Trouble with power supply of my RiscPC

I fear that the power supply of RiscPC is getting problems. The RiscPC is not starting, after a while it is starting but reboots after 20 seconds or something like this.

Voltage at power supply unconnected:

+ 5.06V
+ 11.45V
- 12.1V

It is the 103W version.

As I am owning a two slice case I could assemble myself an alternative e.g. from cap rail power supplies but repairing the old supply if possible would be better. Is there a circuit diagram available for the 103W version?

Many thanks in advance!

I had the problem a few years ago, it was a 220nF capacitor near the input that had dried out. No more problems since…

Hoping this is the source of the problem, it looks like it, the capacitor is to filter out unwanted frequencies and the magnetization can be affected by other devices on the same network.

On the Stardot site a pack was made up of diodes and capacitors for the RPC power unit.

I haven’t seen a circuit diagram of the power units.

I had the problem a few years ago, it was a 220nF capacitor near the input that had dried out.

I don’t think there is any such thing as a 220nF capacitor that can “dry out”. Only electrolytic capacitors can dry out. Capacitors of that low value are never electrolytic.

I don’t think there is any such thing as a 220nF capacitor that can “dry out”. Only electrolytic capacitors can dry out. Capacitors of that low value are never electrolytic.

As you say, more likely to be a 220uF and likely a 300V rated low ESR type, as the high frequency spikes out of the switched mode would probably cook standard devices fairly soon.
Not had to “play” with any of that stuff for over 20 years, for which the nerves in my arms are extremely grateful

Non elecrolytic capacitors do “dry out”. A common problem on the Honeywell central heating controller is the main dropper capacitor failing. I replaced the cap in mine just last month. There are plenty of videos on YT, for instance . Mine was the big blue cap that was faulty.

Your YT instance showed that the main dropper capacitor was OK, and that the electrolytic reservoir capacitors had dried out. That makes much more sense.

Also bear in mind that the Risc PC’s power supply would emphatically not be using a dropper capacitor.

There are plenty of videos on YT, for instance .

Have you changed that link since yesterday? I’m getting a link to Big Clive fixing a Honeywell controller this morning, but it was John B fixing a similar unit yesterday. The John B video was the dropper capacitor that had failed (according to John – I didn’t see him check the failed component, and his explanation of the 48V relays was a bit fuzzy to say the least¹), whilst as Dave says, Clive found that the dropper was fine and it was failed electrolytics to blame.

Non elecrolytic capacitors do “dry out”.

They don’t, actually: it’s a very different failure mode. It might be viewed as semantics by a non-electronic-engineer, but we’re talking safety-critical parts here, so it’s not.

That “big blue cap” looks as if it’s a safety-rated X-Cap, and they do not “dry out”. Their capacitance can reduce, but it’s due to their requirement to “self-heal” in the event of internal breakdown of the insulation: effectively some of the metal film around the failure point fuses away so that insulation’s integrity is retained, at the expense of the loss of some surface area and hence capacitance.

The X-Caps are designed for use “across the line” or similar, where there’s a lot of energy from the supply on hand to do bad things if there’s a fault. The alternative to that is what the old Rifa caps used to do in the Beeb power supplies: fail catastrophically, with smoke and flame — curiously, that’s frowned upon by the standards that one must follow in order to CE or UKCA mark something these days.

This isn’t directed at Jeff or Thomas, but as this is on a public forum it’s worth noting that when attempting to replace a dropper capacitor, the replacement must be an X-Capacitor. You will find similar parts of similar values and sizes which are not X-Caps (and will probably be cheaper), but used as a dropper they can fail end-of-life in various dangerous ways. And if you (in the generic sense, not Jeff or Thomas directly) don’t know what I’m on about, it’s probably best not to be repairing mains electrical stuff based on YouTube videos.

Your YT instance showed that the main dropper capacitor was OK, and that the electrolytic reservoir capacitors had dried out.

True, but the dropper capacitor could also fail in the mode I refer to above, and that seems to be fairly common too. A lower value of dropper cap would result in less energy getting through the supply, but depending on the loading pattern of the device whacking a bigger reservoir cap on could mask that – for a while, at least. So the fix in the video might work, but whether it’s fixed the problem or the symptom is another matter.

In fact, I’m fairly sure a trawl back through Big Clive’s channel will reveal quite a few such faults. All this really shows is that you should never assume anything about a failed circuit: always get a meter (or scope) out and test your hypothesis, because if I’ve learned anything over the years, it’s that the one time you assume what has failed, it will probably be something else.

¹ I’ve not looked at the Honeywell unit, but I’d suggest that a more satisfying explanation for the 48V relays is that the dropper reduces the supply voltage to circa 48V (because capacitive droppers regulate current, not voltage, and as John notes, higher voltage relays draw less current and so work better on dropper circuits). There’s probably then a separate regulator to get down from 48V to whatever the micro needs.

it’s probably best not to be repairing mains electrical stuff based on YouTube videos.

<applause>

it’s probably best not to be repairing mains electrical stuff based on YouTube videos.

I’m sort of inclined to suggest that the sentence should be shortened / amended.

“it’s probably best not to be repairing mains electrical stuff, unless you’re suitably qualified.”

That’s not job protection¹ that’s friend protection. On a good day mistakes hurt, on a bad day you don’t get to comment on what it felt like.

¹ As opposed to the electricians and their red book + city & guilds, which they ignore more than observe.

“it’s probably best not to be repairing mains electrical stuff, unless you’re suitably qualified.”

