This video was a great one! I'd love to know if the retro chip tester finds those RAM chips are actually fully dead. Or perhaps the CPU barfing out like over 6 volts fried all the RAM chips? Although the CPU output 6 volts seems really weird!?!?! Like what failure mode turns a CPU into a step-up converter? I mean, obviously not for real, but still... weird! Looking forward to part 3!
Yeah, the entire set being dead combined with the burning-hot SAM that he just kind of glossed over makes me worry that some kind of overvoltage going on and he's going to just fry the new set if he keeps using it.
@@Meshamu I'm guessing maybe he doesn't want to use it because he wants to show a diagnostic process without the use of custom tools? It certainly would be more satisfying to see them checked on a tester.
When you're pressing enter, not only does it appear to crash, but you get the S in the top left AND a relay click, so it's not crashing, but it's somehow jumping to execute the cassette load function which puts an S in the top left of the screen and turns on the cassette motor (hence the relay click trying to allow current to turn on the cassette motor. Interesting...
One thing ive learned in troubleshooting, be it computers or cars, is that the first thing you dismiss as "it cant be that" is the first thing you should test. Never rely on your gut feeling, always test the "known" goods before go swapping parts around. I had the exact same type of issue you had here with an old VW golf, where the first thing I dismissed because they looked new (spark plug wires), ended up being the problem, but i didnt realize it until i had tested and swapped every other possible part first, wasting a bunch of time and money. All because I assumed that i knew that one part "couldn't be bad". Its an easy mental trap to get into. So now when I think "It cant be that", I test it to be sure.
I've mostly had the opposite problem: replacing parts I thought had to be bad and it ended up not solving the problem. I had a car that got to where it sometimes wouldn't turn the starter over when I turned the key to start, and eventually never would start. There was no click or anything. I spent a lot of money to replace the starter solenoid, the ignition switch, and the neutral safety switch assuming that had to be the problem after the last attempt failed. I finally got a wiring diagram and went through checking every connector in the starting circuit. It ended up being a burnt terminal in the connector near the transmission, one connector away from the neutral safety switch. It cost me about a dollar for a box of terminals.
My first idea was that +6V overshoot was an indicator of a bad voltage regulator or something that could kill all the RAM chips. But then my hunch is that the RAM chips you are putting in are different from the ones that were already in and somehow they are incompatible. If you put all of them back the machine would work as before. Probably has something to do with the timing or some similar properties that can be different on a chip. While the C64 can accept different timing, maybe the COCO is more sensitive in that way.
it was cpu overvoltage sam and sam sent off errors and destroyed all ram at once it unbelievable see that machine does that i never own one but i did watch other video of someone fixing coco 2
Yeah, that 6V overshoot bothers me. You're not getting that from a chip that only has +5V as it's power. Something is leaking power onto the buss. It's possible that the CPU isn't bad, but struggled to overcome the excess voltage/current.
@@russellhltn1396 I agree that 6V seems odd, but it's not that much more than 5V. My guess would have been higher than spec VCC or something else also driving the bus.
@@jnharton TTL requires VCC be 5.0V +/- 0.25V. 6V would be way off and is a half volt above the absolute maximum input voltage. I'd think that would show on a power supply check. I'm thinking it's a defective chip that has access to a higher voltage.
Two other pieces of info you may find helpful in the future: First, inverted @'s on the display very often indicate problems with the DRAM. When you originally fired the board up, those were present amongst the other garbage on screen. It's a good indicator to look there first. If you'd have known, you could have possibly saved yourself some time. Second, you can find more listings for the SAM chip if you look it up under the 74LS783 identifier rather than SAM, Synchronous Address Multiplexer or MC6883.
That's what I did for the SAM -- I saw "SAM" prices were outrageous but the 74LS783 prices were pretty cheap. I guess the Chinese sellers hadn't caught onto that. :-) As for the @ symbols, I'll need to see how it acts with the ROM removed now the RAM is working.
@@adriansdigitalbasement some years back, early 2000s, i got a few of them, cant remember under which number, for 50p each from a seller here in the UK, doubt they have them at that price now, or at all 😉
33:30 It looks to me like the 6th RAM chip on the board is now replaced by the original 5th RAM chip on the desk. And right after 34:24 the 6th, "untested" RAM chip on the desk gets an X mark. Probably doesn't matter but I had to get this out of my system 😉
Yeah there is a bit of a continuity issue there -- I noticed it during editing :-) In the end, all of the chips were bad but yeah eagle eyed folks will notice it!
28:38 - What if these chips were "bad" from the factory. They say 64k but maybe they are broken 64k chips but the first 16k works. So they popped them in and sold this unit as a 16k system. Someone else before you came along, saw they were 64k chips, put the jumper to 64k, and it didn't work. Have you tried setting it to 16k with those "bad" chips in there? I've heard stories of systems (probably from you!) being sold with bigger ram chips that were half dead, but worked in the system because the system didn't use the bad parts of the RAM.
The Wikipedia article does there were 32k CoCo-1 models using eight "half-bad" 64kbit chips so that might be what's going on here, it also seems to say there was no 64k variant of the CoCo-1 which further suggests this might be the case. It could even be that the HIGH/LOW jumper is set wrongly so it uses the faulty part of the chips instead of the working part. Another computer that uses faulty 64k chips is the 48K Spectrum.
@@Torbjorn.Lindgren the 'faulty' 64k ones in the spectrum werent sold under the original 64k type number but marked with their own number plus a suffix indicating whether row or column 'dropout' and which 'polarity' 1 or 0 , and not only used in spectrums but likely many other devices, certainly some non sinclair made external 32k rampacks used them, i have 2,
@@BrainSlugs83 yep, 4 possible combinations, and 4 types of chip from 2 manufacturers , one had a dud column, one a dud row, and of those it could be a high or low dud, good thing is they can be replaced drop in with a 'good' 4164 without worrying about links if its a 128 cycle refresh 4164 type, if a 256 cycle type, it may depend on what the links are set to,
Hi Adrian, just a thought, did you check the speed rating of the DRAM chips you removed in case someone had fitted a complete set of slow chips? Might explain why it sort of worked when all of them were fitted together, but didn't when one was operating at a different speed to the rest. As someone else suggested a RAM tester would confirm they are indeed faulty. Jeremy
Another piece of evidence for this is that the machine failed slightly differently (The "OK" text changed on the repeated portion) when he had only one known good RAM chip in, making it look like the good chip was actually bad. I suspect the same thing has happened in reverse, causing him to misidentify the original RAM chips as all bad, when something different was actually at play.
Occam's Razor. My bet is that the original Motorola 4164's were some of the "half-bad" RAM chips Tandy used to be able to sell "32K" CoCo's. Each one you put back in with the good MT & TI RAM introduced errors because they were known to have one or more errors in the top or bottom half of their address space. If just one of those Motorola chips' problem was with its address lines rather than the data storage or data lines, that could explain the addressing problems you were getting. I'd wager that if you put the Motorola chips back in and set the machine to 32K, you'd be able to get it to work. You might have to switch the jumper telling it to use the high or low bank, but I bet you'd get it to work. Whoever changed it to 64K probably didn't know about Tandy's little cost-cutting trick with the half-bad chips.
Sinclair did the same thing with the earlier batches of 48K ZX Spectrums, with the 32K "high memory" that wasn't in the 16K models being made up of these types of faulty chips. They weren't from Motorola, but I think it was NEC.
Other possibility, of course, is that some nozzle set the jumpers for 16K at some point and fried all the 64k chips with +12v. Then tried to go back to 64K, only to find that he'd killed his computer.
I had a CoCo 1 as a kid and I can confirm, messing with those jumpers can fry the computer! Out of idle curiosity after the CoCo 1 was no longer my "daily driver" machine, I did the opposite - flipped the jumper to 64K to see if I could get "extra RAM" I got that first screen with the inverted @ symbols, and even moving the jumper back, the CoCo was gone.
That is honestly the most likely explanation. I just don't see how else all of the chips could be bad. As for setting the 16k machine to 64k, some people have said that could damage the 4116 DRAM chips. Interesting -- it's an experiment I have have to, although it would suck to sacrifice 4116 DRAM. I'll think about it.
4164s would almost definitely die if set to 16k as if using 4116s as that would put 12v on the 5v pin...this was and may still be a problem with some ebay 4116s that were actually skimmed and relabelled 4164s! several people had issues with these '4116' burning up so i got a few and done tests on them, tested perfectly ok as 4164,
back in the day… pre Y2K… we used to modify [mix and match] all sorts of different computers together to make our own custom systems! (especially laptops for rugidizing)… this episode took me back to the future!!!
Got it figured out at 30:02 min in... It was showing /+, which is OK minus 32... (ASCII of "O" is 79... minus 32 is 47, which is ASCII for "/"), same for the plus sign, the ascii of K minus 32 is the ASCII of the + sign....
