PixInsight Forum (historical)
Software Development => New Scripts and Modules => Topic started by: mschuster on 2015 August 10 08:16:15
-
WavefrontEstimator Version 1.19: Script for testing the on-axis optical quality of telescopes.
WavefrontEstimator estimates the on-axis optical wavefront of a telescope from long-exposure defocused stellar images. Exposure differences between an intra-focal image and an inverted (rotated by 180°) extra-focal image of the same bright star reflect local changes in the curvature of the wavefront. WavefrontEstimator measures the defocused image exposure differences, reconstructs the wavefront, and provides a diagnosis of wavefront aberrations. WavefrontEstimator relies on long combined exposures of at least 100 seconds and 100,000 e- to average out the effects of atmospheric turbulence and provide sufficient signal-to-noise ratio. The telescope must be in thermal equilibrium.
The estimated wavefront defines a map of optical phase on the aperture plane, normalized to zero mean phase. Non-zero estimates correspond to wavefront aberrations and lower image quality.
WavefrontEstimator resolves wavefront deformations or corrugations at a maximum spatial frequency that depends on the degree of defocus, the aperture diameter and focal length of the telescope, and the observation wavelength. The maximum corrugation spatial frequency measured is denoted the corrugation resolution. Typical values range from 10 to 15 cycles per aperture diameter.
Call for help: If you would like to help test this script, please let me know!
For more information and examples, please see the script's documentation.
Thanks,
Mike Schuster
-
yes i would be interested in trying this - however, will it work with a reflector that has a secondary obstruction? the defocused stars will look like donuts rather than the pure circular diffraction pattern (is there a name for that?)
thanks
rob
-
Hi Rob,
Yes, the script should work with a secondary obstruction as long as it is circular and centrally located (i.e. the usual case). The documentation PDF has examples of these ring shaped Fresnel diffractions in section 6.
What is the aperture diameter, focal length, and obstruction diameter (an estimate)? I can run more tests given this info.
Thanks,
Mike
-
Hi Mike,
Thank you so much on behalf of me and all PixInsight users, and congratulations on an excellent design, programming and documentation work.
-
Hi Rob,
Yes, the script should work with a secondary obstruction as long as it is circular and centrally located (i.e. the usual case). The documentation PDF has examples of these ring shaped Fresnel diffractions in section 6.
What is the aperture diameter, focal length, and obstruction diameter (an estimate)? I can run more tests given this info.
Thanks,
Mike
Hi Mike, the OTA is a Ceravolo 300 in f/9 configuration, so 2700mm focal length, 300mm aperture and 165mm obstruction. i would probably wait until the moon is back in the evening sky before getting the test images, so it will be a little while before i can experiment with this.
thanks,
rob
-
Hi Rob,
Thanks. Moonlight is OK. Please make sure the telescope is in thermal equilibrium. This is very important. I forgot to mention this requirement in the documentation. You have an excellent site with good seeing and stable temperatures which will make testing easier.
Assuming PL 16803, I set gain 1.4 e-/DN, pixel size 9 micron, defocus exposure 45,000 e-. With corrugation resolution set to 10 in the exposure estimator tool, required defocus distance is 8.8 mm. Is your focuser capable of moving both inside and outside at least this much from the position of best focus?
Assuming a separate guide scope (i.e. tracking system is not compromised by defocusing like it is on my setup), you should be able to use long exposures. At 100 seconds the estimated target star magnitude is 5.1 with an R, G, or B filter, or 6.3 with an L filter. The script wants ~100,000 e- total or more, so you need at least two intra-focal frames and two extra-focal frames. These magnitudes are estimates, you should double check that the entire defocused image lies within the linear operating region of the detector. You also will also need a master bias for calibration. No other processing is necessary (i.e. no flat fielding and no other pre-processing).
I will run synthetic tests.
Thanks,
Mike
-
Hi Mike - yes, PL 16803 is correct. we have an optec focuser which has a pretty long throw so i think we can easily reach 8.8mm inside and outside of focus. i will check. i don't remember what the step size is.
we are using an external guidescope so we can do long exposures while defocused.
normally we use RBI pre flash - do you think this should be turned off for the purposes of this script?
thanks,
rob
-
Ok good.
