Olivetti PCS11 from 1992 - PSU fault

[jms2]

I realise this is off topic because the machine in question is a 386 PC - but in my defence, it's an Olivetti so does have some kind of vague link to Acorn . It's also rather a rare and interesting machine - it's a tiny and neatly designed desktop, apparently using early 90s laptop-type components. Here's a photo (I pulled the innards out of the case and didn't want to put them back in for the photo - also all the PSU shielding is missing in this photo, it does exist in reality!):

Most of the volume is taken up by the PSU, hard drive and floppy.

The problem with this machine is that the PSU appears to be dead, which is why I was asked to take a look at it. I have been able to establish that the fuse is not blown and the high voltage side of the PSU seems to be OK (as far as the transformer). I don't know much about PSUs but I am aware of how potentially dangerous they are. I also found this superb article on Ken Shirriff's website, so my research is based on this: https://www.righto.com/2021/05/teardown ... upply.html

The transformer and all the surrounding components look OK:

Unlike in Ken's post, this one doesn't seem to have the control IC for the transformer on the main board. Instead I think it is on the daughterboard (the UC3842 chip on the right), and I assume the "firebreak" style gap is to achieve separation between the high voltage and low voltage sides?

There is a second daughterboard on the front, which has an LED and a momentary pushbutton which is the on-switch (and also Reset I suppose):

The symptoms are interesting and I hope might give a clue to what is going on. If I switch it on at the mains, The LED glows weakly and the fan runs, but only for about a quarter of a second. Then both stop. Then they start up again, and this weak pulsing repeats up to 5 times before the machine goes totally dead. If I then switch it off and back on again, nothing happens. However if I leave it for a few minutes, I can get the sequence to repeat. There doesn't seem to be any output from the PSU, although I haven't attempted to measure this during the "pulsing" stage.

So my question is, does this feel like it might be repairable by fitting a new control IC, or should I abandon all hope? There's almost nothing about these machines that I could find online, except other people finding broken PSUs and trying to fit a new one instead. But are PSUs really not repairable?

[1024MAK]

Yes the UC3842 is the control chip. The white chip near the “fire break” is the opto-coupler / opto-isolator. And yes, the “fire break” separates the mains / primary side from the output / secondary side.

The symptoms sound like one or more of the electrolytic capacitors have failed.

The first one to check, is the “start-up” capacitor on the mains side. This is the electrolytic one connected between pins 7 (+) and 5 (-/GND) of the UC3842 (Cvcc in the datasheet schematics). It will have a value of between 100 µF and 220 µF. Note pin numbers relate to the package of chip used in the large typical schematic as shown in the datasheet. If a different chip package is used, the pin numbers may be different.

If you do remove this capacitor, at the same time check the high value “start-up” resistor(s), typically 100 kΩ (shown as Rstart in the datasheet schematics). Note that this resistor sees a high value, so best if any replacement has a suitable voltage rating.

If replacing both of these components does not solve the problem, try replacing all the electrolytic capacitors on the output / secondary side.

The datasheet for the UC3842 is here

Any replacement electrolytic capacitors should be 105°C and may need to be low ESR types.

Mark

[jms2]

Thanks Mark, very useful indeed!

I'll get some parts unsoldered and tested. It's nice to know I have at least a bit of a chance of success!

[jms2]

I have made some progress on this, but it's inconclusive so far.

The first one to check, is the “start-up” capacitor on the mains side. This is the electrolytic one connected between pins 7 (+) and 5 (-/GND) of the UC3842 (Cvcc in the datasheet schematics). It will have a value of between 100 µF and 220 µF. Note pin numbers relate to the package of chip used in the large typical schematic as shown in the datasheet. If a different chip package is used, the pin numbers may be different.

In this case, the chip is in the 14-pin SOIC package, so the pin numbers are a bit different; but I was able to find Cvcc, which is the small electrolytic below the really big capacitor on the right of my second photo. The big capactor is for smoothing the rectified DC I think, and is 100uF. I desoldered it (as much to improve access if anything else) and found it was OK. I didn't desolder Cvcc, but it's marked 220uF and testing it in circuit I got close to that value.

If you do remove this capacitor, at the same time check the high value “start-up” resistor(s), typically 100 kΩ (shown as Rstart in the datasheet schematics). Note that this resistor sees a high value, so best if any replacement has a suitable voltage rating.

This appears to be a very physically large resistor, marked as 64k, which when tested in-circuit gives 67.4k. I could remove this to test it properly but it seems OK.

If replacing both of these components does not solve the problem, try replacing all the electrolytic capacitors on the output / secondary side.

I did manage to find this photo online (pretty much the only piece of information available relating to fixing this PSU). The author said that these capacitors are in the feedback circuit and renewing them is a cure for a dead PSU. These seem as good a place as any to start, although they do look a rather odd package.

