How to find short?

[Kaio Macedo]

Hello guys, I'm having trouble finding short ...

Even using a digital multimeter and power supply.

A test I use is to check the resistance between the positive and the negative battery connector.

But it has not proven functional, as to energize the digital power supply accuses short (high current consumption when off), belying the first test with the multimeter.

Another test is to inject make voltage points having 0ohm resistance with respect to GND, but in some cases not work, especially when the short is located in an integrated cicuito.

Does anyone know another method, and if possible an explanation of the case, thank you.

[MattHollands]

I have to say I'm not sure I understand what you've tried so far.

If you don't have a thermal camera, tracing down a short can be difficult. How many components are on the shorted rail - are there any components that are suspicious? If there aren't many components, you could try taking components off one by one to trace down the issue.

A picture of the board in question might help us to suggest a good approach.

[Kaio Macedo]

I have to say I'm not sure I understand what you've tried so far.

If you don't have a thermal camera, tracing down a short can be difficult. How many components are on the shorted rail - are there any components that are suspicious? If there aren't many components, you could try taking components off one by one to trace down the issue.

A picture of the board in question might help us to suggest a good approach.

No, it's not a specific case, is that in some cases it is really hard, as in the case of integrated circuits, which have to remove to confirm the problem.

I'm starting now, but thermal camera is a device that will buy.

The opening of the purpose of this topic is to know if someone uses some other technique to find short.

[jeroen79]

If no specific case is given, a few generic methods:

Trial and error:
Unplug parts until the current draw normalizes.
Test removed components individually.

Thermal:
Look or feel for things that get hotter than they should.
Use thermal camera, IR thermometer, thermocouple or fingers.
Smell anything burning? See magic smoke?

Follow the current:
Measure the voltage drop along wires or PCB traces.
Maybe use sensitive voltmeter or differential amplifier.
Put a current clamp around suspect wires.
Use non contact hall sensor or logic current tracer.

[rs20]

+1 to the "follow the current" approach. If you're frustrated with using your multimeter in resistance mode, replace the multimeter with a power supply set to 1/2 amps at 0.5V max voltage, and then use a decent voltmeter to follow the voltage gradient. The short itself will be at some voltage X, all points on the low side of the short will be < X, all points on the high side > X. The points with voltage closest to X in value will necessarily be in close physical proximity to the location of the short, and you can trial-and-error your way in like that.

[Brumby]

Thermal:
Look or feel for things that get hotter than they should.
Use thermal camera, IR thermometer, thermocouple or fingers.

Louis Rossmann has a simple technique.  Really cheap, too.

Spray the board with isopropyl alcohol, power it up and see where it evaporates faster than anywhere else.

[xavier60]

I find shorted MLCC capacitors on laptop mainboards by passing 500ma into the shorted rail from my bench supply. I reference my Fluke 87V near to where I have connected the negative of the bench supply to the ground plane. I then go looking for the highest voltage reading on the ground plane. The shorted part will have the highest voltage reading on its ground end. 10uV resolution or better is needed.

[xavier60]

This photo shows example readings at and near the shorted part.

[kripton2035]

you need some kind of milli ohm meter to search for this.
some to build here : http://kripton2035.free.fr/continuity-repos.html
good luck.

[Gyro]

The 'follow the current' approach has always worked fine for me, even on really complex boards.

Even a cheap 3 1/2 digit DMM has a 100uV resolution on its 200mV range. Set the current limit on the PSU in the 100mA to 1A range (depending on how 'chunky' it is) and voltage limit less than 0.5V. Follow the voltage gradients on the the GND and the Supply rails across the board. Measure from the supply connection points on the PCB (not the supply leads). With a cheap meter and 1A current you will get 100u ohm resolution, which is ample.

Even with supply and ground planes it's easy to home in on the short location. Follow the gradients on both, you may find that you get a better resolution on supply traces than you get on GND. Once you home in on the area of the short , just go round and measure the voltage across every component that receives power. That, and a magnifier, will locate the fault really quickly.

