From: "James Lerch" 
To: 
Cc: "ATM List" 
Reply-To: "James Lerch" 


Greetings All,

This is going to be a 'little' long winded, if your not interested in all
this Robo garbage, just hit that 'delete' key now :)  (BTW, I'm writing
this during a caffeine induced intoxication, so pls excuse my wandering
thoughts!)

If your still reading this, I'm afraid this post is actually going to need
a table of contents, so here it goes:

Chapter 1, General Robo Comments
Chapter 2, Couder Mask vs Robo results Chapter 3, Robo Vs RoboC by Dale
Easton Chapter 4, Robo Vs Hartmann by James Burrows Chapter 5, Robo Vs.
Robo Blink test Chapter 6, Robo Vs. Field tests
Chapter 7, Summation


** Chapter 1, General Robo Comments

The original goal of Robo was to create a faster and less subjective mirror
test for use in our Telescope Optic Class.  Our ATM class currently has
upwards of 15 students, all at various stages of Grinding, Polishing and
Figuring.  Over the last few years its not uncommon to have 5 or more
students in a loop doing a few rounds of figuring, washing, drying, and
testing.  Prior to Robo this was a most unfavorable environment, with lots
of eye strain and varying results, and very few tests being accomplished.

With Robo, mirror testing has turned into a "Plop, point, and
click" affair, with highly repeatable results given in less than 5
minutes (Accuracy is yet to be determined, more on that later)

The most recent thread on Robo started as a result of obtaining different
KE results based on image intensity, so let me speak to this for a moment.

Originally Robo used a Slitless Foucault source.  During this time my
normal test was done with a bright input image source and very low 'Shade
of Gray' value. My current hypothesis is this gave fairly accurate results.
 However, during a recent test it was noted that going to a darker input
image, or Higher "shade of gray" gave a larger value for total KE
readings, which was rather disturbing.

My current hypothesis for why the above was noted is based on non-uniform
illumination as a change in the "Virtual Slit Width" in the
slitless test setup that results from a changing input image brightness or
changing "Shades of Gray"

Since my recent transition to a Slit source Foucault setup, changes in
input image intensity or "Shades of Gray" have greatly reduced
the change in KE readings.  As Bob May pointed out, even in a slit source
based test, one of the knife edges is still not being used when at a zone
null.  However its my opinion that the slit source greatly reduces
non-uniform illumination, and therefore greatly increases the test
accuracy.

One more note on my current light source setup.  I use a Sanded flat LED,
which is pushed back in its holder about an 1/8" (the holder is a
piece of 1/4" thick plywood with a hole drilled in it and a friction
fit for the LED)

With the LED pushed back in its hole an 1/8", I then covered the hole
with a piece of 'frosty' cellophane tape (Scotch tape?).  The knife edges
are held flush to the face of the cellophane tape, with one knife edge
being long enough to be used as both part of the slit and the knife edge
for the camera.

In essence my light source is the result of a 'Frosty' LED illuminating a
'Frosty' piece of tape.  This new setup seems to give VERY uniform
illumination, which appears to be a good thing!



** CHAPTER 2 (Robo vs Couder Mask Eyeball testing)

This whole SAGA started as a result of testing a sample optic sent to me by
Carl Zambuto.  When Robo's results were radically different than Carl's
results, Robo was considered to be in error.

To test this consideration, I made a Couder mask that replicated the Couder
mask that Carl used for testing the optic.  An image of this mask may be
seen here:
http://lerch.no-ip.com/atm/Comp/Mask.jpg (36KB)

I then took the Optic and the Mask to our ATM lab, and had our most
experienced Couder Mask Foucault test operators run the numbers using our
'Old School' test setup.  Amazingly the results were almost Identical to
Carl's test results done on the opposite coast!

The resulting Couder mask numbers (an average between the West coast and
East Coast test results) may be viewed here:

http://lerch.no-ip.com/atm/Comp/Couder.gif (21KB)

For those that don't want to follow the link, the image shows that the
optic is essentially perfect with a 0.99 Strehl

For comparison, here are the results from Robo (using the 8 sets of KE
readings from the most recent "Shades of Gray" experiment)

http://lerch.no-ip.com/atm/Comp/FigXp_SOG_Ave.gif (22KB)

The above results have a standard deviation (n-1) of 0.0028" and show
the optic as strongly under corrected with a Strehl of 0.796.

