Click here if you are having trouble logging into the forums
Privacy Policy |
Please read our Terms
of Service | Signup and
Troubleshooting FAQ | Problems? PM a Red or a Green Gu.... uh, User
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
A "discussion" with a gentleman over the sensitivity & accuracy of the standard knife edge test, known as Foucault led me to believe he was in gross error. I recently completed a 6" F/13.63 mirror, that I tested thoroughly 4 times with naked eye, and 4.5x50mm scope attached to the test stand. More of this later.
The disagreement was, he insisted that Foucault was not accurate to closer than .010" accuracy at the ROC with reads. I felt that the consensus was this equated to about 1/10th wave as approximate (swa) guess.
I knew this was, if not grossly incorrect, was too fast a reply. F/D plays a large part in assessing tolerance of mirrors. Also, .010" was, and has been bantered around both in literature and in various places as "accepted norm". This may appear to be true if some details are accepted as "good enough". Those being-
Construction of toy testers as wooden blocks sliding on rods, with poor or springy KE arms with various loose tolerances.
Testers as a unit, being fairly light, and too easily moved by careless testing.
Accepting everything as read, being"the final word". I call this "the earth is flat" syndrome. I also believe that various pieces quoted as sources, are "cross referenced" and so have an inevitable bias as to what's possible. In particular, the .010" accuracy limit seems to be 5th hand and is repeated ad nauseum.
***********************
On to the test of that long mirror. Through data obtained on hundreds of mirrors, I have the following range of accuracy of readings with -MY- tester.( It consists of a lathe cross slide complete with adjustable gibs, on a flat steel top, which is welded to a column affixed to a transmission bell housing. To that, there is welded a tripod. Total weight around 100# or more. It doesn't move easily, nor flex when pushed on hard)
ROC range Tolerance (max + - value, typically -within- said value)
to 60" .001"
60-120" .002"
120-160" .003"
200-300" .005-.015" respectively
*********************************
The "nominal" test of this mirror revealed the efforts at making predominantly a sphere from the start. There exists a roll off at the edge, which equates to +.002" at the 4th zone, and .012" at the outer zone. Approximately 3/4" ~ of the outer part of the mirror is outside of the nominal sphere in the central 4.5"
It has to be remembered that this spherical section is a null with the KE test, and has all the inherent characteristics of a null. You see all the surface at once, astigmatism is glaringly obvious if present, and I had repeatability in this area down to .001" as long as I stuck with either eye, or scope throughout the test sequence.
The outer 3/4" of mirror, I bagged as +.002", and +.012" respectively in the outer 2 zones. Repeatability was within .0025" when average of all 4 tests were reduced.
Shadowing was very delicate and with just a cursory glance, the mirror appeared to be a perfect sphere. Only under close scrutiny over the period of an hour and a half could all the finer points outlined above be ascertained. Data on following post later.
--------------------
So many projects, so little time!
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
Test data as follows:
Spherical section comprising central 3 zones taken as .026" The 4th zone showed a very slight increase of ROC to the tune of registering +.002"~ average on the tester. The outer zone was reading +.012" average throughout the tests Tabulated as follows:
1 2 3 4 5
.026 .026 .026 .028 .038 (PVW= 1/37th, 1.00 St., max RTA=.1837)
Of course, a sphere being relatively "simple" to make, reduces as follows on my spreadsheet.
PVW Strehl max rta
1/17 1.00 .3216 (this all without refocusing from above nominal data)
When refocused -.0015", results are as follows:
1/25 1.00 .2902
**************************
I'd like to cover the nominal measurements of this mirror as tested with the spherical information.
When taking off the correction at the outer zone equates to a change of the wavefront error, magnitude being 1/17-1/37 = .05882-.02703, = .03179(1/31.5 additional error across the .012 delta change. No refocusing.
The refocused sphere amounts to adding smaller error, equating to:
1/25 - 1/37 = .04000-.02703, = .01297(1/77)
Taking the average of these equates to about 1/44th additional error induced my retaining a spherical figure, over the range of .012" delta of the longitudinal abberation. I did this to get an average between the 2 extremes.
**************************
Lets suppose that the addition is linear across the total delta. I know it isn't exact, but for brevity's sake to make a point when splitting hairs-
If the test is reviewed above, the tolerance found for the outer zone readings are found within .0025" (remember, this is my tester)
If we take .003" as the basis of acceptable tolerance with my test rig then approximately 1/4th of the additional error can be taken as a rough sensitivity with the setup I use.
Over that .003 total variance of my reads, the sensitivity of this KE tester exceeds 1/100th wave. This may be a moot point with this particular mirror. ML tolerance for the outer zone is + - .044". It might be argued from his point of view it's needless. But long mirrors aren't the only thing. Short mirrors exist as well.
This long mirror favors the characteristics that KE provides quite well. It maximises the accuracy potential of the test. There is -NO- other test whatsoever that can approach this precision existing today that I'm aware of with long mirrors Interferometers/AC included.
***************
But the blanket statement that KE is only good for .010" gross tolerance of readings/ accuracies no better than 1/10th wave PVW is bunk, plain and simple. That means the literature is incorrect. It all depends on how precise -YOU- want to be/ what's acceptable to you. I'm posting this to show what's possible with careful and diligent testing, good experience, and proper equipment; and what is factual first hand information. Regards, Mark
--------------------
So many projects, so little time!
Edited by Mark Harry (10/23/09 08:06 AM)
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
An old pic of my tester, with hundreds of mirrors to its credit. I have made a few changes, so I'll try to update this shortly. The battery is just in the background, not a part of the apparatus.
--------------------
So many projects, so little time!
Edited by Mark Harry (10/22/09 04:50 PM)
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
The slit has been replaced with a pinhole. There is a place for a scope and bracket to be bolted behind the blade. There is absolutely no play with the KE itself, and is operated by a floating pushrod/thd rod arrangement. Somewhat crudely executed, but very effective.
