From: Dale Eason 
To: Jeff Anderson-Lee , atm{at}shore.net
Reply-To: Dale Eason 


--0-1435785844-1057627259=:53734
Content-Type: text/plain; charset=us-ascii


That is what my version does except...  It also finds the outer most zone
first before finding the inner.  The it divides this distance into 15
zones.  The problem is finding a good method of finding the center.  I have
started using the classic zone balance method to do that.  And the down
side is that you only know the zone radii to the best of you resolution. 
At the moment for me it is about 300/mirror diameter.

I used this method instead of the classic zone balance method because it
was taking me along time to find the balance point for each zone.  With the
flip and diff method the only searching is for the first and last zones and
a lateral knife position that give you a constant illumination for each
measurement.  So to begin with my algorithm was much faster than James on
my machine.


Definition:  Zone balance method.  The classic way of measuring the zone,
move the knife until both zones left and right  dim at the same time and
rate.

The way I determine the lateral knife setting is this.  I average a strip
of pixels across the center of the mirror and adjust the knife until the
average matches a set point.  The set point is configurable by the user but
I have found that I get the same results with a wide range of set points. 
I make sure that intensity curve of the strip is within the CCD pixel
intensity range. Making sure that it is not clipped nor is it zero. Jeff
Anderson-Lee  wrote:
 Dale Eason wrote:
Concerning MC analysis for both numbers. I don't think you would do both. 
You would only do the ones that you measure with the least certainty.  For
instance in the classic Foucault test it would be the longitudinal numbers.
 For the flip and diff you know the longitudinal numbers to the best of
your positioning system but are a little unsure of the zonal radii.
This leads me to something I've been mulling over recently.  Why does
automated Foucalt testing have to mimic the manual method?  Here's a
possible alternative:

1) Find the farthest in null point (which might of might not be the central
zone if you have a central hill).
2) Step out a measured amount and take several images that are averaged
together to reduce noise.
3) Use flip and diff to compute the zone radius (or radii?) 4) Repeat 2 and
3 until the zone radius reaches the edge of the mirror.

All stepping will be in one direction (after step 1), minimizing the
longitudinal KE measurement error.  The only tricky part is knowing where
to place the KE laterally in order to take the measurements.

Comments? Thoughts?

Jeff Anderson-Lee






---------------------------------
Do you Yahoo!?
SBC Yahoo! DSL - Now only $29.95 per month!
--0-1435785844-1057627259=:53734
Content-Type: text/html; charset=us-ascii

That is what my version does except...  It
also finds the outer
most zone first before finding the inner.  The it divides this
distance into 15 zones.  The problem is finding a good method of
finding the center.  I have started using the classic zone balance
method to do that.  And the down side is that you only know the
zone radii to the best of you resolution.  At the moment for me it
is about 300/mirror diameter.  
 
I used this method instead of the classic zone balance method because it
was taking me along time to find the balance point for each zone. 
With the flip and diff method the only searching is for the first and last
zones and a lateral knife position that give you a constant illumination
for each measurement.  So to begin with my algorithm was much
faster than James on my machine.
 
 
Definition:  Zone balance method.  The classic
way of measuring
the zone, move the knife until both zones left and right  dim at
the same time and rate.
 
The way I determine the lateral knife setting is this. 
I average a
strip of pixels across the center of the mirror and adjust the knife until
the average matches a set point.  The set point is configurable by
the user but I have found that I get the same results with a wide range of
set points.  I make sure that intensity curve of the strip is
within the CCD pixel intensity range.  Making sure that it is not
clipped nor is it zero.Jeff Anderson-Lee
<jandersonlee{at}sbcglobal.net>
wrote:





 Dale">mailto:atmpob{at}yahoo.com">Dale
Eason wrote:

Concerning MC analysis for both numbers.
I don't think you would do both.  You would only do the
ones that you
measure with the least certainty.  For instance in the classic
Foucault test it would be the longitudinal numbers.  For the flip
and diff you know the longitudinal numbers to the best of your positioning
system but are a little unsure of the zonal radii. 

This leads me to something I've
been mulling over
recently.  Why does automated Foucalt testing have to mimic the
manual method?  Here's a
possible alternative:
 
1) Find the farthest in null point
(which might
of might not be the central zone if you have a central
hill).
2) Step out a measured amount and
take several
images that are averaged together to reduce noise.
3) Use flip and diff
to compute the zone
radius (or radii?)
4) Repeat 2 and 3 until the zone
radius reaches the edge of the mirror.
 
All stepping will be in one
direction (after step
1), minimizing the longitudinal KE measurement error.  The only
tricky part is knowing where to place the KE laterally in order to take the
measurements.
 
Comments? Thoughts?

 
Jeff Anderson-Lee
 
 
 

Do you Yahoo!?
http://pa.yahoo.com/*http://rd.yahoo.com/evt=1207/*http://promo.yahoo.
com/sbc/">SBC Yahoo! DSL - Now only $29.95 per month!
--0-1435785844-1057627259=:53734--

--- BBBS/NT v4.01 Flag-4