In a message dated 12/3/98 12:06:53 PM US Eastern Standard Time,
[log in to unmask] writes:

> Doug and Terry @ CSL, are you listening?

Sure, when someone yells loud enough <grin>.

Steve refers to a conversation I have at least once a week.  First, the units
micrograms of NaCl equivalence per square inch is NOT the same as micrograms
of sodium or micrograms of chloride per square inch that ion chromatography
gives.  Ionic conductivity machines, such as the Omegameter, do not give you a
measure of either sodium or chloride.  As soon as we get the bugs worked out
of our web page, I'll post my standard text of why this is.

> I've enjoyed all the wonderful technical analysis regarding the differences
>  between Ion Chromatography and NaCl equivalence, which comes from the bulk
>  ionic conductivity "process control" procedure.  Many past discussions and
>  research projects have established that the bulk ionic process control is
> NOT a quantitative lab methodology, but more akin to using a piece of
hanging
> rope as a weather forecasting tool.

I don't know that I would go THAT far, but I agree with the principle.  The
ROSE test, in all its various forms was designed for process control - it was
designed to look for Changes.  If I had a 5 last week, 5 yesterday, 5 this
morning and this afternoon I get a 10, SOMETHING changed and I better go found
out why.  MIL-SPECs not withstanding, it was never really intended to be a
measure of "is this GOOD, is this BAD".  The EMPF, as part of the IPC Ionic
Conductivity Task Group, conducted an extensive study a few years ago and
showed the large problem with repeatability or accuracy with any of the
machines tested.  So there is extensive data available that agrees with
Steve's assessment that ROSE testing is not a good quantitative lab method,
but more a general "what ballpark am I in".

>
>  The NaCl pass fail was devised in 1972, and as such, doesn't really apply
to
>  today's designs.  Some of the ion species being detected in your IC tests
do
>  not contribute to field failures.  Many field failures occur at chloride
>  levels that pass the bulk ionic.

I had the fortune to work with both Mr. Hobson and Mr. Denoon at NAC, prior to
Mr. Hobsons death and Mr. Denoons retirement.  Both were amazed at what the
military did with their simple ballpark test.  The original ROSE test was
developed in the 1972 time frame.  The accept/reject work, based on some of
their work correlating ROSE to SIR, was done a little later, in the 1975-76
time frame.  The automated ROSE tests were developed in 1978.  Dr. Jack Brous,
now retired from Alpha Metals, was also heavily involved in this work.  ALL of
the initial work, accept/reject parameters, and control parameters were
established with high solids rosin fluxes.  The parameters were based on the
solubility of those rosin residues.  As such, the historical military pass-
fail criteria are not valid for other fluxes, which have totally different
chemical structures and totally different solubilities and electrical
characteristics.  But, logic never stopped the military before......

About half (or more lately) of the work in our lab is tracking field failures
back to the ultimate problem in manufacturing.  In a majority of the cases we
see, the assemblies were perfectly acceptable when the historical ROSE limits
were blindly imposed.  But the ROSE testing does not often point out either
problem chemical residues or a localized concentration of flux residues.  Our
monthly column in Circuits Assembly magazine points out many such instances.

>
>  If some customer has blindly imposed this spec on you, you need to educate
him.

Heavens, yes.  This applies to both bare boards and finished assemblies.  The
two questions I usually ask are:  "Why are you imposing this specification and
level on the manufacturer?" and "Why do you feel that this will protect you
from contaminated bare boards or assemblies?".  The responses are most often:
"Ahhhhh........"; "I don't know, we've always done it this way"; and "because
I don't have anything better that I can use".

The third response is somewhat valid.  For specifying board cleanliness, ROSE
is FAR more widely available than ion chromatography and MOST people have some
familiarity with ROSE.  Much less expensive too.

>
>  If your fellow employees have imposed this requirement, you need to
challange
>  the validity and develop a test and limit that provides real value to your
>  company and its customers.