RBI pre flash should be fine. I say leave it ON. As long as the defocused image is well-exposed (i.e. ~50% full-well) the shot noise in the image will easily overwhelm RBI pre flash leakage noise (if any).
Thanks,
Mike
-
Wow,
Quite a tool.
-
Mike,
Hell of a script. I will test with my secondary system (Takahashi e130D and STT8300) next time I am out under the stars, hopefully around Sept 12. For your test purposes, would it also be helpful for me to test with my CDK12.5 and F16M?
Best,
Jim
-
Mike,
I am assuming that you need to be able to get an adequately defocused image on both sides of focus. I ask because with my e130D/STT8300 combo, I have very little inside focus to work with.
Thanks,
Jim
-
Hi Jim,
Yes, you will need to image on both sides of focus. The pics below show example parameters for both setups with 5 second L filter exposures. You may not be able to get +/- 2.3mm on the Tak. If not, maybe +/- 0.75 mm is possible, with corrugation resolution set to 8? Tests on the CDK would be very helpful also.
When you capture a frame, if you can run PI while you are capturing, you might want to first double check that it is not over- or under-exposed, then do a bias-subtract using ImageCalibration, and finally run the Exposure measurement tool. It will measure defocus diameter and exposure. Use this to double check your focus position and exposure time. It may take a couple iterations. Take care to get inside and outside focus diameters as equal as possible (within a few %). This all can be a bit of a pain, especially the first time or if focusing manually.
You will need at least 100 seconds total on each side of focus. I like to capture more, like 300 seconds each side. It is critical to capture enough to average out seeing effects. Otherwise the script will confuse image intensity variations due to seeing with optical aberrations and give garbage as a result.
Thanks,
Mike
Takahashi:
(http://mschuster.zenfolio.com/img/s5/v133/p1440105477.png)
(http://mschuster.zenfolio.com/img/s3/v45/p1440106741.png)
Planewave:
(http://mschuster.zenfolio.com/img/s10/v99/p1440105529.png)
(http://mschuster.zenfolio.com/img/s8/v15/p1440105490.png)
-
i have to remember to work on this tonight. the weather at SRO has been really bad for a month due to a huge fire burning just to the south...
rob
edit: phooey, looks like it's going to be cloudy tonight...
-
Hi Mike
Have you taken 1.15 down from your Dropbox? I tested with 1.12 last night and got very good results. Was struggling a little bit at the beginning since I fed the script uncalibrated frames to start with and it would not converge. Fixed that with a resized version of my master bias and it all worked fine.
I'm testing an (very old 1991) Astro-Physics 152mm f/7.5 refractor working at f/8.5 or so with a Moravian G4-16000.
Roberto
-
Hi Roberto,
New version 1.15 is now available in Dropbox. Testing continues for a few more days, and then it will be released as an official PI update.
Thank you so much for your post, I am really glad to hear the script worked OK!
Mike
-
Dear Mike,
This is a fantastic script - I tried it on Saturday. It worked really well (other than the result - some collimation to do... ).
One question: my focuser is not adjusted well at the moment and the focus positions are not repeatable. I captured images about 12mm either side of the focus position, but there is an error in that of about 0.1mm.
How sensitive are the results to the exact locations of the images?
Thank you very much for a wonderful tool.
Colin
-
Thank you so much Colin! I am very glad that it worked!
Focus distance on either side should be equal to within 5%. The script automatically adapts to differing focus distances. So you are well within the allowed range.
Thanks,
Mike
-
Dear Mike.
First off, congrats on your script. I have two questions:
1) I am planning to use your script to create a PSF that models my optical train and feed that model to Pixinsight for deconvolution purposes. How do you think I could do this?
2) I guess the 180º filp your script does it automatically. I just have to provide extra and intra focal images. How do you get 5% max difference in focus between extra and intra? (Maybe measuring PSF external diameter?)
Thank you in advance,
Sam
-
Hi Sam,
1) The script’s PSF estimate is likely too optimistic for deconvolution purposes. The estimate characterizes “on-axis optics only” and does not account for several aspects that typically increase Strehl diameter on target captures such as atmospheric turbulence, focus drift, sensor tilt, off-axis aberrations, frame registration interpolation, and light scatter.
2) The script’s Exposure tab provides helpful tools. First estimate a target defocus diameter, then repeatedly adjust focus and measure the result until defocus diameter is within tolerance. The script does the 180º flipping.
