TECHNET Archives

February 2000

TechNet@IPC.ORG

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Brian Ellis <[log in to unmask]>
Reply To:
TechNet E-Mail Forum.
Date:
Mon, 28 Feb 2000 12:15:41 +0200
Content-Type:
text/plain
Parts/Attachments:
text/plain (610 lines)
Per-Erik

Thanks for your support. I knew, when I wrote what I did, that I was opening a big can
of worms and that many would be shocked and a few thinkers, like yourself, would agree,
at least partially, with what I admit were my controversial statements.

However - and, unfortunately, in my state of semi-retired bliss, I cannot substantiate
what I say with references as I have been forced to reduce my library from an ad-hoc
room lined with shelves to one small bookshelf - I think I may be able to supply the
answer to the 100 megohm dilemma. It was simply the practical limit that electrometer
valve (tube, west of the Atlantic) instruments could measure (do I remember right? Was
it the Philips EF80?) before the FET was invented? At that time, most people used the
Megger, being far cheaper than electrometer systems. This was a magneto-type device
where you turned a handle until a clutch slipped, when it was supposed to be generating
a very rough waveform 100 V or 500 V, depending on the model. A double moving coil
without hairsprings would measure the volts and amps and indicate the insulation
resistance. The meter scale was non-linearly calibrated and the 100 megohm mark was the
first one, about 1/2 mm from the infinity mark and the same again from the 50 megohm
mark. The crunch was that turning the handle transmitted sufficient movement to the
instrument, even if you pressed it hard onto a table, that the meter needle would
oscillate by about +/- 2 mm, so the 100 megohm pass/fail criterion was little more than
an inspired guess. It is also because the Megger was available in 100 and 500 V
versions that these voltages were used in all the early specs - and to hell with all
notions of voltage gradients. This is what I mean about inherited legacies. Those were
the days, my friends, and, unfortunately, I am old enough to remember having used these
methods! :-(  For you younger guys, you will be able to post on the Technet, in due
course, that we used to interconnect with a putrid system called the printed circuit
and we actually used lead solder to fix the the components, believe it or not: your
turn cometh! :-)

Now, coming back to the clean slate: how can we get this moving? Unfortunately, as an
old has-been, I am no longer financed by anything other than the pension I live on plus
the odd consultancy job. My moving and shaking is therefore more or less limited to
this beautiful island. Unless... However, my first imperious suggestion is that we coin
new terms to replace all the old ambiguous ones and define them absolutely with
scientific rigour. Then, before we can start thinking about new standards, we need a
scientific study of what all these phenomena are all about and the contributions of
such things as the ions in the substrate materials, the effects of the lack of
homogeneity (reinforcements), the effects of curing the substrate resins, the effects
of humid air (far from negligible), etc. Parallel with this, we should try to correlate
reliability with both the theoretical and practical parameters. Only then should we
start to formulate standards. Believe me, guys, what I know about SIR - and I venture
to suggest the same applies to my illustrious colleagues on this thread - is just the
veriest tip of the iceberg. But I am stating a fact and not just blowing my trumpet
when I state that I know a helluva sight more about the subject than 80% or more of the
guys who use the technique, slavishly adhering to standards that date from cable
insulation resistance, but not understanding a thing. OK, this is a zillion dollar
programme I'm talking about but it will all have to come out in the wash, sooner or
later (especially when the EU start issuing Directives on the subject!!).

If I can help, I will.

Per-Erik, if you happen to be thinking of taking your holidays in Cyprus this year,
please let me know and we could get together (this is not a spam for island holidays!!)