**************
One thing I must add about using either just my eye, or the scope. The deltas, and repeatability was exactly the same value wise. There was no difference. There were things I could see with only the scope, and higher contrast with just my eye.
One thing where the scope differed- The nulls, and zonal measurement readings over 161.58" ROC was .020" shorter than just my eye -AS A WHOLE-.
I also did some calculations of misalignment with the KE stage to the optic. 2 degrees amounts to an inconsequential difference in the readings (~.0001" over total correction)where it would affect error assessment. And that represents a rather HUGE amount (almost 6" at this ROC!), considering the flatness of my floor, and the initial aim of the stage itself.
I posted this as food for thought. I'm sure there are those who'd dismiss it prematurely. Their choice. It took me about 3 years to reduce my errors to the values stated with this tester. I suppose one of the wooden block testers could be built that could provide .010 repeatability in a couple evenings. I thought the effort I put into the tester paid smaller returns the longer I spent on it. But at this point, when I test with it, the results are what they are.
If I recall correctly, the ML standard varies linearly. That being so, if this mirror could be ascertained to be tested to 1/100th precision, and F/4.5 could be fabbed to 1/30th wave sensitivity easily; not taking into account the increased sensitivity at shorter ROCs.
Enjoy, Mark
--------------------
So many projects, so little time!
|
kfrederick
professor emeritus
Reged: 02/01/08
Posts: 660
|
|
HEY MARK where is the ronchi ? You showed me alot when i flew up .thanks for the info .kf
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
Scope/eye tidbits- Well, as I mentioned, the deltas were exactly the same with either eye or scope. The scope magnifies approximately 4.5x. It spreads the light over a roughly 20-fold bigger image plane by area, and contrast is drastically reduced. I feel that a magnification more like 2.5x might be about the right compromise to facilitate looking at the glass and obtaining the information needed. As it was, between both ways, The need of using a scope on a small mirror across a large distance I see as being a good idea. The scope excelled in determining width of zonal areas, and edge defects. I wish I could take photos, but my wife's camera tries to focus on the slit, so it never comes out worth a darn.
M.
--------------------
So many projects, so little time!
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
"where is the ronchi ?"
****************
(I stuck it in your flight bag when you weren't lookin'!)
M.
--------------------
So many projects, so little time!
|
mark cowan
Vendor (Obsidian Optics)
Reged: 06/03/05
Posts: 2157
Loc: salem, OR
|
|
Hi Mark,
I'm reading through this very interesting thread a bit slowly, but allow me to contribute the following:
On a well corrected smooth mirror Foucault is geometric in execution, save the very edge of the mirror where diffraction effects alter the zonal reading slightly (and this can be modeled and so corrected for).
The inherent accuracy of Foucault absent other errors typically encountered (human bias and limitations; mechanical - tester slop, inadequacy, inaccuracy; lack of thermal equilibration; inadequate support of mirrors under test) is extremely high.
For a fused silica 14.7" f/4.5 that I tested very carefully manually both visually and digitally (the latter having less bias) the agreement on the applicable areas was better than 1/40th wave PVW. The manual testing was quite repeatable, typically <.05mm on the longitudinal, fixed source, in line I think with what you're showing.
I haven't analyzed the repeatability on the digital system, but expect to as I'll be using the full-up automated version of the same technique in production soon. Here's a link to the reduced data comparison - it makes my point better than any number of words.
Now one can ask whether both of the curves are in error, and it will take further validation through a calibrated IF to answer that, or at least to pin another error bound on the data.
Best,
Mark
|
Mike I. Jones
Carpal Tunnel
Reged: 07/02/06
Posts: 1572
Loc: Fort Worth TX
|
|
I tend to agree with Mark that after making over 55 mirrors myself and testing probably over a hundred other mirrors over the past nearly 40 years, I have developed a fairly keen eye for zonal testing and can usually repeatedly call zones to within 0.003"-0.005" for mirrors up to f/8 (f/16 at COC), at least in the outer zones. For f/4 mirrors and narrow edge zones I can usually repeat outer zone readings to within 0.002". The central zones are always cranky to read and my scatter increases toward the center, but their reading tolerance increases anyway. I visited about this very thing with Carl Zambuto off and on while he was making the McDonald Observatory 18" f/8 mirrors and he's pretty much of the same opinion regarding his zonal reading repeatability.
My 20 millidollars: to repeat zone readings to below 0.01" during a long work session requires:
(1) A precision tester with ball-bushing ways or slides, real micrometer spindles, and zero backlash (springs and/or magnets)
(2) A slitless tester and nice bright LED light of your preferred color and variable brightness (I like yellow the best),
(3) A REALLY steady test path with zero turbulence, often achieved by making a tunnel with bubble wrap right up to the test stand, and
(4) Jitter-free mirror and tester stands, preferably with remote-control az/el motors on the mirror stand to eliminate the endless back/forth trips during alignment.
I don't EVER want to take away from the simple wood and bolt-based testers so many people use and are rightly proud of - I built one myself way back at my start in ~1971, then bought one of those old University Optics testers and used it for years, until I got lucky at a ham swap meet and found the 2-axis precision stages with micrometers I use now.
But to punch through the 0.01" and 1/10 wave accuracy/precision barrier does take some fancier and more precision equipment, as well as a practiced eye that comes from doing a lot of mirrors, and independently cross-validation testing them on the sky, with autocollimation and/or with an interferometer.
Mike
|
John Carruthers
Skiprat
Reged: 02/02/07
Posts: 2269
Loc: Kent, UK
|
|
I agree that 0.010" may be a fair estimate for some testers but I would hope to do better. My tester is built around a micrometer head for advancing the knife. Several (read 'many') tests improve accuracy too. Repeatability is all and it was satisfying to see my values duplicated on another test bench at a different location with a different opperator.
--------------------
Jc
ATM 10" F6.1, 1/25th wave spec (max wavefront error +/- 1/12.6 in zone 4 of 6, sodium light )
6" F7 spec
127mm F9.4 Refractor
10 x 50 bin
ETX80 (finder)
Canon 20D
PST
DSI 1
and a curious mind
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
I have a new pic showing the scope mounted
--------------------
So many projects, so little time!