At this point, with such a large deviation in test results Robo was
considered in error.


** Chapter 3 (Robo vs RoboC by Dale Easton)

In an effort to eliminate Software as the culprit, Dale Easton provided me
his version of Robo written in C++.  Dale's software uses an entirely
different approach than my methods.  Dale uses Fixed KE positions and
Solves for the resulting zone null radius.  He also uses a different method
to define the zone null radius than I do (I'm not entirely clear on the
method, but that is irrelevant at the moment)

After running his version of RoboC on my hardware, I obtained the following
test comparison image.
http://lerch.no-ip.com/atm/Comp/Dale.jpg (225KB)

In the above image, the top graph is Dale's RoboC results and the bottom
graph is from one of my SOG tests.

Both test results show a strongly under corrected optic, with Dale's code
giving a Strehl of 0.736 compared to My Robo result of 0.716.

Of interest is the close Strehl reading and nearly identical Surface error profile.

However, this doesn't eliminate my Hardware, which could be in error.


**  Chapter 4 (Robo Vs Hartmann by James Burrows)

This chapter is devoted to hopefully eliminating my hardware as the source
of the deviation between classic Couder Mask testing and Robo testing.

Towards this end I performed a Hartmann test using a 7 zone Hartmann mask
with 10mm wide by 20mm tall openings, each separated by 10mm (excluding the
center 20mm radius section of the mirror)

The test was performed with a non-lasing laser diode as the source, and a
Vesta web-cam CCD as the sensor.  While this test was rather difficult to
setup accurately (keeping everything aligned, square, and rotationally
oriented was difficult) I was able to get two data sets, one inside of ROC
and one Outside of Roc.

The results of these two tests may be viewed here:
http://lerch.no-ip.com/atm/Comp/Hart_Inside.gif (107KB)
http://lerch.no-ip.com/atm/Comp/Hart_Outside.gif (107KB)

In each of the above images, the top half of the image is the test result
showing surface profile of the optic, and its numerical results.  The
bottom half of each image shows the correlation between the Actual Hartman
mask intensity profile (in green) Vs the Simulated profile (in Red)  Of
interest is that the Correlation is greater for the test conducted inside
ROC than for that of Outside Roc.

For those that don't want to download the images, the results are as follows:

Inside Roc = under corrected, 0.679 Strehl Outside Roc = under corrected,
0.824 Strehl

Of interest (for me anyway) is the surface error profile, especially when
compared to the Robo SOG experiment results seen here:
http://lerch.no-ip.com/atm/Comp/FigXp_SOG_Ave.gif (22KB)

My interest in the comparison of the surface error profile is the high
correlation between all three plots. All three show a high center, a valley
near the 65% radius, a small hill near the 75% radius, another valley near
the 85% radius, and finally a turned up edge.  I must ask, what are the
Odds of this being just a coincidence, especially with the entire lack of
correlation when compared to the profile of the classic Couder mask
results?


** Chapter 5 (Robo Blink Test)


A suggestion was proposed from this list that perhaps aiming for a matched
"Shade of Gray" was inappropriate, and a more appropriate
simulation against Classic Eyeball Couder mask testing could be obtained by
using the "Blink Method".  Towards this goal I came up with the
following experiment:

A) When Robo called a zone null by matching a 'shade of gray' the following
process was implemented.

B) A search was performed at 0.001" increments at +/- 0.015"
distances from where robo called the zone null

C) At each search location, the code laterally moved the knife edge in to
the return beam until both sides of the fixed zone radius intensities were
at 0

D) The code then stepped the KE out of the return beam in 0.000125"
increments (the smallest step possible with my hardware).  After each step
out of the return beam, 30 frames of intensity data were averaged together,
and the left and right radius intensities were recorded to a file for post
processing.