In the CoCo 1, the original expansion was only up to half the chip. Tandy put "bad" (but partially working) RAM in the systems and sold them as working machines. However, this machine is now wired as 64K, requiring fully working RAM. If the repack was in mid-work OR if the person was unaware of how Tandy saved money by using partially non-working RAM this could be the issue. You were working under the assumption that the setting for 64K meant that the RAM was all working as 64K - not necessarily a good assumption.
I loved this review/repair of a Coco1. The Coco2 was my first computer, and I still have it. Everything Coco still has a warm place in my heart, so I loved that you were able to get this one working....and considering the outcome of the issue, it was similar to a problem I had after I did a 64K upgrade a few months ago. Coco for ever.
Great diagnosis video of my starter computer. I remember having this exact issue... seems either one RAM chip goes or they all do. I wound up replacing all of them to fix a similar issue back in the day.
I'm just shocked it broke at all. I had several Cocos back in the day, and they were all 100% bulletproof. I abused the hell out of them as well; they spent years neglected in a 120° attic before I dusted them off. I wouldn't know what to do if one had broken on me.
Thank you for the effort that you put into finding that mischievous fault. I really like watching the process of narrowing-down a fault. I never would have guessed that all of the RAM had failed. I was thinking that perhaps a bit of wire fell into an edge connector socket and bridged two address lines. No.. I was wayy off!
Nice repair. Retro Chip Tester would help out testing those DRAMs. Also I think the BackBit Chip Tester has some Coco chips it can test too. I find the BacBit to be cool to keep om my workbench for it's size and the ability to test custom chips for Commodore, Atari, Amiga, Etc...
Another great video, thanks Adrian. As you stated, the coco is a very simple system. This makes repair easier than most, as the cpu will run regardless of the video system, and vice versa. The “crash” you were seeing looked like it jumped to a tape loading routine. If you type “cload” the screen clears and “S” it shown in the top left to show its searching the tape. Did we hear the relay click? It’s likely. You made some pretty accurate assumptions on the way the machine works. Fair play. Interestingly enough, the 6847 is used in the acorn atom with the same inverted video :)
(Posting before having watched all the video, so this may be covered at a later point.) First two pins on the 6809E mentioned are "E in" and "Q in" - the clock input signals; there's two, as the 6809 uses a pair of overlapping signals for timing, with the "Q" signal leading the "E" signal by a a quarter-phase of the signal.
yep, Q as in quadrature, the pair of lines give you four discrete states that are used to cycle the processor around the address,fetch, execute etc. states. To fully confirm they are functioning you should scope both simultaneously and verify the phase difference. With this technique you need to quadruple the apparent frequency (800 kHz?) to give you a comparison with say, the Z80 cpu which uses a simple single phase clock, so 3.2MHz versus 4MHz typically used in Z80 of the period.
I did a lot with 6809 systems (running "mini UNIX" Microware OS-9 operating system) in the 1980s. The Hitachi HD6809 (or the faster HD63B09 or HD63C09) were our preferred choice. Better than the original. I still have the development system (running the same CPU and OS) in my basement. It could be a challenge to make it boot again, mostly because of the floppy drives. The Hitachi CPU even has some additional features, like added/extended registers and enhanced multiply and division operations. But to produce "mainstream" 6809 code, I did never use these additional features.
Agreed. I have often fallen asleep while watching Adrian on my TV from my recliner. Not because he is boring or uninteresting (far from it) but because he is so soothing with his commentary. I think I have sleep learned entire Apple repair series and had to go back and rewatch while awake. 😆
It's taken me this long to realise that putting a *Colour* Computer motherboard and a black-and-white TV in the same case isn't as big a win as I initially thought.
Back in the late 70's, I worked for a guy running a business out of his basement. We made cables like the one you had trouble with (a standard as everyone knows). I can't recall if there was a jig or manufacturer tool to install the ends, or we used a vise. I believe we used a purpose-built tool. Digitize Inc is still in business today...
I have a hand tool for installing those - used it just a few weeks ago for making a cable to connect a Model 1 to the expansion unit. It has a reversible insert, one side for pin connectors like in the video and the other for edge card connectors. No idea where I got it - had it for probably 35 years. Most likely Radio Shack had them since that's about the only place I got stuff at the time.
Now, I bought a dead Oric-1 some years ago that had more than a smattering of dead chips. It turned out that what had happened was that due to its unusual supply regulation arrangement it had effectively been fed the full 9V and fried over half its ICs. From that point on I learned to check the supply rails *FIRST*.
I remember a tech channel that introduced me to a fantastic chip tester, which I bought and assembled. It is a lot easier that flipping chips… wait.. it was your channel I saw that on! 😊 Why didn’t you use that?
I predicted multiple dram chips would be bad, but never imagined 8 of them would be! It would be interesting to understand how the machine was able to function at all, like mapped through the bad chips.. amazing...
You probably won't believe me, but I actually did predict ALL of the behavior almost perfectly. That ALL of the RAM chips are dead. That replacing one chip will cause the second OK to become garbled. etc Here is my reasoning: The text is repeating exactly once thus one address line is bad. 32 chars per line, 4 lines, it is probably A7, although that depends on how it is wired. The CPU outputs 6.15V on at least one of the lines. Not sure how it manages to do that but it is BAD not only for the CPU but for other chips as well. And it probably fried all of the RAM at once since the address goes directly to the RAM - only select logic goes through SAM. The DRAM of the time was particularly fussy and fragile, especially to voltages, and it didn't have many protections, so it makes sense that it died first. So it we are seeing the perfect copy of text that could only mean that ALL of the bits have exactly the same issue, so ALL of the memory chips have A7 gates completely fried and internally stuck high. When ALL of the bits are high, the code and data gets "overriden" by another portion of said code/data, so it we're lucky it might somewhat run and display correctly albeit duplicated. When we replace one chip, one bit is corrected and NOT duplicating thus that one bit is missing from the duplicated text if we are also lucky to avoid code corruption. If we replace all but one, the code or data gets corrupted in that one bit - we would see some characters each 4 lines are off by 1/2/4/8... ASCII codepoints, i.e. A becomes B or D or , etc. But the code gets corrupted so much that it doesn't run (that's pure luck it didn't happen with single bit correct). At some point of the video I was experiencing some sort of...umm... epiphany? looking at somewhat confused Adrian and knowing what will happen next. It was kinda bizarre sensation, like I actually could see into the future. Anyway since only a single address line is bad, the computer can probably run perfectly if the dead address line is switched with the highest one. The computer will just have only half of the RAM capacity (32K?) available to it (unless its ROM just expects all 64K to be present)
@@jwhite5008 That makes sense with regards to the text on the screen. I wouldn't say I predicted all of them were bad, but it did seem like something on the address bus went bad (likely the CPU) and took out one of the address lines, looked like A7 or A8 since the text repeats after 128 or 256 characters at a glance, in at least some of the other chips. I still don't understand why the thing crashes with any one of the bad chips in it but runs with all eight bad chips though. That's really weird.
It actually does makes perfect sense. The chips probably received a voltage much higher than what they were designed for (something like 5.5V max) on one of the address lines - like you say most likely A7 or A8. Some of their row/column select logic gates inside the RAM chips were fried and now act as if the line is always high (or always low - doesn't matter) irregardless of actual input. The reason why the second OK was garbled when one good chip is inserted is because one chip = one bit is now NOT faulty, the duplication "omits" one bit from each character, i.e. that one bit displays what it was supposed to rather than duplicated data. So the characters on 4th row - which include that one working, non-duplicated bit - get "garbled" - but actually the replaced bit is fine - it is all of the other bits that are wrong. The reason why the system sort-of-partially-works with all bad RAM, and with all good RAM, but fails if one of the chips is bad is that when all of the RAM is bad, the data is written in wrong spots, but the bytes of data themselves are not corrupted - only misplaced. Thus it overwrites the other data that should be there. However, if you write the data and read it before it happens to be overwritten that way you do read valid data. For example, if there is a bunch of text strings in the memory, each terminated by zero-character, in the "misplaced RAM scenario" the computer will read wrong strings but still terminated, whereas in "partially misplaced (garbled) RAM scenario" it will not find that terminator (because the terminators of two text blocks are unlikely to align with each other) so it keeps reading, overflows the buffer and tramples some random memory. Similarly, although most code is executed from ROM, if some code is copied to RAM, it is copied correctly, but the execution rolls over and skips portions of the code. When one bit is wrong the code is completely random and probably hangs. While it may seem vert bad, (and it is), the CoCo is much much less complex than PCs nowadays. There is a chance that the values that were written are read before they are replaced. So by that chance the computer is somehow able to survive enough to display OK prompt and let you type stuff. Now when some but not all of the chips are replaced with good ones, the bit-by-bir corruption occurs - similar to what happened with text when one chip was replaced. Some bits are read from one place while others are read from another. I believe the latter this is much worse than when all bits of the data is correct but misplaced because it tends to be more random-looking values. Of course there is a lot of luck involved - it easily could be that the machine wouldn't boot with either or all of the chips being bad, but that is not what happened. And it was probably a very lucky coincidence that it was able to boot with only one good chip. The instability with all bad chips is probably because they either sustained too much damage overall, or not enough to fully destroy all of the A7 gates in every chip.
