leads to a web page, the forum can’t display your picture. It does work if the actual address of the image is used.

Alternatively you can use the thumbnail link (for forums) that some sharing sites provide.

We don’t have a problem with links to commercial sites or products as long as they are relevant to the subject. So replacement parts, add-ons, expansions, test gear, software etc. are all perfectly okay. And in fact it’s preferable to provide a short description and a link so that it makes it easier for others reading the forum.

Back to your machine. I’m assuming you have all the eight ‘lower’ DRAM chips fitted. There’s good news and not so good news.

I was expecting the image to be static.

The fact that the image is now static shows that the moving pattern was due to (at least some) signals from the Z80A making it to the ULA. It’s impossible to say if IC27 was working correctly, or is partly faulty. I still think it is partly faulty because the diagnostic board was unable to access the ‘upper’ DRAM, which is completely independent of the ULA.

However, it’s not the typical image that is normally shown when the Z80A is inactive (disconnected from the ULA, missing or faulty).

So it’s possible that there is a problem with the ULA or the connections between it and the ‘lower’ DRAM.

I do think it’s worthwhile getting a replacement module for IC27.

Mark

Hi Mark! Thanks for the tips. Sorry, I didn't notice the add files option at the bottom. I'll do that going forward.

I went ahead and got that ZXMUX: https://myretrostore.co.uk/product/zxmux/

I have bought from him before, and I got the chip tester I'm waiting on from him. Nice guy and he's got some neat stuff. Also sells on eBay, but I like to avoid that when I can.

Once I have the chip tester and the ZXMUX, it'll eliminate a lot of uncertainty. (can I really have 20+ bad chips?? that would be a bummer.)

I do have another ZX, but it's actually in much worse shape than this one. It's a series 2 with a big old lifted trace going to the Z80. That one has a ULA50102E version ULA, and I'm not certain it's in great shape. It has some really weird stuff going on with a 74LS00N soldered on top of it (pic attached). I'm hesitant to put it in the 6a... Can a bad ULA break other stuff on Spectrum? (the ULA in the 6a is a 6C001E-7)

If it's a problem with the connections between the ULA and the ‘lower’ DRAM, how would you recommend troubleshooting? I have tested continuity on all lower and upper ram. I have an oscilloscope, and I'm seeing activity on the address and data lines, so the no memory thing is really confusing. If the ZXMUX doesn't do it, I guess the next step is to dig up details on what the correct waveforms for the data and address lines are supposed to be and do a more detailed probe. Would you tackle this differently?

Thanks again for your time and advice, Mark.

astroturtle wrote: ↑Fri Mar 21, 2025 9:32 pm can I really have 20+ bad chips??

If the board suffered power supply problems, yes. The 4116 (or equivalent) "lower" DRAM chips often die if the +12V is present but the -5V is missing.
And if a 4116 dies and the +12V or -5V leaks out to other parts of the board, that can damage other chips.

Or if there was a misaligned interface/expansion on the edge-connector...

astroturtle wrote: ↑Fri Mar 21, 2025 9:32 pm I do have another ZX, but it's actually in much worse shape than this one. It's a series 2 with a big old lifted trace going to the Z80. That one has a ULA50102E version ULA, and I'm not certain it's in great shape. It has some really weird stuff going on with a 74LS00N soldered on top of it (pic attached). I'm hesitant to put it in the 6a... Can a bad ULA break other stuff on Spectrum? (the ULA in the 6a is a 6C001E-7)

That's actually a ULA5C102E. It's the first, original design of ZX Spectrum ULA. It's the same type used on most issue one boards.

This ULA version has some bugs. Most importantly, it has a problem with the I/O contention circuitry. This affects the Z80 when trying to perform I/O instructions, particularly keyboard reads from machine code running in RAM. The "fix" is the addition of the "cockroach" mod (a 74LS00 chip hand soldered with wired connections on top of the PCB, or later on, a mini PCB with the chip on this).

ULAs are backwards compatible, as in a later version can be used in an earlier board. They are not generally forward compatible. It's unlikely a ULA50102E would work in an issue 6A board because of the changes in the DRAM timing.

As issue two boards with a ULA5C102E and a "cockroach" are not very common, I would leave the ULA and the "cockroach" as it is.

astroturtle wrote: ↑Fri Mar 21, 2025 9:32 pm If it's a problem with the connections between the ULA and the ‘lower’ DRAM, how would you recommend troubleshooting? I have tested continuity on all lower and upper ram. I have an oscilloscope, and I'm seeing activity on the address and data lines, so the no memory thing is really confusing. If the ZXMUX doesn't do it, I guess the next step is to dig up details on what the correct waveforms for the data and address lines are supposed to be and do a more detailed probe. Would you tackle this differently?