Brian

Tegehall Per-Erik wrote:

> Brian, Doug, Graham,
>
> I have just read last weeks discussion with great interest since this is my
> favorite subject.
>
> I agree that using SIR for qualifying a flux with a bare board with copper
> comb patterns is of little value. Passing such a test is no guarantee for
> that it will not cause SIR problem on a product when combined with other
> contaminants (as shown by Adams et al., Circuit World, Vol 20, No 2, pp
> 41-44, 1994) which I think you all agree with. Still, the test could have
> some value since it could be used to sort out bad flux candidates early in
> the developing process before too much money has been used up. To assure
> that a flux will not cause any reliability problem on a product, it must be
> evaluated using a test board representative for the materials and processes
> that will be used for manufacturing the product, as required in J-STD-001B,
> Appendix D. Thus, I am a devoted supporter of the methodology in Appendix D
> using representative test vehicles. BUT...
>
> Something is seriously wrong with Appendix D. In the minutes from the
> Soldering Subcommittee at PC/SMTA Electronics Assembly Expo '98, it was
> concluded that "the industry appears to be ignoring the materials
> compatibility part of 001B more and more. Few have used 001B Appendix D for
> process qualification and those who have used it have not passed the test".
> This means that either can we expect poor reliability of today's electronics
> or else is the test method and the acceptance criterion not relevant. I
> believe it mainly is the latter, although that does not excluded that many
> products might not have an acceptable cleanliness.
>
> What is then wrong with the method and the criterion? I cannot but come to
> the same conclusion as Brian. The problem is that there is practically no
> scientific base for neither the test method nor the criterion. The whole
> business is so full of distorted definitions, misunderstandings,
> non-scientific practice and criteria that the best thing would be, to use
> Brianīs words, "start again with a totally clean slate".
>
> There are many things that can be questioned. I will give some examples.
> First, just measuring SIR is a problem since the measured currents are very
> low. Perhaps more important is the fact that the measuring of SIR will
> affect the measured value. In 1996, Chan showed that the contamination of a
> board with 10 microgram per sq. inch table salt caused a decrease of the SIR
> from 10 Gohm to 5 Mohm. When a bias of 100 V DC then was applied to the
> contaminated board, SIR increased to 40 Mohm after 60 seconds, to 1 Gohm
> after a few hours and to 8 Gohm after 100 hours. This is the reason for the
> practice of having a stabilization time of usually one minute. Despite that,
> this will create some uncertainty in the measured value. Another interesting
> point is that after 100 h of applied bias, the SIR was about the same for
> the contaminated board as for a clean board. This means that you cannot
> discriminate a rather highly contaminated board from a clean one, not what
> you would expect of a good method. The increase in SIR is brought about by a
> depletion of ions due to the DC bias causing migration of the ions. By
> reversing the bias, the ions will migrate in the opposite direction and
> cause a decrease of SIR. Hence, many methods prescribe that measuring
> potential should have reversed polarity and the measurements should be taken
> after a stabilization time of one minute. In the investigation performed by
> Chan, he showed that after reversing the bias, it might take one hour before
> the SIR is affected and 10 hours before minimum in SIR is reached. So using
> reversed measuring bias is a good method to assure that you have no idea of
> what you have measured.
>
> Then we have the very high polarization and measuring voltages prescribed by
> most standards. Is that really relevant for products using voltages of 5 V
> or less? Brian has written many good papers about this.
>
> What surprise me most is that, through the years, I have seen practically no
> discussion of the relevance of the acceptance criteria in standards. The
> criterion in Appendix D is that the SIR measured after 96 and 168 hours
> should be minimum 100 Mohm (for non-coated assemblies). The criterion is the
