I may have missed this if it was mentioned above, but what is the voltage between earth and the chassis? Also, between the earth (assuming the scope is earthed -is it?) and the point that drew sparks? I suggest measuring with a high impedance scope probe.

Yes, scope is earthed. The voltage between earth and the point that drew sparks was 11.6VDC (which is the correct voltage there). I also was finally able to insert the 1k resistor between ground and chassis and measured 28mVAC. Other attempts to measure AC voltages between say ground and chassis (when connected) or to the spark point kept falling towards 0mV, so the 28mV might be something.

Not only does the spark to a high impedance not make logical sense & the circuit appears to work, what's making this sooo much worse for me is is when I leave it on for 20 seconds or more I can start to smell something but I cannot identify where it's coming from either through smell or touch. It's as if this thing is like ready to blow. Very unnerving.

 

Thank you for posting the schematic. The 0.8 ohm reading you got in post #27 confirms that there is no break in the earth path on the mains wiring. The only other thing I can think of on the mains earth wiring that could cause a problem is if there was something else in the building with the neutral wire on something else being swapped with the earth in error so it was drawing current via the earth wiring causing a small AC voltage between different points on the earth wire. You could check for this reote possibility by measuring the AC voltage instead of the resistance between the points I suggested in post #25. I think what you found in post #31 could lead to solving the problem.

Ya, but I think we ruled out that possibility when I tested the 2nd device that uses the similar supply. I'd assume that if some 3rd device had the neutral/line swapped, the 2nd device would spark too? I did check the AC voltage as your described and nothing, but also, I wasn't able to duplicate the problem in post #31 so I dunno.

I install a diode on the input to the 5 volt regulator (With a large value decoupling capacitor after the diode.) when it is feeding a microcontroller if the raw DC supply is also feeding a motor (Or similar load) that draws high current pulses which causes the raw input voltage to drop for a short time. In that case the capacitor after the diode supplies the 5 volt regulator but can not try to hold up the raw supply.

Interesting! Well the 12V supply is for relays so that sorta counts, no?

 

Re the unexpected low resistance reading in post #31. Are either of the LM 78 series regulators clamped to a grounded heat sink with an insulating washer ? If so could the insulating washer be damaged ?. Also are any of the large electrolytic capacitor casing possibly shorting to ground ? If it is possible the case is connected to the negative terminal of the capacitor.

Les.

 

Re the unexpected low resistance reading in post #31. Are either of the LM 78 series regulators clamped to a grounded heat sink with an insulating washer ? If so could the insulating washer be damaged ?. Also are any of the large electrolytic capacitor casing possibly shorting to ground ? If it is possible the case is connected to the negative terminal of the capacitor.

Interesting you should mention that and yes both are using insulating washers...I'll check that and the caps here in a bit.

On this version I moved the regulators from their previously position and nearly blocked the power switch with the heatsinks (like I had to remove the switch and press on the faston terminals, then put the switch back in place). So I had to use some shorter stamped fin type heatsinks but the mounting clips weren't in the correct position so I had to remove them. Currently, the heatsink(s) are floating.

I do have some new heatsinks coming tomorrow that fit. And while I'm at that I will replace the regulator just to see...i mean the problem is pretty much confined to just 10 components or so. I might just replace everything.

 

Ok, so all of the cap tops show ~1M to ground and I verified the heatsinks are floating. I found another clue though...

I repeated what I had been doing but this time with the scope off. As usual, the scope probe sparked and I removed power. But grabbed my meter and set it to DC volts and also got a spark to the meter probe! The meter measured like 9V between the diode and chassis.

Doesn't this seem to suggest the problem is not ground current related since the device sparked to the meter and that the device was off at the time?

