Solved : PSU cutting out?

After months of operation without incident, my Mendel90 has just started to stop suddenly.

I am printing from sd card, so it's definitely not a connection issue.

At the moment it lasts between 1 to 15 minutes before just cutting out and stopping dead. It looks like the power drops out for less than a second, and the Raspberry Pi I have attached resets.

The electronics is Melzi/PSU I got with the kit. Really no idea how to debug this one?

The only difference I can think is it's probably the warmest weather I've printed in - could the stepper drivers overheating cause this? The whole of the bottom of the board is quite hot to touch.

EDIT : Turned out to be a loose connection inside the PSU.

Edited 2 time(s). Last edit at 08/11/2013 01:25PM by marcosscriven.

Interesting... I have had similar issues for quite some time, I had always assumed it was due to noise on the mains. I have found that turning on my benchtop grinder seems to trip the supply, also starting a print where heatbed, hotend and all the steppers are running trips it too (so I now heat the bed for ~10sec before starting a print to prevent this) but other times it stops for no apparent reason.

Does the PSU go into some sort of safety mode where the main fan turns off and the supply needs to be switched off for ~20sec before it will turn back on normally? All this happens from prints via the SD card and in a plug which only my laptop and printer are connected to.

Yes, if the mains dips too low the PSU switches off to protect itself and doesn't recover unless you turn it off for a while. I noticed this recently when we had a series of brown outs.

If it happens when loading up the 12V rail it probably gets over-voltage on one of the other two. In which case more dummy load on the 5V and or 3.3V may fix it.

If the motor drivers get too hot they shut down for a while, which gives a load of steps missing on an axis. It wouldn't shut down the PSU.

---

[www.hydraraptor.blogspot.com]

After leaving the PSU off over night, I tried again this morning (with the added benefit of ruling out the room being too hot).

This time it stopped even before it made it to its starting position (which is at least good in the sense I didn't waste any filament)

You mention this is probably due to a mains voltage drop or brownout, but I've not noticed any dimming of lights in the house. The electrics are only a few years old, and it's in fairly central London (not that London is immune - I just mean city infrastructure rather than some overhead wires in a remote country house).

How would you recommend testing the actual cause? I have a multimeter, but I don't think it would be anywhere near fast enough to monitor the voltages. Add to that it's fairly sporadic (from a few seconds to 15 minutes or so), it's fairly challenging to regain confidence in it.

If it's a case of just trying out higher dummy loads - what extra/total loads would you suggest?

Edited 1 time(s). Last edit at 08/10/2013 04:24AM by marcosscriven.

One other thing I just noticed:

Looking through the PSU grill at the back, I can see two large shiny T-shaped bits (perhaps simple heat sinks). As the PSU is on it's side, the top one is at right angles, but the bottom one looks as if it's sagging about 10 degrees. I've never looked before so I don't know if that's normal.

Maybe something has dislodged? I can't really imagine there's something wrong with the electricity supply here (esp. over a few days), rather than something wrong with the PSU.

I just did some basic measurements on my multimeter across the two terminal on the main power block:

- 12.45v when the machine is switched on and idle
- 12.35v when the machine homes
- 12.05v when the bed/hot end start to heat up (bed 60, hot end 220)

At that point the head starts to move to X0, Y0, Z0.05 as the hot end heats up. I don't see the voltage drop below the 12.05v mentioned above, but as it moves to the starting point again I saw the PSU just reset.

Presumably then it's just a case of replacing the PSU? Can anyone recommend one? I don't relish the prospect of soldering all those PSU wires together to a tip again - is there a simpler way?

EDIT: The other thing that's less than ideal about hacking up the PSU wires is the warranty is presumably voided - would it not be better to connect to female ATX plugs, and do the soldering on those? Then in a case like this I can just send the PSU back and get a new one. Also, when you got the new one you could simply plug it back in!

Edited 3 time(s). Last edit at 08/10/2013 06:30AM by marcosscriven.

The problem is you need to connect to all the connectors so the sockets would cost a lot compared to the PSU.

Try measuring the 5V and 3.3V rails when the 12V has max load. If they are too high then it needs more dummy load. Try doubling them.

As power supplies age the capacitors lose capacitance so the ripple increases. If it was close to over voltage then it might then trip. I have had one running 24/7 since the start of kit production and had no problems with it.

---

[www.hydraraptor.blogspot.com]

Thanks for responses. I checked the 5V and 3.3V rails.

On 5V:

Starts at 5.15V
Bed starts to heat up: 5.26V
Motor starts to move to home position: 5.37V and cuts out

I did this twice, and the figures were the same. I'm not sure how accurate the last measurement is, because the multimeter is still changing as the PSU cuts out.

3.3V rail was rock solid at 3.408V the whole way through.

So the question is, is 5.37 too high? That's 7.5% over.

Would this do as a 20R resistor (not sure if that's a standard size that will still screw through the holes in the Mendel90): http://uk.rs-online.com/web/p/panel-mount-fixed-resistors/6150482P/

Unfortunately they only do them in 10s

Maplins does 10W wire wound resistors: http://www.maplin.co.uk/10-watt-wirewound-resistor-2181 individually - but I'm guessing you went with 50W ones for a reason?