@@jwhite5008 that all sounds very logical, and likey is the case. To see it happening in real time, and actually still function, albeit partially, is very interesting. Makes me wonder, if any one pair of those chips were placed out of the order they originally fried in, would it still function? I'm thinking not but may be wrong... I.e. chip one in socket two and vice versa, etc.
@@Quickened1 yes - all 8 RAM chips had exact same issue. All of them had same address bit stuck (probably A7) - which is why they displayed uncorrupted data at wrong addresses- this is what initially surprised Adrian.
Your scope should have runt pulse triggering, it really makes sense to make use of it here. Also in case no DRAM chips are at hand, switching them around and observing the pattern change can help too, especially if you have an idea about how far they are apart in terms of adressing
power short on data lines ? for instance it lived in a botched together setup say at some point a power wire made contact with the data lines , straight through all ram and cpu.
I know he always says "MT Ram" but that company is still around today, MT was just the initials for Micron Technology. In the 1990s they made some of the highest quality RAM sold through their direct to consumer brand Crucial Technology.
I usually look at the datasheet for the chips, since some designers label the pins with "what they do in their circuit," and not what they actually are. Besides, they will be more legible.
Those connectors like on the keyboard cable are called IDC connectors. Officially, IDC means Insulation Displacement Connector (as in you don't have to strip the wires) but I like Ben Heck's name better: "It Didn't Connect."
I took basic programming in college. I have written many programs. Including a klindergarden one. My first daughter used my program and was moved to fiirst grade, Also in 1st grade she did great. I would love to program in basic again. I am currently programing in Turbo Pascal also.
Amazing! Who could have expected 8 out of 8 faulty modules… and a kind-of-working system ONLY when all of them were plugged in! 😂 what are the chances! Great video as usual!
Rockby has their specials out. They're selling 68B09EP processor, which may interest those that want to build or play around with this wonderful old 8-bit processor. Also selling 8154 chips which work great with the SC/MP CPU, or an old 6502, 6800 or similar processor if you want a little bit of RAM and IO for a small project.
I own a Minipro model TL866A programmer. It tests RAM chips and 74xx series logic chips all day long like nobody's business. It reads PROMS so you can check them, too! I've got the diagnostic card with full harness kit but I now need a dead test cart as well. And a good OEM original 1541 test floppy.
You better check the TL866A manual. It does not check any DRAM, nor SRAMs < 16kbit, nor 1200 logic ICs, so it is quite useless to test hardware from the 1970s or 1980s. It does also not read 1702 or 2708 EPROMs.
@@herbertschuster9088 Wow. I musta been lucky, so far! It has helped my business repair an Amiga 2000, a Commodore 128, two Commodore VIC20's and two Commodore 64's, two medical devices, A plastic injection moulding machine control, a plastic wrapping machine control, A flash chiller controller board, a trash compactor controller, a motorcycle ride height controller board and 3 industrial machines. I agree that it will leave you wanting with MANY older or proprietary IC's. Like the National Semiconductor ones on a machine I am rebuilding right now. But it is FAR from completely useless and has shaved DAYS off of many reverse-engineer style no-schematics diagnostic jobs.Oh, to find the source of that machine's crazy behavior, faults and errors! Three axis servo motor controllers with encoders, an HMI, translation cards, drives, main controller...impatient customer with "status check" texts and calls...I'll use ANY help I can get!
It's great watching you work on a system that you're less familiar with and working it out from the basics - really helps me to learn how to troubleshoot. Especially find you talking through how a scope trace should/shouldn't look very useful!
I was thinking the same; I'd assume that it's unlikely that a bunch of components just suddenly fails. And that makes me worry if the reason for the surge is still there, and could destroy more components. Maybe it would be a good idea to check if there are any loose threads from some wires hanging near other wires in the power supply. Though, the cause might have been temporary, like at some point something conductive had fallen in the power supply to briefly bypass something. But still, if that's the case, that something could be still somewhere there inside, and could do the damage again. (I don't remember if in an earlier video Adrian already cleaned the powersupply -- if that's the case, the culprit might have got removed at that point.)
I too was thinking that it was a power spike that killed the RAM and CPU. Maybe they were more sensitive to the spike and that was the only reason they were killed or maybe the other chips had more passives in line that dampened the spike?
Hi Adrian, a tip I learned over the years: Don’t use Deoxid on Potentiometers. The stuff is so aggressive that it will be the cause of trouble itself after some time. I use an inert cleaning spray instead. If you do use Deoxid, then always clean afterwards with a few good sprays of the inert cleaner to get the Deoxid out. Switches can also be lubricated with real contact lubricant spray after cleaning.
@@DavideNastri Contact cleaner is mostly the same as Deoxid, so good for switches. Also here make sure to clean afterwards with a a safe inert cleaner.
I use Kontakt 60 as Deoxid and Kontakt WL as inert cleaner. Also I use Kontakt 2000 as lubricant for switches. Pretty expensive, but it woks and lasts.
I heard there are 2 types of Deoxit, the liquid one in a bottle is 100% concentration and can cause problems if not thinned out, and the spray can that is 10% that is usually fine.
For the several dozen commenters claiming the 6V killed the RAM, there was no 6V. Adrian measured the 5V rail, it was spot on. The 6V came from the ringing by the mile long ground path, because he put the ground clip on that IC pin which is grounded through a long, thin wire. So it was a probing issue, the 6V wasn't really there.
The first thing you should do is test ALL chip voltages at their supply pins with a DC volt meter, AC volt meter AND a scope to check for value, noise and ripple. For instance, bad caps can maybe cause high ripple which can cause overvoltage/undervolatge (e.g. voltage is 4V-6V with 2V ripple and DVM reads 5V) can maybe cause DRAM to be unstable? Drawing more power (e.g. fitting two ROMS or different chips) can cause more ripple and thus change the symptoms. Maybe those RAM chips were OK? The 6V on one address line didnt just come from nowhere!
I wonder if the specs of the "good" and "bad" RAM chips are the same. As they are socketed, the originals could. have been swapped out for a different spec'ed devices. Are the access times the same? It would be interesting to compare their respective data sheets. They may even have different voltage range specs. As you say, strange they were all "bad", which would point to a spec difference.
It reminds me Timex 2048 that I repaired. It (probably) received a hot plug of cartridge. CPU, ROM and bus buffers were fried. Also SCLD got slight hit. Here are no buffers so CPU and memory could got damaged this way.
Hi, When you checked the address signel, we see clearly at 27:01 that there is a glitch in the address. If you mesure it again whith the new RAM chip, it may be different
I don't know your oscilloscope but it looks like when you reset the system, the scope just doesn't refresh the screen because it's not triggered. If you put it the scope on auto-trigger, you'll probably see a horizontal line on any data/address line while you reset the CoCo. The RAM problem looks to me like there is some sort of subtle difference between the old and new chips, not necessarily defective chips. Maybe different RAS/CAS timings that cause bus contention? Anyway, great video! My guess was that the problem was something with the 6847. Great repair!
Maybe something in that messy case momentarily dumped a higher voltage into the RAM through an address line, or something. It's conceivable that the other chip that handles the multiplexing is just robust enough to have handled the hit and kept chugging along, while the fragile RAM chips got fried.
One final thing and I'll shut up. To my knowledge (and CoCopedia agrees) Tandy/Radio Shack only ever sold a 64K CoCo 1 very late in the 1's lifecycle when they started selling a white case model CoCo/TDP 100 and it'd be a shame to think that someone caniballized one of those to make this repack. There are just so few of them around. I really hope this was a 4/16/32K gray model that was upgraded to 64K.
Well the original 64kl RAM chips are Radia-Shack branded ICs, and this motherboard only has 16k/64k jumpers on it, so I'm assuming this was a 64k board. Don't worry though, I have some info on this machine now and it was assembled with warranty repair spare parts. Hopefully we'll have some more info in the next part.
hi there was a 8 track 1 inch reel to reel someone have done kind of the same made 2 decks out of one i have seen recording decks that have been jazz up
At about 7:15 you probe the crystal input and note the frequency is off and blame the scope. Note that adding the scope load of 10 pF to that circuit WILL change the frequency anyway. I wonder if that overvoltage you saw on the CPU somehow caused the DRAM failure?