Amen, brother, amen.  But just as many people don't know about the valid and
invalid uses of ROSE, far fewer have the background to go about determining
how to determine "how clean is clean" and "how do I set up valid control
numbers".

Delphi Delco Electronics, here in Kokomo, Indiana, has begun to specify bare
board cleanliness based upon an initial qualification using ion chromatography
and process monitoring using both the standard ROSE test and also a modified
ROSE test (which the IPC is working on developing as a TM-650 method).  It is
their C-7000 specification.  Surprisingly, most vendors have not battled the
imposition of this kind of specification for board cleanliness.  Many had to
scramble up the learning curve in a hurry, but it is leading to better product
overall.  Of course, with the volume that Delco does, what Delco wants, Delco
gets.  Delco took the IC qual and ROSE monitor route, because thier testing
showed far too much variation in the board residues coming into their no-clean
assembly operation.  A HUGE amount of research went into that spec.  I have
recommended to a number of clients that when looking for a fabricator, find
one that is supplying to the Delco C-7000 spec.

Realizing that a nickel-gold board will have different residues than a tin-
lead board than a palladium board, etc., they do break down cleanliness
requirements depending on mask and metalization combination.

>
>  Does anyone want to offer a sane set of limits to be imposed on components
based on IC testing???

Depends on your definition of "sane" <grin>.

In our opinion, there is no longer any Golden Numbers, or a single cleanliness
measure that can be blindly applied to all items.  As an example, CSL has 4-5
cleanliness recommendations for bare boards, depending on what the base
material is, the metalization, and the kind of mask.  We have 3-4 cleanliness
recommendations for assemblies, based upon whether it is high solids RMA, low
solids fluxes, or water soluble fluxes.  For components, we have another half
dozen, depending on the application.  Hybrids are much more sensitive than
other components.  A simple 7400 14 pin plastic dip has a far different
sensitivity to contamination than will a cerdip op-amp or other high impedance
device.  Some components, such as transformers, are relatively "dirty" by ion
chromatography, but it is due to how they are manufactured and those residues
usually don't hurt the board.

So when I am asked "how clean do I need to be?", the answer is always "it
depends".  It depends on the application, it depends on the materials used, it
depends on the end-use environment, it depends on the density of the assembly,
etc.

I am often asked to defend our cleanliness recommendations.  "How do you KNOW
this level of chloride/bromide/sulfate is bad?", usually when trying to
convince a fabricator or assembler that there is a problem.  Our
recommendations were developed over time from our failure analysis work.  If
we see corrosion and metal migration on assemblies, and the associated residue
level is X, and after the residue is reduced by cleaning to a level of Y, and
the problem goes away, then the critical point was somewhere between X and Y.
Over time, a histogram develops with two populations, one where the residue
level is in the region of Y and no problems occur, and another in the region
of X, where electrochemical failures occur.  Our recommendations are generally
the line between the two populations.

We call our recommendations "guidelines" because they should not be rigidly
applied to all instances.  We use them as a starting point until other testing
shows a different accept/reject level.  Some clients have residue levels that
are clean by our standards, but still have problems due to high density or
other design issues.  Others have residue levels that lead me to believe
"these should be failing almost immediately", yet have no problems in service.
Cleanliness needs to be determined on a case by case basis.

Well,  I hope some have found this of use.

Doug Pauls
Technical Director
Contamination Studies Laboratories
Kokomo, Indiana

################################################################
TechNet E-Mail Forum provided as a free service by IPC using LISTSERV 1.8c
################################################################
To subscribe/unsubscribe, send a message to [log in to unmask] with following text in the body:
To subscribe:   SUBSCRIBE TechNet <your full name>
To unsubscribe:   SIGNOFF TechNet 
################################################################
Please visit IPC's web site (http://www.ipc.org) "On-Line Services" section for additional information.
For technical support contact Hugo Scaramuzza at [log in to unmask] or 847-509-9700 ext.312
################################################################