Computer PSU turning off when other circuits switched on/off

[Peter SE14]

I have a self built PC and I'm having trouble with it switching off when other circuits are switched on and off. I've already swapped the PSU from a cheapish brand to a Seasonic Focus GX.

I did try connecting via a UPS but that seemed not to help, the PSU would turn off one in three times when the UPS was self tested, but that could be due to the UPS being consumer grade and I'm unsure what exactly what (or where) the surge suppression is in the circuit (although I suspect it's a MOV type). I'm probably going to send the UPS back as it is supposed to cope with a PFC type PSU.

The thing that does work is a PD Devices surge protector (RS 206-7522) in a 13 amp plug and that has stopped the PC from tripping when switches are turned off and on. The mains sockets themselves test okay with a Kewtech 107 and it's loop impedance test is fine.

So while I'm happy I've fixed my current issue, this leaves me wondering if I have some sort of electrical fault or just a very fussy computer which doesn't like light switches.

Is worth getting an electrician in to try to diagnose things? Would a new consumer unit with surge protection built in help?

I've attached a picture of the current consumer unit.

Consumer_unit

• Peter SE14
• 3 Oct 2020

 

[DetlefSchmitz]

What other circuits - could it involve fluorescent lighting?

 

[JohnW2]

So while I'm happy I've fixed my current issue, this leaves me wondering if I have some sort of electrical fault or just a very fussy computer which doesn't like light switches.

I would think almost certainly the latter. I've never heard of this happening even when large loads on the electrical installation are switched on/off, even without any 'surge protection'. One might hypothesise that some fault in the electrical installation's earthing might somehow be implicated, but since your Kewtech shows an acceptable loop impedance, that's more-or-less ruled out.

Do you mean that the PC's PSU is shutting down? If so, I think the first thing I'd look into would be the possibility an imperfect (high resistance) connection in the signal line (maybe at the connector) that turns the PSU on/off (pin 16 of an ATX connector).

Kind Regards, John

 

[Peter SE14]

Well there are various fluorescent lighting fittings potentially involved. This is mostly in a basement so flourescents have been used to help with head room (although I have installed replacement led flourescents in existing fittings).

I can check to see if there is anything odd with the PSU powerswitch connectivity with a multimeter. I've just realised I've no idea how the soft power switch on a modern PC works... time for some research I think.

 

[JohnW2]

Well there are various fluorescent lighting fittings potentially involved. This is mostly in a basement so flourescents have been used to help with head room (although I have installed replacement led flourescents in existing fittings).

As I said, I've never heard of anything having the effect you are experiencing bit if you are saying that the tubes in the 'fluorescent fittings' are now LED tubes, that would make it even less likely to be an issue.

I can check to see if there is anything odd with the PSU powerswitch connectivity.

Assuming that the PSU is working normally (and you've tried changing it, so we assume that is the case), it will remain on so long as pin 16 of the connector is pulled low, and it will only go off if the voltage on that pin rises towards +5V. If there were a high impedance connection in the path to that pin (or at that pin) it's just conceivable that any transient 'spikes' around might be raise the potential high enough to switch the PSU off.

Kind Regards, John

 

[DetlefSchmitz]

Without going into detail with what the OP has installed, I'll simply relate my experience with fluorescent lights (probably not for the first time). I once bought an electronic time switch to turn on a light somewhere, and it worked fine. One day I turned off the under-cupboard (fluorescent) light in the next room, and the time switch in the first room exploded (black smoke marks all up the wall). I was sent a replacement, which died in the same way for the same reason. So it seemed that turning off a fluorescent light with a particular faulty design, caused a sufficient spike in the mains to severely disrupt a connected appliance in a different room.

Presented for your consideration.

 

[Adam_151]

Definatly something odd going on here, I could quite beleive a cheap and cheerful PSU beig affected by switching noise etc, but for it to continue with a seasonic is odd as they are far from the cheaper end of the scale! The fact that a surge suppressor helps does point to it being switching noise though. I would have guessed earthing issues but that appears to have been ruled out, Could the noise be entering the computer another way, perhaps through a network cable or similar. Can you disconnect everything else from the PC and see if it still does it (does it do it everytime w switch is flicked, or would you have to leave it a while to see if its cured it)

Could the PC case be picking up an earth at a different potential from anywhere else? (i.e. sat against a radiator) etc

 

[JohnW2]

Without going into detail with what the OP has installed, I'll simply relate my experience with fluorescent lights (probably not for the first time). I once bought an electronic time switch to turn on a light somewhere, and it worked fine. One day I turned off the under-cupboard (fluorescent) light in the next room, and the time switch in the first room exploded (black smoke marks all up the wall). I was sent a replacement, which died in the same way for the same reason. So it seemed that turning off a fluorescent light with a particular faulty design, caused a sufficient spike in the mains to severely disrupt a connected appliance in a different room. ... Presented for your consideration.

