Hi again, Brian, et al.,
Thanks for further insights into Contaminometer operation. Nice to know that is helps interpret measurements of standard deviation as well, that's another critical part of process control. Understanding the measurement process is fundamental in any process control effort. Believe I've said this in the past, but am convinced it's true.
Brian, when you say out-of-tolerance, do you mean out-of-control? One of the things about SPC is that if a process is in good shape, you get the warning before anything goes out of tolerance. If a process is bad, you can get out of tolerance but still be in control, in which case the process needs to be revamped. To be precise, I'm thinking about control charts for individual measurements.
Which leads to another thought. Gary, you say the data are not normal. Does this mean, the data taken as individual measurements over a long time? I'm assuming that's the case. If a process were getting out of control, it would not produce normal data. By the way, how did you determine it's not normal? There are numerous tests for normality in statistical literature and in software packages. And I recall a book by Hahn and Shapiro, published in 67, the title of which I forget, which had a lot of plots of data randomly drawn from a normal distribution. On the normal probability paper, ideally the plots would have been straight lines, but they got some pretty wavy looking things - and these from a normal distribution.
Contamination data, strictly speaking, would not follow a normal distribution, since the numbers cannot go below 0. Something I will have to be careful about.
If your subgroup data, such as you might get if you measured 5 boards from a lot, are not normally distributed, it doesn't make much difference since you would plot the means of subgroups on a control chart and such means would follow very closely a normal distribution.
And there are transformations that will convert non-normal data to normal so you can plot them. "SPC Methods from Quality Improvement", Chapter 9, by Quesenberry, published by Wiley has some discussion.
Lou Hart
----------
From: Brian Ellis[SMTP:[log in to unmask]]
Sent: Tuesday, March 28, 2000 3:04 AM
To: [log in to unmask]
Subject: Re: [TN] process control of final cleanliness in a no-clean world
Lou, Gary
I agree that a single SD limit can be misleading. This is why I put the
sentence in the plural and added the word 'significantly'.
I also understand that, with an Ionograph, it may be difficult to obtain
a true Gaussian curve because of the very nature of its operation adding
bananas to the apples and pears. We had a similar problem with the
dynamic Contaminometer. With an Omega Meter or similar, providing you do
not stop the test before near-asymptote, it would be easier. On the
static Contaminometer or, at least, the better models, we had a system
which, if near-asymptote was not reached within 15 minutes, it
mathematically analysed the resultant curve into its component
exponential curves and then extrapolated them together to infinite time,
thus obtaining the true asymptotic value (as well as giving us a lot
more info by examining the shape of each component and hence the
contaminant type). Each type of board had its own file. By the time we
had ca. 20 results in a file, it approached a normal curve with a
constant process: with 50 results, it was nearly always good
(occasionally with a single maverick -- unexplained -- value). If a
subsequent single value was apparently out-of-tolerance, a vague warning
appeared on the print-out of the individual test. If another one was
out-of-tolerance in the same direction, then the warning was stronger
and a third one stated a probability of the process going wrong. If the
SD widened significantly over a number of tests, then another warning
appeared. So, yes, I believe SPC can be useful, especially if it is
automatic like this.
We had the same algorithms in the dynamic Contaminometer, but they were
less precise because there were more variables in the measurement
process, despite the fact that we automatically had the same base-line
conductivity for all measurements. I believe that this variability was
due to more leaching from the ionics in the laminate which were more
poorly controlled than the soldering/cleaning process. We noted it
especially when changing from batch to batch of a given board,
especially with different makes of laminate but sometimes even with the
same manufacturer but, presumably, different lots.
Brian
Lou Hart wrote:
>
> TechNetters, (and Brian), I suggest looking into statistical process control techniques for something like this. A standard Shewart chart would be the thing. In fact, I plan on doing it myself in the very near future, thanks to Brian's earlier suggestions on cleanliness testing techniques.
>
> If you take 1 standard deviation as a "control limit" you'll have lots of false alarms. The purpose of SPC is not so much to tell you when to do something, but rather when not to do something, to leave things alone. A while back, a technetter said he'd made a chart, but found points always inside the limits. That's exactly the way things you want things to be. Go to work on something else and don't waste time tinkering with a well-behaved process. Unless it is producing bad stuff, in which case you need a new process.
> Lou Hart
>
> ----------
> From: Brian Ellis[SMTP:[log in to unmask]]
> Sent: Saturday, March 25, 2000 8:12 AM
> To: [log in to unmask]
> Subject: Re: [TN] process control of final cleanliness in a no-clean world
>
> Wade
>
> I agree with the others but would add one minor point. Qualify your
> process as "safe" according to your needs. Measure the ionic
> contamination with whatever instrument you will on, say, 20 boards from
> a similar batch, keeping the test going to near-asymptote (say less that
> 1% conductivity change in one minute). Note these final values and
> calculate the mean and standard deviation (ideally, the distribution
> should be a neat bell-curve). For that type of board, run regular tests
> in production: if they fall significantly outside the mean +/- SD
> (either higher or lower), then your process is probably going off the
> rails. Note that a lower reading of contamination may be just as
> dangerous as a higher one, because it signifies a change in the process
> parameters.
>
> Brian
>
> "Oberle, Wade" wrote:
> >
> > Dear technetters,
> > How do all of you in the no-clean world monitor and control board
> > cleanliness. In the days of OA flux, we used an Ionograph or Omegameter to
> > monitor our 'cleaning' process. Do some of you still use an ionic
> > contamination tester or do you use SIR testing or nothing or what?
> >
> > Thanks in advance for your advice.
> >
> > Wade Oberle
> >
> > ##############################################################
> > 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
> > ##############################################################
>
> --
> Brian Ellis
> Protonique SA
> PO Box 78
> CH-1032 Romanel-sur-Lausanne, Switzerland
> Voice: +41 21-648 23 34 Fax: +41 21-648 24 11
> E-mail: [log in to unmask]
> URL: Technical and consultancy divisions:
> http://www.protonique.com
> Web services division:
> http://www.protonique.com/webserv
>
> ##############################################################
> 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
> ##############################################################
--
Brian Ellis
Protonique SA
PO Box 78
CH-1032 Romanel-sur-Lausanne, Switzerland
Voice: +41 21-648 23 34 Fax: +41 21-648 24 11
E-mail: [log in to unmask]
URL: Technical and consultancy divisions:
http://www.protonique.com
Web services division:
http://www.protonique.com/webserv
##############################################################
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
##############################################################
|