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
2 pics of the mirror test "rack". It's a massive column mounted on floor and joist bearings, made out of ash hardwood, and 2x6s, glued and screwed together. It has a hinged spindle mount so as to accomodate a mirror mounted on a backer plate directly off the machine polisher. It also has bare glass mirror rack shown in use.
When the spindle mount is used, if there's a section of the mirror I don't touch, the repeatability is -WITHIN- .001" whatever the ROC is with any subsequent reads.
Mike mentioned about taking away from the wooden block testers, and that is the last thing I intend to do. -HOWEVER-, if you want to track just what the result of figuring actions are doing to the glass, you -HAVE- to have this repeatability available to make correct assessments of the actual wear taking place. This will quickly show how scetchy and inaccurate most all figuring information is in the printed literature that's available. I can't stress this fact enough, sorry to say. If correct information was available, it would eliminate a lot of tail-chasing, IMO.
Mark
--------------------
So many projects, so little time!
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
The spindle mount.
--------------------
So many projects, so little time!
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
Mike wrote:"and independently cross-validation testing them on the sky, with autocollimation and/or with an interferometer."
Mark C. wrote:" will take further validation through a calibrated IF to answer that,..."
*********************************
I agree with testing them in actual end use, but with the case of this mirror, where the central 3/4 of the mirror's a sphere, AC will not show a null. Also, the 4th zone defined as +.002 delta, will result in a movement of the stage of AC setup in the vicinity of half, to 1/4 thousanth of an inch; depending on the light source being fixed or moveable(generally the latter) AC might see it, but I doubt it.
IF reference optics are considered mighty fine if 1/20th surface or better. I'm not sure it could possibly reveal any additional worthwhile info with this particular mirror. Not sure -if- the calibration required is possible...? I doubt it. There would have to be a point of reference used...?
*******************
I was spoiled with IFs. I could set a block of lenses in a rack, center the IF pattern, go to the monitor and stab a button on the keyboard. The phase shift would sweep across the optics, and BANG, there were the results of error, Strehl, RMS, etc. in about 1 second.
But when I came home the realisation of affording such a setup was way outside my wallet. I chose what was the best test as far as ease of use, reduction information available, least amount of reference optics, and highest sensitivity available within reason. KE made the most sense, especially when coupled with my previous polishing experience. A compromise nonetheless, but it also is fairly easy to understand conceptually and mathematically; a big plus.
All I can say, it works for me very well. I can see astig, detect anomolies well below the threshhold of other similar caliber home-brew test methods, track specific and targeted polishing regimens, and measure slope error very accurately. Might not be everybody's cup of tea, though.
********************************************
Referring to the first mentioned "discussion" I would not think it proper that my commentary be condemned on that aspect alone, nor accuracy potential doubted based solely on their particular experience and capabilities; and their use of antiquated and biased 5th hand information as references for their justification. I hope the gentleman comes to that realisation in the end.
--------------------
So many projects, so little time!
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
(sorry for the rant!)
M.
--------------------
So many projects, so little time!
|
kfrederick
professor emeritus
Reged: 02/01/08
Posts: 660
|
|
HI Mark How much more senitive is the knife edge as the distance increces .NICE setup
|
RossSackett
professor emeritus
Reged: 08/17/07
Posts: 691
Loc: Memphis, TN
|
|
Mark,
You've tramped through a lot of territory here that I have no qualification to comment on, but I will mention that I completely agree with the advantages of a heavy, solid base for the Foucault tester and smooth linear motions. I anchored my tester on a granite tile, and use an X-Y table designed for a milling machine for my motions, both for finding ROC and cutting in. I use a 0.001" micrometer head for the actual readings rather than the divisions on the handle wheels. With calm air, a dark room, and a 96" ROC mirror close to ambient my repeatability was around 0.003", not far from your experience. I also found that I was more reliable nulling by first darkening the zone, then cutting out looking for the first glimmer of light (rather than starting full-bright and cutting in for the first hint of a shadow). (I think I posted on that over the summer.)
I realize this isn't really what you were on about, though. 
Ross
--------------------
"A craftsman relies on science when the state of knowledge allows it, tradition and experience when it does not, and makes art whenever he can."
12 scopes from 4.25 to 18" and a 24" in progress. 12 ATM awards. Webpage: http://stardazed.com/ Some more scope pix at http://www.flickr.com/photos/8315630@N04/
Anagrams: Amateur astronomer = A mature moon-starer; Dobsonian maker = Debonair as monk
|
Gary Fuchs
scholastic sledgehammer
Reged: 05/22/06
Posts: 867
Loc: Easton, PA, USA
|
|
Mark (et al?),
Much of this is over my head but I have a "toy tester" (I guess two if you count the AC) and probably the means to make something a little better so have a question:
How do you align the tester so that the KE stage is moving parallel to the optical axis of the particular mirror for that test session? Or do you need to? In other words so that at both extremes of its (longitudinal I think) travel it's parallel.
Quote:
but with the case of this mirror, where the central 3/4 of the mirror's a sphere, AC will not show a null.
Anyone:
This puzzles me. Why wouldn't AC show a null for the central 3/4 of the mirror, or for that matter, the entire mirror, given the focal ratio mentioned?
From my very limited AC experience it seems, with a fine screen, say 250 lpi, that it is very sensitive to edge errors - they stand out very clearly. I wonder how that resolution compares to what you can evaluate with KE, either visually or with the automated system described?
Thanks,
Gary
|
Mike I. Jones
Carpal Tunnel
Reged: 07/02/06
Posts: 1572
Loc: Fort Worth TX
|
|
Gary,
I test carriage alignment by first running the longitudinal stage in and out (axially) with the KE cut halfway in. If the KE shadow stays put during axial translation, it's aligned. If the shadow translates laterally during axial translation, that tells you which way to turn the stage. You can then turn the KE 90 deg. and repeat the process to get vertical alignment. I probably should have added good alignment to the list of 0.01"-busters I posted earlier.
