Using an IR camera to find a short on a PCB - looking for tips & tricks

[cvanc]

Hi all-

Looking for tips from those who've been here before.  I need to find a short on a PCB that's dragging down a +/-15V rail.  Here's a few details:

1- Very generic architecture; a single center-tapped secondary feeding a discrete bridge, then into a 7815 & a 7915.
2- Both ends of the secondary have one of these resettable fuses http://www.wondhope.com/smd/smd2920/smd2920-050.pdf
3- As soon as power is applied the fuses heat to about 250 degrees F
4- The inputs to the regulators only rise to about 3 or 4 volts
5- All 4 rectifiers test good in-circuit; all electroltyics have been replaced (they were no-name junk but changing them didn't alter the symptoms)
6- The PCB has a fair number of opamps hanging across these rails

I assume one of the opamps has faulted, or maybe one of the voltage regulators.  But the fuses drop so much voltage that the downstream stuff does not heat up enough to show on my E30bx (hacked as usual).

I'm thinking about jumpering around the fuses and slowly bringing it up using a Variac on the AC input.  I'll watch for PC traces & chips to glow on my camera and kill the power as soon as I have seen the hot spot.  But this has some risk, and i thought I'd ask here if other approaches might be available.

Any inputs and suggestions are most welcome.  Thanks.

[boB]

I cases like this, I like to drive the rail voltage with a power supply limited to something less than the 3.3V or 5.0V or whatever and a couple of amps current limit and do kind of like you're saying....   Watch for smoke while watching the temperature.  This, just to get some dissipation in whatever is hogging the current.

Sometimes the bad part will blow itself up, but if it goes open, then you can't find it and it's too late.  Hence the limited current.

boB

[DaneLaw]

Here can you get an idea of the example you can get on PCB circuitry with a budget thermal camera at 320x240 (my case sub300USD)

I have it on a cheapo-microscope stand to I can easily adjust that hysteric depth of field, but for live-coverage the framerate becomes a usefull factor but havent tested it to find faulty shorts yet but do reckon it will be a helping hand with this level of details for current flow....'




First clip with above stand, the other two not.

[Terry01]

Post a few decent photos of the PCB in question and the chances are the guys here will be able to point you to the most likely culprits. Like someone else said i'd use a dc power supply and start low volts & amps and just bring both up slow. You may even follow the rail with a meter too and it may help you find the short before it has a chance to burn a component out.

Good luck however you go and hope you find the problem and get it fixed easy as possible.

[Vipitis]

to me this looks very great.

[digsys]

I've used both tricks on many occasions - ie made up a ~2-5mm dual probe to trace voltage drop (using very sharp pins), plus the thermal camera.
A hint for the thermal method - freeze the PCB if you can, then apply power ... even low power usage tracks will stand out like varicose veins :-)

[eKretz]

Yup, if you don't have a good thermal contrast you have to create one. You can also just stick a board in the fridge for 30 seconds then check it under power. Don't leave it in the fridge too long because you don't want condensation when it's taken out. You can also run a fan on it while testing... The hot spots will stay hot since they're creating heat. Many ways to skin that cat.

[mikeselectricstuff]

If a device is drawing excess current chances are it's toast anyway, so overriding any current limiting is unlikely to kill anything that wasn't dead anyway.
A device will light up on a thermal camera way before it smokes, so using a PSU with variable current-limit, and winding up the current limit until something lights up will usually get you to the problem very quickly.

[cvanc]

OK I finally got around to fixing this device and I thought I'd share the method I used.  IR imaging was the key to a fast diagnosis.

First, the item is a small stereo audio DAC made by Emotiva, it's model 'Stealth DC-1'.  It has a decent reputation.  I got it cheaply in 'for repair, has fault' condition.  Overall 1 is a photo of the main board with the lid & front panel removed.

I verified it was not passing any signal and had excessive DC at the outputs.  However, it did boot and the front panel display & controls seemed to behave as they should.  Probing the power supplies I found the +/-15V rails were way low.  So I got out the IR camera (a hacked E30bx) and took a look.

The power transformer has 3 secondaries and each of them has these SMD 'resettable fuses' on them (see post #1).  They were getting very hot which essentially opened the circuit and allowed only a trickle  of current to pass.  See image Hot fuse.

The only way to troubleshoot I could think of was to short out these resettable fuses and externally provide current limiting that I could control.  My goal was to stay safe & not burn anything up but I needed to see what was getting hot downstream of the fuses.  They were in the way of an easy diagnosis.

[cvanc]

So I put some temporary jumpers across them (photos Fuse link 1 and 2), and got out my trusty light bulb current limiter (Limiter).  I have incandescent bulbs in several different wattages for use in the limiter, depending on the fault mode current I feel safe in allowing.  Because this was a low power device I used the smallest bulb I have - 15 watts.

With the 15W lamp in series with the AC input I turned power on while watching the FLIR.  In just a few seconds one opamp (an LM4562 in the headphone amp stage) got really hot (Hot LM4562).  It quickly got to over 180°F (remember this is with external current limiting active).  Nothing else on the board was remotely close temperature-wise.  So I yanked it and put in a new one.

I kept the limiter in circuit and powered back up - the temperature was much better.  So I bypassed my limiter and now (with full power applied) the new 4562 was running a much more sedate 125°F (Normal 4562).  And that was all it took.  I verified proper operation and put it back into service.

It's an interesting failure, though... the LM4562 is in the headphone stage as I mentioned, but it does not drive the headphone jack directly.  It drives a stereo pair of BUF634 buffer amps (Headphone amp).  So it's pretty well protected from any bad loads that might get connected to the headphone jack.  My buffers were both perfectly fine, only the driver opamp had failed.  So I'm not sure exactly what made this failure happen.

[Fraser]

Excellent result 

My thermal cameras have been excellent for tracing components that are either getting too hot or not getting hot when they should ! That information can help with further checks, depending upon the scenario.
I have found whole sections of embedded PC's that are effectively in an 'off' state and cool when they should be warm or hot.... the cause .... the section is being held reset by a fault and it only remains to chech why the Reset state is active in that area alone. I have also found IC's getting warm in equipment when they should be in a low power standby or off state. The cause was the failure of the power control circuit, in one case a failure in the Power MOSFET control circuit on one sub system supply rail.

I love thermal cameras in PCB inspection or thermal profiling. Sometimes such can be VERY enlightening, especially when looking at PSU circuits. Cost cutting by the OEM can lead to some PSU's that are guaranteed a limited life through excessive heating of components. If caught before failure, action can be taken to reduce temperatures and so stress on components.

Even a relatively low resolution thermal camera can help in better understanding what is getting hot, or not, on a PCB Far quicker and likely less painful than the old finger tip test I used to practice before I had thermal cameras or IR thermometers.

Fraser

[floobydust]

...  So I'm not sure exactly what made this failure happen.

You get an ESD discharge plugging in headphones.
Walking across a room with the headphones and jacking-in is enough you can even hear the discharge in the headphones, there's enough energy there.
A -ve transient can go past the output stage and damage the driver stage.

Well-designed audio gear has protection TVS diodes at headphone jack outputs.
 

[cvanc]

You get an ESD discharge plugging in headphones.

Yeah, it must've been something like this.  The person I got it from said it failed while he was fiddling with the cabling & connections.

[cdev]

I have some LCD coated plastic sheets (sort of like a mood ring) that turn different colors based on temperature.

This stuff will show you what on a PCB is getting hot instantly.

Its thermal imaging on the cheap. It costs very little and it will tell you exactly what on a PCB is getting hot, fast.

[GeorgeOfTheJungle]

Sometimes it's super easy: