ZX Spectrum Initialisation and Investigating Initialisation & Memory Faults using a Logic Analyser

[@worroSfOretsevraH]

I absolutely love your videos, can't wait for the next one.

[@solidhit2]

Your Videos are pure gold!

[@volodimirkun]

It moght take some time for me to relly dig into detailes of this video, but any way - COOL. For a professional engeneer it might all be obvious, but for a nooby like me it is so helpful, man!

[@RetroMarkyRM]

These videos are outstanding!!! Are you planning going into similar depths about the commodoe 64 and/or BBC micro. I´d love to hear your exceptional wisdom on those.

[@JoulesperCoulomb]

Perhaps, but I'm working on something rather different first, so it won't be for a while.

[@RetroMarkyRM]

ok, good luck with that. look forward to your return whenever that is :) stonking channel.

[@pcuser80]

Such a great video!

[@xyzzy450]

Have you done any work with the Watford Electronics SPDOS disk interface? Cheers, Trevor

[@povvercrazy]

Fantastic videos, not sure if you still monitor this chat still, but recently got hold of an old issue 2 spectrum that was sold as 16k but appears to actually be a 48k model with Oki upper ram, followed your guides to preliminary checks, everything seems ok, powered up, again seems ok screen displays as I believe it should, but I'm using a bench supply set at 9v limited to 750mA at first and I got an OL after a few seconds so I disconnected, as per your comments I'm presumed it was the capacitors charging, so reconnected the supply and this time it doesn't go OL, but here's the problem the amperage display on my bench power is constantly fluctuating, all voltages I measure appear accurate and within tolerances, so my long winded question have you come across this before ? would I be correct in thinking its capacitor fault? I have ordered a set of replacement caps anyway. Also it has 74ls158 in IC25 and 26, and IC26 has leg 11 cut, and leg 10 bent and connected to the socket in place of leg 11, does this seem ok? Ive tried to google it but cant find any info about this maybe im not looking hard enough :) I was amazed at how clean this speccy is zero corrosion or crime inside or out, looks like it was brought in 1982 then wrapped in plastic and hid away perfectly stored from the elements.

[@JoulesperCoulomb]

It's normal for a ZX Spectrum's input current to fluctuate quite a lot, depending on what the computer is doing. I do recommend you replace the original electrolytic capacitors. Old capacitors are a very common cause of faults in ZX Spectrums and can lead to the lower RAM being damaged. When it comes to use in a ZX Spectrum, a 74x158 can be considered to be the same as a 74x157. The former has additional inverting buffers on its outputs, but this doesn't matter in the Spectrum circuit. The modifications that have been made to IC26 are because the internal arrangement of the 32 Kbit OKI RAM ICs is not the same as that of the similar TI RAM ICs and so addresses must be formed slightly differently. You should leave this as it is so long as any such OKI upper RAM ICs are fitted. It will not stop you from using fully functioning 4164 ICs to replace any faulty upper RAM ICs, should that be necessary. Issue 3 and later ZX Spectrums have links that can be connected to allow OKI RAM to be used without this modification. Does the computer initialise as a 48K model? If not, most likely one or more of the upper RAM ICs is faulty. You probably could use the Memtest program available at zx.zigg.net to identify which.

[@povvercrazy]

thanks for the reply, I've been using your memtest programming it's very useful, when running on my speccy after a re-cap i got 16 errors for upper ram!! i notice i had a weird mod to ic26, pin 10 was bent into where pin 11 should be and pin 11 was cut off! after examining the schematics i noticed on an issue 3 (mines a 2) for oki upper ram pin 11 should be connected to A14 line and pin 10 ground, so i got a new 157 wired it up correctly well as an issue 3 and it sorted most the problems out, running memtest only gave 1 error which lead me to ic 20 being faulty, swapped that out now I'm getting an error saying i now have a n error with ic17 seems I'm chasing chips haha think i might just replace the whole lot :)

[@TheMadmagik]

whats the name of your logic analyser and software?

[@JoulesperCoulomb]

+MAGIK I was using an Open Bench Logic Sniffer and the OLS client software. The latter is riddled with annoying bugs, but it gets the job done.

[@Livebirdboxcatterall]

Hi, really loving watching your Spectrum repair videos. However I have run into a snag and wondered if you could give me a pointer. I have a 48k which is running as a 16k. Games crash when loaded, print peek 23733 comes back as 127. I then entered print peek 23732+peek 23733*256 which give me a number less than expected. Poking 85 and 170 to one above that number led me to ic 21 which I changed, Then the test showed ic 18 and then ic 22. I have changed all 3 but still have memory issues and it is now pointing to ic 15. Is it possible that they have all gone bad or should I be looking at something else. I know it's hard to say without having the thing in your hands but a pointer would be great. Thanks again.

[@JoulesperCoulomb]

I've certainly repaired a few ZX Spectrums that had six or seven faulty upper RAM ICs. If you are able, load and run the Memtest program available at zx.zigg.net and tell it to test the upper RAM. You'll probably then be able to work out which ICs are faulty from one set of test results. If the computer's simple RAM test (that it runs at power on) doesn't reliably detect all of the upper RAM as faulty, then it might attempt to use some of it for the BASIC operating environment, leading the strange behaviour and crashes. Sometimes (not always), connecting pin 5 of IC23 to the 5 V supply rail can get round this because it'll make sure all upper RAM is non-functional. If you like, this connection can be removed once the computer has initialised. Take great care not to let the connecting wire contact anything else.

[@Livebirdboxcatterall]

Thanks for the reply. I first again entered the print peek + peek command as above which gave 32775, I poked both 85 and 170 to 32776 and they came back as 85 and 170. I ran the memtest and straight away got 16 errors at 32768 through 32783. If I poke 170 to 32768 it comes back incorrect with ic 20 at fault. I will go ahead and change it over. I wish I'd bought one of the all in one modules now and removed them all. Thanks again for your reply.

[@Livebirdboxcatterall]

Replaced ic20 having already replaced ic 15. Ran the memtest again and straight away had the same, failed 32768 through 32783. However I can poke all of those addresses with 85 and 170 and they all return 85 and 170. What am I missing here?

[@JoulesperCoulomb]

RAM faults don't always show up when there are just a few reads and writes with long delays between them, as is the case when they are due to the BASIC POKE and PEEK commands. The Memtest program performs blocks of writes and reads in a tight loop. Use the values reported by the Memtest program to work out which bit positions have errors.

[@Livebirdboxcatterall]

Indeed, and given my beginners knowledge in understanding the results I wrote a simple routine which pokes 170 to 32783 prints what it finds and loops. The result was page after page of 170 with the odd 171 or 234 which points to ic 15 or 21 which I have already changed out. It also now returns 128 on command print peek 23733 which apparently show a 16k spectrum, earlier it returned 127 which apparently shows a 48k working as a 16k. I seem to of hit a wall and will need to have a browse of forums and such. Thanks again..