I've started making an adapter for a laser pointer on my tester to try to speed the alignment up but haven't finished it. Even if that works well I'd still double check it with the process above.
Mike
|
DAVIDG
Carpal Tunnel
Reged: 12/02/04
Posts: 1985
Loc: Hockessin, De
|
|
One also needs a method to check the test procedure as a whole to determine if one has any bias in doing the test. How one person judges equally nulling of the zones can very differently from the other and one method can be correct and the other not. Without a cross check to making sure that there is no bias, one can be systematically adding errors to the results and not know it. These erors can both come from the person doing the test, the equipment and the procedure.
One common mistake is to move the knife edge close to were the ideal reading should be and then try to find a null. If you want to check your repeatablity you need to do the test and have someone else record the values. Then take the whole setup down, set it back up and do the readings again. You need to do this at least three times, ten is better and look at the standard deviation of each group of zone readings. One also should start at different zones, not just the center or the edge. In theory it shouldn't matter were you start, if the data shows it does, then you have a bias.
- Dave
--------------------
Homemade 'scopes 8"f/7,6" f/5", 6"f/4, 4.25" Schief. 60mm Coronagraph,60mm H-alpha system, 4.25" White-light Solar Newtonian,solar spectroscope, 4.5" f/16 Schupmann Medial refractor, 14 Stellafane awards 7 in optics
Engineering = Taking what you have and making what you need.
|
mark cowan
Vendor (Obsidian Optics)
Reged: 06/03/05
Posts: 2157
Loc: salem, OR
|
|
Quote:
if you want to track just what the result of figuring actions are doing to the glass, you -HAVE- to have this repeatability available to make correct assessments of the actual wear taking place.
I agree entirely - and I think this is often overlooked. The idea that you'll measure it carefully at the end but just slop along until you get to the end doesn't hold water. To predict figuring action you need to be able to measure figuring action, to the resolution you expect to finish the mirror to. Otherwise you get stuck in a cycle of correcting the corrections. 
Best,
Mark
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
"This puzzles me. Why wouldn't AC show a null for the central 3/4 of the mirror, or for that matter, the entire mirror, given the focal ratio mentioned?"
Because the central 3/4 of this mirror is a -sphere-. Not a paraboloid. AC nulls on a paraboloid only.
M.
(I could see that the central 3/4s of this mirror was a sphere with my KE)
I also pointed out that with a misalignment of 2 degrees, which equates to almost 6" of deviation from the axis,(L,R,U,D) shrinks the measured values by a factor of 1/10,000" over the total correction value of .055" (inconsequential) That's a lot of leeway, and a gross misalignment. The measurements would still be valid.
M.
--------------------
So many projects, so little time!
|
Gary Fuchs
scholastic sledgehammer
Reged: 05/22/06
Posts: 867
Loc: Easton, PA, USA
|
|
Thanks Mark,
Quote:
"This puzzles me. Why wouldn't AC show a null for the central 3/4 of the mirror, or for that matter, the entire mirror, given the focal ratio mentioned?"
Because the central 3/4 of this mirror is a -sphere-. Not a paraboloid. AC nulls on a paraboloid only.
M.
I'm still puzzled. My understanding is that AC gives a null with a paraboloid because the (proper) paraboloid is returning all the parallel rays to one point - focus, or I guess practically speaking, to within a close enough distance that we can't tell the difference visually looking at a star.
I forget or probably never really understood the theory so please bear with me technical-wise.
And that at a certain point where the ROC of the mirror is long enough, the difference between spherical and parabolic is small enough, in terms of where the returned rays fall, that we can't see a difference. I haven't tested it - don't want to pull out my Schief primary - but it would seem that at or near anyway the f13 or so focal ratio of your mirror the AC test would show a null, for the center, if not the entire mirror.
No reason I can think of not to use KE or Ronchi for that but beyond the 3/4 I'll bet AC with a fine (~250 lpi) screen will be pretty good.
Gary
|
Wes James
Post Laureate
Reged: 04/12/06
Posts: 3452
|
|
While I am too inexperienced to hang with you big dog mirror makers, and meaning no disrespect to those who are, I AM familar with quality control processes and the accuracies needed to ensure repeatability and tolerances to make a good mirror. My only experience making mirrors are that I attended the DelMarva mirror making weekend this past year/refiguring a mirror using Dick Parker's double pass autocollimator- and after getting my 4.25" Schief optics to the polish/figuring stage utilizing the foucault test. In other words, I have no firsthand experience figuring a mirror using a mask! However, even with that limited experience, I can't understand why anyone would still utilize zonal masks for parabolizing a mirror- when you absolutely see the surface on an A/C tester. (I'm speaking on a large scale procuction/commercial basis here, not the amateur ATM'er making a few mirrors). It was so easy- actually seeing where you needed to work than having to read zones and carefully measure, hoping you can achieve the same shade in on each zone the same every time. With a double pass autocollimator, you see it right before your eyes- a parabola is a null... what could be easier? No having to accurately and repeatedly match identical shades of grey in different zones... you get the tester adjusted- which only takes a few moments, and look at your surface. As it actually is. TDE's, hills, zones, ridges, all clear as can be. It has been said that the only way that the DelMarva mirror making workshop can turn out so many mirrors on a 3 day weekend is because of the speed and ease of setup of Dick's A/C tester and Steve Swazey's Ross Null test bench he brings, also a null test for a parabola. Time is also money for a commercial mirror maker, is it not?
So, my question is, for commercial mirror makers to be using masks and knife edges, there must be a reason... Other than the size of the flat required- but if you're doing it for a living- one should be able to make one. A smooth flat is the main requirement, not flat to a ridiculous wavelength. An A/C tester tests at half the distance of a knife edge, too... what could be easier- and more accurate? What am I missing?