It looked like you had two address lines shorted together, with the repeating text further down the screen( down memory). I'm not surprised you couldn't find it by looking at the address lines unless you were looking for two identical signals on the address lines. I am surprised you couldn't duplicate the fault by placing the memory back in the same socket. I'm wondering if that memory is highly oxidized, because some of it was working well enough to show in screen memory.
Good thing that when whatever happened, it had the CPU and those 8 RAM chips blow first to protect the other electronics... ;-) It's not only amazing that it was a CPU and all the RAM, but that whatever was big enough to do that didn't also blow caps/resistors/etc too... Wow!
You need to check the power supply to the DRAM as they have multiple voltages feeding them. I can't remember if they have a -5 or +12 rail but it could be off and out of spec for the Motorola chips but the MT is OK with it. Also that tarnishing (not rust) from pins can be removed with a pencil eraser. DON'T sand them as you could remove the plating.
Hi Adrian, did you try the retro chip tester on these "bad" RAM? The Motorola RAM may have timing issue that need to work with the same brand/spec, seems mixed with MT RAM may cause problem.
Probably worth re-capping the board at some point. Particularly around the DRAM. I'd just do the electrolytics for a start, they're usually relatively easy.
29:21 - Question, (may sound stupid but correct me if Im wrong) why you didnt take out all chips, and check them with multimeter in diode mode? While out of circuit, Red@GND leg, black on each leg, and make a chart, then go to next chip. Is this method OK for this type of IC circuit? If all chips are the same, all measurements should be the same.
The Ram chips went bad. At some point there was a short that killed one or more of the address lines. and it would kill all the lines on each chip that would lean toward a power spike that fried all the ram..
FWVLIW: A wild guess: Apart from the usual "What's wrong with the PSU?" I suspect some kind of 'jolt' was fired across the bus and took out the CPU along with the RAM. What 'jolt'? I wonder if there's an issue with the cartridge port extender / the floppy controller / plugged in backwards or it just fell out at some point powered..? Regardless I'd check all of the above before plugging in - in particular I'd try powering up the floppy controller unconnected to a computer and check for shorts before using it. Not sure if just continuity testing it would be enough if one of the IC's got fried to a dead short across unexpected pins. As an aside I wonder if this machine was assembled with 1/2 bad RAM and the jumpers set to something less than 64k because it worked. A previous owner saw the 64k IC's and changed the jumper(s) to 64k and 'broke' it... Maybe putting in the original RAM and setting to a lower configuration might 'fix' it again..? I don't know this system at all so I don't know if that might hose something if the wrong RAM and jumper setting is used - watch out!
I think a power surge may have been the cause, or it may have been hooked up to 220V. Another possibility is tyat therewas a cascade failure from the 6.9V that the CPU was putting out on the address (Or was it data?) line, damaging multiple ICs on the board.
The bad RAM should not make a difference -- that original CPU was dead, not progressing from reading the first byte from the ROM. The RAM isn't even needed for that to work as we should have at least been seeing it trying to execute code before crashing due to bad RAM.
I never knew the Radio Shack computers ever had anything but a Zilog processor. The 6800 processors are Motorola. My TRS80 Model III had a Z80 processor.
Your info about the need for this computer to work with a 6809E can be interesting for French hobyists who try to repair a Tandy CoCo because the Thomson MO5, a quite common computer used in schools, also use this exact model of CPU, so they can scavenge a MO5 to get this CPU.
I see that I'm not the only one who was sitting here in front of the screen thinking "why don't you put them in the Retro Chip Tester?!?" Hope mine will be finished soon (and working then...)
Noticed that when you were seeing the double-OK failure you only had the single ROM chip installed, but when you were seeing the full-screen failures you had the extended ROM as well. More likely it's a mix-and-match timing issue like others have commented, but I wondered if the reason you weren't seeing the partially-working OK failure is because the extended ROM demanded more from RAM and didn't work at all with the failure.
Could it possibly be electrostatic discharge hitting one of the address lines and knocking the ram out. I know that having that motherboard near a television unit might be exposing it to large amounts of ESD. Could also explain why the CPU was fried. I mean that high voltage CRT is only inches away from a motherboard and I know how much those kinds of black and white TVs cracked and popped when you turn them off or on a lot of the time
Not a computer, but a DPS video timebase corrector, I inadvertently opened a solder jumper that provided the clock signal to a bank of Ram, but after one turn on realized my mistake, reconnected jumper, corrupted image on next turn on, and found out that nearly all the ram chips in that bank were now bad. So maybe something like this happed to the CoCo, a bad clock signal?
One guess I have is that this failure was caused by something plugged into the cartridge slot. I don't see any buffer chips or anything else to protect the rest of the system from it and 12v is right next to some CPU signal lines. My other guess would be one of the multiple power rails feeding the RAM chips went way out of range at some point and damaged them.
I looked up the mc68908 datasheet (as I could not find one for 8040517 if my life depended on it) which is what you probably should have done the moment you hit the memory pin with the 6.15V level. In the datasheet the Absolute Maximum Ratings for VDD is -0.3 - 6.0 Volts (-1 - 6V for a standard 4164 memory chip). To me that indicates 6 volts plus a protection diode drop. That got my alarm bells ringing. If that is the case, I am not entirely surprised that none of the RAM chips survived. But good show, that you ultimately chose to brute force the issue.
I used a logic pulser with a logic probe a lot in those days. It identifies stuck pins to ground or Vcc. Today, I use a thermal camera and a scope most of the time, I can see which chips are acting up just by their temperature. My understanding was that Radio Shack built the CoCo from Motorola's application notes and never paid Motorola for any engineering work, since everything was off the shelf parts! So I don't know what this SAM chip is all about. Also, you never showed the access time for the original bad RAM chips. They could've been replaced by the previous owner trying to fix it with slower RAM! The only time I've seen all 8 chips go bad, was due to power surge.
CoCo repack, an odd idea? I thought the main feature of the CoCo was colour. kind of lost here. I vaguely did something similar, in that I put a model 1 into a model 4 case, however I didn't trash a M4 to get the case, I wandered down to my local Tandy store and ordered it as a spare part.
On my first job. I had a great dram tester. Put chip in. Turn knob to match speed or above. Press button. Green for good. Red for bad. 99% reliable tester.
This video was a great one! I'd love to know if the retro chip tester finds those RAM chips are actually fully dead. Or perhaps the CPU barfing out like over 6 volts fried all the RAM chips? Although the CPU output 6 volts seems really weird!?!?! Like what failure mode turns a CPU into a step-up converter? I mean, obviously not for real, but still... weird! Looking forward to part 3!
Yeah, the entire set being dead combined with the burning-hot SAM that he just kind of glossed over makes me worry that some kind of overvoltage going on and he's going to just fry the new set if he keeps using it.
Came here to post this. I’d love to see if the retro chip tester pro can find the bad chips.
Very odd, so where was that 6V coming from?, that would have been my quest as soon as I saw it.
I was a bit surprised he didn't jump right to the chip tester, when he suspected bad chips.
@@Meshamu I'm guessing maybe he doesn't want to use it because he wants to show a diagnostic process without the use of custom tools? It certainly would be more satisfying to see them checked on a tester.
When you're pressing enter, not only does it appear to crash, but you get the S in the top left AND a relay click, so it's not crashing, but it's somehow jumping to execute the cassette load function which puts an S in the top left of the screen and turns on the cassette motor (hence the relay click trying to allow current to turn on the cassette motor. Interesting...
I'm 55 and started out with the Radio Shack TRS computers when they first came out !
Fun video, brings back great memories !
One thing ive learned in troubleshooting, be it computers or cars, is that the first thing you dismiss as "it cant be that" is the first thing you should test. Never rely on your gut feeling, always test the "known" goods before go swapping parts around. I had the exact same type of issue you had here with an old VW golf, where the first thing I dismissed because they looked new (spark plug wires), ended up being the problem, but i didnt realize it until i had tested and swapped every other possible part first, wasting a bunch of time and money. All because I assumed that i knew that one part "couldn't be bad". Its an easy mental trap to get into. So now when I think "It cant be that", I test it to be sure.
I've mostly had the opposite problem: replacing parts I thought had to be bad and it ended up not solving the problem. I had a car that got to where it sometimes wouldn't turn the starter over when I turned the key to start, and eventually never would start. There was no click or anything. I spent a lot of money to replace the starter solenoid, the ignition switch, and the neutral safety switch assuming that had to be the problem after the last attempt failed. I finally got a wiring diagram and went through checking every connector in the starting circuit. It ended up being a burnt terminal in the connector near the transmission, one connector away from the neutral safety switch. It cost me about a dollar for a box of terminals.
