Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Foucault Unmasked - New Foucault Test Software

  • Please log in to reply
324 replies to this topic

#301 mark cowan

mark cowan

    Vendor (Veritas Optics)

  • *****
  • Vendors
  • Posts: 9,582
  • Joined: 03 Jun 2005
  • Loc: salem, OR

Posted 09 May 2020 - 07:45 PM

So it doesn't use the conic it calculates for anything?  OK, I see what you mean now. 

 

Because of the waves of undercorrection  in my example the surface error curves all looked much the same.   Sorry for the misunderstanding.

 

Then if the ROC adjustment simply had +-1.5 mm as does FigureXP, instead of +-0.13 as FU does,  it'd be fine.  That would allow flattening the curve in the portion of the mirror I'm interested in.  Plus of course some way to specify a central obstruction like SixTests.

 

Thanks!  bow.gif


Edited by mark cowan, 09 May 2020 - 07:48 PM.


#302 Dale Eason

Dale Eason

    Surveyor 1

  • *****
  • Posts: 1,574
  • Joined: 24 Nov 2009
  • Loc: Roseville,Mn.

Posted 10 May 2020 - 01:47 AM

Sorry for the side topic split here about James Lerch and FigureXP.  I think figureXP could be modified to help this sofware.  I have just accidentally discovered a link to James Lerch through a You tube comment.  

I'm trying to reach him and hope I can.  I see his you tube itself has not been updated for 10 years but today I saw him post on a truck fixing you tube so I know he is around.

 

I hope the OP does not mind.  I think it could help his software as well.  

 

Dale


  • brave_ulysses, LarsMalmgren and Pinbout like this

#303 ckh

ckh

    Surveyor 1

  • -----
  • topic starter
  • Posts: 1,963
  • Joined: 21 Feb 2015
  • Loc: Arizona

Posted 10 May 2020 - 12:52 PM

I'm working on the hole/obstruction enhancement.  Might take a few days.  


  • brave_ulysses likes this

#304 mark cowan

mark cowan

    Vendor (Veritas Optics)

  • *****
  • Vendors
  • Posts: 9,582
  • Joined: 03 Jun 2005
  • Loc: salem, OR

Posted 10 May 2020 - 03:15 PM

Thanks!  

 

Briefly continuing Dale's excursion OT, FigureXP was Figure45 by Dave Rowe in a Windows wrapper.  At that time either Dave or James added the ML display to it as it doesn't exist in Figure45 (circa 2000).  Figure45 is DOS and won't run on my W10 machine at all, but there's no source associated with the zip file, attached here:

 

Attached File  fig45.zip   59.88KB   10 downloads

 

There is a Doc file describing the use, testing, and contributors:

 

Attached File  Figure.doc   41KB   5 downloads

 

Might be of some interest to somebody. 


  • LarsMalmgren and ckh like this

#305 Leonardo Priami

Leonardo Priami

    Vendor - Telescopi Italiani

  • -----
  • Vendors
  • Posts: 71
  • Joined: 25 Jan 2007
  • Loc: Italy

Posted 11 May 2020 - 05:17 AM

I'm working on the hole/obstruction enhancement.  Might take a few days.  

Thanks!!



#306 tommm

tommm

    Viking 1

  • *****
  • Posts: 688
  • Joined: 16 Dec 2015

Posted 11 May 2020 - 10:49 AM

 

 

I'd also like to see (more work) a ML envelope display with the ability both to change the conic and the ROC.  I use FigXP for this and almost never the surface plot as the information ML produces tells you exactly how to correct the mirror based on zone-zone variances.  

 

I think I sent Carl a copy of a spreadsheet I made back around 2002 with the ML mathematics in it (also Foucault based on Holleran's area centers).  The ML math is pretty simple. This is the write up on it that I put in the spreadsheet, with the S&T reference:

 

A Graphical Approach to the Foucault Test
Millies-Lacroix
Sky and Telescope, February 1976, pgs 127-129

 

In Lunettes et Telescopes, Paris, 1935, A. Danjon and A. Couder recommend that a mirror be considered finished if it satisfies this double criterion:
1) The geometric image of least aberation in the plane of focus should not exceed the size of the theoretical Airy disk.
2) The maximum wavefront error must not exceed a quarter of a wavelength of light, and the defects should be much less than this over most of the surface.