While it could be argued that it's been being done this way forever, and one can produce a good- or a great mirror with the foucault tester, we also used to be able to make calculations with a slide rule just fine, thank you very much (I still have one!)... and they still work... and build cars on an assembly line without robotic machines. We used to do fine with snail mail... why do we need email? The only thing that I see you can't do with an A/C bench that you can with a foucault tester is come up with an exact number at which to rate a mirror- but if you have a better mirror easier, who cares if it's 1/10 wave or 1/20 wave? I'd rather have a mirror that I know visually is a great mirror- than have a mirror that's rated at some number which may or may not be really accurate. And with the issues of repeatability and judging shades of grey identically- however many times one must to evaluate a mirror's zones? I believe the mechanical issues can be resolved to a reasonable accuracies, but I fail to see how the visual judgement issues can be resolved to the accuracies we're looking for here.
I hesitate- due to my lack of experience to even ask the question- feeling I don't have the "dues paid" thing, but the question continues to nag me. So I'll ask it anyway!
--------------------
Wes
Atlantic Beach, FL
Some bino’s from Miyauchi 5x32 Binon's up through Garrett 20x110 Signature's,
Some telescopes from a Stellarvue 80mm NHNG up through a couple of 8” reflectors…
And a wonderful 4.25" Delmarva Shiefspiegler!
Some good friends, made here on C/N.
Oh- several cats and a wonderful wife!
Anyone want a cat???? :-O
"When your work speaks for itself- Don't Interrupt" -Gamble Rogers
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
"but just slop along until you get to the end doesn't hold water. To predict figuring action you need to be able to measure figuring action, to the resolution you expect to finish the mirror to. Otherwise you get stuck in a cycle of correcting the corrections."
***********************
Funny thing you brought this up! I had looked at the mirror the previous night. It was oblate, with a little edge issue left after smoothing it off with pitch.
Next day, I started to polish it again, for a period of about 40 minutes. When I had about 10 minutes to go, I diluted the cerium, placed 1 drop of black rouge on the mirror, did an oval with an amplitude where the center of the glass was hanging right on the edge of the same size polisher and back to COC. I made 4-5 ovals while moving the mirror through about 2" of offset. I did this for perhaps 4-5 complete turns. Then, I turned it over, and did some 1/2 oval strokes with about 1.5" of offset total for about the remainder of the 10 minute interval. Otherwise, I never looked at the mirror on the rack. The results are what's reported initially in the post.
This is what -CAN- ocurr if results of wear from polishing is predicted and executed correctly. That is the gist of this posting, as well as how a simple test apparatus can help one accomplish such a feat. You saw right through that! 
Oh, what a difference 15 + years make! A big transition from a 30k point phase-shift IF to KE. It was like going back to the stone age, but in actuality, the KE has a lot of serious potential, if utilized correctly.
Mark
--------------------
So many projects, so little time!
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
Wes, an AC will only show a null on a paraboloid, and any other shape, such as a sphere, ellipse, or hyperboloid one loses the advantage of a null.
On top of that, AC does -NOT- have the same sensitivity in measuring longitudinal abberation- in fact, none at all, with a paraboloid. As you have noted, it's not as straightforward to assess a numerical value to the errors present. (At least with the sensitivity I'm comfortable with.)
This has been hashed out before numerous times. Seems like about half a dozen times a year, on average. So I'm not going to get into all the ramifications and write a book here.
IMO, everyone here that favors AC, most all recommend getting hold of a surplus shed piece of doubtful origin and quality. Or a few do make their own flat. The only flats I see that are adequate are made by optical shops, and guys like Dave Kelly. They take the time to make it RIGHT. To make a flat with say a few 10's of fringes of power can't adequately be assessed as being smooth enough. Absolutely not!
But I don't want to discourage what you like or want to do. If you like AC, and don't want to bother with the math, go for it. I'm simply showing what can be accomplished with no other optics to get in the way of high sensitivity and accuracy. And to dispell myths and misquotes about KE.
M.
--------------------
So many projects, so little time!
|
rwiederrich
Goldfinger
Reged: 11/17/05
Posts: 8307
Loc: Bremerton Washington
|
|
Mel Bartel even suggest to use your eyes instead of some lengthy math equasion to determine fine parabaloids by way of visually comparing a computer Ronchigram to your own mirrors ronchigram. He contends that your visual comparison is very accurate and will result in an accurate parabaloidal figure.
Just another way.
Rob
--------------------
www.goldmtobservingcenter.com
A great place for amateur astronomers, and ATM's to come and enjoy their hobby.
http://tech.groups.yahoo.com/group/HomemadeRefractorTelescopes/ My homemade refractor group.
www.vimeo.com/6014031
|
Wes James
Post Laureate
Reged: 04/12/06
Posts: 3452
|
|
Mark-
Thanks for taking the time to reply. This still does not answer the question of why the foucault/knife edge is still favored in the commercial mirror making environment when A/C is so much faster and easier? I don't believe the commercial environment would not be using a surplus shed flat!
And, you're right, it seems the topic does get brought up not infrequently. I was asking in a slightly different vein, though- as pertains to the production shop/environment.
Best~
Wes
|
Gary Fuchs
scholastic sledgehammer
Reged: 05/22/06
Posts: 867
Loc: Easton, PA, USA
|
|
Quote:
Gary,
I test carriage alignment by first running the longitudinal stage in and out (axially) with the KE cut halfway in. If the KE shadow stays put during axial translation, it's aligned. If the shadow translates laterally during axial translation, that tells you which way to turn the stage. You can then turn the KE 90 deg. and repeat the process to get vertical alignment.
Thanks Mike - I'm going to try that.
Gary
|
mark cowan
Vendor (Obsidian Optics)
Reged: 06/03/05
Posts: 2157
Loc: salem, OR
|
|
Quote:
Referring to the first mentioned "discussion" I would not think it proper that my commentary be condemned on that aspect alone, nor accuracy potential doubted based solely on their particular experience and capabilities; and their use of antiquated and biased 5th hand information as references for their justification. I hope the gentleman comes to that realisation in the end.
