Re: [ATM] holographic null test on round robin mirror C

["matt" <mariusrf@bellsouth.net>]

-----Original Message-----
From: Andy Saulietis <iss@pvtnetworks.net>
To: matt <mariusrf@bellsouth.net>; atm@atmlist.net <atm@atmlist.net>
Date: Friday, February 04, 2005 11:41 AM
Subject: Re: [ATM] holographic null test on round robin mirror C


>Matt writes,
>
>I was referring to the classical Hartmann test in which a real
>star is used, the imaged spots are both inside and outside
>focus at a precisely known separation distance, and simple
>geometry is used to calculate the best focus spot diameter.
>
>>exactly local wavefront slope deternimation seems to be accurate enough
for
>>the largest professional telescopes , since they're using it for wavefront
>>sensing in adaptive optics.
>
>This will not scale well for small apertures, see below.
>
>>a Hartman mask or a small lenslet array in front of a webcam style CCD is
>>all it takes. The size of the defocused pattern can be made arbitrarily
>>small, by getting the system closer to focus . Even on a 640x480 size
webcam
>>ccd, with 6um square pixels, the pattern would fit. Defocus each
>>spot/subaperture to 100um diameter, and you could still fit 10x10 to 20x20
>>subapertures . Calculating spot centroids even for large blur size is a
>>photometric problem similar to doing star photometry, where centroids are
>>calculated to subarcsec accuracy although seeing blurs star images to
>>several arcsecs . If exposure is not saturating the image and still in the
>>linear area, centroid calculation works .
>
>The problem is that even with  centroiding to 1% of the spot diameter,
>the resulting slope errors are too large to be useful for mirror testing
>and figuring. I used to design star trackers for NASA, and know
>that centroiding  at these accuracies is not a trivial problem,
>requiring extensive calibration models for each pixel in the
>detector array.
>
>
>>today's largest telescopes all use derivatives of the Hartman, for example
>>Shack Hartman, where the mask is replaced by a lenslet array . This is the
>>most widespread adaptive optics wavefront sensor . There are a few other
>>ideas for wavefront sensing, like curvature sensing using a vibrating
>>mirror, or the pyramid wavefront sensor, shearing interferometry , etc.
The
>>Shack Hartman is dominating now for active and adaptive optics systems.
>
>Jim Burrows has a functional Hartmann derivative test but it's not
>the classical type..he uses the overlapping spot diagram interference
>pattern to determine slopes. Perhaps he can weigh in on the subject?
>He and I participated in the development of the modified Harmann
>test, and the analysis that just using a CCD to do a classical Hartmann
>test was not practical. Actually, I was hoping that it *would* work
>because of it's inherent simplicity and the potential ability to
>test *any* conic surface. Finally, the Shack-Harmann test may
>be the one of choice..but is it practical for an ATM to use?
>
>
>

regarding Shack-Hartman, the maximum wavefront slope over a subaperture is
dictated by the ccd geometry and lenslet array focal length.
Any combination is possible, just changing the lenslet with a different
focal length one, the sensor can be reconfigured for a wide range of
wavefront slope measurements.

As far as anecdotal evidence about how scalable for small apertures this is,
I have measured an 8" mirror with a 32 subaperture cardboard Hartman mask
placed over the telescope aperture , a cheap ccd at prime focus and the
analysis software . This also convinced me of how inexpensive the method
really is . This mask samples the 8" mirror at points spaced 1.5" apart and
it's 2 dimensional . I was not pushing any limits in doing this .

I have a precision metal Hartman mask to be used at prime focus that cost me
close to $10 , and a quartz lenslet array (I used it in a Shack Hartman
sensor ) of 10x10 subapertures that sells in the $200 range . The mask or
lenslet array is the only expense, assuming one has the camera and computer
. There are numerous sources and prices are ranging from close to nothing up
to a couple hundred $ . The interesting part is the sensor can be calibrated
with just a pinhole source and some basic alignment techniques , no
expensive optics required. No flats, no xyz linear stages, no spherical
refernce elements .

best regards,
matt tudor

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