 

You need do give a bit more detail in the description of events. You say " The meter measured like 9V between the diode and chassis." but you don't say which diode and which end. The fact that you got a spark to the meter probe suggests some kind of high voltage low current discharge. I have been assuming that the spark with the scope was between the scope probe ground lead (NOT the probe tip.) and the chassis. Is this assumption correct ? I suggest repeating insulation check between the transformer secondary (With the secondary and primary disconnected from the rest of the circuit.) and the laminations / mounting clamp, the primary and laminations / mounting clamp and between the primary and secondary using an insulation tester rather than the multimeter. In the UK with 240 volt mains we would set the insulation tester to 500 volts. I don't know what testing voltage would be used is you are in a 120 volt region.

Les

 

You need do give a bit more detail in the description of events. You say " The meter measured like 9V between the diode and chassis." but you don't say which diode and which end. The fact that you got a spark to the meter probe suggests some kind of high voltage low current discharge. I have been assuming that the spark with the scope was between the scope probe ground lead (NOT the probe tip.) and the chassis. Is this assumption correct ?

Hi Les, my apologies for not being more specific. No, the spark is occurring from D46's cathode to probe TIP. I earth the scope probe via it's alligator clip to the ground wire (close to C85's ground lead) that connects circuit ground to the chassis...I make this connection prior to powering on the device. When I measured the "9V between diode and chassis", I have the DMM probe negative clipped to the chassis ground (so it's out of the way) and the positive probed sparked a couple of times again at D46's cathode; the meter registered 9VDC. This again would be a correct voltage in that I had powered the circuit down already.

That's what is making this so weird based on my level of electronics: the circuit appears to be working correctly in that the measured voltages are correct with respect to chassis, but we're getting the spark from an innocuous point to a high impedance scope probe TIP. 11.65VDC (expected voltage at D46's cathode) should never cause a spark to probe tip & 9VDC certainly should never cause a spark to a DMM, especially when all ground connections check out. And we're getting a slight burning smell after about 10-20 seconds but any heat dissipates quickly enough that I can't locate it's source via smell or touching components. It just doesn't make sense.

I'll attempt to test the transformer again but I don't have an insulation tester. Any recommendations on a decent one that won't break the bank?

 

I built the insulation tester that was in the June 2010 edition of Silicon chip magazine but I find using my old second hand windup "Megger" more convenient. (It is quicker than putting batteries in the electronic one.) It is something that is used very infrequently so I don't leave batteries in the electronic one.The fact that the spark is to high impedance probes is very strange. I would only expect that to occur with very high voltages. (And I can't see any way voltages that high could be generated in that power supply.)

Les.

 

I wonder a bit now if that spark might be static electricity not at all related to the items at all. That is one explanation, it may not make any sense, I am not sure if it does or not. But if the spark did not repeat then it may be a possibility.

 

OK, I am pretty sure I found it...it's both dumb and sneaky at the same time.

I got the new heat sinks today and replaced the 12V regulator. Whenever I work on a power supply I always perform a resistance check on all of the supply nodes to ensure no place that should have power is grounded. I got to D46's cathode and briefly saw 0.2 ohms on the meter before it jumped to 800k or so. I checked a few more times getting only a high impedance before it dropped to 0.2 ohms again.

I got out my loupe and saw the problem: the top layer is a ground pour and just below D46's lead there's a small bit of exposed copper. So it looks like what was happening is as I was probing I was either pushing the lead down or bridging the gap from ground to lead with the probe tip. For whatever reason I recall lowering the ground pour isolation distance on this design so that decision likely contributed. And, this is the first time I've ever ordered a black PCB which makes this very difficult to see.

After that turned the supply on for about a minute without issue. Tomorrow I'll put a dab of epoxy over that spot and raise D46 up off the board a bit...hopefully that takes care of it!!

I'll report back but thank you everyone who helped, I really appreciate it!

 

It only needs to be a class-X cap if it is directly across the MAINS.

Actually this isn't true. I used to use a normal ceramic across the secondary until I studied the IEC 60950 and 60065 standards. The 2005 version 14.2.3 says:

Capacitors or RC-units across a secondary winding of a transformer with MAINS
frequency output, the short-circuiting of which would cause an infringement of the requirements
with regard to overheating, shall withstand the tests for subclass X2 capacitors or RC-units as
specified in IEC 60384-14, Table II.