You know, I've heard that same thing so many times, and yet I know it just ain't so. Somehow, and god knows how or when, this gem of "KE accuracy" got ingrained and passed on. I did once see an old OSA article that analyzed KE testing and came up with around that number. I read it, and the methodology was flawed, IMO - the skill of the Foucault operator was what was setting the accuracy, not the test itself. But the conclusion sticks.
On the other thing - yes, IF won't validate the accuracy beyond a certain point (though luckily it will do it to a point good enough for most people). So how can you validate it, should you want to? Star testing will reveal maybe 1/40th wave PVW error, if you're lucky. Autocollimation isn't going to show you anything you can't see with the KE, SFAIK (I'm not an AC fan, yet.), used properly.
Best,
Mark
|
kfrederick
professor emeritus
Reged: 02/01/08
Posts: 660
|
|
Very powerful test .
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
Wes wrote:"...A/C is so much faster and easier? "
***************************
Well, not really. It cannot be set up faster or easier than KE with the actual test.
M.
--------------------
So many projects, so little time!
|
DAVIDG
Carpal Tunnel
Reged: 12/02/04
Posts: 1985
Loc: Hockessin, De
|
|
Quote:
Wes wrote:"...A/C is so much faster and easier? "
***************************
Well, not really. It cannot be set up faster or easier than KE with the actual test.
M.
A friend ask me to test a mirror for him by Autocollimation. Here is my setup that took me about 10 minutes to put together since this a 10" f7.2" mirror and doesn't fit on my normal testing table. I didn't have to measure ROC, make a mask, make any knife edge reading and enter them into a computer program or graph them. So there is no error from them in the test, yet one look told me that this mirror is very good. With 250lpi ronchi grating the lines are very straight indicating a very well corrected mirror.
- Dave
--------------------
Homemade 'scopes 8"f/7,6" f/5", 6"f/4, 4.25" Schief. 60mm Coronagraph,60mm H-alpha system, 4.25" White-light Solar Newtonian,solar spectroscope, 4.5" f/16 Schupmann Medial refractor, 14 Stellafane awards 7 in optics
Engineering = Taking what you have and making what you need.
|
Gary Fuchs
scholastic sledgehammer
Reged: 05/22/06
Posts: 867
Loc: Easton, PA, USA
|
|
Quote:
Wes wrote:"...A/C is so much faster and easier? "
***************************
Well, not really. It cannot be set up faster or easier than KE with the actual test.
M.
From what I've seen, set up, from scratch, with an uncored rectangular flat that looks at half the mirror at a time, and I've only used mine with that kind of flat, does take (usually) a bit longer than regular Foucault KE or Ronchi.
Once it is set up though, using an index mark on the mirror, I find I can put the mirror on the stand and have the image immediately or with a slight tweak to the tester stage.
The original arrangement I tried, using a diagonal so that the entire mirror, save the obstructed center, could be viewed, was certainly more time consuming for me to set up, though again, once aligned I could usually get the image with a minor adjustment.
Now my equipment is pretty crude and while I've seen the test done a number of times and received I think really good instruction, still, I'm working on my own and who knows what I could be doing better that a more experienced and knowledgeable person might see. KE takes a bit of experience to get right too.
Now the above applies to my type and particular tester.
Dick Parker's tester, that I've seen and used at the DelMarva weekend uses a cored flat and I think is as easy, or maybe easier to set up than KE.
Here's Dick's tester in use:
The other day I looked at an someone else's mirror (with AC) and just couldn't get it aligned correctly. My son came down and had it in a few seconds. At Delmarva Dick Parker and Dave Groski (and others) were swapping out mirrors of varying sizes and f#s all day and things were moving pretty quickly.
I think that the difference in set up time, which I would think would be the time to get a returned image with either test, is, with a bit of experience (like seeing it a few times) not going to be significant. Walking back and forth between the mirror stand and tester I can get the image with Foucault in 30 sec or less, but once in awhile it takes a bit longer. Maybe I set the mirror down wrong or the tester got bumped and the laser spot is somewhere behind me on a shelf or passes right through the viewing opening and I think it's really to the side of the tester. A few more seconds in any case.
When you factor in the time you spend with Foucault KE and zone readings, and the fact that you need to make at least several readings to see where your mirror's at, and all the stage movements, fixing the mask accurately, taking and noting indicator readings - which really would comprise a complete test - then AC is certainly way ahead.
I'll also echo what Dave and Dick and others have said that using AC (or Ross Null, or any other test) doesn't mean not using Foucault KE or Ronchi or something else. You need to take care with each.
Now AC is going to cost more because of the flat, but, as I think Wes pointed out, for a pro this shouldn't be the obstacle it might be for an amateur. Time is money?
As to the subject, it would seem to me that AC testing is closer to viewing a star than a reduction of readings and all the attendant opportunities for error to creep in and so more likely to be more accurate as well as faster and easier.
Gary
|
mark cowan
Vendor (Obsidian Optics)
Reged: 06/03/05
Posts: 2157
Loc: salem, OR
|
|
Mark H. said:
Quote:
On top of that, AC does -NOT- have the same sensitivity in measuring longitudinal abberation- in fact, none at all, with a paraboloid. As you have noted, it's not as straightforward to assess a numerical value to the errors present.
I'll leave it to him to amplify (or simplify, as needed) that, but since this thread is talking about testing accuracy, that is definitely an accuracy issue. It may not have made immediate sense to some people, though.
When you want or need numbers, there are several ways to go. But to flatly assert that "AC is easier and faster" when it doesn't give numbers, is less sensitive to aberrations (according to Mark), and the standard Foucault test takes only a few minutes to set up and conduct (if you have the experience of course) is questionable.
