Zel's Fleet Blog - BX, Jag XJ-S, Sinclair C5, Mercedes 208D & 230TE, AC Model 70.

[RichardW]

I put one of those, 20Mb, in an Amstrad PC 1512 in about 1990!!

[Zelandeth]

Here was this afternoon's task.





Every electrolytic cap on there changed.

...Then I had to change the mains filter caps back again. The new ones it turns out were too tall. Though those I was least worried about, they're Elna branded caps which I've never had an issue with. There was no evidence of leakage, and they test fine.

Despite the rear of the board looking horrendous at first glance, aside from a bit of surface tarnishing, everything looks fine.



So if everything works I'll re-paint the rear of the board where I've removed the solder mask. Knowing whether it works or not remains to be seen. It was going on 2200 by the time I got it fully reassembled, and I couldn't face the idea of having to traipse outside in the pouring rain to the garage to reset the RCD/Circuit breaker if there's an issue with ground leakage or something in the primary side goes pop. So testing that is a job for tomorrow. I still have my doubts, as I have a feeling that the electrolyte may have soaked into the board itself. We'll see.

A package did arrive in the late afternoon for the T5200 that I had forgotten was on the way.



Which contained a nice new old stock 80387DX-20. So no empty sockets here any more.



Which seems to work just fine.



I've never actually done a direct comparison with the same machine with and without a math co-processor before (as I've never actually fitted one before) so was curious to see what impact it has on the benchmark results.

So, NPU benchmark results...

Before: 100.6K Whetstones.
After: 1660.8K.



Yeah, that's quite a difference. Exactly what difference that would make in the real world I'm not sure. Just nice to know that the machine has pretty much all of the main factory options fitted now.

Webpage for the T1200 is about 95% done, I just need to finish the final summary and check I've got all the images properly linked etc. Wonder what the odds of me finding a math co processor for that are...

[CitroJim]

Magic I know from experience just what a difference a maths co-processor makes. Back in the late 80s when my brain used to work fairly well I wrote some HF Radio propagation prediction software in C. It was very maths heavy and displayed the results across a 24 hour period graphically.

On a standard 286 PC it ran very slowly. That same PC with a co-processor fitted and it truly flew!

Here's hoping your repaired PSU is good on test...

[xantia_v6]

I would be surprised if an IC was ever commercially available to fit the T1200 copro socket. See some discussion at
https://forum.vcfed.org/index.php?threa ... mhz.70514/

I recall that I tried to find the source of a compatible chip circa 1990 and drew a blank, although the NEC microprocessor catalogue may have given a part number and some "preview" data. I think that the CMOS version of the 8087 was obsolete before it could be put into production.

[Hell Razor5543]

There are a couple of things you might want to try (if you haven't already done so); shadowing the BIOS, and (provided the memory banks are occupied with the same spec memory) memory interleaving. A friend was dubious about the first one, but then we did a simple test; running a fractals image. When we shadowed the BIOS it better than halved the render times. Later, when we enabled memory interleaving, there was a similar improvement.

[Zelandeth]

There are a couple of things you might want to try (if you haven't already done so); shadowing the BIOS, and (provided the memory banks are occupied with the same spec memory) memory interleaving. A friend was dubious about the first one, but then we did a simple test; running a fractals image. When we shadowed the BIOS it better than halved the render times. Later, when we enabled memory interleaving, there was a similar improvement.

Sadly I don't think there's a huge amount of options there as the BIOS on these systems is very simplified - and I don't think there are any options for memory handling beyond selecting whether there is any expanded memory installed. I will double check though.

-- -- --

It turned out that on testing the T3200SXC power supply had survived its brush with death. Powering it on resulted in a machine which was powering up and clearly POSTing just fine. Sadly it hasn't (as I expected) sorted the display issue. The panel drive levels were clearly all way out of spec. You could just about make out the display inverted if you viewed it from just the right angle.

While I had it in bits I grabbed a photo of the innards while everything was out of the way.



I'll be annotating that somewhere down the line for use on the webpage for this system when I write it. Huge amount of the motherboard is usually buried with everything installed.