 

On page 141 in How to Make a Telescope, Jean Texereau gives the residual longitudinal error at the center of curvature as:
Lc = m - c - r2/R   where m is the measured value, c is a constant subtracted from the measured values for all zones, r is the mean radius of the zone, and R is the radius of curvature of the mirror.

 

He also states that the longitudinal error at focus, Lf, is one fourth the above error, or:  Lf = Lc/4, and shows that from the geometry the transverse aberration is to a good approximation:

 

lf = Lf * r/f = (r/4f)(m - c - r2/R)  from the above equations.

 

Millies-Lacroix imposes the first above condition, lf < q, on this equation, where q is the radius of the Airy disk given by q = 1.22*l*f/D,  f is the focal length, l is the wavelength of light (taken as 5600 A), and D is the optical diameter of the mirror. This gives:

 

r/2R(|m - c - r2/R|) < q, where |m - c - r2/R| denotes the absolute value of the difference as some values will be negative, some positive, depending on the slope of a particular zone relative to that for a perfect parabola. 

 

Solving for m - c - r2/R:

 

m - c -r2/R < + 2Rq/r for m - c - r2/R > 0, and m - c - r2/R > - 2Rq/r for m - c - r2/R < 0

 

Millies-Lacroix then plots the values m - c - r2/R and +/- 2Rq/r.  The values  m - c - r2/R must fall within the envelope given by +/- 2Rq/r to satisfy the above inequality.  If the mirror is a perfect parabola the aberation is r2/R, so m - c - r2/R will be zero for all measured values, and thus plot as a straight line on the x axis.

 

It's easy to create the plot in a spreadsheet, but a lot more work I expect for Carl to add it to his code.


Edited by tommm, 11 May 2020 - 10:51 AM.

  • John Lightholder and ckh like this

#307 ckh

ckh

    Surveyor 1

  • -----
  • topic starter
  • Posts: 1,963
  • Joined: 21 Feb 2015
  • Loc: Arizona

Posted 11 May 2020 - 11:29 AM

Thanks for the formulas.  I'll try to look into that later.

 

Carl



#308 ckh

ckh

    Surveyor 1

  • -----
  • topic starter
  • Posts: 1,963
  • Joined: 21 Feb 2015
  • Loc: Arizona

Posted 14 May 2020 - 05:57 PM

There is a new hole/obstruction feature in FU that appears (to me) to work. You can enter the diameter of the hole or obstruction under the Diameter field in the main window.

 

I'd like a volunteer to try it out. If you are interested please send me a private mail.

 

 

Note:

 

I've been checking results with Sixtests (written by Jim Burrows). Both programs agree very closely on Strehl Ratio, RMS, best fit ROC. 

 

However, in SixTests on the Plot page there is a option for "Parabola" and another for "Conic". I assume Conic shows the "best fit conic constant'.  Whereas FU gets -0.789 for best conic (in one of my test sets in FU with a substantial hole), SixTests gets -0.853. 

 

FU gives a slightly better Strehl and RMS using the FU best conic value than using the Sixtests best conic value. Also the programs differ in the best conic without and with the obstruction, but otherwise agree. 

 

FU without hole 0.788, with hole 0.789.

SixTests without hole 0.779, with hole -0.853.

 

I continue to look into this, but best conic is the only parameter that is an issue.

 

Carl


Edited by ckh, 14 May 2020 - 06:21 PM.


#309 mark cowan

mark cowan

    Vendor (Veritas Optics)

  • *****
  • Vendors
  • Posts: 9,582
  • Joined: 03 Jun 2005
  • Loc: salem, OR

Posted 14 May 2020 - 06:27 PM

In Sixtests "Conic" option calculates the best fit conic constant, "Target" is the same screen but you can choose to set the ROC and/or the conic.  Note that the ROC changes slightly between "Parabola" and "Conic" selection as well (in my test data anyway).


  • ckh likes this

#310 ckh

ckh

    Surveyor 1

  • -----
  • topic starter
  • Posts: 1,963
  • Joined: 21 Feb 2015
  • Loc: Arizona

Posted 15 May 2020 - 12:17 PM

Mark,

 

I'm thinking about the RoC offset slider issue. Do you want to be able to make any chosen place on the surface plot flat (0 slope) by moving the slider?