They changed this to be even more inclusive (and vague) in a later version: "Capacitors operating at mains frequencies".

The reason makes sense if you think about it. The immediate secondary still represents a continuous supply of energy like mains, so if that cap were to short, then the winding current will only be limited by the impedance of the winding, wiring and shorted cap. We can't be certain this would blow the primary fuse or that there's a thermal fuse, and so the secondary is free to keep drawing current until it melts or catches fire.

The standard is ambiguous about whether the cap coming after a secondary fuse negates the requirement, however, I'd rather be safe.

 

I am trying to imagine the benefit of a capacitor across a transformer secondary when the voltage is mains frequency. Nothing comes to mind. For an inverter, probably. But not in a normal application.

 

I am trying to imagine the benefit of a capacitor across a transformer secondary when the voltage is mains frequency. Nothing comes to mind. For an inverter, probably. But not in a normal application.

I mentioned it earlier, but what it does is lower the resonant frequency of the winding so that any ringing the rectifier diodes cause when they switch is lowered in frequency as well, making it harder for that hash to couple into nearby circuits. There's a secondary advantage that ringing's loop area is reduced as it can return to its source through the X cap, rather than have to travel through the winding (which again makes it easier to find it's way into more circuitry). But that said, you're right; for most applications (save audio, precision measurement, and maybe medical devices) the cap is pointless.

 

One valuable thing shown by this whole incident is the value of having intentional test points in a system instead of probing around. The problem was a mechanical problem, not an electrical circuit issue. A separate pad for probing would have avoided the problem completely.

 

One valuable thing shown by this whole incident is the value of having intentional test points in a system instead of probing around. The problem was a mechanical problem, not an electrical circuit issue. A separate pad for probing would have avoided the problem completely.

Funny, I wrote "ADD TEST POINTS!!!!" in my build notes after figuring out the problem. And as stupid as this one turned out to be, there has never been a problem I've had in electronics that has not revealed some valuable lessons. There were a few here:

1. Add test points
2. Decreasing the ground pour isolation may make the PCB look better on the screen, but it doesn't take component leads or inaccurate scope probing into account.
3. Black PCBs are likely best for proven designs because they make exposed copper (and traces for that matter) hard to see.
4. When a problem manifests that appears to be defying the laws of physics, you should first ask how the known laws of physics apply to what you are observing, rather than approaching it as if the physics had suddenly broken down.

 

Yes, scope is earthed. The voltage between earth and the point that drew sparks was 11.6VDC (which is the correct voltage there). I also was finally able to insert the 1k resistor between ground and chassis and measured 28mVAC. Other attempts to measure AC voltages between say ground and chassis (when connected) or to the spark point kept falling towards 0mV, so the 28mV might be something.

Not only does the spark to a high impedance not make logical sense & the circuit appears to work, what's making this sooo much worse for me is is when I leave it on for 20 seconds or more I can start to smell something but I cannot identify where it's coming from either through smell or touch. It's as if this thing is like ready to blow. Very unnerving.

28 mvAC across 1k is 29 uaAC I don't think that is anything to worry about.

The point that drew sparks was measured to be 11.6 VDC. Did you scope that point to make sure it did not contain any AC?

 

The point that drew sparks was measured to be 11.6 VDC. Did you scope that point to make sure it did not contain any AC?

Ya, it did not contain AC. We solved the problem I believe above. There was a little spot where the solder mask was scratched off under the lead of the diode, exposing either the lead or the probe to the ground pour causing a short when probed.

I really appreciate your help!

 

OK, I am pretty sure I found it...it's both dumb and sneaky at the same time.

So, @Ian0 nailed it in post 8. Well, not exactly, but close enough.

Bob

 

So I wasn’t far out, back in post #8.

 

So I wasn’t far out, back in post #8.

Yep, and if you had known the pad he was probing was surrounded by ground plane, I suspect your answer would have been right on.

Bob