I did have a go at repairing the display, replacing every electrolytic capacitor in there. Which was precisely as much of a pain of a job as I expected it to be. I'm really not set up for dealing with tiny components like these.



Especially as this is one of those things that was never designed to be serviced. Several things (like that small PCB) were pressed together on assembly (AND soldered!) making it basically impossible to disassemble non destructively. I did what I could though.

Sadly it didn't work. We went from an unusably washed out display to no display. Can't say I was massively surprised. It was a bit of a long shot trying to repair this display, but it was worth a try. Especially with it being a very early TFT panel rather than a DSTN panel which would have been expected back then.

I'm not sure what signals are carried to the display panel, if it was actually just VGA analogue data it *might* be possible to retrofit another panel to get the machine back to fully functional, but with there being 27 conductors rather than the 15 on a normal VGA cable, even allowing for several power rails in there that seems a lot of extra conductors. That's a job for another day though.

Otherwise though the system seems to be back up and running, albeit on an external display.



A little while ago I mentioned the fact that the 80386SX CPU on this board comes with a fairly hefty performance penalty over the DX variant because it only has a 16-bit rather than 32-bit data bus. So a lot of clock cycles are essentially "lost" because it has to do bank switching to handle the full memory address range.

Now conveniently I benchmarked an 80386DX machine yesterday.



This also runs at 20MHz, so all things being equal you would expect the numbers to be identical.



Ouch...so yeah, you're taking roughly a 30% performance penalty with the SX rather than DX version of the CPU. I imagine that would be particularly apparent if you're doing memory intensive work.

Bit of a shame I couldn't get the display going, especially as this computer is incredibly clean - even more so than my modern laptop. However those are the breaks in this hobby sometimes. In fairness it's probably one of the lower value machines to me so I'm not going to lose too much sleep over it.

The condition of the power supply though has been a bit of a wake up call, and I will now be fully re-capping the PSUs in the rest of the mains powered Toshibas I think, irrespective of if I've had any prior issues with them (or whether I spotted anything during a recent visual inspection). That's something I'd only ever really had major issues with on the battery powered machines before - I'd rather spend a few hours for the peace of mind now than having to mop up the mess in the future.

[Zelandeth]

Had a couple of hours today where I was actually feeling vaguely human and the weather was looking decent. So I said screw it to the stuff around the house that needed doing and instead pulled TPA out of winter hibernation.



Started first try, didn't even smoke today. Think she was out briefly once in February, but certainly hasn't had a decent run since November or December. It was February when I fitted the new voltage regulator, and I've never done a test run since then.

Until today anyway.



Charging system looks to be behaving so much better now. Jumps up to 13-14V pretty much as soon as you're above 1500rpm or thereabouts, and stays there rock steady. To get really meaningful charging before I basically had to be going fast enough to have the CVT maxed out. Now if I'm moving I have good system voltage.

Hard to say as I wasn't out for a massively long run and it's been a while since I had the car out, but it honestly felt like she was pulling better at high revs, just felt as though she was holding 60 easier than usual. I guess it's possible that the generator was bleeding off some power at the top end (I was seeing nearly 16V on the voltmeter at that sort of speed), which isn't now happening as it's being properly regulated.

Only thing I noted that wasn't too happy after the hibernation was the handbrake felt really gummy. It's definitely at the front end of things rather than rear of the cantilever mechanism under the floor. I can't quite tell if it's the cable itself that's sticking or the lever itself. I've lubed both up and it seems better, hoping it will free off fully with a bit more use. While I had the grease out I also gave the door runners a thorough greasing...I hadn't realised how bad it had got. First time I opened it afterwards I just about threw the door off the front of the car! You can slide it with one finger now. I really do need to properly sort the rollers on the door at some point...that's not massively high on the priority list though to be honest. Priority one right now of course is getting the lovely newly repaired hub fitted.

We've got a guest staying for the new couple of weeks though, so not expecting to get a chance to do much of anything for a while. I know they're going to want to see TPA though, so seemed all the more reason to pull her out of hibernation.



