Date: 5 Nov 92 15:55:21 GMT
From: "William H. Jefferys" <bill@bessel.as.utexas.edu>
Subject: Hubble's mirror
Newsgroups: sci.astro,sci.space

In article <1992Nov3.213906.886@mrdog.msl.com> dhl@mrdog.msl.com (Donald H. Locker) writes:
#Now that I know a little about mirror-making, I'd like to hear again
#how the Hubble mirror contractor messed up the figure of the main
#mirror.  I understand it has spherical aberration, but wonder how
#[Rockwell?] managed to do that.

The Hartford Courant had a Pulitzer Prize winning
series of reports that go into gory detail on this.
They may still have copies available. It's entitled
"Hubble error: Time, money and millionths of an inch."
Write the Hartford Courant, 285 Broad St., Hartford
CT, or call Corporate Affairs Manager Sylvia Levy
at (203) 241-6431.

Bill

Date: 7 Nov 92 07:18:37 GMT
From: Henry Spencer <henry@zoo.toronto.edu>
Subject: Hubble's mirror
Newsgroups: sci.astro,sci.space

In article <PXiye-Go3@lopez.marquette.MI.US> stick@lopez.marquette.MI.US (Stick,CommoSigop) writes:
>>... put briefly, they
>>fouled up the test, performing it incorrectly, and never
>>checked by any independent method.
>
>   According to Dr. Steve Maran, who works on the HST project at the
>Goddard Flight Center, and who was recently a guest lecturer at my college,
>none of the above is true.  The company that ground the mirror did it
>exactly to the specs they were given.

You're sure that's what he said?  It's *not* what Lew Allen's review board
found (and documented in detail).  The grinding people at Perkin-Elmer did
indeed produce an essentially perfect mirror to the wrong spec.  But it was
Perkin-Elmer that botched the spec, by mis-building the reflective null
corrector that was used to measure the mirror shape, and Perkin-Elmer that
ignored three successive hints that something was wrong with the RNC.
(First, the error made it impossible to build the RNC without a slight
design change; the change was made without anyone asking why it was
necessary.  Second and third, results from two other measurements were
disregarded because the RNC was thought to be superior.)
-- 
MS-DOS is the OS/360 of the 1980s.      | Henry Spencer @ U of Toronto Zoology
              -Hal W. Hardenbergh (1985)|  henry@zoo.toronto.edu  utzoo!henry

Date: 9 Nov 92 14:52:54 GMT
From: "William H. Jefferys" <bill@bessel.as.utexas.edu>
Subject: Hubble's mirror
Newsgroups: sci.astro,sci.space

In article <PXiye-Go3@lopez.marquette.MI.US> stick@lopez.marquette.MI.US (Stick,CommoSigop) writes:
#In <yssPTB4w165w@bluemoon.rn.com> gerry@bluemoon.rn.com (Gerard M. Foley) writes:
#
#>dhl@mrdog.msl.com (Donald H. Locker) writes:
#
#>> Now that I know a little about mirror-making, I'd like to hear again
#>> how the Hubble mirror contractor messed up the figure of the main
#>> mirror.  I understand it has spherical aberration, but wonder how
#>> [Rockwell?] managed to do that.
#>> 
#>It wasn't Rockwell (I forget axactly who it was, but it was an
#>otherwise reputable New England outfit) and put briefly, they
#>fouled up the test, performing it incorrectly, and never
#>checked by any independent method.
#
#   According to Dr. Steve Maran, who works on the HST project at the
#Goddard Flight Center, and who was recently a guest lecturer at my college,
#none of the above is true.  The company that ground the mirror did it
#exactly to the specs they were given.
#
#   The specs were wrong.

If my friend Steve Maran really said this, then he's 
incorrect. The specs were fine. What happened was that
Perkin-Elmer mis-manufactured a critical piece of
equipment, a reflective null corrector, that was used
to figure the mirror. A field lens was located 1.3 mm
out of its proper position. The reasons for this having
happened are rather complex and involved. They are 
described in gory detail in the Hartford Courant
report series that I posted information about earlier.
I personally attended many days' worth of reviews where
these events were found out, so I have reason to believe
that what I say is accurate.

The poster to whom you responded had it essentially right.

I don't think that Steve would be as far off base as
this, and I suspect that you have misunderstood him.