[jms2]

Here's a better view of the two capacitors (the whole board is the other way up in this photo). They seem to have two pads at one end, but also there's some indication of a third attachment point at the other end. Maybe its glue?

[daveejhitchins]

They just look like ordinary capacitors in a 'case' for surface mount e.g. the two legs are brought out at one end and maybe the case, at the other end, is use to hold-down the package ???

Dave H.

[jms2]

After much research and asking around, I have established that these are simply electrolytics in an obsolete package. The same type is used on certain Sega Game Gear boards apparently.

[jms2]

I changed the two capacitors indicated above, using normal (but low ESR) electrolytics, and the PSU now runs!

There is still a remaining problem though. The PSU provides +5V, +12V, -12V and 0V signals and the levels of these look good. But there is also a signal which I think is "POWER GOOD", which should go high when everything is stable. On this machine it never goes high, even though the voltages seem fine.

I haven't renewed any of the low voltage capacitors apart from the two above, so would I be right in thinking that renewing them might fix this problem?

[1024MAK]

No idea. I’ve never needed to investigate a “power good” line not working correctly.

Have you been able to load test the +5V output with a test load (I use 12V 35W or 50W) quartz halogen lamps or 12V 21W car brake light lamps. The “power good” signal may be driven from some circuitry that needs to see a load on the main output.

Mark

[jms2]

I’m measuring the voltages with the PSU connected to the main board, so hopefully that should be providing the correct load.

I couldn’t find anything much online about missing “power good” signals, but I did find a generic circuit diagram on which the signal comes from an LM393. No idea if that’s what’s used here (I put the case back together) but if it is, I could try renewing that.

In the meantime I wonder if faking this signal will cause the machine to boot.

[1024MAK]

I know that PC (as in IBM compatible PCs) SMPSUs have power good signals.
In some systems, this signal enables the motherboard to indicate to the OS that the mains has failed so that it can quickly stop certain functions before the power runs out completely.

For testing purposes, yes, try faking the signal.

The LM393 is a comparator chip. I use them in some of my own designs (monitoring the state of battery systems). Before changing it (or whatever other comparator or op-amp) is being used, it’s worthwhile testing the voltages on the input pins. It may be that a voltage reference chip has failed.

It could still be a capacitor that is faulty, but if all the outputs are in specification (within 5% at least and with no discernible ripple on a ‘scope) with the PSU supplying the computer board, it’s probably not one of the larger capacitors. Are there any smaller electrolytic capacitors that have not been renewed?

Mark

[jms2]

Plenty to go at there! I'll have a play over the weekend and see what I can find.

[jms2]

Faking the 'power good' signal does indeed allow the machine to boot!
I'll have to fix it properly now of course.

[jms2]

It's been a long while since I posted about this machine. It has been very frustrating. Essentially, by faking the "power good" signal manually I was able to get it to boot fairly reliably. I decided to park the PSU issue and try to get the hard disc working. I found that, coincidentally, the machine had the same 40MB Conner drive as found in some A5000s. In this one the head was effectively glued to some rubber bumpers that had melted, but even having fixed this problem (by removing the rubber bits) the machine still didn't recognise that it had any kind of HD installed. This persisted after I plugged in a CompactFlash adaptor instead. So a revisit to the PSU seemed necessary.

Close examination revealed a few "interesting" things:

1) The "Power Good" circuit is on the ceramic daughterboard, which, being virtually impossible to solder, means that even if I could figure out what's wrong with it, I wouldn't be able to fix it.
2) The Power Good signal is derived, not from the 5V output, but from the 12V output. Exactly how, I wasn't able to figure out. But instead of outputting the desired 5V signal, what it produces is about 0.2V.
3) The 12V output was low, down at 11.4V.

To me this seemed encouraging and logical - the lack of a Power Good signal was actually accurate, because the 12V signal was low. This might also explain why the hard disc wasn't being detected, as that uses 12V. I identified an electrolytic capacitor which was used to smooth the 12V output, tested it, and found its capacitance value was low (39uF vs 47uF). So I fitted a new one. The 12V output is now.... 11.1V.

I tested the other capacitors on the low voltage side, but they were all OK. Annoyingly, the -12V output is fine (as are the 5V outputs).

In frustration at ever being able to repair this PSU, I have looked at alternatives. But they are either very expensive (PicoPSU) or too big (the Olivetti PSU enclosure is really small). It's frustrating that it seems to be so close to working, and yet not quite...

Any thoughts welcomed!

[1024MAK]

Was the PSU connected to a load when you tested the +12V voltage?

And by load, I mean the main (presumably the output that has the highest current capacity) output, which is normally the +5V rail.

The other common problems are faulty rectifier diodes or dry joints on the transformer.