> same independent of the type of application and severity of the service
> environment. It is quite obvious that that is not how it should be. Also,
> bearing in mind the result found by Chan that a board contaminated with 10
> ug/sq. inch had about the same SIR as a clean board after 100 h, it seems
> strange to only have requirements for SIR measured after 96 and 168 h and
> completely disregard the first 96 hours.
>
> The 100 Mohm requirement for SIR is not new. I have tried to track the
> origin of this requirement and hoped for finding the reasons for why this
> has been chosen as acceptance criterion, but with little success.
> Interestingly, in an IPC technical paper from 1985 (IPC-TP-542) it is
> mentioned that the practice at that time was to require a SIR of 100 Mohm
> but then only as a requirement for initial resistance. Why was that changed?
> I cannot get rid of a feeling that this was done because it was to hard to
> pass that requirement and not because it was found to be more relevant. Just
> like the SIR requirements for printed boards once was lowered because
> otherwise it was to difficult to pass the requirement for HASL boards when
> they become popular to use. So how do we know that the present criterion is
> relevant? Since I failed to track the origin of the 100 Mohm requirement,
> but also the reason for changing from having a requirement on initial
> values, I would appreciate if someone could shed some light to this.
>
> Much more could be said about the non-scientific approach of the present
> standards for SIR testing but I think this contribution already has become
> to long as it is. To sum up, I give Brian my vote for starting with a new
> totally clean slate.
>
> Per-Erik Tegehall
> IVF
>
> > ----------
> >
> > Graham
> >
> > You asked for a reply, so here goes:
> >
> > 1. Where I'll be going, the heretic will be burnt sooner or later!
> >
> > 2. I'm glad you said British and not English: otherwise, I'd be after you.
> >
> > 3. Keep moving? I was 35 years in Switzerland. If that does not show
> > staying power,
> > please tell me what does. At least, in Cyprus, I get the number of  days
> > of sun
> > tripled, so that I can look after my orange trees.
> >
> > 4. Target to hit - I was always that: the price for being outspoken. I
> > must shamefully
> > admit that, over the past 20 years, I have become an even easier target,
> > having put on
> > some 30 kg of pure fat and not an ounce of muscle.
> >
> > 5. IC v. SIR v. ionic contamination testing: each of them serve their
> > purpose and they
> > are certainly not interchangeable. With IC, you can measure the quantities
> > of
> > **specific** ions, provided you have the appropriate columns/detectors. It
> > does not
> > measure ALL ions. For example, if you set it up to measure, say, Cu, Sn
> > and Pb as
> > cations and Cl, Br and SO4 as anions and somebody neatly salts your
> > assembly (e.g. from
> > an onshore storm or from the salt aerosol thrown up by tyres on a salted
> > road), then
> > you will measure nothing. You need numerous columns to get the overall
> > picture. It is
> > also very difficult to measure some organic anions. On the other hand,
> > ionic
> > contamination testing will give an indication of the aggregate of all the
> > ionic
> > species. It has the supreme merit of being the only practical process
> > control method.
> > SIR, as I've indicated in earlier postings, measures an aggregate of
> > different
> > resistances in parallel, some of which are contamination related, some of
> > which are
> > not. There is great confusion as to the definition of SIR. It is the
> > resistance, as the
> > name implies, of the surface insulation of a substrate. It is NOT
> > detecting the
> > presence of dendrites. That is ECMR (electrochemical migration
> > resistance). Neither
> > will give you a picture of the reliability of a given assembly due to
> > contaminants, at
> > least without many qualifications. To illustrate this let me present a
> > hypothetical
> > situation:
> > A flux (or paste) is qualified using one of the standard IPC test patterns
> > and the
> > overall fluxing/soldering/cleaning process is seen to be OK. On the basis
> > of this,
> > production starts. One of the components is a large fine-pitch QFP with a
> > very small
> > standoff. First of all, residues enter under the component. They are