Bill Jefferys

Date: 9 Nov 92 22:41:08 GMT
From: "William H. Jefferys" <bill@bessel.as.utexas.edu>
Subject: Hubble's mirror
Newsgroups: sci.astro,sci.space

In article <1dh7e9INN481@rave.larc.nasa.gov> claudio@nmsb.larc.nasa.gov (Claudio Egalon) writes:
#Henry Spencer is right. Although it was sometime ago that I read an 
#article in Physics Today- which explained very nicely the real messed 
#up-, I still recall that there was three diferent instruments that were 
#used to measure the curvature of the primary mirror. Two of them 
#used refraction principle and one of them used the reflection 
#principle (don't remember the details so I may commit some  
#mistakes. If so, I apologize). The results of the other two instruments 
#that used the refraction principle to measure the mirror curvature 
#matched up but... the measurements produced by the one that used 
#reflection, reflective null corrector, was different and, since it was 
#thought to be superior to the other two,  no one bottered about 
#finding out why the discrepancy between the measurements... of 
#course untill they found out that there was something wrong with  
#Hubble. 
#
#*However* as  I recall (and I may be wrong) the error in the RNC 
#came along because the WRONG surface in the instrument was 
#providing the reflection and not that because the instrument was 
#built improperly. Aparently there was an scratch in a surface that was 

No, the reflective null was built improperly. Here's what
happened: The field lens was supposed to be aligned to the
cell by using a precisely machined Invar rod, whose length
was traceable to NBS standards. A laser was used to align
the end of the bar to the cell. In order to guarantee that
the laser was reading just the end of the bar (which was
rounded), an end cap with a small hole was placed over the
end of the bar. The end cap extended 1.3 mm further than
the end of the bar, and was painted black so that there
would be no return to the laser from the end cap. Unfortunately,
as it turned out, a fleck of the black paint got scraped off,
and a significant return from the end of the bar was obtained.
Thus, the bar was positioned too far away from the laser, and
the other end of the bar (which was positioned against the
field lens) was 1.3 mm from where it should have been.

In order to misposition the field lens, they people who did
this _had to shim out the lens cell with washers_!!! The
cell had been properly built, but when the lens didn't fit
where they thought it should (in the wrong position), they
kludged it to _make_ it fit.

You are right that the refractive null and the inverse null
agreed with each other, but not with the reflective null. Why
the engineers didn't see the red flag waving by 2 out of
3 vote, and the fact that they had to shim out the lens position
in the cell, is beyond me.

#suposed to be black, for absorbing incoming radiation, and it was this 
#scratch that worked as the WRONG reflecting surface. So,  that how 
#the wrong prescription came about. Another reference for the 
#investigation on the problem of the mirror was the report put out by 
#Lew Allen's review board (I guess I have a copy of it at home and I 
#am confindent that you people can get one from NASA but I do not 
#know whether they  will charge you or not). Maybe, not oddly 
#enough, there were other complaints about the handling of the  
#project  like people did not like to be bothered with criticism (read 
#peer review)... So aparently, the problem was both technical and 
#arrogance.
#
#Claudio O. Egalon

Bill

Date: 11 Nov 92 14:52:35 GMT
From: "William H. Jefferys" <bill@bessel.as.utexas.edu>
Subject: Hubble's mirror
Newsgroups: sci.astro,sci.space

In article <1992Nov8.232547.27419@ke4zv.uucp> gary@ke4zv.UUCP (Gary Coffman) writes:

#According to the NASA investigation of the Hubble mirror problem, there
#was nothing wrong with the specs. The backup mirror built to the same
#specs by Kodak is *perfect*. The problem was that Perkin-Elmer bollixed
#up the testing of the figure of the mirror and ignored the results of
#a Foucault test that showed the mirror to have spherical abberations
#because their primary test was supposed to give more accurate results.
#Unfortunately, they installed a fixture backwards when making the
#primary measurements and got the *wrong* result.

I'm pretty sure there was never a Foucault test; that
there hadn't been one became an issue after the mistake
was discovered. This was brought up at NASA reviews I
attended. After all, the interferometric test was supposed
to be so accurate that a crude test such as the Foucault
test wouldn't tell you anything you didn't already know
(or so they thought).

The "fixture" that was used backwards was the Invar rod;
It was indeed used backwards, but that was not the chief
reason for the mis-manufactured null corrector. This was,
as I posted earlier, the laser return from the end cap, 
which had had a fleck of black paint scraped off.

