Finding power - gnd shorts

Faced with the same problem I had no choice but to use a dremel tool to do a binary search to isolate the problem to a particular half, then quadrant, then octant, then ....

Of course a Real Man(sm) would just blow the short out with AC line voltage.

Un bel giorno rickman digitò:

Increase the current until you see something heat up or melt. :)

--
emboliaschizoide.splinder.com

I've heard a story of some big expensive board that had similar problems, I believe dust or something on the films for the pcb made some tiny shorts.

some brave guy just took the biggest PSU he could find and puff they all worked

-Lasse

can

ng

can

his

find

A small compass

If you were a really old fart like me, you'd have a dusty old Toneohm on one of the back shelves of the lab. Ask around.

If this is a 4-layer board, take a careful look at any mounting holes. I forgot to do that on a recent design and the inner layers came right up to the plated-through mounting holes for one module.

I had to drill out the mounting holes and warn the end user to mount the module with nylon screws until the next board revision.

Mark Borgerson

but you have checked anyway, right ?

Normally the 4 wire 'microvolt gradient' method is enough to narrow down the location.

That, and a highlighted view of the PCB layers, showing just the two offending nets, can give you a smaller set of 'candidiate locations' for a short.

Find the lowest voltage diff point, and then vary the current 2:1 and get a milliohm value from the dV - that also gives a clue of what can be the cause, and possible solutions.

Copper or solder whisker faults can be cleared with a fusing current, applied in the right place :)

-jg

Ground is a plane as well? If so, most of the power will be dissipated in/around the short.

Spray the board (or section by section) with cold spray until you get a thin layer of ice (just a white haze). Then run a current big enough to get some power in the short. If you're lucky, you will see a spot defrosting more rapidly then the rest of the board. Try to spray evenly, uneven spraying can defrost in different times too.

The fancy option would be a heat camera, but I suspect you don't have one sitting on a shelf. ;-)

Maybe filming the defrosting board and playing it back slowly will help?

--
Stef    (remove caps, dashes and .invalid from e-mail address to reply by mail)

Finding the short is easy:

1) Turn up the PS current limit until until the voltage jumps and the current drops to 0. 2) Examine board for charred spot.

;)

We used to use a current probe to find shorts on blank PC boards.

--
Grant Edwards                   grante             Yow! In 1962, you could buy
                                  at               a pair of SHARKSKIN SLACKS,
                               visi.com            with a "Continental Belt,"
                                                   for $10.99!!

Rick, what ever you do, we want to see it on You Tube ;-) Looking forward to that, Best Regards, Dave

First try gradient. Dump in constant current, switch the DVM to the

200uV range and probe at 1/2" or so distance. IOW the probes "walk" behind each other in lockstep. Move the trailing one for greatest gradient, then keep going until you see a steep drop or reversal. That should lead you to the "sink". Of course this won't work well if there is more than one short. I once had a whole series of tested (!) boards that had four shorts each.

If this fails try to get a hold of a camera that is somewhat infrared sensitive. Snap a pic in the dark, load onto PC and stretch the histogram to wazoo. Sometimes that shows a distinct hot spot or possible more than one. The temperature at the short is usually a lot higher than elsewhere on the plane.

If you are near San Francisco maybe John Larkin lets you put it under his FLIR camera. That ought to show it.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

... snip ...

Have you taken an unstuffed board and tested that assumption?

--
 [mail]: Chuck F (cbfalconer at maineline dot net) 
 [page]: 
            Try the download section.

Is this multi-layer or 2 layer?

I have seen creases/folds on inner layers (8 layer board) cause this type of problem.

If you still have any blank boards I would test them FIRST!

I have seen bare board testing that basically follows the netlist for continuity, but does NOT check for track to track or track to plane shorts.

First obvious thing is does ANYTHING get noticeably warm/hot.

Without knowing the circuit you could have a combination of faults (reversed components, interplane short, wrong component on a power amp can cause some combinational effects).

Let alone a drop of solder from a soldering iron bridging two power rails I once did by accident.

When you measure 1mV have you just moved one probe. Moving both probes gives to nearest and further away points on the two tracks to determine voltage gradient. The gradient method and how tone ohms basically work (the closer the two probes are to the short or one point, the lower the impedance, giving usually a higher frequency o/p).

See everybody elses comments as well.

--
Paul Carpenter          | paul@pcserviceselectronics.co.uk
    PC Services
 Timing Diagram Font
  GNU H8 - compiler & Renesas H8/H8S/H8 Tiny
 For those web sites you hate

I had this some years ago with my very first 4 layer PCB. Aaaargh! I took a board and sawed it in half: the short should be in either the left or the right half. It was in both. So I sawed the two halves in half. The short was now in all four pieces....

Eventually I took a belt sander and removed the outer PCB layers, thus revealing the fault: the VCC layer has VCC connections, while the ground layer had both VCC and ground connections. Armed with this, and the Gerbers, I went back to the manufacturer... who was forced to confess that some underling had decided that the two layers were meant to be merged. So I eventually got my clean PCBs, but it took several weeks all told.

JS

If it is a lot of board, and they are not too expensive, Try more current. 1 amp is a lot. Something should be getting warm. If it is not the board, that leaves wrong / bad components, backwards parts, and a schematic error.

But tested to prove that they conform to the supplied CAD data. Some CAD systems will pass ERCs but if there are any manually split planes problems can be in the Gerbers.

Geo

A couple of people have mentioned the heating effect but if you only have a maximum of 10mV and 1A that is only 10mW. Even with a 10A limited supply, 100mW is not much dissipated over a couple of copper planes?

Geo

A short is ohmic, so with a current of 10A, the voltage increases to 100mV, power goes up to 1 whole Watt (P = I^2 * R). Enough to see at least some heating effects. But as mentioned, with my method of freezing the board a little, some luck is involved as well, it does not always work.

--
Stef    (remove caps, dashes and .invalid from e-mail address to reply by mail)

Did they also claim they had been 'bare-board tested' ? ;)

They may well have passed, as many testers learn from a sample!

-jg