My first idea was that +6V overshoot was an indicator of a bad voltage regulator or something that could kill all the RAM chips. But then my hunch is that the RAM chips you are putting in are different from the ones that were already in and somehow they are incompatible. If you put all of them back the machine would work as before. Probably has something to do with the timing or some similar properties that can be different on a chip. While the C64 can accept different timing, maybe the COCO is more sensitive in that way.
it was cpu overvoltage sam and sam sent off errors and destroyed all ram at once it unbelievable see that machine does that
i never own one but i did watch other video of someone fixing coco 2
Yeah, that 6V overshoot bothers me. You're not getting that from a chip that only has +5V as it's power. Something is leaking power onto the buss. It's possible that the CPU isn't bad, but struggled to overcome the excess voltage/current.
@@russellhltn1396 I agree that 6V seems odd, but it's not that much more than 5V. My guess would have been higher than spec VCC or something else also driving the bus.
@@jnharton TTL requires VCC be 5.0V +/- 0.25V. 6V would be way off and is a half volt above the absolute maximum input voltage. I'd think that would show on a power supply check. I'm thinking it's a defective chip that has access to a higher voltage.
Two other pieces of info you may find helpful in the future: First, inverted @'s on the display very often indicate problems with the DRAM. When you originally fired the board up, those were present amongst the other garbage on screen. It's a good indicator to look there first. If you'd have known, you could have possibly saved yourself some time. Second, you can find more listings for the SAM chip if you look it up under the 74LS783 identifier rather than SAM, Synchronous Address Multiplexer or MC6883.
That's what I did for the SAM -- I saw "SAM" prices were outrageous but the 74LS783 prices were pretty cheap. I guess the Chinese sellers hadn't caught onto that. :-) As for the @ symbols, I'll need to see how it acts with the ROM removed now the RAM is working.
@@adriansdigitalbasement some years back, early 2000s, i got a few of them, cant remember under which number, for 50p each from a seller here in the UK, doubt they have them at that price now, or at all 😉
just had a look on their ebay page, not listed now ...
theres some of the LS type listed from china at not too bad price £6something, but postage to UK £44 😲 sacre merde!
@@andygozzo72 Cricklewood had them at last look for £12
33:30 It looks to me like the 6th RAM chip on the board is now replaced by the original 5th RAM chip on the desk. And right after 34:24 the 6th, "untested" RAM chip on the desk gets an X mark. Probably doesn't matter but I had to get this out of my system 😉
Yeah there is a bit of a continuity issue there -- I noticed it during editing :-) In the end, all of the chips were bad but yeah eagle eyed folks will notice it!
28:38 - What if these chips were "bad" from the factory. They say 64k but maybe they are broken 64k chips but the first 16k works. So they popped them in and sold this unit as a 16k system. Someone else before you came along, saw they were 64k chips, put the jumper to 64k, and it didn't work. Have you tried setting it to 16k with those "bad" chips in there? I've heard stories of systems (probably from you!) being sold with bigger ram chips that were half dead, but worked in the system because the system didn't use the bad parts of the RAM.
The Wikipedia article does there were 32k CoCo-1 models using eight "half-bad" 64kbit chips so that might be what's going on here, it also seems to say there was no 64k variant of the CoCo-1 which further suggests this might be the case. It could even be that the HIGH/LOW jumper is set wrongly so it uses the faulty part of the chips instead of the working part. Another computer that uses faulty 64k chips is the 48K Spectrum.
Good thinking 99.
@@Torbjorn.Lindgren the 'faulty' 64k ones in the spectrum werent sold under the original 64k type number but marked with their own number plus a suffix indicating whether row or column 'dropout' and which 'polarity' 1 or 0 , and not only used in spectrums but likely many other devices, certainly some non sinclair made external 32k rampacks used them, i have 2,
I think the ZX Spectrum did that to cut on costs, but the jumpers needed to be set for each chip to choose which bank was the good one...?
@@BrainSlugs83 yep, 4 possible combinations, and 4 types of chip from 2 manufacturers , one had a dud column, one a dud row, and of those it could be a high or low dud, good thing is they can be replaced drop in with a 'good' 4164 without worrying about links if its a 128 cycle refresh 4164 type, if a 256 cycle type, it may depend on what the links are set to,
Might be an idea to run them thru your chip tester
Hi Adrian, just a thought, did you check the speed rating of the DRAM chips you removed in case someone had fitted a complete set of slow chips? Might explain why it sort of worked when all of them were fitted together, but didn't when one was operating at a different speed to the rest. As someone else suggested a RAM tester would confirm they are indeed faulty. Jeremy
I was curious if he was going to put all that bad ram in the retro chip tester.
Yeah, I was thinking exactly the same thing, that the RAM might actually be okay, but just the wrong speed.
Wasn't one of the address lines on the CPU high at 6 V? Possible overvoltage on one address line damaging CPU and all DRAM chips?
@@0x8badf00d Well he only checked the +5 rail so anything could be happening on the other rails and appearing as weird "ram" issues...
Another piece of evidence for this is that the machine failed slightly differently (The "OK" text changed on the repeated portion) when he had only one known good RAM chip in, making it look like the good chip was actually bad. I suspect the same thing has happened in reverse, causing him to misidentify the original RAM chips as all bad, when something different was actually at play.
Occam's Razor. My bet is that the original Motorola 4164's were some of the "half-bad" RAM chips Tandy used to be able to sell "32K" CoCo's. Each one you put back in with the good MT & TI RAM introduced errors because they were known to have one or more errors in the top or bottom half of their address space. If just one of those Motorola chips' problem was with its address lines rather than the data storage or data lines, that could explain the addressing problems you were getting. I'd wager that if you put the Motorola chips back in and set the machine to 32K, you'd be able to get it to work. You might have to switch the jumper telling it to use the high or low bank, but I bet you'd get it to work. Whoever changed it to 64K probably didn't know about Tandy's little cost-cutting trick with the half-bad chips.
Sinclair did the same thing with the earlier batches of 48K ZX Spectrums, with the 32K "high memory" that wasn't in the 16K models being made up of these types of faulty chips. They weren't from Motorola, but I think it was NEC.
Other possibility, of course, is that some nozzle set the jumpers for 16K at some point and fried all the 64k chips with +12v. Then tried to go back to 64K, only to find that he'd killed his computer.
I had a CoCo 1 as a kid and I can confirm, messing with those jumpers can fry the computer! Out of idle curiosity after the CoCo 1 was no longer my "daily driver" machine, I did the opposite - flipped the jumper to 64K to see if I could get "extra RAM" I got that first screen with the inverted @ symbols, and even moving the jumper back, the CoCo was gone.
That is honestly the most likely explanation. I just don't see how else all of the chips could be bad.
As for setting the 16k machine to 64k, some people have said that could damage the 4116 DRAM chips. Interesting -- it's an experiment I have have to, although it would suck to sacrifice 4116 DRAM. I'll think about it.
very easy to fry 4116 chips, if their -5v isnt present, it'd likely kill them even if 5v was put on the 12v pin , as with 4164 pinout
@@irinotecanhcl well sadly, we now know it wasn't REALLY gone... Just some bad 4116's..
4164s would almost definitely die if set to 16k as if using 4116s as that would put 12v on the 5v pin...this was and may still be a problem with some ebay 4116s that were actually skimmed and relabelled 4164s! several people had issues with these '4116' burning up so i got a few and done tests on them, tested perfectly ok as 4164,
back in the day… pre Y2K… we used to modify [mix and match] all sorts of different computers together to make our own custom systems! (especially laptops for rugidizing)… this episode took me back to the future!!!
Got it figured out at 30:02 min in... It was showing /+, which is OK minus 32... (ASCII of "O" is 79... minus 32 is 47, which is ASCII for "/"), same for the plus sign, the ascii of K minus 32 is the ASCII of the + sign....
Nice!
A guest! Unprecedented! Looking forward to it.
In the CoCo 1, the original expansion was only up to half the chip. Tandy put "bad" (but partially working) RAM in the systems and sold them as working machines. However, this machine is now wired as 64K, requiring fully working RAM. If the repack was in mid-work OR if the person was unaware of how Tandy saved money by using partially non-working RAM this could be the issue. You were working under the assumption that the setting for 64K meant that the RAM was all working as 64K - not necessarily a good assumption.
I loved this review/repair of a Coco1.
The Coco2 was my first computer, and I still have it. Everything Coco still has a warm place in my heart, so I loved that you were able to get this one working....and considering the outcome of the issue, it was similar to a problem I had after I did a 64K upgrade a few months ago.
Coco for ever.
Great diagnosis video of my starter computer. I remember having this exact issue... seems either one RAM chip goes or they all do. I wound up replacing all of them to fix a similar issue back in the day.
I'm just shocked it broke at all. I had several Cocos back in the day, and they were all 100% bulletproof. I abused the hell out of them as well; they spent years neglected in a 120° attic before I dusted them off. I wouldn't know what to do if one had broken on me.
