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


Hi Eric,

Here's the "short" version on how Robo does its stuff.

For each zone radius we have a left and right side (just like a Couder
mask). Robo grabs a pixel intensity count for the area that is under the
calculated position for both sides of each zone.

Next Robo establishes the "Shade of Gray" value for what I call
the "Control Zone".  The Control Zone is simply the side of the
zone radius that is opposite the knife edge.  In the case of my Foucault
platform the knife edge is on the right, so the Control Zone is on the
left.

So when starting a test on my platforms, the first thing Robo does is move
the knife edge laterally in or out of the return light cone to get the left
side of zone 1 into the "Shade of Gray" range (in my case a pixel
intensity value between 30-60).

Once the Control zone is in the right "Shade of Gray" range, a
simple comparison is done against the remaining zone.  The results of this
comparison can be 1 of 3 things:

a) the remaining zone is BRIGHTER than the control zone, in which case Robo
moves the platform longitudinally TOWARDS the mirror and starts all over
again

b) the remaining zone is DIMMER than the control zone, in which case Robo
moves the platform longitudinally AWAY from the mirror.

c) the remaining zone is in the "Shade of Gray" region and of
equal intensity to the control zone, in which case Robo calls this the zone
null, records the platform position and moves on to the next zone.

Since all the above is done in real time from a live video feed, Robo takes
advantage of this and does a lot of noise reduction by averaging a lot a
frames prior to making any decisions.

For example:
A) prior to moving the knife Laterally to set the shade of gray, a minimum
of 5 frames of video data are averaged together.

B) prior to moving the knife Longitudinally to find a zone null, a minimum
of 15 frames of video data are averaged together

C) prior to proclaiming a Zone Null, a minimum of 30 frames of video data
are averaged together

Just to make things a little less noisy, for each half of a zone, in each
and every frame of video data, A group of pixels are sampled and averaged
together.  By sampling a group of pixels centered on the zone radius, we
reduce noise induced from the video source, as well as hopefully avoid the
inadvertent measurement of micro-ripple (dog biscuit or whatever you want
to call it ;)

When all of the above is rolled together, and each zone null is measured
multiple times, Robo does a fine job of obtaining consistency in null
measurements.

The problem is I can get consistent, but significantly different readings
by simply changing the intensity of the video feed (changing the light
source brightness, or changing the camera's sensitivity).  I can also
change the resulting readings by altering the "Shade of Gray"
value the software uses for the control zone.

Common Sense tells me that the Brightest image source combined with the
lowest "shade of gray" will create the narrowest of a
"Virtual Slit" in the Slitless Foucault test, and that this
should be the most accurate of measurements.

Unfortunately the above hypothesis flies in the face of the classic
eyeballed Couder mask Foucault test done by Carl Zambuto and independently
confirmed by several eagle eyed Foucault operators at our local Mirror Lab.

At the moment I sit here rather confused.  Robo's goal was to eliminate the
subjective measurements of "shades of gray".  However Robo was
based on the theory that a given "UN-MASKED" mirror surface had
only one null position for each zone radius, independent of light source
intensity, or the shade of gray used.  This theory seems to be flawed,
which doesn't bode well for Robo's future!

Hope this helped Eric!

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


----- Original Message -----
From: "Eric Royer" 
To: 
Cc: "ATM List" 
Sent: Monday, May 05, 2003 2:35 PM
Subject: Re: ATM Robo-Foucault, Image intensity, and Changing knife edge
readings


>
> Hello James,
> I'd like to help if I can but I'm not sure to understand how you tell from
> the image that the KE is at a null position for a given zone. Can you
> explain how you do that ?
>
> > #2 Changing the 'Shade of Gray' the software uses to set lateral knife
> > position
> How do you use the 'Shade of Gray' in your algorithm ?
>
> Eric
>
>
>

--- BBBS/NT v4.00 MP