 

Thanks,

Carl

 

PS: I was unable to PM you.



#311 mark cowan

mark cowan

    Vendor (Veritas Optics)

  • *****
  • Vendors
  • Posts: 9,582
  • Joined: 03 Jun 2005
  • Loc: salem, OR

Posted 15 May 2020 - 03:20 PM

PMs should work now.  :)

 

Yes, the use for the slider is to flatten any part of the surface plot.  Is that what you mean?  A automagic "make this area flat" would be cool though.


Edited by mark cowan, 15 May 2020 - 04:02 PM.


#312 ckh

ckh

    Surveyor 1

  • -----
  • topic starter
  • Posts: 1,963
  • Joined: 21 Feb 2015
  • Loc: Arizona

Posted 15 May 2020 - 07:35 PM

The point of RoC offset is to choose a polishing strategy that will get you to the desired conic with a slightly different RoC from the best fit RoC.

 

A very large change in RoC may be necessary to make given zone flat (match the desire conic). E.g, suppose there is a steep slope defect near the center. If I base the RoC offset slider range on such a bad case, it may be so large that small slider offsets will be too sensitive.

 

Figure XP uses an offset range about twice as large as FU does. Because of that, the slider becomes rather sensitive near the middle, so you are forced to use the arrow keys to move the slider by very small amounts.

 

I'm not keen on that "automagic make it flat" approach because I don't think it would be clear to most users who are used to sliders in other programs to set the RoC offset.

 

I'm thinking of making the slider exponential. Near the center of the range, it would have a small effect on RoC but a much larger effect toward the extremes. 



#313 mark cowan

mark cowan

    Vendor (Veritas Optics)

  • *****
  • Vendors
  • Posts: 9,582
  • Joined: 03 Jun 2005
  • Loc: salem, OR

Posted 15 May 2020 - 10:23 PM

Are you just explaining this in general or talking to me in particular?  I use the ML-envelope graph to determine specific adjustments but I never get that from the surface profile since the former is much more informative about error tolerances in that it presents a weighted error curve.

 

FigXP ROC adjustment is 10x your amount for the example I showed.  It appears to be +-1.5mm regardless of ROC.

 

The slider should be linear across the range.  Allowing the user to select a range would accomplish more than fiddling with the sensitivity.

 

I don't much care for a automatic method I just couldn't figure out what you were asking about.  :shrug:

Thanks for your work, I can try the beta version if you like.



#314 ckh

ckh

    Surveyor 1

  • -----
  • topic starter
  • Posts: 1,963
  • Joined: 21 Feb 2015
  • Loc: Arizona

Posted 16 May 2020 - 09:49 AM

OK, I'll consider that.



#315 chantepierre

chantepierre

    Explorer 1

  • -----
  • Posts: 76
  • Joined: 27 Mar 2020

Posted 22 May 2020 - 03:29 AM

Hello,
I just wanted to notify that Foucault unmasked works fine under Mac OS and Linux by using wine.

Maybe you could add this to the Readme on GitHub while not officially supporting those platforms, it happens to work.
  • LarsMalmgren and ckh like this

#316 hamishbarker

hamishbarker

    Messenger

  • -----
  • Posts: 413
  • Joined: 26 Apr 2017

Posted 23 May 2020 - 05:29 AM

Hi,

 

The installer for Foucault Unmasked now resides on GitHub.  To access the installer and readme file goto https://github.com/c...ucault-Unmasked.

 

You will see two files listed:

 

README.md

Setup_Foucault_Unmasked.msi

 

Click the green button that says "Clone or Download" and then click "Download ZIP".

 

Note that the Profile Mode control in the upper left corner of the application main window is not implemented. It has no effect.

 

Carl

hi carl,

 

late to the party, but thanks so much for developing this! I will try it out on my (possibly) finished 22" f4.6 mirror. per my measurements using figure xp it's pretty close, about 1/6 wave wavefront P/V. On star test it's not as sharp a snap focus as I would hope, but I don't have confidence in my foucault testing to have confidence in what the biggest error is and therefore decide how to try to figure it out, so to speak.

