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


At 02:27 PM 7/21/2003 -0400, James Lerch wrote:

>Ok, Let the FUN begin 
>
>For those that want "Just the Facts" here's everything I know
at the moment:
>http://lerch.no-ip.com/atm/2ndTry/lerchtest1.ZIP (300KB from RF Royce)
>

First - and unfortunately - second impression: The agreement between Robo
and these interferometric results is not so good. Here are estimates of
surface RMS from Robo and Royce:

My estimate from reprocessed images: 23.8 nm.

James' original data (pooled): 20.0 nm.

Royce interferometry:   47nm
   minus coma                    46 nm
   minus coma & astig.   38 nm.

A foucault test is insensitive to primary astigmatism, which based on the
interferometry appears to be the second most serious defect on this mirror
(and enough by itself to disqualify it as "diffraction limited").
Removing coma and astigmatism terms still leaves a residual RMS of 38nm,
which is about 50% larger than I estimated.

In slightly more detail, here are my estimates of 4th through 10th order
Zernike coefficients from Robo and from the interferometric report
(spherical terms only):

         My reprocessed          James Original          Royce
4               -9.1                            2.1             -33.8
6               -14.9                           -12.1           11.3
8               -13.2                           -12.9           4.6
10              -9.2                            -9.1            ?

Sorry about any misformatting. These are all in nm on the surface. The
first column are my estimates that I posted back around the first of the
month. The second is my estimate of the Zernike coefficients from James'
original posted data. The third column is from Royce's report, rescaled to
surface errors in nm.

Basically, James' measurements agree reasonably well with my reprocessing
of the raw data except for the total correction error. I read the mirror as
slightly overcorrected; James read the overall correction as essentially
right on. We both agree there are higher order defects needed to fit the
apparent rolled edge.

By contrast the interferometry shows a huge amount of overcorrection with
higher order terms having the opposite sign of what I predicted. Overall
this isn't even close to reasonable quantitative agreement.

One thing that bothers me about the interferometric report is that the
summary notes a rolled edge starting as much as 3/4" in from the edge.
That's exactly what the foucault test shows, but there's no real sign of it
in the graphs or numbers provided with the report (unless it shows up in
the missing 10th order coefficients). Instead the interferogram shows
overall overcorrection.

Another thing that bothers me a little is the decision to do the
interferometry at the center of curvature without any nulling optics. Does
anyone familiar with interferometry know if that could be a source of
systematic error?

Right now I'd suggest that James' first task should be to figure out why
there's such a huge discrepancy between his estimate of overall correction
error and the interferometry. I'm not 100% convinced the problem is with
the foucault data, but the burden of proof has to lay on Robo-foucault to
show it's right. The higher order discrepancies are troubling too, but they
can wait.

Mike Peck


------
Michael Peck
mpeck1{at}ix.netcom.com

--- BBBS/NT v4.01 Flag-5