Oops !

[Dave]

Standard prototype nightmare. You assemble a multi layer board with lots of
high pin count SM devices, plus lots of decoupling caps and so on. Final
check before power on - AAAAAAggghhh ! fractional ohm short between the
ground and power planes !!. The pcb is (probably) ok, because you did some
basic checks on it before you started assembling (and it was 'tested').

So, does anyone have any magic recipes for recovering from this ?

It may be a solder bridge or two, or a bad component (I've seen an 0603 cap
that was a solid short.)

Does measuring resistances help ? - ground/power planes are pretty low
impedance to start with

Even with reasonable tools, removing and replacing high pin count (100 -
200+) devices is somewhat risky.

I seem to remember that HP had a 'current sniffer' that enabled you to
inject pulses and track them. Would that work with a plane/plane short ?
Could you do something similar with a signal generator and a Spectrum
Analyser with a suitable probe ?


Dave

 

[[email protected]]

A four-wire ohm-meter can be useful in localising the short. You need
something with a lot of digits, because the resistance of the trace
leading to your short isn't all that high, and you will have to be able
to detect the decrease in this resistance as you walk along the trace
towards the short (or the increase as you walk away from it)

Even with a good ohm-meter it isn't a quick or easy job.

 

[Paul Burke]

fractional ohm short between the
ground and power planes
So, does anyone have any magic recipes for recovering from this ?

Start with the keys-under-the-lamppost method. Goggle at it for ages
with a jeweller's loop. Remove components that are easy to get off.

If that doesn't find it, try connecting a beefy power supply to the
rails, at a suitably low voltage (below the normal supply) and see if
you can blow the thing out, or at least see where it gets warm- assuming
it's a solder blob and not a VCC/ground pair swapped somewhere.

But the best bet with these things is, at least the first time, assemble
it incrementally and test as you go along.

Paul Burke

 

[Rob]

Start with the keys-under-the-lamppost method. Goggle at it for ages
with a jeweller's loop. Remove components that are easy to get off.

If that doesn't find it, try connecting a beefy power supply to the
rails, at a suitably low voltage (below the normal supply) and see if
you can blow the thing out, or at least see where it gets warm- assuming
it's a solder blob and not a VCC/ground pair swapped somewhere.

But the best bet with these things is, at least the first time, assemble
it incrementally and test as you go along.

Paul Burke

I've used this method with a current limited bench supply when all else has
failed. Set the volts low and current to zero. Crank up the current a bit at
a time looking/feeling & sniffing for a result.
rob

 

[Tony Williams]

A four-wire ohm-meter can be useful in localising the short. You
need something with a lot of digits, because the resistance of
the trace leading to your short isn't all that high, and you will
have to be able to detect the decrease in this resistance as you
walk along the trace towards the short (or the increase as you
walk away from it)

Yes. I managed to locate a shorting whisker with
a Keithley 197, which resolves to the milliohm.
As you say it is painful, but when you are
desperately desperate.......

 

[Jeroen Belleman]

Standard prototype nightmare [...] fractional ohm short between the
ground and power planes !!.

Use a suitably limited power supply to put an ampere or so through
the short and use a uV meter to trace the current flow. That should
locate the short to a few mm. Don't try to blow it out with too much
current. The short is not necessarily the weakest link.

It worked for me.

Jeroen Belleman

 

[Spehro Pefhany]

Standard prototype nightmare [...] fractional ohm short between the
ground and power planes !!.

Use a suitably limited power supply to put an ampere or so through
the short and use a uV meter to trace the current flow. That should
locate the short to a few mm. Don't try to blow it out with too much
current. The short is not necessarily the weakest link.

It worked for me.

Jeroen Belleman

And in the future, consider buying electrically tested boards for
complex prototypes. There's a good possibility the short is hidden
under a component body.