Absolutely super video. One of the best retro repair videos I have seen.
Thank you for the effort that you put into finding that mischievous fault. I really like watching the process of narrowing-down a fault. I never would have guessed that all of the RAM had failed. I was thinking that perhaps a bit of wire fell into an edge connector socket and bridged two address lines. No.. I was wayy off!
Nice repair. Retro Chip Tester would help out testing those DRAMs. Also I think the BackBit Chip Tester has some Coco chips it can test too. I find the BacBit to be cool to keep om my workbench for it's size and the ability to test custom chips for Commodore, Atari, Amiga, Etc...
Another great video, thanks Adrian. As you stated, the coco is a very simple system. This makes repair easier than most, as the cpu will run regardless of the video system, and vice versa. The “crash” you were seeing looked like it jumped to a tape loading routine. If you type “cload” the screen clears and “S” it shown in the top left to show its searching the tape. Did we hear the relay click? It’s likely. You made some pretty accurate assumptions on the way the machine works. Fair play. Interestingly enough, the 6847 is used in the acorn atom with the same inverted video :)
(Posting before having watched all the video, so this may be covered at a later point.)
First two pins on the 6809E mentioned are "E in" and "Q in" - the clock input signals; there's two, as the 6809 uses a pair of overlapping signals for timing, with the "Q" signal leading the "E" signal by a a quarter-phase of the signal.
Was thinking the same thing.
yep, Q as in quadrature, the pair of lines give you four discrete states that are used to cycle the processor around the address,fetch, execute etc. states. To fully confirm they are functioning you should scope both simultaneously and verify the phase difference. With this technique you need to quadruple the apparent frequency (800 kHz?) to give you a comparison with say, the Z80 cpu which uses a simple single phase clock, so 3.2MHz versus 4MHz typically used in Z80 of the period.
I did a lot with 6809 systems (running "mini UNIX" Microware OS-9 operating system) in the 1980s.
The Hitachi HD6809 (or the faster HD63B09 or HD63C09) were our preferred choice. Better than the original. I still have the development system (running the same CPU and OS) in my basement. It could be a challenge to make it boot again, mostly because of the floppy drives.
The Hitachi CPU even has some additional features, like added/extended registers and enhanced multiply and division operations. But to produce "mainstream" 6809 code, I did never use these additional features.
I always look forward to your videos. You are to vintage computer repair as Bob Ross was to art.
Agreed. I have often fallen asleep while watching Adrian on my TV from my recliner. Not because he is boring or uninteresting (far from it) but because he is so soothing with his commentary. I think I have sleep learned entire Apple repair series and had to go back and rewatch while awake. 😆
@@michaelkramer5199 I know exactly what you mean.
I wouldn't be surprised if Adrian had a cult following of students and night shift workers.
It's taken me this long to realise that putting a *Colour* Computer motherboard and a black-and-white TV in the same case isn't as big a win as I initially thought.
The SAM (System Address Multiplexer) chip was also used in the Dragon 32 & 64 computers (they were a Welsh clone of the TRS80 Co-Co).
20:51 - I'm guessing A7 is stuck low (assuming 4 lines of 32 characters each before it repeats)
Back in the late 70's, I worked for a guy running a business out of his basement. We made cables like the one you had trouble with (a standard as everyone knows). I can't recall if there was a jig or manufacturer tool to install the ends, or we used a vise. I believe we used a purpose-built tool. Digitize Inc is still in business today...
There is a tool designed to clamp the contacts securely and accurately, but a bench vise is almost as good.
I have a hand tool for installing those - used it just a few weeks ago for making a cable to connect a Model 1 to the expansion unit. It has a reversible insert, one side for pin connectors like in the video and the other for edge card connectors. No idea where I got it - had it for probably 35 years. Most likely Radio Shack had them since that's about the only place I got stuff at the time.
@@EddieSheffield I picked up one of those from eBay a year or three ago. I use it all the time to make custom-length IDE, floppy, SCSI cables.
Now, I bought a dead Oric-1 some years ago that had more than a smattering of dead chips. It turned out that what had happened was that due to its unusual supply regulation arrangement it had effectively been fed the full 9V and fried over half its ICs. From that point on I learned to check the supply rails *FIRST*.
Awesome! More 6809 Coco/Dragon videos please!
I remember a tech channel that introduced me to a fantastic chip tester, which I bought and assembled. It is a lot easier that flipping chips… wait.. it was your channel I saw that on! 😊 Why didn’t you use that?
He's even asked himself in a previous fix video why he didn't use it sooner!
I predicted multiple dram chips would be bad, but never imagined 8 of them would be! It would be interesting to understand how the machine was able to function at all, like mapped through the bad chips.. amazing...
You probably won't believe me, but I actually did predict ALL of the behavior almost perfectly.
That ALL of the RAM chips are dead.
That replacing one chip will cause the second OK to become garbled.
etc
Here is my reasoning:
The text is repeating exactly once thus one address line is bad.
32 chars per line, 4 lines, it is probably A7, although that depends on how it is wired.
The CPU outputs 6.15V on at least one of the lines.
Not sure how it manages to do that but it is BAD not only for the CPU but for other chips as well.
And it probably fried all of the RAM at once since the address goes directly to the RAM - only select logic goes through SAM.
The DRAM of the time was particularly fussy and fragile, especially to voltages, and it didn't have many protections, so it makes sense that it died first.
So it we are seeing the perfect copy of text that could only mean that ALL of the bits have exactly the same issue, so ALL of the memory chips have A7 gates completely fried and internally stuck high.
When ALL of the bits are high, the code and data gets "overriden" by another portion of said code/data, so it we're lucky it might somewhat run and display correctly albeit duplicated.
When we replace one chip, one bit is corrected and NOT duplicating thus that one bit is missing from the duplicated text if we are also lucky to avoid code corruption.
If we replace all but one, the code or data gets corrupted in that one bit - we would see some characters each 4 lines are off by 1/2/4/8... ASCII codepoints, i.e. A becomes B or D or , etc. But the code gets corrupted so much that it doesn't run (that's pure luck it didn't happen with single bit correct).
At some point of the video I was experiencing some sort of...umm... epiphany? looking at somewhat confused Adrian and knowing what will happen next.
It was kinda bizarre sensation, like I actually could see into the future.
Anyway since only a single address line is bad, the computer can probably run perfectly if the dead address line is switched with the highest one. The computer will just have only half of the RAM capacity (32K?) available to it (unless its ROM just expects all 64K to be present)
@@jwhite5008 That makes sense with regards to the text on the screen. I wouldn't say I predicted all of them were bad, but it did seem like something on the address bus went bad (likely the CPU) and took out one of the address lines, looked like A7 or A8 since the text repeats after 128 or 256 characters at a glance, in at least some of the other chips.
I still don't understand why the thing crashes with any one of the bad chips in it but runs with all eight bad chips though. That's really weird.
It actually does makes perfect sense.
The chips probably received a voltage much higher than what they were designed for (something like 5.5V max) on one of the address lines - like you say most likely A7 or A8.
Some of their row/column select logic gates inside the RAM chips were fried and now act as if the line is always high (or always low - doesn't matter) irregardless of actual input.
The reason why the second OK was garbled when one good chip is inserted is because one chip = one bit is now NOT faulty, the duplication "omits" one bit from each character, i.e. that one bit displays what it was supposed to rather than duplicated data. So the characters on 4th row - which include that one working, non-duplicated bit - get "garbled" - but actually the replaced bit is fine - it is all of the other bits that are wrong.
The reason why the system sort-of-partially-works with all bad RAM, and with all good RAM, but fails if one of the chips is bad is that when all of the RAM is bad, the data is written in wrong spots, but the bytes of data themselves are not corrupted - only misplaced. Thus it overwrites the other data that should be there. However, if you write the data and read it before it happens to be overwritten that way you do read valid data.
For example, if there is a bunch of text strings in the memory, each terminated by zero-character, in the "misplaced RAM scenario" the computer will read wrong strings but still terminated, whereas in "partially misplaced (garbled) RAM scenario" it will not find that terminator (because the terminators of two text blocks are unlikely to align with each other) so it keeps reading, overflows the buffer and tramples some random memory.
Similarly, although most code is executed from ROM, if some code is copied to RAM, it is copied correctly, but the execution rolls over and skips portions of the code. When one bit is wrong the code is completely random and probably hangs.
While it may seem vert bad, (and it is), the CoCo is much much less complex than PCs nowadays. There is a chance that the values that were written are read before they are replaced. So by that chance the computer is somehow able to survive enough to display OK prompt and let you type stuff.
Now when some but not all of the chips are replaced with good ones, the bit-by-bir corruption occurs - similar to what happened with text when one chip was replaced. Some bits are read from one place while others are read from another.
I believe the latter this is much worse than when all bits of the data is correct but misplaced because it tends to be more random-looking values.