Bill

Date: 14 Nov 92 15:03:48 GMT
From: "William H. Jefferys" <bill@bessel.as.utexas.edu>
Subject: Hubble's mirror
Newsgroups: sci.astro,sci.space

In article <1992Nov9.125222.1@mdcbbs.com> rivero@mdcbbs.com writes:

#  The error was introduced into the main mirror because the setup used to
#test it was missing some spacers which were planned to locate the
#test rig at the correct focal point. Because the spacers were not present,
#the mirror was figured perfectly, but to the wrong focal point, outside
#the range of the Hubble's focusing apperatus. This error would have shown
#up had the entire telescope been tested as a whole prior to launch. Sadly, 
#due to budget constraints, this was never done.

Baloney. The mirror has spherical aberration due to
the fact that the null corrector was mismanufactured.
To mismanufacture it, P-E had to put some spacers
IN that should NOT have been there. Hubble can 
focus, but the spherical aberration prevents all
the light from being in focus at one time.

I'll agree, the error would have been caught if the primary
and secondary mirrors had been tested together. There are
lots of simple tests that would have caught the error.
None were used.

Bill

Date: Sun, 15 Nov 1992 00:14:02 GMT
From: Henry Spencer <henry@zoo.toronto.edu>
Subject: Hubble's mirror

In article <83625@ut-emx.uucp> bill@bessel.as.utexas.edu (William H. Jefferys) writes:
>I'll agree, the error would have been caught if the primary
>and secondary mirrors had been tested together. There are
>lots of simple tests that would have caught the error.
>None were used.

This is rapidly becoming a contender for Technological Myth Of The Decade.

If you look at what actually happened, rather than the popular mythology,
the fact is that three independent tests were run, and TWO OF THEM DETECTED
THE ERROR.  At which point, Perkin-Elmer management decided that the third
was more trustworthy than the other two, and ignored the two failed tests.

The *only* test that would have detected the error so unmistakably that
it couldn't have been ignored would have been an end-to-end test doing
real imaging with primary and secondary together.  This would *not* have
been a particularly simple or cheap test, thanks to various complications
like gravitational distortion of the primary, and it would have involved
a substantial risk of surface contamination.  NASA's decision not to do
it was probably a mistake, but it wasn't a particularly stupid or obvious
mistake, especially in a project that was behind schedule and over budget
already.  The mirrors had already been tested, successfully, in far more
sophisticated ways.  (NASA didn't know that the primary had flunked two
out of three tests.)
-- 
MS-DOS is the OS/360 of the 1980s.      | Henry Spencer @ U of Toronto Zoology
              -Hal W. Hardenbergh (1985)|  henry@zoo.toronto.edu  utzoo!henry

Date: Mon, 16 Nov 1992 01:39:53 GMT
From: Henry Spencer <henry@zoo.toronto.edu>
Subject: Hubble's mirror
Newsgroups: sci.astro,sci.space

In article <1992Nov15.170247.18454@murdoch.acc.Virginia.EDU> gsh7w@fermi.clas.Virginia.EDU (Greg Hennessy) writes:

>#This is rapidly becoming a contender for Technological Myth Of The Decade.
>
>Err, correcting Dr. Jefferys on something to do with HST is best done
>very carefully. Dr. Jefferys is the PI on the Fine Guidance Sensors...

To the best of my knowledge, the only statement in my posting which is
not a solidly-established fact -- verifiable by reference to the Allen
panel's report -- is my opinion that P-E management, having ignored two
indications of trouble (actually three, if you count the fact that the
reflective null corrector unexpectedly required extra spacing washers),
would have disregarded the results of other simple tests too, unless
they were absolutely unmistakable (i.e. an end-to-end imaging test).

If I am in error -- certainly possible -- I'd appreciate hearing about
it, preferably with enough detail so I can verify it.
-- 
MS-DOS is the OS/360 of the 1980s.      | Henry Spencer @ U of Toronto Zoology
              -Hal W. Hardenbergh (1985)|  henry@zoo.toronto.edu  utzoo!henry

Date: Tue, 17 Nov 1992 21:27:04 GMT
From: Henry Spencer <henry@zoo.toronto.edu>
Subject: Hubble's mirror
Newsgroups: sci.astro,sci.space

In article <1992Nov16.033555.26144@murdoch.acc.Virginia.EDU> gsh7w@fermi.clas.Virginia.EDU (Greg Hennessy) writes:
>#To the best of my knowledge, the only statement in my posting which is
>#not a solidly-established fact . . .
>
>I was referring to your statment implying that Dr. Jefferys was
>contributing to the "technological myth of the century." Correcting
>Dr. Jefferys on matters of the HST is like correcting Dennis Ritchie
>on matters of C.