Best regards,
Spehro Pefhany

 

[Boris Mohar]

Standard prototype nightmare. You assemble a multi layer board with lots of
high pin count SM devices, plus lots of decoupling caps and so on. Final
check before power on - AAAAAAggghhh ! fractional ohm short between the
ground and power planes !!. The pcb is (probably) ok, because you did some
basic checks on it before you started assembling (and it was 'tested').

So, does anyone have any magic recipes for recovering from this ?

I just recently tracked down external layer power to ground short on a fully
populated board. Just set your power supply max voltage below the circuits
operational voltage in case the short accidentally clears. Set the current
limit to few hundred mA. Probe the voltage drop across every decoupling cap.
The lowest voltage drop will indicate that you are getting close to the
location of the short and it will very likely not be the decoupling cap but a
nearby IC pin. So far I have not had a single shorted ceramic cap nor a
shorted board due to poor manufacturing.

 

[Winfield Hill]

Tony Williams wrote...

Yes. I managed to locate a shorting whisker with a Keithley
197, which resolves to the milliohm. As you say it is
painful, but when you are desperately desperate.......

Most of those meters are remarkably wimpy when it comes to test
currents, i.e. 10mA. For this purpose I made a small 0.5A
current source that's limited to 0.5V maximum output to avoid
any possiblility of damaging a semiconductor. A 500mA current
makes it easy to see the increasing drop along a conductor as you
trace toward the short, using an ordinary handheld multimeter.

.. 10 ohms, 5W
.. +5V o----/\/\---+----o test-
.. ,--|<|--' lead
.. RTN o---+------------o jacks
.. Schottky

 

[John Larkin]

Tony Williams wrote...

Most of those meters are remarkably wimpy when it comes to test
currents, i.e. 10mA. For this purpose I made a small 0.5A
current source that's limited to 0.5V maximum output to avoid
any possiblility of damaging a semiconductor. A 500mA current
makes it easy to see the increasing drop along a conductor as you
trace toward the short, using an ordinary handheld multimeter.

. 10 ohms, 5W
. +5V o----/\/\---+----o test-
. ,--|<|--' lead
. RTN o---+------------o jacks
. Schottky

Or just a lab supply set to 0.5 volts or so, with a suitable current
limit.

John

 

[Joerg]

Hello Dave,

Standard prototype nightmare. You assemble a multi layer board with lots of
high pin count SM devices, plus lots of decoupling caps and so on. Final
check before power on - AAAAAAggghhh ! fractional ohm short between the
ground and power planes !!. The pcb is (probably) ok, because you did some
basic checks on it before you started assembling (and it was 'tested').

So, does anyone have any magic recipes for recovering from this ?

The "pcb is (probably) ok" statement makes me worry. What's the
confidence level there? Take another bare PCB from the stack and measure it.

Ok, here are the tricks I used: Supply it with VCC at a few amps. Then
take a millivoltmeter and solidly connect one probe to where VCC hits
the board (board side, not on the wire or other side of a connector).
Tap along the board with the other probe and look for a dropping
gradient until you find the "valley". It'll be shallow so patience is
required.

Another one that some consider a dowsing rod method: Take an EMCO near
field probe for H-fields. If you don't have one make a few loops of 1/2"
diameter at the end of a coax. Then ride a few hundred kHz on top of
VCC, 10-20mVpp is enough. For feeding in use very good coax such as quad
shield and lots of toroids slid onto it. Hook up to an analyzer or comm
receiver and sniff along. Toroids on that coax, too. #77 material or
something. Any short that isn't part of a plane but represents a small
"trace" will light up.

Then there is the IR camera trick. Usually a bridge is the least
conductive part of the whole scenario and tends to get warmer than its
surroundings.

Don't do the car battery trick to burn it out. You or someone else can
get hurt and if it was the PCB itself the stench can be awful.

Regards, Joerg

 

[Jim Thompson]

Hello Dave,


The "pcb is (probably) ok" statement makes me worry. What's the
confidence level there? Take another bare PCB from the stack and measure it.

Ok, here are the tricks I used: Supply it with VCC at a few amps.

