Re: [ATM] Faucault testing questions

[Mark Holm <mdholm@telerama.com>]

Long experience by many atm's and professionals shows that a properly 
done Foucault test on a 10 inch f/6 mirror is capable of producing a 
reliable indication of a good mirror.  Theory or no, practical 
experience says it works.  Now, that does not mean it is an easy test 
for a beginner to learn.  Ken may be making some mistake, or his tester 
may introduce a reading error.

Ken,

How repeatable are your test results?  If your reading uncertainty is 
high for some reason, even as many as nine reading sets averaged does 
not necessarily damp out the noise to the needed level.  One thing you 
might try is using the median reading values rather than the mean.  
Statisticians have shown that for small sample sizes (nine is considered 
small) the median is less likely to be thrown badly off by an outlier 
than the mean.  In fact, for a small sample, the sample median is a 
better estimator of the population mean than the sample mean is.

If your reading uncertainty is high, perhaps there is something about 
your tester or method that is causing trouble.  Perhaps your micrometer 
or carriage is producing jumpy movement.  Perhaps your knife edge moves 
back and forth a little when it should be moving only sideways.  Perhaps 
your tester is sitting on a table that can flex somewhat forward and 
back as you lean on it.  I learned to keep my body away from the table 
while reading the test.  Perhaps you just have difficulty reading the 
shadows.  Is your tester situated so that you have your head in a 
comfortable position?

Is just one set of readings widely out from the rest?  If you have 8 
sets indicating a similar situation, and one set indicating trouble, 
that one may be in error.

Does the no mask Foucault test indicate a smooth surface, or do you have 
zones.  It can be hard to get reliable readings when there are fairly 
steep zones.  The shadows don't quite behave themselves.  If the zones 
are near the edge, there can be a pretty big effect on rms results.  A 
turned edge is quite difficult to measure reliably.

Does one zone have much more variation in readings than the others?

Jim Burrows asked about sigma values.  If you used Frontsix, it 
calculated that for you.  I think Andreas' program does too.

Also, be careful that you entered the zone radii correctly.  It is an 
easy thing to mess up.  Maybe go back and recompute them to be certain.  
Sounds like you used couder.exe to make your masks.  Run it again and 
make sure you recorded all the zone center radii properly.

Is there a trend over the time it took you to make the nine reading 
sets?  Do the readings start out showing overcorrection and move toward 
undercorrection?

How much time did you let the mirror cool off before reading.  Plate 
glass is known to require a fairly long cooling period before readings 
become stable.  Is the temperature in your testing room stable?  If it 
happened to be on a trend when you did the test, the mirror may have 
been lagging, and thus not showing its true figure.  For instance, my 
home has a programmed thermostat that sets the temperature down several 
degrees at about bed time.  On a cold night, the house temperature drops 
for an hour or so before the furnace comes on again.  That several 
degree drop could take a mirror perhaps two hours to recover from.

Also, since I am the author of Frontsix, I am interested to know what 
kind of trouble you had with it.

-- 
Mark Holm
mdholm@telerama.com


_______________________________________________
ATM mailing list http://www.atmlist.net/