Since I've been known to do the latter, I don't see why I shouldn't do
the former. :-)  Actually, I don't immediately remember having had
occasion to correct Dennis on the net, but I have found bugs in both
his code and his book.
-- 
MS-DOS is the OS/360 of the 1980s.      | Henry Spencer @ U of Toronto Zoology
              -Hal W. Hardenbergh (1985)|  henry@zoo.toronto.edu  utzoo!henry

Date: 19 Nov 92 20:40:53 GMT
From: "William H. Jefferys" <bill@bessel.as.utexas.edu>
Subject: Hubble's mirror
Newsgroups: sci.astro,sci.space

In article <BxqDzI.B1q@zoo.toronto.edu> henry@zoo.toronto.edu (Henry Spencer) writes:
#In article <83625@ut-emx.uucp> bill@bessel.as.utexas.edu (William H. Jefferys) writes:
#>I'll agree, the error would have been caught if the primary
#>and secondary mirrors had been tested together. There are
#>lots of simple tests that would have caught the error.
#>None were used.
#
#This is rapidly becoming a contender for Technological Myth Of The Decade.
#
#If you look at what actually happened, rather than the popular mythology,
#the fact is that three independent tests were run, and TWO OF THEM DETECTED
#THE ERROR.  At which point, Perkin-Elmer management decided that the third
#was more trustworthy than the other two, and ignored the two failed tests.

The tests to which you refer are the refractive null and
the inverse null, in addition to the faulty reflective null.
All three of them are based on the same basic interferometric
technology, and are not independent in the sense that I 
meant. You are correct, had P-E decided to investigate the 
discrepancies between the reflective null and the other two 
devices (which agreed with each other) the problem would have
been discovered. Had the optics people who built the reflective
null realized that there was a problem when they found themselves
shimming out the lens cell for the field lens with washers, the
problem would have been found. Had they measured the position of
the field lens with a PLASTIC RULER, it would have been found.
There were lots of red flags flying, and they were all ignored.

By an independent test, I meant a geometric-optics test such as a
Hartmann test that could have, fairly cheaply, detected the problem. 
I certainly don't want to be accused of promoting a Technological 
Myth; I am personally quite cognizant of the technical issues involved,
owing to my long association with the HST project. But I am using
'independent' in a different sense than you. 

Bill

Date: 25 Nov 92 05:15:05 GMT
From: "William H. Jefferys" <bill@bessel.as.utexas.edu>
Subject: Hubble's mirror
Newsgroups: sci.astro,sci.space

In article <BxxLo9.I6H@zoo.toronto.edu> henry@zoo.toronto.edu (Henry Spencer) writes:
#In article <1992Nov17.121839@cs.man.ac.uk> mario@cs.man.ac.uk (Mario Wolczko) writes:
#>Any idea why an end-to-end test would have been more susceptible to
#>gravitational problems than the null corrector test?  Couldn't both be
#>performed with the primary flat on its back?  And why are there
#>more risks of surface contamination?
#
#Adding tests means moving the mirrors around and doing work in their
#vicinity, which automatically increases the risk of contamination.
#The problem there isn't the nature of the test, but simply the fact that
#it's yet more fiddling with the mirrors, when you would really like to

#handle them as little as possible between manufacturing and launch.

Contamination was a major concern because even a small
amount of organic contamination (a little oil, human
evaporant, or whatever) would seriously degrade UV
performance.

#I would guess that the null-corrector tests were done with the mirror
#flat on its back; it would seem the obvious approach.  The problem with
#gravitational distortion -- I would think -- is simply that it requires
#doing the test in a vertical orientation, which considerably complicates
#the test facility (if for no other reason, because you need a vertical
#shaft of considerable height to mount everything in).  That is, it's not
#something you could expect the P-E optics shop to be able to rig up with
#equipment on hand; it might even need a special building (although there
#are reports that the USAF already has one).

Right. The interferometric tests with the null corrector were
done with the mirror horizontal, in a large vacuum chamber.
The mirror was mounted on a special "metrology mount" that
compensated for gravitational effects by specially adjusted
weights. I doubt that an _in vacuuo_ two-mirror test could
have been performed[*]. A crude "sanity check" test could have
been done; such a test, if planned (it was not) would have
given the whole test process more visibility & might have 
forced the powers that be to recognize that there was a
problem. The problems with the null correctors were too
easily swept under the rug.