Edited 1 time(s). Last edit at 08/10/2013 07:54AM by marcosscriven.

I thought the dummy resistors were 15W? I bought myself a couple from RS for the power supply of my second printer.

Interestingly it has the same brand power supply as my mendel90 except it is rated for 650W, and with the same dummy loads doesn't suffer from any dropout issues. It is connected to a RAMPS board so the arduino draws its power from my laptop which maybe helps with the situation? But it also runs the same amount of steppers, a heatbed and hotend as my mendel90. Which makes me wonder why would my mendel90 be more susceptible to power dropouts.....


Stock No.
159764 HS15 Al house wirewound resistor,10R 15W
159742 HS15 Al house wirewound resistor,4R7 15W

I just added a 10R 10W resistor in series with the existing gold 10R, and it hasn't made a difference unfortunately. All it's done is move the lower extreme of the 12V rail to 11.95V, rather than 12.05V. It didn't affect the max of the 5V rail much either - only 0.02V less, at 5.35V.

The first attempt to print failed right away, and the second attempt failed just as it was finishing the first layer.

I watched the voltages as it was printing, and I didn't see anything special (i.e. the voltage at the time of failure was within the range I'd seen during the rest of the print).

Any ideas what the next step is?

Edited 1 time(s). Last edit at 08/10/2013 12:16PM by marcosscriven.

One factor I've observed in computer PSU units is that they get hotter if not kept clean. The cooling fan gets quite inefficient with a layer of dust on it. This could be a factor given the correlation with recent hot weather. I open the psu up to clean it which can be dangerous if one is not familiar with precautions to take. Leaving it overnight without power before opening it and not touching anything inside with fingers and dusting it with a non metallic brush should be safe.

You need to put it in parallel to increase the load. Putting it in series will make it worse, which it seems to have done. I was going to say I could send you another 10R but it seems like you already have one.

5.37V does seem too high, I think 5% is the normal limit for a 5V rail. So it does look like an over voltage shut down. Why you would need it I don't know because 99% of machines don't. I think one other customer said he had a problem with this and fixed it with double the dummy load. They are very cheap power supplies though, so it seems there are some variations between them. Or it could be due to variation in the bed resistance.

I could provide a bigger dummy load, but then it might give 12V over voltage when there is no load on it, so might need a third dummy on that as well. At some point it is time to give up and use a more expensive PSU, but then I would have to increase the kit price. You need to pay about four or five times the price for a decent PC PSU and possibly ten times for one I would use to actually power a PC. In general a 3D printer is a lot less demanding as the voltage doesn't really matter unless it is too low because all the things that use the 12V rail do their own regulation.

The kits I ship now use a different make 500W PSU because the Alpine one is no longer available.

---

[www.hydraraptor.blogspot.com]

Doh - I misunderstood load to be resistance. I just now tried in parallel (and double checked with the meter the resistance is 5 Ohms).

With that setup, my voltages on the 5V rails were:

Starts at 5.12V (compared to 5.15V without the extra resistor in parallel)
Bed starts to heat up: 5.24V (compared to 5.26V)
Motor starts to move to home position: 5.36V and cuts out (compared with 5.37)

So it still doesn't seem to have done the trick, and doesn't seem to have made much of a difference to the voltages. You mentioned that it may be because the capacitance has reduced, and that would increase ripple. When the load is constant, I don't see the voltage jumping around though, which presumably I would if there were significant ripple (if I correctly understand ripple that is).

I might disconnect it and take the hoover to it (without opening it up, for now) - it doesn't look dusty, but it's worth a try getting rid of some dust.

After that though I assume to last thing to do is replace the PSU?

One other thing I just tried - instead of running a print, I just repeatedly ran G28 (so both bed and hot end are off). 5V rail never went over 5.17V, and after a few times the PSU dropped out.

It's almost imperceptible the time the PSU goes off - the PSU fan barely slows, and the only way I know it's switched off is the blip down to 0 or on the multimeter.

The ripple I refer to is high frequency ripple due to the rail being rectified from the switching frequency and then smoothed with capacitors. The output of a switch mode power supply is never pure DC, it has a saw tooth waveform on it of typically around 100mV and a frequency more than 20KHz, perhaps 80 - 150kHz for an ATX, smaller high efficiency ones in MHz.

Blips out? Sorry, I thought we were talking about it shutting down and needing to be switched for some time to run again. Mixed problems in the same thread confusing!

I have never known one to dip out and recover, they normally shut down when over voltage, over current or mains under voltage. A dip would normally be a bad connection somewhere, perhaps triggered by vibration when you run the motors.

How have you connected your PI? What happens without the PI? Does the Melzi reset? Where are you measuring the blip on your voltmeter? Can you induce it by tapping the PSU, waggling the mains lead, or output leads.