> > partially removed
> > during the cleaning. The resultant diluted liquor is pushed to one side by
> > the air
> > knifing, where it slowly evaporates off but not before it has dissolved
> > some ionic
> > species from the component itself. As the volume diminishes, the liquor
> > will, by
> > surface tension, concentrate itself in the spaces between the QFP leads,
> > exactly where
> > the resultant dried contamination will cause the most harm. No amount of
> > SIR testing
> > will forecast the reliability when this happens - and happens it does.
> > Furthermore, the
> > substrate used may be the same as the IPC test coupon, in that they are
> > both FR-4, but
> > that does not mean that the epoxy chemistries are the same and the
> > reaction between the
> > flux and the substrate may be VERY different, even by more than an order
> > of magnitude
> > from make to make of FR-4, let alone other substrate types. Why this
> > difference? Simply
> > because the physico-chemical bonds between the flux residues and the
> > substrate are
> > different: this depends much on the epoxy chemistry and the MW of the
> > surface
> > molecules. I have observed half an order of magnitude difference between
> > different
> > batches of the same substrate and, I suspect, possibly a difference on
> > whether the
> > laminate was pressed on the outside or the inside of a packet.
> >
> > In other words, there is no way that you can predetermine the reliability
> > of an
> > assembly. We have imperfect tools, each one of which may give us a small
> > piece of the
> > jigsaw puzzle, but a lot of the pieces are missing from the puzzle box.
> >
> > The big weakness of SIR is the time it takes to conduct a test. In most
> > factories, the
> > assemblies would already be delivered to the customer before a test is
> > terminated. It
> > is therefore useless as a process control tool. However, in 1988, I did
> > propose an
> > overnight non-destructive SIR test for cleaned assemblies, but this never
> > caught on.
> >
> > Retsina is actually a soldering flux. However, the clean slate is proposed
> > to rid
> > ourselves of all the bad legacies SIR and ECMR have been associated with,
> > and still
> > are. Basically, SIR is a function of the substrate and not of the
> > contaminants and,
> > until this realised, confusion will reign - for ever and ever. I know I'm
> > not alone in
> > this school of thought: there are several well-known names within our
> > industry who
> > agree with me.
> >
> > Now, to set the ball really rolling: a) the de Bono type corrosion tests
> > are similar to
> > ECMR in that they do not reproduce the conditions obtained in practice and
> > b) what is
> > the influence of substrate chemistry and contaminant chemistry on the
> > electrical noise
> > generated?
> >
> > Going from Latin to French, bonne chance (it's lucky I cannot e-mail you
> > in Greek!)
> >
> > Brian
> >
> > Graham Naisbitt wrote:
> >
> > > Good Morning everyone,
> > >
> > > Doug, Graham is not more restrained than you - I've been out for a
> > couple of
> > > days visiting customers! However, burning the heretic Brian, is perhaps
> > a
> > > little strong - even for you. Always bear in mind that Brian is a
> > British
> > > exile first (to my knowledge) to Switzerland now to Cyprus. SIRBrian has
> > to
> > > keep moving because that makes him a more difficult target to hit!!
> > >
> > > Scott, sorry I had to get that bit out of the way first.
> > >
> > > You state:
> > >
> > > > I have been observing this thread with interest.  We run standard SIR
> > > > coupons monthly through our production floor to monitor our cleaning
> > > > processes (SMT, wave, DI wash, water-soluble flux chemistry, standard
> > > > production processes and handling).
> > >
> > > Can you please inform me the method you use, the type of coupon and is
> > the
> > > test pattern overmounted with components?
> > > Are you running this test using coupons from each process stage or
> > simply on
> > > "completed" coupons?
> > > >
> > > > My question:  Is all of this necessary?  Would it be possible to gauge
> > > > acceptable cleanliness levels based on another type of measurement,
> > such
> > > as
> > > > Ion chromatography?  Or does IC miss things that SIR picks up?
> > >
> > > Is it necessary? Well what is the level of deviation you have seen from