[snip]

I had a 6V/100A supply. On bare screened thick film hybrid boards one
little pulse was all it took to remove "invisible" shorts ;-)

...Jim Thompson

 

[Joerg]

Hello Jim,

I had a 6V/100A supply. On bare screened thick film hybrid boards one
little pulse was all it took to remove "invisible" shorts ;-)

A tech who did that at one client needed a new shirt afterwards. Thank
God he wore glasses.

Regards, Joerg

 

[Jim Thompson]

Hello Jim,


A tech who did that at one client needed a new shirt afterwards. Thank
God he wore glasses.

Regards, Joerg

Pinhole faults only go "plink" ;-)

...Jim Thompson

 

[Jim Thompson]

Hello Jim,


If you happen to know it's a pinhole fault, that is. Non-pinhole faults
have a tendency to go "KABOOM".

Regards, Joerg

I was being somewhat facetious... plink "adjustment" was made after
ohming and microscope viewing ;-)

...Jim Thompson

 

[Joerg]

Hello Jim,

Pinhole faults only go "plink" ;-)

If you happen to know it's a pinhole fault, that is. Non-pinhole faults
have a tendency to go "KABOOM".

Regards, Joerg

 

[[email protected]]

I recently helped one of our junior engineers troubleshoot this problem
using this technique. Turns out the fellow neglected to read the part
of the data sheet about the thermal pad on the bottom of the quad flat
pack (many of the new ICs have them) - he had vias that were shorting
out to this pad.

Tom

 

[quietguy]

I pesume you have checked it with an ESR meter to eliminate the possibility of
a reversed semiconductor?

Next step (after visual inspection, esp for bridged pins on any of the SM ICs)
would be to isolate sections, if possible, to narrow the field a bit before
you start removing bits

David - who is sympathetic

 

[Richard H.]

So, does anyone have any magic recipes for recovering from this ?

It may be a solder bridge or two, or a bad component (I've seen an 0603 cap
that was a solid short.)

Try reflowing, at least all the high-pin-count / low-pitch parts. Use
plenty of liquid flux (e.g., an RA flux pen). If that doesn't work,
repeat and poke around with some desoldering braid to see what you might
wick out from behind the pins.

At least it worked for me. I brought up the first board of a proto AOK
step by step, got cocky with the second one and populated it fully,
complete with a power short. Spent a lot of time under a loupe,
probing, scratching with picks, etc. until I was convinced it was a bad
PCB. Reflowing the fine-pitch parts solved the problem. (Along with a
little solder wicking above the foot, to draw away excess solder but not
kill the joint under the foot.)

Even with reasonable tools, removing and replacing high pin count (100 -
200+) devices is somewhat risky.

Not much risk, but the parts probably aren't cheap and labor isn't
either. The air bath + air pencil system at http://www.zeph.com works
really well. The board takes no damage, and with some care the parts
can even be reused.

HTH,
Richard

 

[Dirk Bruere at Neopax]

Try reflowing, at least all the high-pin-count / low-pitch parts. Use
plenty of liquid flux (e.g., an RA flux pen). If that doesn't work,
repeat and poke around with some desoldering braid to see what you might
wick out from behind the pins.

At least it worked for me. I brought up the first board of a proto AOK
step by step, got cocky with the second one and populated it fully,
complete with a power short. Spent a lot of time under a loupe,
probing, scratching with picks, etc. until I was convinced it was a bad
PCB. Reflowing the fine-pitch parts solved the problem. (Along with a
little solder wicking above the foot, to draw away excess solder but not
kill the joint under the foot.)




Not much risk, but the parts probably aren't cheap and labor isn't
either. The air bath + air pencil system at http://www.zeph.com works
really well. The board takes no damage, and with some care the parts
can even be reused.

HTH,
Richard

Pump a load of current through it and see where the smoke comes from?

--
Dirk

The Consensus:-
The political party for the new millenium
http://www.theconsensus.org