E) The code continued to step the KE out until both sides of the zone
radius had exceeded an intensity of 200.

F) The code then repeated the process until all 31 longitudinal positions
had been recorded (15 locations closer to the mirror, 1 at the
pre-determined null position, and another 15 positions further away from
the mirror)

G) This test took nearly 3 hours to complete! (not including post processing time!)

F) Post Processing went as follows.
 For each longitudinal position, the difference for each Lateral position
was calculated (Left zone intensity - Right zone intensity).

 The difference was entered into a spread sheet, such that a Positive value
indicated the Left zone was brighter than the right zone, and a negative
value indicated the opposite (Left > right).

 Once all this data was entered into the spread sheet, the longitudinal
location that had the total value of lateral knife edge values closest to
zero was considered zone null.

 My assumption is that this would in essence resemble the "Blink"
version of
the classic Couder mask test since a full range of lateral position
intensity differences were recorded.

For those interested, here's the spread sheet that was created:
http://lerch.no-ip.com/atm/Comp/Brute_Force.xls  (90Kb)

The results of this test showed that when Robo called a null using a
matching shade of gray, the 'Blink' method showed a identical answer for
each zone within 0.001" inch.

In other words it didn't disprove the normal Robo code results.  (Nothing
ventured nothing gained?)


** Chapter 6 (Robo vs Field Tests)

Our local ATM group had a Public observing night last Saturday.  During
this event 6 telescopes that were tested with Robo were present.  (we do
not have an OTA to test Carl's mirror yet, but we are working on this)

During this Public observing night we did some extensive field testing of
these 6 telescopes using the Ronchi-Star test and Suiter's defocused
Star-Test (with Suiter's book in hand!)

Here's a list of the optics tested

14" F/4.5 (Commercial Mirror)
12.5" F/6
11" F/5.8
8" F/8
8" F/6
6" F/5
6" F/4

While an entire book could be (and has been) written on Field testing
astronomical telescopes, I shall only give the following comments (since
field testing is entirely subjective at this time)

"Mirrors that tested near perfect with Robo (0.95 or greater Strehl)
had Ronchi Lines with no perceivable curvature inside or outside
focus"

"Mirrors with lower Strehls showed curvature in the Ronchi lines, and
that curvature matched the Over/Under corrected predictions by Robo"

"Of all the mirrors, two were 'Critically' Star tested, (the 6"
F/5 and the 14.5" F/4.5).  The 6" showed near perfect correction
and the 14.5" showed over corrected, both results concur with Robo,
Suiter's illustrations, and the predicted defocused structure given by
FigureXP's star test sim."

Now, does any of the above mean anything, honestly NO!  We went into this
field testing with the fore knowledge of 'what to look for' on each mirror.
It is highly possible that this fore knowledge led us astray!


** Chapter 7 (Summation)

#1 Classic Couder Mask testing (using identical Couder Masks), conducted by
skilled Foucault test operators (whose results are nearly identical) show
the Zambuto Optic as nearly perfect (0.98 Strehl)

#2 Two versions of Robo-Foucault software, and Jim Burrow's Hartmann test,
show the optic as under corrected with a Strehl somewhere between 0.7 and 0.8.

#3 Field testing of optics previously tested by Robo-Foucault seem to agree
with Robo-Foucault's predictions

The above presents a conflict between the classic method of ATM optical
testing using a Couder mask (which has Ten's of decades of history
supporting it's results) and more recent testing methods with only a few
years of history supporting their results.

The question is which method is Accurate?

To hopefully solve this dilemma, Carl's optic is currently on its way to
Bob Royce of R.F. Royce optical, for testing with Mr. Royce's Double Pass
AutoCollimation Ronchi Test.  As I understand this test, it will tell us if
the Optic is perfect or not.

(Humor mode on!)
Anyone wanna place some bets? 
(Humor mode off!)

Well I guess I've wasted enough bandwidth for the moment.  All that's left
to do is anxiously sit and wait for the test results from Mr. Royce!


Take Care,
James Lerch
http://lerch.no-ip.com/atm (My telescope construction,testing, and coating site)

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