I always forget how much this little car makes me smile. I'd had a pretty rough morning and in the last hour before I wandered into the garage was quite shaken up by a run in with a very aggressive driver in a BMW threatening me at knife point for having the audacity to be surprised to find them charging at 40mph the wrong way through a one way system. Felt a lot better after getting TPA out for a run on a nice sunny afternoon when the roads were actually pretty quiet though.

I had the van out for an hour or so a couple of days back, which I think was the first time she's been more than a mile from base this year. Do want to give her a really good run though - takes quite a while to get that old lump fully up to temperature! Need to consult my records but I'm pretty sure she's due a service too.

[CitroJim]

Excellent and pleased you're on the mend Zel A few joyrides in the Invacar will do you the world of good and accelerate your full recovery no end

[Hell Razor5543]

I hope that you are OK after your 'discussion' with that BMW driver? I don't suppose you got the reg so you can report him to the Police? You never know, they might have caught it on camera (depending on where it happened).

[CitroJim]

I hope that you are OK after your 'discussion' with that BMW driver? I don't suppose you got the reg so you can report him to the Police? You never know, they might have caught it on camera (depending on where it happened).

I hope so too... Complete psychopaths like that need to be subject to the full force of the law and then jailed for an indefinite period... It's worrying that such individuals are loose on our streets.

But then, having said that. I met one just over three years ago in an Audi TT...

[Zelandeth]

I hope that you are OK after your 'discussion' with that BMW driver? I don't suppose you got the reg so you can report him to the Police? You never know, they might have caught it on camera (depending on where it happened).

Not interested. As they didn't actually try to gain entry to my vehicle nor take an actual swing at me they class it as antisocial behaviour and nothing more. The whole carrying an offensive weapon thing they won't follow up without video or photographic evidence. Provided by me. They won't approach the operator of the (CCTV controlled...) car park in question unless something had actually gone down.

Likewise not interested in the fact that the plate on the car comes back as a VW Passat rather than a BMW 3-series...it's taxed, MOTd and insured so won't ping any ANPR cameras...

Remember, this is Thames Valley Police we're talking about. My experience with them to date has shown them to generally be second only to Milton Keynes Council in terms of being utterly useless (at best, downright dangerous at the other end of the scale).

Moving on to more positive matters...

Some of you may recall that a good few months back among a whole load of computers I was gifted by a near lifelong friend included as a star attraction a very tidy Apple II Plus. Which while sound, had some major stability issues. Up till a few days ago I had been struggling a little to figure out how to approach it. My hunch was that we had some duff memory, but I didn't really know how to confirm that. Until a couple of days back when I stumbled across a video on YouTube by Adrian's Digital Basement where he mentioned a diagnostic utility called Cillin II. He was mainly using it for disk mechanical diagnostics in the video, but mentioned in passing the other features it packed in - including a memory test feature. Cue a frantic scramble on my part to find a disk image - which thankfully was readily available. Which I could use easily thanks to the Floppy Emu. Finally I was set up for some actual diagnostics to try to get this old crate working properly again probably for the first time since the 90s.



Hey look, we have memory errors!



These were all clustered between addresses $8212 and $BDDE. Which is quite helpful as that narrows it down to a handful of chips on the board, on the upper row of the group in the white rectangle below.



Cue a bit of old fashioned deductive reasoning and experimentation.

Hey would you look at that...



A clean memory test - after running for about 30 minutes. Which is a good 15 minutes longer than I've had it running even a BASIC prompt without falling over to date.

Additional good news...



That's another clean memory test for the 16K of memory on the language card - which I'd previously not been able to even get the machine to boot with installed.

The culprits?



Two dead memory chips with a stuck bit. One from the motherboard, one on the language card.

This was quite exciting news as it meant I could actually start doing some proper testing of the hardware as it was for the first time running in a stable fashion - and had the software to do it.

Disk drives are an obvious target, I have six of them. Four Apple Disk IIs, one Super IV and one Cumana (as usually seen attached to BBC Micros).

The Cillin diagnostic suite actually has some pretty advanced disk testing facilities. Step one is to format a disk for use...



Which went without incident.

Probably the most useful test I think is one which is labelled as the random read/write test. This randomly seeks and reads/writes to every sector on the disk - a good workout for the mechanism which may have been dormant for a couple of decades, and a good way to weed out any mechanical issues.