Mark

[Phipli]

Some power supplies don't individually regulate each voltage, but regulate all voltages based on feedback from one. On a computer, you'd probably regulate the 5v to protect the logic. So... It is common that the 12v is low if the 5v rail is not very loaded. To be honest, 12v doesn't need to be very accurate because it is usually powering motors like disks and fans. 11.4v was probably OK. 11.4 is exactly -5% which feels like where you'd perhaps put a PWR Good threshold? Not sure.

Like 1024MAK asked, was the PSU loaded? Was it loaded realistically on multiple rails?

[jms2]

That's a very good question. When I tested it post-modification it wasn't loaded. I'll put the machine back together (easier said than done) and then re-test the voltages when they are under load. Thanks!

[1024MAK]

Oh, and typically a +12V rail and -5V rail are specified to have a tolerance of +/-10%.

So 10.8V to 13.2V (although most PSUs manage to be a lot closer to 12V).

Mark

[jms2]

Good news, up to a point. When under load, the +12V line is now at 12.04V! So that problem is solved.

Sadly the Power Good signal is still not working. I am thinking that a workable solution might be to use a 555 timer circuit to switch on this signal after say 500 ms.

Of more immediate concern is the fact that the BIOS now thinks that not only the hard drive is missing, but also the floppy drive (which had previously worked OK, allowing me to boot into DOS). Further investigation needed, but all I have done is unplug them and plug them back in again. I think the connectors were keyed so I can't have got them the wrong way round. Hmm.

[jms2]

I fixed the problem with not recognising the drives - it took me absolutely ages, but basically the CMOS ram was corrupt and, unlike an Acorn machine, you can't simply hold down Delete or some other key on boot to wipe it. You need a special utility program. Once I had found this, and wiped the memory, things started working and the machine was able to boot into Windows!

I also implemented the "fake Power Good" signal using a 555 timer circuit. I used the "delay before turn on" circuit here, taking the raw 5V and 0V from the PSU and using these to power the 555 timer. I picked a capacitor and resistor combination which gave roughly a 0.5 seconds delay. I took my "power good" signal directly from pin 3, assuming that the 470 ohm resistor shown on the linked page was needed to protect the LED in that project.

This circuit worked... but it wasn't perfect:

1) It didn't always work. Sometimes I would have to switch the machine on and off again twice before the circuit kicked in.
2) I noticed the 555 was getting quite warm.
3) Finally, it failed, presumably because the 555 was getting overheated.

Should I be using a resistor in series with the output, or some other refinement?

[1024MAK]

If you want it to ‘reset’ properly if the machine has only been off for a short while, you need a diode to discharge the “timing” capacitor. A 1N4148 or 1N4001 type will do. Connect the anode to the capacitor negative lead. Connect the cathode to the capacitor positive lead.

Edit: see correction in my next post…

The original/ordinary NE555 version should really have a power rail decoupling capacitor connected between Vcc (+5V) and 0V/GND close to the chip. In this circuit, the value is not critical, suggested values are 47µF or 100µF.

I presume the NE555 is overheating because it’s not happy with whatever it’s feeding. A NE555 has an absolute maximum output current of 200mA from pin 3. But if this pin is subjected to a higher voltage than the supply voltage to the NE555, it will not like that.

You need to do some tests. What is the voltage on the ‘Power Good’ circuit (with the NE555 disconnected)?

Do you know how much current flows into or out of the ‘Power Good’ circuit?

Also, what value resistor did you use?

Mark

[jms2]

Thanks Mark, this is all very helpful. It has taken me longer than I expected to come back with some answers, because I discovered something I hadn't expected. This is going to make this quite a long post - apologies for this but I feel I might be getting closer to a solution!

First, the easy stuff to answer: I used a 5k1 resistor and a 100uF capacitor. In the example on the webpage, they used the same capacitance and a 47k resistor, for a delay of 7 seconds. By dividing the resistance by 10 my plan was to get roughly a 0.7 s delay, and this seemed to work OK.

Regarding the question of what voltage we are feeding into: I had simply assumed that the input would be 0V and was of high impedance, such that presenting a 5V signal there would bring the CPU out of reset. However... if I now cut the wire leading to this input (so I'm not doing anything to it), and switch the machine on, I find the voltage at this pin sits at just under 5.1V. The nominal 5V input to the board (and what I'm using to power the 555) is slightly over 5V, so essentially the same.

I'm very confused by what's going on here. My practice up until now of connecting this pin to 5V (at first manually, or later by using the 555) has been the only way to get the machine to boot, but now I'm not sure why this worked. I started out by reading several rather confusing accounts, complete with diagrams, on this thread. The people on there are discussing how to use a different PSU with the machine - which is kind of similar to what I'm doing, except I'm using an original PSU but one which is slightly broken such that the circuit that feeds pin 22 is not working.