---

[www.hydraraptor.blogspot.com]

Sorry - seems I misused the the description 'cut out' then.. I've never had to reset the PSU - it appears to just stop very briefly and start again. (Sorry to have caused any confusion !)

I did have the raspberry pi powered across the 5V dummy resistor - and when the 'blip' (or whatever you call it?) happened, it would reboot. To rule that out I switched backed to powering the RPi independently.

Once I did that I could see that the Melzi resets during the blip. (And of course the RPi would now stay on)

I've not had cause to touch the electronics before this started happening, and I can't see anything loose - that was one of the first things I checked.

I'll go over the block terminals tomorrow and make sure all the screws are tight. The only other connection to check then is the soldering on the two dummy loads - but they were screwed in well so not sure why / how they could suddenly have a connection issue.

I don't think it will be the dummy loads. They would not cause a dip in both rails. Maybe a shut down, or just the 12V rail too low.

Perhaps a loose mains connection, or a loose connection inside the PSU itself if both rails are going at the same time. Possibly the mains switch. Is there any crackle when it dips?

---

[www.hydraraptor.blogspot.com]

Ok... so after all that, I rechecked absolutely all the connections (both on the Melzi, and at each of the motors/switches etc). No stray wires, all withstood a little tug without coming out.

I removed the second dummy load (so back to the original config), powered up, then tried something simple... tapped the PSU with my hand.

Lo-and-behold, most of the times I tapped, I saw the dip - I'm pretty sure it's something inside the PSU as I tried tapping with the same force all over the machine, and it only happens when I do it right at the back of the PSU. There's no crackle at all (presumably with that q. you're trying to establish if it's a short somewhere?)

I'm happy to open up the PSU if someone can give me guidance on making sure the capacitors are all discharged - I know the larger ones can hold their charge for days if not weeks. How do you do it?

EDIT: I just completely removed the PSU from the Mendel90, just to isolate it, and it's definitely the PSU. Tapping or shaking the unit lightly causes the fan to briefly slow down, and I can in fact hear some very quiet zinzulation if I put my ear near to it. (Somewhat gingerly for fear of a bang!)

Edited 4 time(s). Last edit at 08/11/2013 05:57AM by marcosscriven.

Here's a pic of the back of the PSU - the bottom of the picture is towards the floor, and to the left is the fan

You can just about see on the lower left the T-shaped bracket I mentioned above:

"Looking through the PSU grill at the back, I can see two large shiny T-shaped bits (perhaps simple heat sinks). As the PSU is on it's side, the top one is at right angles, but the bottom one looks as if it's sagging about 10 degrees. I've never looked before so I don't know if that's normal. "

Do you think this is the issue? Should it be at an angle? Has it sagged because it's on its side?

On second thoughts, probably worth me spending £20/£30 rather than blowing myself up with a capacitor.

Would this do? [www.amazon.co.uk]

It's £30 inc VAT with free delivery. It's the cheapest one I could find on Amazon that didn't have at least one review saying it had blown up.

I can't find any specs as to its rating on each rail though.

EDIT: I quite like the idea of a salvaged server PSU from ebay: http://reprap.org/wiki/Power_Supply

EDIT2: Or even this: [www.ebay.co.uk]

EDIT3: Looks like the 500W Alpine PSUs can still be had from Dabs for £11.98: [www.dabs.com]

Edited 3 time(s). Last edit at 08/11/2013 06:05AM by marcosscriven.

The same type of PSU I have here has the bottom heatsink at an angle but not as much.

The reason is asked about crackle is that loose mains connections usually arc and can be heard.

It does seem like the PSU has a loose connection somewhere.

Option 2 is probably a good option as the wiring is a lot simpler and you don't need dummy loads. XBOX 203W PSUs reputably work well and are cheap second hand.

---

[www.hydraraptor.blogspot.com]

I'm going to try the XBOX 203W route - chiefly because you can also get female sockets , allowing me to unplug the printer easily.

Is it worth me sending back my old one for post mortem? Poor thing only lasted seven months!

Capacitors in a psu are usually discharged in a few minutes so just make sure it is turned off and left off for a while before opening it. [www.overclock.net]

The problem with sending it back is the postage costs more than the PSU. As I don't use that type any more there is little to be gained.

Capacitors in a working PSU will discharge fairly quickly but in one that has shut down abnormally, or failed they can stay charged longer than it takes to dismantle it. Experienced engineers would avoid touching the HV section and short out the big caps with a screw driver, just to be on the safe side.

If you are curious you could take it apart to see if there is anything obviously wrong with it.

---

[www.hydraraptor.blogspot.com]

Thanks Ralph for that link - gave me the confidence to open it up - I left it for a few hours and made sure the capacitors (including the 'X' ones mentioned) were discharged.

The only thing I could find loose was a grounding wire from the socket/on-off switch assembly to the PSU PCB. Everything else seemed secure.

I removed a little dust, put it back together, switched on, and no matter how much I tapped or shook it, seemed fine again.

So I've connected it back up to the Mendel90 and it seems to be running fine again... what a palava.

Thanks for the all the help.