What about the control lines:
ULA pin 35 (RASL) to 4116 pin 4 (RASL)
ULA pin 1 (CASL) to 4116 pin 15 (CASL) via the CIN / COUT link (selection links near IC28).
ULA pin 4 (DRAMWEL) to 4116 pin 3 (WRL)

Continuity test and test with your 'scope. I presume your 'scope has an analogue bandwidth of at least 25MHz?

Are your 4116 (or equivalent) "lower" DRAM chips rated at 150ns (or faster)? Slower types are unlikely to be fast enough for the timing used by the ULA.

The 'scope tests you did do, was this with IC27 removed? Only don't forget, both the Z80 and the ULA generate address signals, and the Z80, the ROM and the RAM drive the data bus. Unless you know which is doing what (by watching the signals on the various control lines), you don't which device is generating the signal on the address line or data bus line. That's the fun of the Spectrum and it's contented memory and I/O system

Mark

Hi Mark!

If the board suffered power supply problems, yes. The 4116 (or equivalent) "lower" DRAM chips often die if the +12V is present but the -5V is missing. And if a 4116 dies and the +12V or -5V leaks out to other parts of the board, that can damage other chips.

I believe that for the 4116s that were originally on the motherboard. I have written those off until I can run them through a RAM tester. I only put the other RAM I have on the board, once I was certain the voltages were correct. Anyway, I'll know for sure in a week or so...

That's actually a ULA5C102E. It's the first, original design of ZX Spectrum ULA. It's the same type used on most issue one boards.

Thank you! I have seen the "cockroach" mod before, but never done like this, so that's why I was confused. Good to know! That poor machine is my next target. The lifted trace to the ZX is pretty bad and long, but I think it can be salvaged. But that is fun for another day!

ULA pin 35 (RASL) to 4116 pin 4 (RASL)
ULA pin 1 (CASL) to 4116 pin 15 (CASL) via the CIN / COUT link (selection links near IC28).
ULA pin 4 (DRAMWEL) to 4116 pin 3 (WRL)

Thanks for this tip! I tried all of these and continuity is good on all of them. For ULA pin 1 (CASL) to 4116 pin 15 (CASL) I got continuity on ULA pin1 to 4116 pin 15. I also got continuity between those chips and the C IN/OUT jumper. No continuity on the H jumpers above it.

In the meantime, I desoldered the Z80 and the PCF1306. (I'm waiting on some dip sockets for these to arrive in the next week or so.) Everything else was already socketed, so I removed all the ICs from the board and did continuity/resistance tests on every data, address line etc., between every IC location on the board. Everything checks out, so at least it is unlikely I have any broken traces. I'm also going to put the board under the microscope later today and do a closer visual check to confirm there's nothing I missed.

The only unusual thing I noticed is pins 2 and 14 on the RAM. Between 2 and 14 on the same chip, I get continuity, as expected. However, between pins 2/14 on one RAM location (eg IC6) and pin 2/14 on another ram (eg IC5) I get a resistance of about 17 Kohm. Same with upper RAM on pins 2/14. I see similar behavior on the series 2 ZX board, but I'm not sure if that's normal?

astroturtle wrote: ↑Tue Mar 25, 2025 12:27 pm The only unusual thing I noticed is pins 2 and 14 on the RAM. Between 2 and 14 on the same chip, I get continuity, as expected. However, between pins 2/14 on one RAM location (eg IC6) and pin 2/14 on another ram (eg IC5) I get a resistance of about 17 Kohm. Same with upper RAM on pins 2/14. I see similar behavior on the series 2 ZX board, but I'm not sure if that's normal?

Assuming you are talking about the “lower” DRAM. On those DRAM chips, pin 2 is the single data input (“D”) and pin 14 is the single data output (“Q”). So to enable both writing and reading they are connected together and then on to one of the data bus lines to the ULA and via a resistor to the Z80A. See this post for the pin-out.

Each DRAM chip only deals with one data line and hence only one bit of an eight bit byte. That’s why there is a group of eight chips.

All other signals on these DRAM chips are in parallel between all eight chips in that group.

It’s similar for the ”upper” DRAM except these connect their data lines to the Z80A directly.

Mark

Whelp, it wasn't the PCF1306P or the RAM. It's the ULA.

I went through every RAM chip I have with the tester (great device btw!) and it turns out I only had 4 bad chips, and none of them were on the ZX. Installing the ZXMUX had no effect and I still got the no RAM loop.

However, I have a friend working on his own 48k and having some similar challenges. In his case, he’s getting horizontal black and white bars, plus some random screen garbage. We swapped ULA’s and his started doing the no memory loop, and mine started showing the black and white bars + screen garbage. *sad face* So time to start saving up for a vLA82, I guess?

I have started documenting these fixes, and as I go along I'm adding to this ZX repair here:
https://www.astroturtle.com/wp/2025/04/ ... he-issues/