Steven Beckman sent the following message asking questions regarding SIR
testing.  As the chairman of the SIR Task Group, I have attempted to answer
them.  I apologize in advance for my long windedness.

Douglas Pauls
Contamination Studies Laboratories
[log in to unmask]


Subj:	SIR Test Methods
Date:	95-09-20 04:17:45 EDT
From:	[log in to unmask] (BECKMAN_SJ)
Resent-from:	[log in to unmask]
To:	[log in to unmask], [log in to unmask]
     
     There is much written, but seemingly little consensus, on SIR Test 
     methods.  I'm using IPC-TM-650, number 2.6.3, as a reference.  Any 
     insight that can be provided regarding the following open issues, 
     in the context of IPC Level 2 testing of printed circuit card 
     assemblies, would be appreciated.

*  I agree that there has been little consensus up to this point on SIR
testing.  The committees have only recently agreed on terminology between SIR
testing vs. electromigration testing.  Many companies have their own test
methods, which work for them.  Add to this the fact that there are dozens of
factors which affect SIR testing, and you can begin to see why there is so
much confusion.  One of the efforts of the IPC SIR Task Group has been to
write a tutorial type of manual or handbook which gets us all on the same
playing field, talking the same lingo.  Then we intend to work on improving
the test as a predictive tool.

Level 2 testing, I assume you are talking about J-STD-001 and not the Phase 2
ODC alternatives testing, has been set to a particular SIR profile,
IPC-TM-650, method 2.6.3.3, Rev A.  This standardizes the test to a
"non-condensing" ramp method.  The appropriateness of SIR testing for
J-STD-001A, Appendix D has been extensively debated, and I could go on for
quite some time.  This might be better done in the next IPC SIR Task Group
meeting in Providence.

     
     1) Bias voltage level, nominal set to -48vDC, but some argue in favor 
     of lowering the voltage closer to operating level, i.e. -5 or -15vDC?

*  This touches on the issue of typical operating voltage seen in service,
vs. the use of a higher voltage as an accelerating factor.  As with any test,
there are advantages and disadvantages to each.  The original 48 volts, in
the IPC Phase 1 ODC project, was based on work done at AT&T.  I have found it
to be a reasonable biasing voltage, certainly more so than the 100s of volts
in military specs.  The only time I would really suggest a 5 or 15 (or any
lower bias) would be when the dielectric material cannot sustain a higher
voltage, as in the case of very fine lines and spaces.

     2) Effect of REVERSING the voltage polarity when going from the BIAS 
     to the TEST sequence, e.g. if bias is -48vDC, going to 100vDC for 
     measurements?

*  The issue of reversing polarity has also been debated.  The use of the
reverse polarity was based on the work of Emory Gorondy, DuPont.  His
technical papers on the subject are part of the IPC technical library.  The
theory was that if you had an electrolytic solution, formed from contaminants
and condensed water vapor, any mobile ions present would gradually migrate to
the respective cathodes and anodes under an electrical bias.  Appling a
measurement voltage of the same polarity might then yield a resistance that
was higher then it could be, since the mobile charge carriers were already
migrated.  By reversing the polarity, the mobile ions would "sweep" back
across, yielding a more representative resistance or indication of
contamination.

I have experimented with both ways of doing measurements.  I have not really
seen a difference between normal bias and reverse bias. The jury is still out
on this issue.
     
     3) Appropriateness of the "ohms/square" conversion principle in 
     establishing insulation resistance thresholds for non-standard 
     patterns, e.g. inter-leafed footprints?

*  I will reveal a personal bias (no pun intended), I don't believe in the
ohms per square concept as it applies to SIR testing.  The concept is taken
from the ASTM methods on sheet resistance.  It's central assumption is that
you have a uniform contaminant across the surface. In most cases on PWBs,
this is not the case.  Some of the SIR research conducted by Duncan Culver's
Telecom SIR group showed that the ohms per square normalization didn't really
work, or that it was not a consistent effect.  The effect of increasing the
number of squares, in my view, is to increase the sensitivity of the pattern
to contamination.  You should still keep in mind the concept of minimum
resistance between any two polarized paths.
     
     4) Pass/Fail criteria:
     
        a) Absolute threshold per IPC (100Mohm) or Bellcore (3Gohm) even    
           with non-standard patterns

*  The issue of pass-fail criteria in SIR testing has been endless debated.
 The bottom line question is "What does this mean to my product in the
field?".  Since I view SIR data as pattern geometry dependent, the pass-fail
criteria cannot be easily compared between the two.  Each approach has good
points and bad points.  I like Bellcore's approach of having different SIR
levels depending on if it is a bare board, solder masked, solder coated, etc.
 It is more realistic than the IPC's approach.  The advantage of the IPC
method is it states a minimum resistance applied to different patterns.  I
don't care what the geometry is, it must be above level X.  The concept of
pass-fail criteria is one to be addressed by the IPC SIR Task Group.

        b) Scaled threshold per "ohms/square" concept

*  I must beg off on this one.  What is a scaled threshold?

        c) Two decade drop in insulation resistance anytime during test
     
*  If this were used as a pass-fail method, you must be certain that the two
decade drop is due to the contaminants on the board and not the test
conditions.  Having a decrease in the water quality of the humidity chamber
can globally kill SIR levels.  In such a case, the boards might be squeeky
clean, yet fail your criteria.  If you don't have an SIR test professional
who knows his (or her) stuff, you can easily fail good product.  I've seen it
happen (elsewhere, of course).  You must also specify where that two decade
drop is measured from.  Is it between minimum values, mean (geometric)
values, etc.?  Is is two decades between initial ambient readings and final
ambient readings or from the initial readings to the first elevated
temperature readings?  Will you have a different set of "drop" criteria for
differing levels of temperature and humidity?  Lots of questions to answer.

     
     Thanks very much.
     
     Steven Beckman
     Lockheed-Martin Commercial Electronics
     603-885-2550 
     
     [log in to unmask]