> > your
> > > SIR results?
> > >
> > > IC will inform you EXACTLY what is on the surface - but, unless you are
> > Doug
> > > it cannot inform you if the product is reliable.
> > > SIR will inform you if the product is reliable, it cannot inform you
> > what is
> > > on the surface that may be causing a reliability problem.
> > >
> > > >
> > > > It's plausible that localized contamination, say underneath a low
> > profile
> > > > QFP, might not raise the overall contamination levels much, but would
> > have
> > > a
> > > > detrimental localized affect.
> > >
> > > Agreed. As I always point out, the industry desire to use a simple pass
> > /
> > > fail ROSE level of <1.5microgrammes/square centimeter NaCl equivalence
> > (in
> > > line with another current TechNet thread that Joyce posed) also implies
> > that
> > > it is acceptable to leave UP TO that level on every square centimeter of
> > > your assembly. If you manufacture fine-line fine-pitch and / or use CSP
> > > devices, then this level will almost certainly prove to be too high. As
> > an
> > > example, I know that certain sections of IBM employ a level < 0.2.
> > >
> > > In respect to Brian's suggestion to "start with a clean slate", I cannot
> > see
> > > why, unless you have discovered a new application for Retsina? The NPL
> > > carried out an extesive research project some years ago, into different
> > > methods of defining "clean, reliable" (my interpretation) electronic
> > > assemblies. They listed some 22 different methods most of which are
> > > destructive.
> > >
> > > The 2 best were considered IC and SIR. There was also a consideration
> > for
> > > Impedence Spectroscopy, but that hasn't materialised yet.
> > >
> > > As I have just stated, there is an application for both methods to be
> > used.
> > > SIR to inform you the level of reliability by its measuring the degree
> > of
> > > current leakage from REPRESENTATIVE coupons processed in parallel with
> > your
> > > production. This at least seem to be what Scott is doing at Benchmark.
> > >
> > > Then if a problem is discovered, IC would be useful in determining the
> > EXACT
> > > nature of the dendritic failure.
> > >
> > > And if my Latin isn't too rusty - Brian suggests he is "I am (either) a
> > > develish (or) diabilical advocate (lawyer or attorney) who is in the dog
> > > house!?"
> > >
> > > As I said in my old latin school book from dumpty-dum years ago: Latin
> > is a
> > > language as dead as dead can be, it killed the ancient romans and now
> > its
> > > killing me".
> > >
> > > I look forward to Scotts reply - and no doubt Brian and Dougs!!
> > >
> > > Regards,
> > > Graham Naisbitt
> > >
> > > [log in to unmask]
> > >
> > > WEB: http://www.concoat.co.uk
> > >
> > > CONCOAT Ltd
> > > Alasan House, Albany Park
> > > CAMBERLEY GU15 2PL UK
> > >
> > > Tel: +44 (0) 1276 691100  Fax: +44 (0) 1276 691227
> > > > -----Original Message-----
> > > > From: Brian Ellis [mailto:[log in to unmask]]
> > > > Sent: Thursday, February 24, 2000 10:36 AM
> > > > To: [log in to unmask]
> > > > Subject: Re: [TN] AW: [TN] components cleanliness
> > > >
> > > >
> > > > Doug
> > > >
> > > > I thought that one would get you to rise to the bait :-) Am surprised
> > > Graham
> > > > hasn't reacted, either.
> > > >
> > > > Seriously, I consider this an important problem as very few assemblers
> > > test
> > > > the
> > > > bare boards they receive, let alone the components. I consider that
> > there
> > > > are
> > > > also many other ignored parameters, some of which I brought up during
> > > > various
> > > > IPC and ISO WG meetings, to be shot down - for the most part - in
> > flames.
> > > >
> > > > Unfortunately, I do not have an answer to all these problems. I'm a
> > great
> > > > one
> > > > for actually putting the test patterns on the real PCBs, as close to
> > > > critical
> > > > components as possible, but even that is less-than-ideal :-(
> > > >
> > > > If the truth be told, I honestly believe that we should forget all we
> > have
> > > > in
> > > > the way of ECMR and SIR standards and start again with a totally clean
> > > > slate,
> > > > beginning with the basic definitions which have become so distorted as
> > to
> > > be
> > > > misleading. There are FAR too many legacies from test methods and
> > > standards
> > > > that
> > > > date from 30, 40 or more years. Not only did we not have an