The final diagram in the thread seems to suggest it's necessary to ground pin 22, creating a reset, and then pull it high via a 4k7 resistor. Apart from providing the resistor, I thought this was essentially what my 555 circuit was doing. But if the pin sits at 5.1V naturally, why is there a need to pull it high, and why via a resistor? I wouldn't expect any current to be flowing.

Here's the very surprising part - having cut the wire, I find that when switching the machine on it sometimes simply boots on its own, without any further intervention (usually this is the first time I switch it on). On occasions when it doesn't do this, I found that grounding the input creates a flow of 0.41mA and makes the machine boot. I didn't try pulling the pin high... because there doesn't seem to be any need to do so, contrary to what I was doing previously.

Does any of this make any sense to you? I get the feeling that any kind of voltage transition on this pin is sufficient, but there must be some logic to it.

[1024MAK]

Actually, having read what I put, I muddled up the diode connections. It needs to go across the resistor not the capacitor. In normal use the diode is reversed biased so does not conduct. But when the power is disconnected, it allows the capacitor to discharge as the diode is now forward biased as far as the voltage on the capacitor is concerned. Sorry about that.

Mark

[jms2]

Just experimenting a bit more, I found that grounding the pin doesn't reliably make the machine boot. And neither does pulling it up, it seems. But if I connect it to 5V zero current flows, kind of as expected given that it was at 5V anyway. This only adds to the mystery of what was making the 555 overheat.

[1024MAK]

Okay, this is getting confusing.

Is this actually a “power good” signal to the main board?

Is it active high or active low? Or rather, immediately on power on, is it high, then goes low, or immediately on power on, is it low, then goes high?

Is this, for want of a better name, an “open collector” (“open drain”) circuit, that is, the power supply just pulls the line to 0V/ground when required?

Does this system have a reset circuit and is this part of the power supply?

Mark

[jms2]

Okay, this is getting confusing.

If you're confused, imagine how I feel! However, your questions have led me to do some more tests and I think I have gained a better understanding as a result.

Is this actually a “power good” signal to the main board?

Actually, I don't think it is any more. Although, I don't perceive any practical difference between a "Power Good" signal which is initially low, then high, or a "Reset" signal which has the same profile.

Is it active high or active low? Or rather, immediately on power on, is it high, then goes low, or immediately on power on, is it low, then goes high?

I'm only measuring this with a multimeter at this point, but essentially, on switch on it goes from low to high immediately and stays there.

Is this, for want of a better name, an “open collector” (“open drain”) circuit, that is, the power supply just pulls the line to 0V/ground when required? Does this system have a reset circuit and is this part of the power supply?

I think you're right, the PSU is meant to provide the reset via this pin (active low). I have just discovered a totally repeatable behaviour, which is that if I ground this pin, it resets the machine. The current that flows is tiny, just 4uA. I think the original PSU is meant to generate this signal, and this is the bit that is broken. The machine will sometimes boot without me doing anything with this pin. (If it doesn't boot, grounding the pin will get it started.)

Previously, when I was under the impression that this needed to be pulled high, perhaps what I was actually seeing was the above behaviour, in other words me applying 5V wasn't actually doing anything, and the machine was just starting on its own!

I'm still not sure why the 555 was getting so hot. Even after disconnecting the output, so it was just powered but not doing anything, it was still getting hot.

[1024MAK]

Where did you get the 555 from?

Mark

[jms2]

Inevitably - somewhere on Ebay. I see what you're getting at.

However, I have put together a new design which is more complex and involves a logic gate to invert the signal from the 555. This was described as a "power on reset" circuit, which as you said is what is really needed here. It has worked very reliably on both breadboard and stripboard (the machine starts every time), and there's not been any sign of the 555 getting hot or even warm. I reckon I must have created a tiny solder bridge or something on the original circuit, and/or it was just a bit too simplistic for what was needed.

Now "all" I have to do is put everything back together. It is definitely possible, because removal of the physical hard drive has liberated quite a lot of space. But it's still really fiddly. Plus, I am reasonably sure I have got Covid, so I'm not feeling particularly up for the challenge!

[1024MAK]

Well, I know that at least some of the NE555 chips (normally with TI markings) from trading sites are either substandard or fakes. And those contained in kits are not safe either (I had one where the discharge pin was open circuit).

I put a reset circuit in this topic. I know it works, as I have the same circuit in use.

But if you have already got a working circuit, fair enough.

I hope that whatever virus you have got, you can fight off quickly. It’s definitely not nice if you have flu or COVID19. Get well soon

Mark

[jms2]

Thanks - I'm sure I'll be OK.

Your circuit is a bit different than the one I adopted, and is simpler - but if this one carries on working, I'm happy!