You don't need to make multiple passes to use Foucault in production. You hardly need to use it at all in the early stages of figuring, where a quick look with a Ronchi grating and grazing illumination to check smoothness with the KE tells you all you need to know. It's big advantage to my mind is that you can read, in one test, the numerical situation on the mirror, and from this you can derive a figuring approach based on deterministic work to apply. (It shares this with IF, and when done correctly tracks IF quite well, as has been demonstrated elsewhere.) And you can do this to get numerical results to well beyond whatever your desired tolerance is, as well as working on the mirror so as to maximize its fit into the M-L envelope, if that's your working preference.
Experience though counts for a lot, as well as the previously discussed issues of tester construction and alignment. With Foucault when you're done you have numbers (such as they are, but in production you always work your numbers to below the level you know from experience produces a high performing mirror). You also have a test result that didn't depend on any additional optics, and is relatively insensitive to mask alignment, actual ROC, and alignment errors - in other words, if you do it even reasonably right, you get a result from measuring and typing in a few numbers that you can rely on.
I can name several uncompromising opticians that could use any test method they'd like, but use Foucault for mirror analysis (plus additional tests to ensure figure of revolution and freedom from astig, and smoothness). Beyond that though it all gets contentious, and I'm not going there.
YMMV - I think the overall care taken in production and testing far outweighs any variation in testing preferences, when it comes to producing high quality optics.
Best,
Mark
Edited by mark cowan (10/24/09 10:11 PM)
|
mark cowan
Vendor (Obsidian Optics)
Reged: 06/03/05
Posts: 2157
Loc: salem, OR
|
|
Quote:
Mel Bartel even suggest to use your eyes instead of some lengthy math equasion to determine fine parabaloids by way of visually comparing a computer Ronchigram to your own mirrors ronchigram. He contends that your visual comparison is very accurate and will result in an accurate parabaloidal figure.
You still need the computer to print out the Ronchigrams to compare to. And you need to be sure you have the defocus from paraxial focus exactly right if you want to match them properly. But yes, it can be done. In any case, you're using your eyes to do Foucault, and not doing any math besides (use FigureXP or SixTests to reduce the data, or any homebrew method you've got).
Mel also has said: "It's time to take a Bath!" Really, you should be looking at Bath IF if you want to get surface maps of the entire mirror with numbers attached - cheap to build, relatively easy to use (Thousands have done it! Well maybe not thousands - hundreds! Well, maybe not hundreds - dozens have done it!) Reliability may vary, but both the Interferometry Yahoo group and the IF Wiki have gobs more information. Time marches on...
Best,
Mark
Edited by mark cowan (10/24/09 10:09 PM)
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
" that took me about 10 minutes ..."
*************************
Swapping mirrors takes me about 90 seconds, generally. I have to admit, when going from a 6 to a 20 takes about a 10 minute interval. Same mirror, usually about 15 seconds to place on the stand, and sit down for a look.
=========================================
It's been stated to me in a private PM by someone that experience doesn't mean a thing when testing with KE.
If I had someone in my shop who had lacked experience, and didn't know what to look for, you can bet I'd be there to point out corrective measures to the crude methodology I'd inevitably see. This might be the issue with KE???? Lots may not have the 'stick time"to be proficient?
I have to say, such a statement is ludicrous! Everyone else I know of in the optical business or who may be of amateur status highly recommends experience above all else as the way to build proficiency in testing.
I can add, with further thought, if they can't handle KE with some measure of decent success, how the heck could they handle AC, null with a reference lens, or interferometry with the additional complexity? KE is about as simple as one can get. An excellent starting point.
Also, the ones with the rep for good mirrors generally use KE through a significant part of the manufacturing process, or entirely so. (manufacturers or amateurs) There is likely a good reason for this.
I think from seeing what's written here, that ones who would know, show that they can read generally down to .003" or so with KE. That's far better than the .010 that's bantered around, and found commonly quoted in the literature. This was my primary thought and motivation. Though sources might be well accepted, they may be dead wrong. The myths die hard! 
--------------------
So many projects, so little time!
|
kfrederick
professor emeritus
Reged: 02/01/08
Posts: 660
|
|
Post deleted by kfrederick
|
Wes James
Post Laureate
Reged: 04/12/06
Posts: 3452
|
|
It's like just about anything else, Mark- practice makes perfect.
Or at least gets you closer! 
Wes
|
Dan McConaughy
member
Reged: 11/11/06
Posts: 81
Loc: LA
|
|
Isn't an interferogram of a mirror or system with a fringe analysis more objective than taking knife edge readings and averaging them?
Given the level of sophistication of at least some mirror makers, I'm more comfortable with the fringe analysis.
Are there ways to 'cheat' with the setup of the interferometric analysis (aside from fraudulent reports)?
Dan McConaughy
--------------------
16" F/12 R-C with optics figured by Paul Jones mounted on Byers Research Grade Series II mount with Gemini GOTO housed in 11.5' Sirius Observatory.
|
kfrederick
professor emeritus
Reged: 02/01/08
Posts: 660
|
|
A ronchi can show alot . post a ronchi pic and people will see all kinds of things
|
kfrederick
professor emeritus
Reged: 02/01/08
Posts: 660
|
|
big thin mirror flexing
|
kfrederick
professor emeritus
Reged: 02/01/08
Posts: 660
|
|
hand held camera. ronchi over a light bulb .3/8 hole
|
mark cowan
Vendor (Obsidian Optics)
Reged: 06/03/05
Posts: 2157
Loc: salem, OR
|
|
Hi Dan,
Quote:
Are there ways to 'cheat' with the setup of the interferometric analysis (aside from fraudulent reports)?
Hmm, maybe. 
Even if you're trying to do it right (nontrivial), stepping outside of the comfort zone (say by testing thin mirrors and supporting them vertically on a sling) can introduce enough test stand error to swamp the mirror's errors. And that's before you do the analysis of compound errors to set some sort of bounds on the overall accuracy. If you take multiple IFs with a mirror rotated systematically, reduce each one, then average the results properly, test stand errors can be eliminated. But a 10 minute test stretches to hours that way.