Of course there is a lot of luck involved - it easily could be that the machine wouldn't boot with either or all of the chips being bad, but that is not what happened. And it was probably a very lucky coincidence that it was able to boot with only one good chip.
The instability with all bad chips is probably because they either sustained too much damage overall, or not enough to fully destroy all of the A7 gates in every chip.
@@jwhite5008 that all sounds very logical, and likey is the case. To see it happening in real time, and actually still function, albeit partially, is very interesting. Makes me wonder, if any one pair of those chips were placed out of the order they originally fried in, would it still function? I'm thinking not but may be wrong... I.e. chip one in socket two and vice versa, etc.
@@Quickened1 yes - all 8 RAM chips had exact same issue. All of them had same address bit stuck (probably A7) - which is why they displayed uncorrupted data at wrong addresses- this is what initially surprised Adrian.
Your scope should have runt pulse triggering, it really makes sense to make use of it here. Also in case no DRAM chips are at hand, switching them around and observing the pattern change can help too, especially if you have an idea about how far they are apart in terms of adressing
power short on data lines ?
for instance it lived in a botched together setup say at some point a power wire made contact with the data lines , straight through all ram and cpu.
I know he always says "MT Ram" but that company is still around today, MT was just the initials for Micron Technology. In the 1990s they made some of the highest quality RAM sold through their direct to consumer brand Crucial Technology.
I usually look at the datasheet for the chips, since some designers label the pins with "what they do in their circuit," and not what they actually are. Besides, they will be more legible.
7:40 that is cool thing! And a very nice white baby monitor/TV )))
Those connectors like on the keyboard cable are called IDC connectors. Officially, IDC means Insulation Displacement Connector (as in you don't have to strip the wires) but I like Ben Heck's name better: "It Didn't Connect."
I took basic programming in college. I have written many programs. Including a klindergarden one. My first daughter used my program and was moved to fiirst grade, Also in 1st grade she did great. I would love to program in basic again. I am currently programing in Turbo Pascal also.
The most flustered I have seen Adrian
Amazing! Who could have expected 8 out of 8 faulty modules… and a kind-of-working system ONLY when all of them were plugged in! 😂 what are the chances! Great video as usual!
Wow got two spammer replies in an hour 😂
@@tony359 I just report them for spam as a matter of course.
Dave always says: "Check voltages first!"😂
Rockby has their specials out. They're selling 68B09EP processor, which may interest those that want to build or play around with this wonderful old 8-bit processor. Also selling 8154 chips which work great with the SC/MP CPU, or an old 6502, 6800 or similar processor if you want a little bit of RAM and IO for a small project.
34:45 - I'm guessing that Vcc got up to 6 or so volts, frying the CPU and all 8 DRAM chips!
possible? did he check it??
I own a Minipro model TL866A programmer. It tests RAM chips and 74xx series logic chips all day long like nobody's business. It reads PROMS so you can check them, too! I've got the diagnostic card with full harness kit but I now need a dead test cart as well. And a good OEM original 1541 test floppy.
You better check the TL866A manual. It does not check any DRAM, nor SRAMs < 16kbit, nor 1200 logic ICs, so it is quite useless to test hardware from the 1970s or 1980s. It does also not read 1702 or 2708 EPROMs.
@@herbertschuster9088 Wow. I musta been lucky, so far! It has helped my business repair an Amiga 2000, a Commodore 128, two Commodore VIC20's and two Commodore 64's, two medical devices, A plastic injection moulding machine control, a plastic wrapping machine control, A flash chiller controller board, a trash compactor controller, a motorcycle ride height controller board and 3 industrial machines.
I agree that it will leave you wanting with MANY older or proprietary IC's. Like the National Semiconductor ones on a machine I am rebuilding right now. But it is FAR from completely useless and has shaved DAYS off of many reverse-engineer style no-schematics diagnostic jobs.Oh, to find the source of that machine's crazy behavior, faults and errors! Three axis servo motor controllers with encoders, an HMI, translation cards, drives, main controller...impatient customer with "status check" texts and calls...I'll use ANY help I can get!
That is so satisfying. I like the inverse video. I think I will fit that switch on my Dragon 32.
It's great watching you work on a system that you're less familiar with and working it out from the basics - really helps me to learn how to troubleshoot. Especially find you talking through how a scope trace should/shouldn't look very useful!
(Not that I don't enjoy the c64 stuff!! I do!)
I suspect the cause of the bad RAM and CPU could have possibly been a power surge. It's the only thing that makes sense to me.
I was thinking the same; I'd assume that it's unlikely that a bunch of components just suddenly fails. And that makes me worry if the reason for the surge is still there, and could destroy more components. Maybe it would be a good idea to check if there are any loose threads from some wires hanging near other wires in the power supply. Though, the cause might have been temporary, like at some point something conductive had fallen in the power supply to briefly bypass something. But still, if that's the case, that something could be still somewhere there inside, and could do the damage again. (I don't remember if in an earlier video Adrian already cleaned the powersupply -- if that's the case, the culprit might have got removed at that point.)
I too was thinking that it was a power spike that killed the RAM and CPU. Maybe they were more sensitive to the spike and that was the only reason they were killed or maybe the other chips had more passives in line that dampened the spike?
I suspect a bad power brick. Excellent fun video!
Hi Adrian, a tip I learned over the years:
Don’t use Deoxid on Potentiometers.
The stuff is so aggressive that it will be the cause of trouble itself after some time.
I use an inert cleaning spray instead.
If you do use Deoxid, then always clean afterwards with a few good sprays of the inert cleaner to get the Deoxid out.
Switches can also be lubricated with real contact lubricant spray after cleaning.
Is a good contact cleaner (I use wd contact cleaner) ok?
@@DavideNastri Contact cleaner is mostly the same as Deoxid, so good for switches. Also here make sure to clean afterwards with a a safe inert cleaner.
I use Kontakt 60 as Deoxid and Kontakt WL as inert cleaner.
Also I use Kontakt 2000 as lubricant for switches.
Pretty expensive, but it woks and lasts.
I heard there are 2 types of Deoxit, the liquid one in a bottle is 100% concentration and can cause problems if not thinned out, and the spray can that is 10% that is usually fine.
Yes, xraytonyb is EMPHATIC about this!
For the several dozen commenters claiming the 6V killed the RAM, there was no 6V. Adrian measured the 5V rail, it was spot on. The 6V came from the ringing by the mile long ground path, because he put the ground clip on that IC pin which is grounded through a long, thin wire. So it was a probing issue, the 6V wasn't really there.
eight for eight bad ram chips ... incredible. I really wanna see those run thru the chip tester!
I'm in love with the CoCo!
The first thing you should do is test ALL chip voltages at their supply pins with a DC volt meter, AC volt meter AND a scope to check for value, noise and ripple. For instance, bad caps can maybe cause high ripple which can cause overvoltage/undervolatge (e.g. voltage is 4V-6V with 2V ripple and DVM reads 5V) can maybe cause DRAM to be unstable? Drawing more power (e.g. fitting two ROMS or different chips) can cause more ripple and thus change the symptoms. Maybe those RAM chips were OK? The 6V on one address line didnt just come from nowhere!
I wonder if the specs of the "good" and "bad" RAM chips are the same. As they are socketed, the originals could. have been swapped out for a different spec'ed devices. Are the access times the same? It would be interesting to compare their respective data sheets. They may even have different voltage range specs. As you say, strange they were all "bad", which would point to a spec difference.
It reminds me Timex 2048 that I repaired. It (probably) received a hot plug of cartridge. CPU, ROM and bus buffers were fried. Also SCLD got slight hit.
Here are no buffers so CPU and memory could got damaged this way.
Hi,
When you checked the address signel, we see clearly at 27:01 that there is a glitch in the address. If you mesure it again whith the new RAM chip, it may be different
I don't know your oscilloscope but it looks like when you reset the system, the scope just doesn't refresh the screen because it's not triggered. If you put it the scope on auto-trigger, you'll probably see a horizontal line on any data/address line while you reset the CoCo.
The RAM problem looks to me like there is some sort of subtle difference between the old and new chips, not necessarily defective chips. Maybe different RAS/CAS timings that cause bus contention?
Anyway, great video! My guess was that the problem was something with the 6847. Great repair!
Maybe something in that messy case momentarily dumped a higher voltage into the RAM through an address line, or something. It's conceivable that the other chip that handles the multiplexing is just robust enough to have handled the hit and kept chugging along, while the fragile RAM chips got fried.
One final thing and I'll shut up. To my knowledge (and CoCopedia agrees) Tandy/Radio Shack only ever sold a 64K CoCo 1 very late in the 1's lifecycle when they started selling a white case model CoCo/TDP 100 and it'd be a shame to think that someone caniballized one of those to make this repack. There are just so few of them around. I really hope this was a 4/16/32K gray model that was upgraded to 64K.
Well the original 64kl RAM chips are Radia-Shack branded ICs, and this motherboard only has 16k/64k jumpers on it, so I'm assuming this was a 64k board. Don't worry though, I have some info on this machine now and it was assembled with warranty repair spare parts. Hopefully we'll have some more info in the next part.
hi there was a 8 track 1 inch reel to reel someone have done kind of the same made 2 decks out of one
i have seen recording decks that have been jazz up
Interesting video, surprised by what was wrong, 8 ram chips. It would be nice to see it put back together again.
At about 7:15 you probe the crystal input and note the frequency is off and blame the scope. Note that adding the scope load of 10 pF to that circuit WILL change the frequency anyway. I wonder if that overvoltage you saw on the CPU somehow caused the DRAM failure?
It looked like you had two address lines shorted together, with the repeating text further down the screen( down memory). I'm not surprised you couldn't find it by looking at the address lines unless you were looking for two identical signals on the address lines. I am surprised you couldn't duplicate the fault by placing the memory back in the same socket. I'm wondering if that memory is highly oxidized, because some of it was working well enough to show in screen memory.
Good thing that when whatever happened, it had the CPU and those 8 RAM chips blow first to protect the other electronics... ;-)
It's not only amazing that it was a CPU and all the RAM, but that whatever was big enough to do that didn't also blow caps/resistors/etc too... Wow!
You need to check the power supply to the DRAM as they have multiple voltages feeding them. I can't remember if they have a -5 or +12 rail but it could be off and out of spec for the Motorola chips but the MT is OK with it. Also that tarnishing (not rust) from pins can be removed with a pencil eraser. DON'T sand them as you could remove the plating.
Another great video, thank you very much for all the work.
Hi Adrian, did you try the retro chip tester on these "bad" RAM? The Motorola RAM may have timing issue that need to work with the same brand/spec, seems mixed with MT RAM may cause problem.
I would have liked to see him test the chips on the chip tester too! That would have been neat.
Probably worth re-capping the board at some point. Particularly around the DRAM. I'd just do the electrolytics for a start, they're usually relatively easy.
I really want that little philips crt so bad. Sucks that that sort of thing is not sold anymore.
29:21 - Question, (may sound stupid but correct me if Im wrong) why you didnt take out all chips, and check them with multimeter in diode mode? While out of circuit, Red@GND leg, black on each leg, and make a chart, then go to next chip. Is this method OK for this type of IC circuit? If all chips are the same, all measurements should be the same.
The Ram chips went bad. At some point there was a short that killed one or more of the address lines. and it would kill all the lines on each chip that would lean toward a power spike that fried all the ram..
FWVLIW: A wild guess:
Apart from the usual "What's wrong with the PSU?" I suspect some kind of 'jolt' was fired across the bus and took out the CPU along with the RAM.
What 'jolt'?
I wonder if there's an issue with the cartridge port extender / the floppy controller / plugged in backwards or it just fell out at some point powered..?
Regardless I'd check all of the above before plugging in - in particular I'd try powering up the floppy controller unconnected to a computer and check for shorts before using it. Not sure if just continuity testing it would be enough if one of the IC's got fried to a dead short across unexpected pins.
As an aside I wonder if this machine was assembled with 1/2 bad RAM and the jumpers set to something less than 64k because it worked. A previous owner saw the 64k IC's and changed the jumper(s) to 64k and 'broke' it... Maybe putting in the original RAM and setting to a lower configuration might 'fix' it again..?
I don't know this system at all so I don't know if that might hose something if the wrong RAM and jumper setting is used - watch out!
Btw, 6801/6803 had the multiply and fill 16-bit arithmetic...
I think a power surge may have been the cause, or it may have been hooked up to 220V. Another possibility is tyat therewas a cascade failure from the 6.9V that the CPU was putting out on the address (Or was it data?) line, damaging multiple ICs on the board.
I'd also be curious to see if, now that you've got it working, replacing the original 6809 or the Hitachi 6309 on the board worked any better.
The bad RAM should not make a difference -- that original CPU was dead, not progressing from reading the first byte from the ROM. The RAM isn't even needed for that to work as we should have at least been seeing it trying to execute code before crashing due to bad RAM.
I wonder if that over voltage issue on the bad CPU trashed the DRAM
Whoop whoop It's working!!! Coleco Adam!!
The cassette port; the "angle connector was replaced with a "straight connector" not the other way around. Ron W4BIN
I never knew the Radio Shack computers ever had anything but a Zilog processor. The 6800 processors are Motorola. My TRS80 Model III had a Z80 processor.
Your info about the need for this computer to work with a 6809E can be interesting for French hobyists who try to repair a Tandy CoCo because the Thomson MO5, a quite common computer used in schools, also use this exact model of CPU, so they can scavenge a MO5 to get this CPU.
I see that I'm not the only one who was sitting here in front of the screen thinking "why don't you put them in the Retro Chip Tester?!?" Hope mine will be finished soon (and working then...)
Noticed that when you were seeing the double-OK failure you only had the single ROM chip installed, but when you were seeing the full-screen failures you had the extended ROM as well. More likely it's a mix-and-match timing issue like others have commented, but I wondered if the reason you weren't seeing the partially-working OK failure is because the extended ROM demanded more from RAM and didn't work at all with the failure.
I think the bad RAM was working well enough to let the CPU limp along, but it would cause certain things (like the start-up message) to not work.
Could it possibly be electrostatic discharge hitting one of the address lines and knocking the ram out. I know that having that motherboard near a television unit might be exposing it to large amounts of ESD. Could also explain why the CPU was fried. I mean that high voltage CRT is only inches away from a motherboard and I know how much those kinds of black and white TVs cracked and popped when you turn them off or on a lot of the time
another great video Adrian really enjoyed it
Marginal ram can do some wacky stuff! Definitely would like to see what the MiniPro thinks of that ram.
Not a computer, but a DPS video timebase corrector, I inadvertently opened a solder jumper that provided the clock signal to a bank of Ram, but after one turn on realized my mistake, reconnected jumper, corrupted image on next turn on, and found out that nearly all the ram chips in that bank were now bad. So maybe something like this happed to the CoCo, a bad clock signal?
One guess I have is that this failure was caused by something plugged into the cartridge slot. I don't see any buffer chips or anything else to protect the rest of the system from it and 12v is right next to some CPU signal lines. My other guess would be one of the multiple power rails feeding the RAM chips went way out of range at some point and damaged them.
Nuking SAMs and CPUs on the CoCo is a hallmark of being careless with cartridge connected devices
Wonder if there was some manner of shielding between the motherboard and whatever points an arc can jump off the old TV electronics...
Hey Adrian, double ckeck and try testing the ram chips with the chip tester... wondering if the chip legs were flakey...
@adrian Check all chips in a ram tester. Maybe be the sinclair trick (all bad chips with half part working) or simply a timing issue.
I looked up the mc68908 datasheet (as I could not find one for 8040517 if my life depended on it) which is what you probably should have done the moment you hit the memory pin with the 6.15V level.
In the datasheet the Absolute Maximum Ratings for VDD is -0.3 - 6.0 Volts (-1 - 6V for a standard 4164 memory chip).
To me that indicates 6 volts plus a protection diode drop. That got my alarm bells ringing. If that is the case, I am not entirely surprised that none of the RAM chips survived.
But good show, that you ultimately chose to brute force the issue.
I used a logic pulser with a logic probe a lot in those days. It identifies stuck pins to ground or Vcc.
Today, I use a thermal camera and a scope most of the time, I can see which chips are acting up just by their temperature.
My understanding was that Radio Shack built the CoCo from Motorola's application notes and never paid Motorola for any engineering work, since everything was off the shelf parts!
So I don't know what this SAM chip is all about. Also, you never showed the access time for the original bad RAM chips. They could've been replaced by the previous owner trying to fix it with slower RAM!
The only time I've seen all 8 chips go bad, was due to power surge.
Actually the 6809 jumps to the address stored in $FFFE/$FFFF (not $FFFC/$FFFD).
I would be interested in seeing those bad ram chips tested in his Chip Tester Pro.
CoCo repack, an odd idea? I thought the main feature of the CoCo was colour. kind of lost here.
I vaguely did something similar, in that I put a model 1 into a model 4 case, however I didn't trash a M4 to get the case, I wandered down to my local Tandy store and ordered it as a spare part.
Tell me how to make one for myself, I have plenty of parts to use. Keep up the great videos, Big thumbs up.
What kind of boombox is it, likes like a RC-M70 for sure, with RX-5600 tweeters and soft touch buttons line in the Aiwa 990 ...
On my first job. I had a great dram tester. Put chip in. Turn knob to match speed or above. Press button. Green for good. Red for bad. 99% reliable tester.