> > understanding
> > > of
> > > > what it was all about, then, we didn't have sufficiently good
> > instruments
> > > > and
> > > > even most of those available today leave a lot to be desired. I have
> > never
> > > > seen
> > > > an in-depth scientific analysis of all the mechanisms that can
> > influence
> > > the
> > > > results. From time-to-time, I've brought up some practical
> > observations,
> > > > over
> > > > the past 20 years, some of which have created discussion and others,
> > > > yourself
> > > > likewise, have done the same. Furthermore, we need to address present
> > > > issues,
> > > > let alone future ones. Who, e.g., has done anything about SIR on HDISs
> > > with
> > > > track spacings down to 25 um?
> > > >
> > > > Advocatus diaboli sum (and probably in dome canis est, by your
> > lights!).
> > > >
> > > > Best regards
> > > >
> > > > Brian
> > > >
> > > >
> > > >
> > > > [log in to unmask] wrote:
> > > >
> > > > > In a message dated 02/23/2000 3:09:10 AM US Eastern Standard Time,
> > > > > [log in to unmask] writes:
> > > > >
> > > > > >
> > > > > >  This is the main reason why SIR testing is, in many cases,
> > useless to
> > > > > >  qualify a flux: you qualify it under lab conditions, totally
> > ignoring
> > > > > >  the fact that its behavious will alter in practice due to it
> > mixing
> > > > with
> > > > > >  external contaminants not present on your test vehicles.
> > > > > >
> > > > >
> > > > > SIR USELESS?  Gasp!! Burn the Heretic!!!!!!!!!!
> > > > >
> > > > > SIR can be useful, even as a lab tool, you just have to use the
> > right
> > > > > combination of materials, processes, environment and data
> > > interpretation.
> > > > I
> > > > > would agree that we have a long way to go before our existing
> > methods
> > > are
> > > > a
> > > > > good and true measure of field performance.
> > > > >
> > > > > Doug Pauls
> > > >
> > > > ##############################################################
> > > > TechNet Mail List 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 web site (http://www.ipc.org/html/forum.htm) for
> > > additional
> > > > information.
> > > > If you need assistance - contact Gayatri Sardeshpande at
> > [log in to unmask] or
> > > > 847-509-9700 ext.5365
> > > > ##############################################################
> > > >
> > > > ##############################################################
> > > > TechNet Mail List 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 web site (http://www.ipc.org/html/forum.htm) for
> > > additional
> > > > information.
> > > > If you need assistance - contact Gayatri Sardeshpande at
> > [log in to unmask] or
> > > > 847-509-9700 ext.5365
> > > > ##############################################################
> > > >
> > >
> > > ##############################################################
> > > TechNet Mail List 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 web site (http://www.ipc.org/html/forum.htm) for
> > additional
> > > information.
> > > If you need assistance - contact Keach Sasamori at [log in to unmask] or
> > > 847-509-9700 ext.5315
> > > ##############################################################
> >
> > ##############################################################
> > TechNet Mail List 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 web site (http://www.ipc.org/html/forum.htm) for
> > additional
> > information.
> > If you need assistance - contact Keach Sasamori at [log in to unmask] or
> > 847-509-9700 ext.5315
> > ##############################################################
> >
>
> ##############################################################
> TechNet Mail List 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 web site (http://www.ipc.org/html/forum.htm) for additional
> information.
> If you need assistance - contact Keach Sasamori at [log in to unmask] or
> 847-509-9700 ext.5315
> ##############################################################

##############################################################
TechNet Mail List 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 web site (http://www.ipc.org/html/forum.htm) for additional
information.
If you need assistance - contact Keach Sasamori at [log in to unmask] or
847-509-9700 ext.5315
##############################################################

ATOM RSS1 RSS2