cctalk Digest, Vol 16, Issue 2

From: Dwight K. Elvey <dwight.elvey_at_amd.com>
Date: Wed Dec 1 16:31:36 2004

>From: CRC <technobug_at_comcast.net>
>
>
>On 1 Dec 2004 09:43:56, Paul Koning <pkoning_at_equallogic.com> commented:
>
>>>>>>> "Jules" == Jules Richardson <julesrichardsonuk_at_yahoo.co.uk>
>>>>>>> writes:
>>
>> Jules> I've got a board here with a short between ground and the +5V
>> Jules> rail (actually not quite dead - I'm getting around 10 ohms
>> Jules> between the rails)
>>
>> Jules> Any useful tips for finding the fault? It's a large board,
>> Jules> multi-layer, lots of silicon on it unfortunately :-(
>>
>> 10 ohms means about 0.5 amps -- sounds like a perfectly respectable
>> amount of current for a board with "lots of silicon" to draw.
>> So... are you sure there is a short, rather than just circuitry?
>>
>> paul
>
>Unless there are passives (like terminators) between +5V and GND, 10
>ohms is a little low. 0.5 amps represents a minimum current that the
>board will pull. An ohmmeter will not normally cause the silicon to
>conduct and hence be counted in the total resistance. Some chips do
>have internal dividers, but the they are nomally in the 10K + range.
>
>I normally use a current limited supply at the required voltage and set
>the limit so that it starts to limit. I then use a 5 digit voltmeter to
>trace where the current is going. This is done by placing one probe on
>the +5V input and then measuring the drop as you trace out the
>distribution path. The path drawing the current will continually drop
>while non-drawing paths will remain at a constant voltage drop.
>
>On multi-layer boards you measure drops to the Vcc pins of the bugs and
>map the drop.
>
> CRC
>
>


Hi
 I use a slightly different method. I use a current limiting
supply and set it to about an amp or two. I place the supply
across one of the power planes ( corner to corner on a solid
plane and end to end for E traces ). I do not place the supply
form power to ground, just across the one plane.
 I then place one lead, of a meter that reads 5 digits on a
200 mV scale, from the opposite rail that is shorted two.
One then simply traces for a null on the plane with the
current flow across it. On a plane, I take
a piece of string to mark the line of the null and then
put the supply on 90 degrees, across the same plane. Where
the new null line cross the old string is the where the short is.
 This method will even work if the short is +1K, if that is
the only current flow. I once showed using it on a burnin
board with a 10K resistor that was in wrong out of some
400 sockets and about 12 resistors per socket. I was able to
zero right into the exact resistor.
 The principle is that one uses the plane as though it were
a variable resistor and the short was the contact. If one
measures along the resistor to the contact, the voltage
will be zero at the location where the contact is made.
This is much easier to read than CRC's method since you
are looking for a null, rather than a stop in the change of
the voltage. It also has the advantage that it won't blow
an intermittent short so you can find it. It is more sensitive
than the HP current probe for power planes ( having used
both methods ). It also is run at a very low voltage so
there is no chance of damaging other parts.
 I publish this method every now and then but most people
don't think about how it works and will tend to forget it.
Such is!
Dwight
Received on Wed Dec 01 2004 - 16:31:36 GMT