If you're making mirrors, any test regime is there to ensure you meet your quality targets. And those mirrors go on to be examined critically in the field, so if the quality target was met and the mirrors perform, the actual testing regime is moot.
That said, I'm building a Shack-cube interferometer just to get full surface analysis on large mirrors down to f/2.5. But it's the last step in a long chain of metrics used for fabrication control, and only to validate production control. But granted, if you accept the test methodology, having the full surface metric showing errors on the glass can be considered "more objective" than just knowing that it passed muster with the optician's usual care in fabrication.
Best,
Mark
|
gregj888
sage
Reged: 03/26/06
Posts: 301
|
|
Mark,
I have a Twyman-Green using a cube beam splitter set up with plane mirrors and a collimated (enlarged) beam to look at my deformable mirror. It would also work with flats, though they would have to be small (50mm).
By removing the collimator and changing the reference surface to a sphere you get a spherical wave front Twyman-Green...
http://www.atmsite.org/contrib/Rowe/bath/rtmc04/img5.html
Might be easier than the shack cube...
BTW, for those that haven't seen David Rowe's presentation, it's worth a look...
http://www.atmsite.org/contrib/Rowe/bath/rtmc04/rtmc04.html
Very interesting topic BTW...
Greg
|
mark cowan
Vendor (Obsidian Optics)
Reged: 06/03/05
Posts: 2157
Loc: salem, OR
|
|
From what I understand the Shack-cube isn't terribly difficult, being a beam-splitter and plano-convex lens (reference on the convex) that focuses just outside the back face of the beam-splitter. Once tested and successfully glued up it's robust, resistant to dust, and compact. The GRIN lens almost touches the back of the beam-splitter, but it's removable for dust control. From what I've seen the Bath works fine but I'd rather not have the inherent astig for what I'm trying to do.
But I'll take a look at Dave's presentation, I don't recall seeing this before. 
Best,
Mark
Edited by mark cowan (10/26/09 03:20 AM)
|
Mark Harry
Post Laureate
Reged: 09/05/05
Posts: 3121
Loc: Northeast
|
|
On another aspect, was mentioned something about an 'objective test'.....
*************************************
Well, a test is a test. The only thing that can be "objective" is the operator. I think all of the tests can be messed up one way or the other; deliberately or through a lack of proficiency.
M.
--------------------
So many projects, so little time!
|
John Carruthers
Skiprat
Reged: 02/02/07
Posts: 2269
Loc: Kent, UK
|
|
anyone have any thoughts on using a webcam instead of the human eye for evaluating grey values in a Coude mask? I would have thought using 255 shades could give a numeric value for better matching?
--------------------
Jc
ATM 10" F6.1, 1/25th wave spec (max wavefront error +/- 1/12.6 in zone 4 of 6, sodium light )
6" F7 spec
127mm F9.4 Refractor
10 x 50 bin
ETX80 (finder)
Canon 20D
PST
DSI 1
and a curious mind
|
gregj888
sage
Reged: 03/26/06
Posts: 301
|
|
John,
Robo Foucault does just that and seems to track IF nicely.
Mark, I have a Bath, the one shown is the Twyman-Green. Replace the lens with a spherical mirror on the other arm of the beam splitter. Concave sphere being easier to test.
Greg
|
rwiederrich
Goldfinger
Reged: 11/17/05
Posts: 8307
Loc: Bremerton Washington
|
|
Quote:
A ronchi can show alot . post a ronchi pic and people will see all kinds of things
I seeeee a small rabbit on an elephant..running through the clouds.... 
Not really.....
The center needs more pushing for sure.....Huge mirror
Rob
--------------------
www.goldmtobservingcenter.com
A great place for amateur astronomers, and ATM's to come and enjoy their hobby.
http://tech.groups.yahoo.com/group/HomemadeRefractorTelescopes/ My homemade refractor group.
www.vimeo.com/6014031
|
mark cowan
Vendor (Obsidian Optics)
Reged: 06/03/05
Posts: 2157
Loc: salem, OR
|
|
Greg,
I see what you're saying. I'll be making an f/2 reference sphere for testing the IF out with, I'll send that out for independent looks and get some traceability. Small plano-convex lens as sourced by, say CVI can be acquired coated/uncoated and one could build two cubes, one for coated mirrors, one for not. The quality is likely to be very good near the center for small lenses with lambda/10 specs, I'll be finding that out. 
Using the "outboard" reference sphere would provide an optional method of making it work.
John,
Quote:
anyone have any thoughts on using a webcam instead of the human eye for evaluating grey values in a Coude mask? I would have thought using 255 shades could give a numeric value for better matching?
I've been working up a high-res digital (unmasked) Foucault tester (not the way RoboFoucault does it, BTW!) that in manual digitization tests showed something <1/40th wave PVW agreement with careful manual measurements. It will measure hundreds of points across the mirror, if desired, stepping equally through the caustic and then back-computing the "zonal" radius. 
Best,
Mark
|
gregj888
sage
Reged: 03/26/06
Posts: 301
|
|
Mark,
The reference sphere for calibration sounds like a great idea.
I'm still working on null optics for the 20" F/3.5. I have a lens for a Ross null but it's only about .3 wave (not nearly enough) though a reasonable check on the KE.
I've done a lot of KE tests but in this range I don't expect to get much better than 1/4 wave (though hope springs eternal for far more).
I have the null setup saved that Mike Jones did for Mark Harry a while back. That might work and I may have lenses on hand for an Offner null setup but need to check them in Oslo.
It might be fun to bring the stuff down to Salem in a few weeks and "play" for a few hours if you are up for that?
Greg
|
|
12 registered and 11 anonymous users are browsing this forum.
Moderator: Don W, Mike I. Jones
Print Thread
|
Forum Permissions
You cannot start new topics
You cannot reply to topics
HTML is disabled
UBBCode is enabled
|
Thread views: 636
|
|
|
|
|
|
|
Previous
Index
Next
Threaded
Edit
Reply
Quote
[Re: 