Agreed. I’d hoped that was covered by “if you don’t know what I’m on about”, but I’m definitely not averse to making the warning more explicit.

On a good day mistakes hurt, on a bad day you don’t get to comment on what it felt like.

And in this context, the mistakes may not bite immediately: it could be what happens when the repair fails one month/ six months/ a year/ five years down the line.

All in all, the safe option would be to look for a suitable PSU module with three voltage rails:
+5V
+12V
-12V

Check the datasheet for minimum current on the rails. Typical 120W modules have a 1.5A min on +5V and 0.6A on the +12 with 80W units requiring minimums of 0.5A +5V, 0.1A +12

Hi,
You are right “dry” is not the right term for a ceramic capacitor, in fact this one had burned….I had forgotten.

I have already repaired switching power supplies. For Thomas’s, regulation and switching work. since the voltage values ​​are correct. On the other hand, it would be necessary to be able to measure these voltages under load on the RISCPC PCB.
I would look at the 400V electrolytic capacitor at the input of the circuit. But be careful to discharge the capacitor before measuring….
This is just advice.

I have a spare RPC power supply, if you can’t find one. Either a replacement from Castle which was used for a week or two or the original that was repaired by a (retired) professional; I can’t remember if I swapped back to the original or not, but I expect I did, to make sure it worked.

I’m not really interested in getting rid of it, but just in case you can’t find one on ebay or whatever.

I found meanwhile time to setup an installation with RiscPC mainboard and power supply. It is still behaving as described above. Power supply under load:

+5.06V
+11.68V – 11.8V (decreasing the longer it is running)
-12.01 V

So I think it is a problem with the power supply as
this never reaches 12 V. Ok backup “battery” may be a problem also but this shouldn’t cause this trouble.

The 12V is nominally 12V, but allowed EU be +/- 7%. Which means 11.2 to 12.8 are the extremes permissible.
The podule backplane and IDE connector are both specified as 11.4V to 12.6V (nominally 12.0V).
[source: RiscPC Technical Reference Manual]

Does the RiscPC itself make much use of the 12V? I think it’s on the backplane and for the harddisc and…? Certainly it seems within the permitted range and it shouldn’t cause a reboot, surely? That would be more like a problem with the 5V?

I started today “reverse engineering” to find out the wiring diagram. I started from consumer/secondary side. I think I am understanding the negative 12V part quite well meanwhile ;-).

5V part and 12V part seem to be designed in a similar way. On examination I found a strange soldering at one pin of a capacitator in 5V range. From top the capacitator itself looks fine. I am not very expirienced but I never saw this before. It is not looking like a cold soldering but it looks different compared to other solderings. I must see what comes out.

Have a look on the Stardot site – recapping the RPc power supply.

The Caps leak – and make a mess underneath them.

Handy doing a diagram of the power supply.

I took out the heavy weapon today afternoon this means I took a scope to trace the problem. It is definitely not a problem of the power supply!

The voltages are entirely stable. However from reasons entirely unknown in the moment the machine resets. Yes, the IOMD launches accurate signal on the reset pin in such cases. But why? It mainly happens at start and may stay for a while. It is entirely independent of the parts fitted (in)to the main board (RAMs, Harddisc Network cards and so on). However from time to time it resets without visible reason.
But if pressing the reset button it is likely that trouble starts again for a while. May be a resistor is broken in range of reset button and so loading the related capacitator fails. I must examine. Does any one have a better idea where it could come from? (IOMD broken!?)

The oscilloscope is the right tool, but were you able to measure under load?
Your measurements on unconnected power supply outputs are a bit limited, in principle I measure on the motherboard, there are circuits powered by a voltage of 5V.
I think there is a possible adjustment on the power supply.

Yes I measured of course at power connector and at various places on the main board.

Since around 20 minutes R85 is my candidate. I bridged it manually with a 47K resistor. IOMD didn’t get hot and machine started. Removing the resistor brought the green flashing of the screen back. I was able to reproduce this two times. However in the moment I don’t have a fitting resistor (SMD or classical workaround). That what I used for bridging is old and too big. Must see but perhaps this is the solution.

I had a similar effect years ago inside the motor management of my car. I searched a long time to locate it. It was also a broken SMD resistor controlling signal strength loopback of an operation amplifier amplifying the crankshaft signal.

Maybe the leakage current of C53 is to big? (40uA on 47k Ohms can allready cause logical 1)

Erich you got it! It was C53! I replaced first R85 but had to find that the reamining parts of original SMD resistor had 47K and there was no real change in behaviour. Afterwards I replaced C53 and now my RiscPC starts properly again. Many thanks to all for the help!

A problem with a capacitator is more likely than with a resistor so this might happen to other old RiscPCs too. Indication:
- Nothing happens after switching power on.
- After a while IOMD will get hot and screen will
start blinking green in fast sequence and also
the keyboard LEDs will blink. The frequency of
blinking will decrease with the time and end
after some minutes.
- RiscPC may start now properly and work perhaps
with resets. IOMD will cool down
- Pressing the reset button will start the trouble
again.

C53 ia 10uF electrolyt capacitator but an SMD one.
It can be replaced by a conventional 10uF electrolyt capacitator but keep an eye on right location of minus pin!

R86 can be measured by a multimeter and should show 100 Ohm. Measuring of R85 is more difficult as C53 will be loaded in parallel. D1 should show a symmetric diagram inside scope component tester with artefacts of diode diagram.