 

Hopefully FU will give me more information which might be useful in working out next steps.

 

Cheers,

Hamish



#317 black_rig

black_rig

    Lift Off

  • -----
  • Posts: 14
  • Joined: 24 Jun 2016

Posted 25 May 2020 - 07:03 AM

I have been using FU and want to report it works very well and makes the process of testing mirrors a breeze.

 

I use a Foucault meter on a We Macro rail, so it takes automatic pictures at a regular interval without human intervention.

It its easy with this setup to take 60 or 100 images in a mater of minutes, all perfectly spaced, at any given interval along the axis.

 

Then I batch process the pictures (rotate, crop, etc...) with this script, available on Github. Then feed the images to FU.

 

Now the macro rail, the optical axis of the camera and the optical axis of the mirror all need to be aligned. If somebody has an easy method to do it I am interested.

Also, I tend to think a slight miss-alignment between those 3 axis does not degrade the Foucault accuracy too much, but would be interested if somebody can point me to some writing on the subject.

 

 

IMG-2698.jpg?resize=768%2C576&ssl=1

 

Edited: it is "We Macro Rail" not "Wee Macro Rail", as I incorrectly first wrote.


Edited by black_rig, 26 May 2020 - 06:43 AM.

  • brave_ulysses and ckh like this

#318 mark cowan

mark cowan

    Vendor (Veritas Optics)

  • *****
  • Vendors
  • Posts: 9,582
  • Joined: 03 Jun 2005
  • Loc: salem, OR

Posted 25 May 2020 - 04:00 PM

It's not easy to dial in but your alignment problems can be mostly solved with a clamped (adjustable) laser pointer attached to the platform.  Once you have determined perfect alignment the laser is set to point at the exact center of the mirror (within a mm or so, verified by imaging the spot) and clamped in place.  You can find this alignment because it returns the smallest value of the mirror's conic constant in FU, and it's tested by measuring with FU series to inrease when the test rig axis is rotated slightly left or right (also a few mm or less displacement of the laser). 

 

Future alignment is easy since you just adjust the platform to point the laser at the center of the mirror.  Note this also applies to up/down centration on the mirror's optical axis as well, so you may need some way to fine-tune that.  Compared to this misalignment of the camera itself is trivial as you can center the image of the mirror in the frame easily, and eyeball the optical axis.

All of that only really matters for fast mirrors, but it matters increasingly the faster the mirror is. 



#319 dogbiscuit

dogbiscuit

    Apollo

  • *****
  • Posts: 1,012
  • Joined: 18 Oct 2018
  • Loc: Barksdale

Posted 25 May 2020 - 05:03 PM

I haven't used Foucault Unmasked, but alignment should be much the same as for visual testing.

During alignment do not move the knife left or right on the tester ways, only move the entire tester.  Put knife at mid fore/aft travel on the tester ways. position entire tester to show null somewhere near 70% zone, for most mirrors not all that critical.

Move knife full aft and rotate entire tester to shoe mirror half illuminated half shadow. Move knife full forward and rotate tester by moving the front of the tester left right to show half shadow half illumination, while trying to move rear of tester as little as possible. repeat until  moving knife from full aft to full forward knife shadow covers exactly half of the mirror at both extremes of travel.  Now move the knife to ROC to see mirror profile and fine tune left right knife position moving the knife on the tester ways.  Check that knife movement through full range of nulls from all zones can be see with no adjustment of knife left right position.

 

Harder for me to type it out than it is to do it.  Takes me about 2 minutes.  Once it's done it stays aligned if the tester and mirror stand is not moved.

 

So that adjusts on the vertical axis.  For the lateral horizontal axis I have just relied on having tester source and knife at same level as the mirror's center and that the tester ways are level, and verify that when moving knife through full fore aft range that the image remains at the same vertical position on the knife.  I guess higher precision could be had turning the knife horizontal and adjusting in the same manner as the vertical axis.



#320 mark cowan

mark cowan

    Vendor (Veritas Optics)

  • *****
  • Vendors
  • Posts: 9,582
  • Joined: 03 Jun 2005
  • Loc: salem, OR

Posted 25 May 2020 - 05:35 PM

Unfortunately that's not enough for very fast mirrors, as FU analysis will reveal.  But for most mirrors it works well within the limits of manual testing by eye.  Digital Foucault gets much more accurate results than testing by eye and so requires even more care in setup.