I'll take that! Each "P" in that image indicates a "pass" result for a successfully written and read back sector on the disk. I hadn't even so much as cleaned or looked at that drive or the disk I used there, so I'm well happy with that. Something I hadn't realised until just now is that aside from the initial head bang when it homes itself is how quiet these drives are. The head seek is nearly silent, the spindle motor is more noticeable than the head positioning motor.

The other five will need to be tested over the coming days.

At least I do actually have a functioning system I can do testing and experimentation with now.



Having an easy way to test memory allowed me to ascertain that I do have a full board worth of functioning memory for my "parts" machine too after scavenging a couple of 4116 chips off a spare language card. A dead memory chip on there (Mostek chip, no surprises there then) right at the bottom of the stack would definitely explain why it wouldn't boot when I last tested it. So this thing may well also live to fight another day yet.



This machine is a compete Bitsa, assembled entirely from scrap and found parts back when these machines were worthless, but it would be really nice to actually get it up and running too. Aside from seeing if it will now boot I'll not likely be doing much investigation on that in the near future though, the main system will be the main target for now.

Quite looking forward to seeing what this machine can do now.

[CitroJim]

Excellent! It's been a while since I've seen so much discreet memory like that Zel!

Fingers crossed it all works after your hard work

[Zelandeth]

Excellent! It's been a while since I've seen so much discreet memory like that Zel!

Fingers crossed it all works after your hard work

Thankfully once I had the memory test software it was pretty simple. The memory is just made up of 4116 chips so nothing too exotic.

I've wanted to see what this can do for a while so it will be interesting to do some experimentation with. If the parts machine will work too that would definitely be an unexpected bonus.

[Zelandeth]

Uuuhh...



That's sub optimal.

That's the gear linkage on the Renault, which has lost interest in remaining attached to the gearbox.

Seems she's been looking out for me though, because I was parked here when it happened.



I need a new bush - well a full set of them, which was already in my mind anyway as she suffers from the usual slightly baggy shift that's common to 80s Renaults. So I'd like to replace all the bushes anyway. For now until I can source them I've done the right thing and bodged it with cable ties. The bungee cord in the photo was my fix to keep it out the way while I drove onto the driveway.

If that had happened while on the move that could have been really nasty.

Back in the world of yesterday's computing, I left the Apple II running the memory test for
a couple of hours while I was doing something else today and it remained rock steady. So it looks like we have licked the stability issue. So here was this evening's entertainment.



Which combined with a diagnostic utility which can thoroughly test these drives would let me find out what works and what doesn't.



Yes, that monitor does have a very tired CRT, though it's not quite as dim as the camera makes it look.

I wound up with out the box a pair of working Apple drives plus the Cumana one second from the top. The other Apple ones don't seem to know what they're doing, just head banging a bunch before eventually erroring out, so I don't think are reading anything (meaning it never finds track zero). So step one there will be to check the heads are actually free to move plus are clean. The Super 5 unfortunately seems to have deeper issues as it locks up the controller card when hooked up. Shame as that's one I'd be curious to see how that one worked given how relatively rare third party drives for the Apple II were.

You can definitely tell the difference in the mechanism between the Apple and Cumana drives when in use, the Apple ones use that odd spiral plastic runner system so head seeking is near silent (most similar sound I can think of is a Zip drive seeking), whereas the Cumana drive uses a conventional leadscrew arrangement to move the head, and therefore makes the usual gronking noises you'd expect.

Taking bets on how many more of the drives I can revive when I have time to dig into them.

[CitroJim]

You were lucky with that gear linkage Zel! I once 'lost' a gear linkage on my original AX back in 1990 going up the A5D (as it was then). Early ones had a habit of doing that after loosing a roll pin. Slightly later ones like Bluebell have a modified linkage to prevent it... Hopefully

Happily, I got to a layby and did a temporary 'get you home' on it...

I'll bet, with a bit of TLC, all those drives will work... They'll just need their mechanisms gently overhauled and their heads cleaned... Floppies were very reliable in their day... And so, indeed, were early hard drives as I recall...