Thanks,
Mike
(http://mschuster.zenfolio.com/img/s9/v18/p1672115157.png)
-
Ok, Mike.
Then... how would you recommend to me to do a PSF estimation for deconvolution purposes in Pix Insight? I have an artificial star and I aim mainly to high resolution lunar/planetary work.
Let me know, please.
Thank you in advance,
Sam
-
Sam,
The DynamicPSF process. Expose a star field in typical conditions, then use DynamicPSF to export a PSF given a selection of stars. See DynamicPSF documentation, Juan's posts in this thread (http://pixinsight.com/forum/index.php?topic=6736.0), and other forum posts and resource tutorials.
Thanks,
Mike
-
Mike,
this is a excellent script. I have tried it last week on my TSA102 and it worked perfect. Thank you very much.
Greetings
Tobias
-
Thank you Tobias! It is really great to see your results presented so nicely! I am also pleased that the script worked with a flattener/reducer, I did not do such a test.
Best regards,
Mike
-
hello Mike,
congrats, just stumbled upon your script - amazing, this is a variety of Roddier test popular in France, they work a lot on it.
I tried this years ago, but couldn't come to reliable results with the software available at that time.
Switched to the use of a Twyman-Green interferometer as I do optical testing regularly now.
As soon as I have some time I'll work through your script and then we can compare the results to interferometry,
if you didn't do that already.
Or maybe you are even able to use one of the real star tests in my blog with your script and compare the results.
http://interferometrie.blogspot.co.at/
There are many entries that have both star test and measurement.
respect and best regards,
Tommy
-
Thank you Tommy,
I am reading your blog, excellent work, a really great resource for anyone interested in optical testing!
The script provides an estimate whose accuracy is ultimately limited by atmospheric turbulence and algorithmic approximations. Long combined exposures help to average out atmospheric turbulence, the plot below shows an example of variability in estimated Strehl ratio for different total exposure times with my FSQ-106EDX in metropolitan observing conditions. Long exposures (at least 100 seconds in both intra and extra-focal stacks) are required for reasonable estimates.
(http://mschuster.zenfolio.com/img/s9/v96/p1935665985.png)
In each column, the white line corresponds to the median of the Strehl ratio estimates, the bottom and top of the box correspond to the 25% and 75% quantiles, and the lower and upper fence correspond to the minimum and maximum estimates.
The script is also limited by algorithmic approximations. On tests with known aberrations (and no atmospheric turbulence) generated by numerical integration of the Rayleigh-Sommerfeld diffraction integral, relative accuracy in wavefront RMS error ranged from 5% to 20% depending on the optic and set of aberrations simulated. For more information, see sections 1.9, 5, and 6 in the script's documentation. For more accurate results, interferometric testing is necessary. I have not done such testing, anything you could do for comparison would be very helpful!
The script requires bias-subtracted frames as input (as well as detector specifications), so unfortunately I can't use images clipped from your blog.
Thanks,
Mike
-
Thanks for this incredibly useful tool Mike.
I ran it against my FSQ-85 and the results are revealing. Up until now, Takahashi have been claiming my problem is down to tilt in the optical train. This gives me some useful results to feed back.
Andrew
-
Thank you Andrew,
FYI: my first new FSQ-106 arrived decollimated, fairly good on-axis, but bad off-axis. I sent Tak a frame, Tak asked me to send it back, I did, guessing it would take a trip to Japan and a while to fix. Two days later a new one arrived unannounced. Very good service.
Second one was vastly better, but a small tilt was evident across the frame made easily visible with a Bahtinov mask, and also consistently measureable in SubframeSelector's FWHM plots. I installed a hardware detilting device in the optical train to solve this problem. I have never been able to figure out the cause of the tilt, could be the detector itself or maybe a cumulative tilt from all the various parts.
Mike
-
WavefrontEstimator Version 1.19 soon to be released as an update.
• Fixed a bug in the Bayer RGB detector support code, the script now supports other non-Bayer CFA type detectors also.
• Supports a wider range of defocused image diameters, from 32 to 320 pixels.
• Tested on a Nikon D800 camera with a Zeiss f/2 lens with expected relative error in the estimated wavefront in the 10 to 20% range.
• Includes several other usability improvements.
Thanks,
Mike