Bill

[*] for reasonable cost, that is.

Date: Wed, 25 Nov 1992 19:23:37 GMT
From: Henry Spencer <henry@zoo.toronto.edu>
Subject: Hubble's mirror
Newsgroups: sci.astro,sci.space

In article <83981@ut-emx.uucp> bill@bessel.as.utexas.edu (William H. Jefferys) writes:
>#... three independent tests were run, and TWO OF THEM DETECTED
>#THE ERROR.  At which point, Perkin-Elmer management decided that the third
>#was more trustworthy than the other two, and ignored the two failed tests.
>
>The tests to which you refer are the refractive null and
>the inverse null, in addition to the faulty reflective null.
>All three of them are based on the same basic interferometric
>technology, and are not independent in the sense that I 
>meant...

Okay, point taken -- more diversity in tests is desirable.  However...

>...Had they measured the position of
>the field lens with a PLASTIC RULER, it would have been found.
>There were lots of red flags flying, and they were all ignored.
>
>By an independent test, I meant a geometric-optics test such as a
>Hartmann test that could have, fairly cheaply, detected the problem...

Ah, but would it have been believed?  That's the crucial problem.  My
impression is that it wasn't "we've got three similar tests, and the
best one says we're okay", but rather "we've got three tests, and the
best one says we're okay".  The trick was not detecting the error --
after all, it *was* detected -- but detecting it in a way that would
have been sufficiently emphatic and unmistakable to make P-E doubt
their beautiful reflective null corrector.

The point of an end-to-end imaging test, in this context, is not that
it's an easy way to detect the problem, but that it's so hard to
argue with.
-- 
MS-DOS is the OS/360 of the 1980s.      | Henry Spencer @ U of Toronto Zoology
              -Hal W. Hardenbergh (1985)|  henry@zoo.toronto.edu  utzoo!henry

Date: 3 Dec 92 22:22:04 GMT
From: "William H. Jefferys" <bill@bessel.as.utexas.edu>
Subject: Hubble's mirror
Newsgroups: sci.astro,sci.space

In article <ByADvE.4nn@zoo.toronto.edu> henry@zoo.toronto.edu (Henry Spencer) writes:
#In article <83981@ut-emx.uucp> bill@bessel.as.utexas.edu (William H. Jefferys) writes:
#>
#>By an independent test, I meant a geometric-optics test such as a
#>Hartmann test that could have, fairly cheaply, detected the problem...
#
#Ah, but would it have been believed?  That's the crucial problem.  My
#impression is that it wasn't "we've got three similar tests, and the
#best one says we're okay", but rather "we've got three tests, and the
#best one says we're okay".  The trick was not detecting the error --
#after all, it *was* detected -- but detecting it in a way that would
#have been sufficiently emphatic and unmistakable to make P-E doubt
#their beautiful reflective null corrector.

I can't argue with that! The tests that were performed
lacked "visibility," that is, adequate oversight by people
not involved with the actual testing to ensure that
the anomalies would have been investigated instead of
ignored. At the time, everybody thought the mirror project 
was in good shape. People's attentions were directed 
elsewhere, because another telescope component was felt
to be at high risk. Thus it was possible for people at a
fairly low level to dismiss the mirror anomalies as inimportant.

#The point of an end-to-end imaging test, in this context, is not that
#it's an easy way to detect the problem, but that it's so hard to
#argue with.

Absolutely correct. And an end-to-end test would have had
"visibility."

Bill

Date: Fri, 23 Jul 1993 14:07:22 -0500
From: "Geoffrey A. Landis" <GLANDIS@LERC.NASA.GOV>
Subject: Hubble solar arrays:  how'd they foul up?
Newsgroups: sci.space

I heard a talk by the guys who made the Hubble arrays at the European Space 
power conference shortly after Hubble flew.  Up until Hubble, the European 
experience in solar arrays was entirely with rigid arrays; Hubble was the 
first lightweight flexible array designed and built in Europe.  The speaker 
said that there was a vibration spec in the procurement specifications, but 
they didn't notice it, and didn't pay any attention to the possible problem.  
He said that they were very eager to get a chance to correct the problem, and 
pressed NASA to put replacement of the solar array on the repair mission.

There's quite a body of experience with flexible arrays in the US, but the 
Hubble arrays were the European contribution to the project...