Goodness!

The 'ballasts' of traditional fluorescents certainly are pretty large inductors, but one would have hoped that they wouldn't have the effect you describe. In addition to the fluorescent probably having had a 'faulty design', I imagine that the time switches must also have been particularly susceptible/vulnerable to 'spikes', since the other electronic things in your house (TV, IT, etc.) presumably survived. In fact, assuming that the fluorescent pre-dated the time-switch by an appreciable period of time (without having killed anything electronic), that might suggest that 'vulnerability' of the time switch's design might have been the main issue?

However, I don't think that alters what I've been suggesting'. If the 'power on' line of the OP's PSU is held solidly to ground (by a low resistance/impedance connection), then it really should be impossible for anything to raise it's potential high enough to turn off the main part of the PSU.

Kind Regards, John

 

[Adam_151]

I agree John, I can't see how the 'pilot' line can be interupted by emi. But maybe teh same is not true of the logic circuits which then cause this to be switched to ground. Also the ATX spec includes a pin called 'power good' this is pulled to 5v when the power supply is stable, if the psu decides its not stable, then this drops, and the computer does what it can to protect itself https://en.wikipedia.org/wiki/Power_good_signal

I think we need confirmation on whats happening with the OPs machine, downs it just power down or does it reset?

I'm still thinking earthing issues with a flakey earth that the kewtech socket tester is failing to pick up

 

[JohnW2]

I agree John, I can't see how the 'pilot' line can be interupted by emi. But maybe teh same is not true of the logic circuits which then cause this to be switched to ground.

Yes, possibly, but if that were a possibility, and given all the EMI around, I would have expected to have heard of such behaviour before, but I haven't - have you?

Kind Regards, John

 

[Peter SE14]

Phew... Lots of thoughts here.
To be specific the machine just dies and needs the "soft power switch" pushing which restarts without an issue, from memory the lan link light is still on at this point.

It's been stable since I started using a surge surpressor plug but that doesn't really help with narrowing stuff down.

For testing purposes I have found that switching the downstairs toiler light off and on a few times seems to cause the PC to turn off (usually when the light is turned off)

I suspect the PSU / motherboard is a bit sensitive as it doesn't cope with the UPS self test very well the failure mode seems to be when switching back to local mains(roughly one in three times) having gone to battery backup this was definitely worse with the previous PSU which just failed the UPS test almost every time.

Given the plug tester gave a safe result on the earth this give me some reassurance it's not an obvious issue. In an ideal world I'd attach a DSO on the PSU outputs but these days there is so much stuff between that everything else these days it's a bit of pain to test. (also I don't have easy access to a DSO these days)

 

[SFK]

Could it be your cable to the computer?
Only asking as last week had to replace cable to monitor, as socket into monitor had poor/loose connection.
Took a while to find this issue.
This cable although newish had highish resistance at this interface that resulted in troublesome monitor operation.
Sfk

 

[Peter SE14]

Well it could be a cable issue and I do have a spare C13-C14 connector (I've been using the one supplied by the UPS vendor)

For the surge suppressor plug I chopped the end of a regular UK lead, so that is a different lead.

However I've had a thought that the two monitors are attached to the UPS so I will probably try with a different C13-C14 connector and the wall warts for the monitors unplugged and try the self test again.

Currently I'm leaning more towards an overly sensitive motherboard. I've just been reading the power control chipset manuals (A Richtek RT8944J apparently) and VRMs which makes me feel out of the loop (last time I did any electronics design was 25+ years ago)

 

[Peter SE14]

Well I swapped over all of the cables and tried the UPS without the monitor wall warts plugged in and sadly my PC died on the third attempt at a UPS self test.

Going to stick with the surge suppressor and send the UPS back unless their tech support has any good ideas.

 

[flameport]

I have a self built PC

Is the motherboard mounted onto metal or plastic standoffs,
are the motherboard fixing screws metal or plastic,
and do they have those small fibre washers under the screw heads or not?

The correct answer should be metal standoffs, metal screws and no fibre washers, so that the motherboard is electrically connected to the case via the screws/standoffs.
Other combinations may cause problems.

The other thing is to ensure continuity from the earth connection of the power supply to the computer case, so no fibre/plastic washers there either.