[bktemp]

Miliohm meters are difficult to use because you need a 4 wire measurement. The easiest solution is the previously described method of applying a couple of 100mAs and following the power traces until you find the point with the lowest voltage. Most multimeters have 1mV or 0.1mV resolution. At 0.5A that gives 2mohms or even 0.2mohms resolution. Good enough for most pcb traces.
I have used this several times and it always worked fine, even with mulitlayer pcbs with large power planes, because a via used to connect to components to the power planes has typically a large enough resistance to produce a measureable votage drop.

[xavier60]

I find it easier to take the voltage readings on the ground print on multi-layer PCBs because the + rails are too difficult to follow.
This method doesn't work very well when the short is inside of a large BGA IC.

[Gyro]

This method doesn't work very well when the short is inside of a large BGA IC.

You're into dodgy territory there no matter what technique you use (X-ray?). If the lowest voltages are on the decoupling caps under the BGA then at least you know that it is the BGA or its mounting.

[senso]

That cap looks cracked..
I wouldn't trust any of those caps, remove all of them and re-check for shorts..
Sorted some laptop boards because for no apparent reason 3 ceramic caps across the battery terminals in the board would go short and the laptop would not turn on, because there was way too much current going trough the first current sense(just after the dc-in) and the EC would kill it without any sign..

[MosherIV]

I have seen people use the  milli-onhm meter at various places I have worked.

A classic one is called 'tone-ohm' which makes varying tones depending on the resistance. Useful since you do not need to look at the meter.

If you have access to 4.5 digit DMM or better, you should be able to traces the short to near where the short is just by looking at the resistance. The resistance goes from a few ohms to between 0.3 to 0.1\$\Omega\$ (basically the contact resistance of the probes - ie you are near the dead short). You need the extra resolution that the digits give you in order to use this technique.

[Gyro]

Toneohms use kelvin lead connection, at least to the probe tips. Unfortunately with a standard 4.5 digit DMM (unless it's a bench one with kelvin [edit: leads]), lead resistance will dominate the reading.

[kripton2035]

a milli ohm meter without a 4 wires measurment is not a milli ohm meter, but almost a deci ohm meter...

[MosherIV]

One of my (ex-)colleagues traced a short on a PCB using an Agilent U3402 (5.5digit DMM) with just 2 wire resistance mode.
As I said, the resistance goes lower the nearer to the short you get.
Turned out to be a PCB manufacturing fault, these were 1 off prototype PCBs so it was not worth paying for the PCB manufacture tests.

Make use of what equipment you have at hand.

[Gyro]

Make use of what equipment you have at hand.

No issue there. It's just that he would have made much better use of all that resolution in voltage mode with a source of current rather than losing much of it with the vaguries of probe contact resistance.

Actually, given that your colleague had access to a 51/2 digit with 4 wire  measurement, he would have been much more sensible to leave it 4 wire resistance mode, connected the 'I' leads to board's the supply connections and used the 'V' lead probes to localize the fault. He really did have the equipment he needed to hand, he just didn't use it to its capabilities, he only needed a couple of clip leads. It would have been more accurate and probably faster, especially on a board using power and ground planes. Just because 2 wire mode worked well enough on one particular board it doesn't mean you shouldn't practice the most appropriate way of doing it.

[Seekonk]

You can go a lot higher current than 500ma,  that is why I have a 75A Xantrex power supply.  It can run all day into a short and be happy.  Traces can generally handle a couple amps and you don't need anything fancy to measure the millivolts.  Most times you can just feel what's bad (if it ain't smoking by that time).

[Neilm]

You can go a lot higher current than 500ma,  that is why I have a 75A Xantrex power supply.

Just be careful you don't get the chip logo burnt into your finger

[matkar]

I used Agilent U1272 hand meter  on ohms setting to roughly estimate the location of a short on a populated board. One time I had no trouble finding two shorts (bad reflow job). 4 wire measurement is not required. You are not interested in absolute value. You need to know whether the resistance has risen or dropped. This can be seen on a 4.5 digit meter.