 

Using a laser to sight it in, as close as possible (if not right on) the mirror axis makes it easy enough.



#321 ckh

ckh

    Surveyor 1

  • -----
  • topic starter
  • Posts: 1,963
  • Joined: 21 Feb 2015
  • Loc: Arizona

Posted 25 May 2020 - 06:25 PM

I have been using FU and want to report it works very well and makes the process of testing mirrors a breeze.

 

I use a Foucault meter on a Wee Macro rail, so it takes automatic pictures at a regular interval without human intervention.

It its easy with this setup to take 60 or 100 images in a mater of minutes, all perfectly spaced, at any given interval along the axis.

 

Then I batch process the pictures (rotate, crop, etc...) with this script, available on Github. Then feed the images to FU.

 

Now the macro rail, the optical axis of the camera and the optical axis of the mirror all need to be aligned. If somebody has an easy method to do it I am interested.

Also, I tend to think a slight miss-alignment between those 3 axis does not degrade the Foucault accuracy too much, but would be interested if somebody can point me to some writing on the subject.

 

Hi black_rig,

 

Thanks for the positive feedback.  I too wanted to do the automation of the KE movement and photo snaps, but haven't gotten to it.

 

Congratulations for succeeding!  Perhaps you could tell us about how you automated it in a new thread.

 

Carl



#322 black_rig

black_rig

    Lift Off

  • -----
  • Posts: 14
  • Joined: 24 Jun 2016

Posted 25 May 2020 - 09:03 PM

Hi black_rig,

 

Thanks for the positive feedback.  I too wanted to do the automation of the KE movement and photo snaps, but haven't gotten to it.

 

Congratulations for succeeding!  Perhaps you could tell us about how you automated it in a new thread.

 

Carl

 

The automation does not have much to it: I just screwed the KE on a We Macro Rail, out of the box, no modification.

Then WE Macro Rail has an app which you install on your laptop, or even phone.

 

Then connect your laptop via USB to the rail, so the laptop can control the rail's stepper motor, and connect your camera to the rail, so the laptop can also control the camera, via the rail.

 

From the app on the laptop, choose an interval in Micron for your pictures. The laptop makes the rail with the KE move, then commands the camera to take a picture, move to the next position, etc... This does a much better job than I can do manually. And it is easy to over sample (hundreds of pictures, which FU does not like too much).

 

If you replace the KE by a Ronchi grating, the app lets you move the rail continuously, resulting in some mesmerizing movies, with the lines in the Ronchi dancing and connecting. 

 

Last I use the We Macro Rail, which is cheap and accurate enough for me (I am working on a f/4.5) but there are other equivalent or maybe better products on the market, you can check the Macro Photo forums to get an idea.

 

Thanks Mark and Dogbiscuit for the tips on alignment. I need to put this into practice.

 

Gee


Edited by black_rig, 26 May 2020 - 06:44 AM.

  • coinboy1 and ckh like this

#323 hamishbarker

hamishbarker

    Messenger

  • -----
  • Posts: 413
  • Joined: 26 Apr 2017

Posted 27 May 2020 - 08:46 PM

My mirror is a bit rough so for the inner zones the software indicates null all over the place. I gave up and went back to the coude mask. Would be good to be able to select or unselect individual images. Also it Seems the only way to remove a loaded image is to clear the whole loaded set and reload.

For mirrors with good smoothness and automated capture, should be great, so many data points.

#324 mark cowan

mark cowan

    Vendor (Veritas Optics)

  • *****
  • Vendors
  • Posts: 9,582
  • Joined: 03 Jun 2005
  • Loc: salem, OR

Posted 27 May 2020 - 10:25 PM

Just select the image and hit delete.  No it's not obvious...



#325 coinboy1

coinboy1

    Apollo

  • *****
  • Posts: 1,355
  • Joined: 03 May 2011
  • Loc: Tulsa, OK

Posted 27 May 2020 - 10:35 PM

The WeMacro rail is intriguing especially with the android control.


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics