Troubleshooting 'shorted' power rails

[John Larkin]

On Tue, 21 May 2019 06:07:18 -0700 (PDT), George Herold
<gherold@teachspin.com> wrote:

On Tuesday, May 21, 2019 at 1:20:37 AM UTC-4, dagmarg...@yahoo.com wrote:
On Monday, May 20, 2019 at 8:37:00 PM UTC-4, Winfield Hill wrote:
dagmargoodboat@yahoo.com wrote...

Bill Sloman wrote:

Microvolt sensitive DVM's are cheaper ...

But a microvolt DVM can't track down a diode-to-ground,
which was the problem.

Yes, it can. If starting from the V+ rail, use the
DVM to measure drops at each fork, and see which one
is carrying current. Voltages on the ones that aren't
don't change. Eventually this narrows down to the
offending item. As you progress past the item, the
voltage drops stay the same, revealing the spot.


--
Thanks,
- Win

I certainly understand the theory, I just don't see this application
of it.

Let's see -- my test current was 30mA, the most I dared without
risking a board full of pricey ICs. That'll make roughly 12uV
across the 400-ish micro ohms of a 100x70mm 1oz. Cu power plane.
Doesn't that sound like a bit of a handful to hand-probe?

It would help if the fault were at constant temperature, but it isn't.
Vf ramps down as the component warms up from the dissipation.

I suppose one could let the board thermally stabilize, then follow the
uV drops, being careful not to touch and warm one's test probes.

In my first case there were five faults (five wrong components
stuffed), which would be mighty fancy tracking / simultaneous
equation-solving with a voltmeter of any kind.

Hitting it with freeze spray while watching Vf would be fast and fun.
A -10C step would be easy to see (maybe even a good time to drag out
an old-school analog VOM .

Cheers,
James Arthur

Hi James I like your technique. It's too bad we don't have better
built in heat sensors (like a snake). (Our lips are fair...
but almost kissing a pcb is going to get one looks from colleagues. :^)

Romancing The Board.

Fingers are pretty good temperature sensors. And RF oscillation
dampers.


--

John Larkin Highland Technology, Inc

lunatic fringe electronics

 

[Tom Gardner]

On 21/05/19 15:43, John Larkin wrote:

On Tue, 21 May 2019 06:07:18 -0700 (PDT), George Herold
gherold@teachspin.com> wrote:

On Tuesday, May 21, 2019 at 1:20:37 AM UTC-4, dagmarg...@yahoo.com wrote:
On Monday, May 20, 2019 at 8:37:00 PM UTC-4, Winfield Hill wrote:
dagmargoodboat@yahoo.com wrote...

Bill Sloman wrote:

Microvolt sensitive DVM's are cheaper ...

But a microvolt DVM can't track down a diode-to-ground,
which was the problem.

Yes, it can. If starting from the V+ rail, use the
DVM to measure drops at each fork, and see which one
is carrying current. Voltages on the ones that aren't
don't change. Eventually this narrows down to the
offending item. As you progress past the item, the
voltage drops stay the same, revealing the spot.


--
Thanks,
- Win

I certainly understand the theory, I just don't see this application
of it.

Let's see -- my test current was 30mA, the most I dared without
risking a board full of pricey ICs. That'll make roughly 12uV
across the 400-ish micro ohms of a 100x70mm 1oz. Cu power plane.
Doesn't that sound like a bit of a handful to hand-probe?

It would help if the fault were at constant temperature, but it isn't.
Vf ramps down as the component warms up from the dissipation.

I suppose one could let the board thermally stabilize, then follow the
uV drops, being careful not to touch and warm one's test probes.

In my first case there were five faults (five wrong components
stuffed), which would be mighty fancy tracking / simultaneous
equation-solving with a voltmeter of any kind.

Hitting it with freeze spray while watching Vf would be fast and fun.
A -10C step would be easy to see (maybe even a good time to drag out
an old-school analog VOM .

Cheers,
James Arthur

Hi James I like your technique. It's too bad we don't have better
built in heat sensors (like a snake). (Our lips are fair...
but almost kissing a pcb is going to get one looks from colleagues. :^)

Romancing The Board.

Fingers are pretty good temperature sensors. And RF oscillation
dampers.

And RF oscillation creators (only in circuits designed
by other people, naturally).
"Amplifiers oscillate, oscillators won't".

 

On Tuesday, May 21, 2019 at 10:40:41 AM UTC-4, John Larkin wrote:

On Mon, 20 May 2019 22:20:32 -0700 (PDT), dagmargoodboat@yahoo.com
wrote:

On Monday, May 20, 2019 at 8:37:00 PM UTC-4, Winfield Hill wrote:
dagmargoodboat@yahoo.com wrote...

Bill Sloman wrote:

Microvolt sensitive DVM's are cheaper ...

But a microvolt DVM can't track down a diode-to-ground,
which was the problem.

Yes, it can. If starting from the V+ rail, use the
DVM to measure drops at each fork, and see which one
is carrying current. Voltages on the ones that aren't
don't change. Eventually this narrows down to the
offending item. As you progress past the item, the
voltage drops stay the same, revealing the spot.


--
Thanks,
- Win

I certainly understand the theory, I just don't see this application
of it.

Let's see -- my test current was 30mA, the most I dared without
risking a board full of pricey ICs. That'll make roughly 12uV
across the 400-ish micro ohms of a 100x70mm 1oz. Cu power plane.
Doesn't that sound like a bit of a handful to hand-probe?

Don't be shy, pump in an amp or two from a power supply set to make
0.5 volts max.

At 0.5V my fault would only suck milliamps. Remember, it was a set
of diode junctions.

At least two of the rails so affected had 100's of mV saw-tooth
waveforms on them, some sort of UJT-ish parasitic relaxation
oscillators being formed somewhere.

Then use a DVM with uV resolution. Probe any pairs of
points where the plane is accessable, parts or vias, and see which
direction the current is flowing.

But personally, first I'd crank in a lot of current and thermal image.


It would help if the fault were at constant temperature, but it isn't.
Vf ramps down as the component warms up from the dissipation.

I suppose one could let the board thermally stabilize, then follow the
uV drops, being careful not to touch and warm one's test probes.

In my first case there were five faults (five wrong components
stuffed), which would be mighty fancy tracking / simultaneous
equation-solving with a voltmeter of any kind.

Multiple shorts calls for amps and thermal imaging.

Fire your assembler!

Not mine to fire. But it was Digikey's wrongly-labelled bags
that occasioned the problem. Digikey has committed the sin of
being so good that people don't think to double-check them.

Cheers,
James

 

[Mike]

On 5/20/2019 4:20 PM, Tom Del Rosso wrote:

dagmargoodboat@yahoo.com wrote:

TDR might be good for finding shorted planes. But boards and parts
are generally so good these days, I've never had that happen.

Yeah I'm pretty good at that.

But seriously how can a TDR find a short on a plane?

30 years ago, I tried to TDR a 10 layer board with multiple
power/ground planes. Didn't have much luck locating the
short, but I did find that I needed more bypass caps.

Borrowed a thermal imager and found the short in short order.

One issue was normalizing emissivity. Somebody suggested
spraying the board with foot powder. That worked GREAT.
Problem was that I couldn't get it off, so the board was
unusable.

 

[Tom Del Rosso]

Mike wrote:

On 5/20/2019 4:20 PM, Tom Del Rosso wrote:
dagmargoodboat@yahoo.com wrote:

TDR might be good for finding shorted planes. But boards and parts
are generally so good these days, I've never had that happen.

Yeah I'm pretty good at that.

But seriously how can a TDR find a short on a plane?



30 years ago, I tried to TDR a 10 layer board with multiple
power/ground planes. Didn't have much luck locating the
short, but I did find that I needed more bypass caps.

Borrowed a thermal imager and found the short in short order.

One issue was normalizing emissivity. Somebody suggested
spraying the board with foot powder. That worked GREAT.
Problem was that I couldn't get it off, so the board was
unusable.

If foot powder worked then there must be something else that would.
Maybe a sheet of some kind of paper on top. Put some fans under the
board to suck the sheet down.

 

And RF oscillation creators (only in circuits designed
by other people, naturally).
"Amplifiers oscillate, oscillators won't".

the version I heard 40 years ago is even more pessimistic..

Amplifier oscillate.
Oscillators oscillate on more than one frequency

 

[Michael Kellett]

On 20/05/2019 21:22, dagmargoodboat@yahoo.com wrote:

On Monday, May 20, 2019 at 2:23:56 PM UTC-4, Martin Rid wrote:
dagmargoodboat@yahoo.com Wrote in message:
I had two cases in a row recently of many-layer power-planedboards with power supplies that wouldn't come up above a diodedrop, and I was able to use a trick devised for the first one tosolve the problem on the second one, too --I set a current-limited supply to ~30mA & left the boardto stabilize thermally. I then looked for the 'hot' partswith an IR thermometer. Worked like a charm.Temp. rise was about 1oC @ 20mW. The IR thermometer? $14well-spent.Cheers,James Arthur

I remember the Shortsqueak. Like a logic probe. Worked on
resistance, and changed pitch as you got closer to the short. I
think Nuts&volts had something similar too.
Cheers

--


----Android NewsGroup Reader----
http://usenet.sinaapp.com/

One of my most-treasured tools is a buzz-box I made ages ago from
an LM3909, that lets you easily hear the difference between a short,
an ohm, and tens of ohms.

But that wouldn't have been much help here.

Since it was acting like a diode, I considered using freeze spray
and watching the voltage until I hit a part that made the rail
wiggle appropriately.

Cheers,
James Arthur

I use a variation of that technique, power the board with a fixed
voltage and watch the current as you warm suspect parts by wafting the
soldering iron about near them. I've often found problem parts like this
- usually power and ground pins swapped but occasionally blown or faulty
parts.

I once managed to get a processor, with swapped power and ground pins,
to work by disconnecting them all, making a copper foil ground plane on
top of the chip and connecting all the power pins with nice thin bare
copper wires. Once was enough !

MK





---
This email has been checked for viruses by AVG.
https://www.avg.com