Subject: | |
From: | |
Date: | Tue, 16 Jan 1996 11:25:30 -0600 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
.int [log in to unmask]
Mike Cutter Asked:
>> 1. In bed-of-nails testing, does forcing the output of a device to a known
state by
driving the e-c junction affect the boards reliability?
Back driving is Usually not harmfull BUT you should always try to design to
avoid back driving. The key work is usually (2 or 3 sigma, nowhere near 6) . If
you design for test and leave the tri-state pin accessible for the tester and
only back drive when you have to, you will be OK. If you ground control pins
and do not add tri state to your PALs, GALs and ASIC, you will get burned.
Factors involved are:
Type of part being backdriven (Gutsy, Wimpy drive).
Type of part being tested and its test time. (Takes longer to test a RAM than a
simple gate).
Number of parts connected to the backdriven pin. (More parts, more time).
Sometimes the tester and test part being tested will oscillate, and Ive seen
10 volt p-p on the output being back driven. Usually on PALs/GALs. Our rule is
to never back drive these parts.
The next factor is false failures and the repairs they cause. When backdriving,
the waveforms get very bad (Ringing, Risetime, Glitches) and this causes GOOD
parts to fail the test. Often good parts are replaced with other good parts. We
know rework is bad for reliability.
>> 2. Is there a bed-of-nails tester that that works strictly as a monitor;
that
does not require each node to be stimulated individually?
I assume you are asking because you dont want a test pad on each net. You can
cluster test blocks of parts with most any Incircuit tester. The problems is
Fault Coverage is unknown and often low and developing these tests are costly
as they are manually written instead of automatically generated. Pay more and
get less. When we have done this, we havent been very happy with the results.
Larry Sternig
[log in to unmask]
|
|
|