Affiliation
American Association of Variable Star Observers (AAVSO)
Thu, 01/30/2014 - 22:31
Hi,
I find it difficult to schedule twilight flats, more so as I use the same telescope for visual observation.
I wonder if anyone has experience with flatfield panels?
In particular the electroluminescence Flatfield Panels adverised by Gerd Neumann:
I use a 28cm SCT (C11) and a 106mm Tak refractor.
Any suggestions appreciated.
John
This is what I am using:
http://www.spike-a.com/flatfielders/
Mounted on the wall directly in front of my c11 in the park position. Seems to work fine.
I have been using a spike-a-flat for doing pictures and am just starting to do photometry. I performed the uniformity test described in App D of this manual for my spike-a-flat
https://www.aavso.org/sites/default/files/publications_files/AAVSO_DSLR…
With my setup I find a of 2.5%. Have you performed a test like this on your spike-a-flat?
Regards,
Cliff Kotnik
Hi John,
I would not specifically speak about those panels but more generally about making flats with diffuser/light source at or at short distance of the pupil of the optics. At this level of the light path what determines the pixel being illuminated is not the position of the rays in the aperture but the direction of them, the angle. Then most other rays emitted by such light source (or diffuser) don't address the imager pixels but are diffused inside the optical tubes. Such lost flux is much stronger than the useful one. The resulting light diffused at lens surfaces, tube walls... eventually reachs the sensor and affects the shape of the flat. The result is very depending of the optical design, I have seen cases where the error was only a couple of percents but others showing several tens of percents. The comparison was made against a flat made using collimated light (like the stars).
This problem is a serious one for us in photometry, I was suspecting it for long time, but a recent remark from Arne pushed me to restart experiments about that. Now I understand the problem but regretfully I have no solution to propose ! I can just say: be carreful, check the flat made at pupil against twilight ones. And, as recommanded by Arne, put the critical stars at/near the center of the field !
Clear Skies !
Roger (PROC)
Roger, You are correct about collimated vs uncollimated light, but there are a couple of things you can do to reduce the affect from a flat field source to make them less than you get with sky flats.
1. Locate the source a few feet from the end of the optical tube but not so far that it subtends less angle than your FOV.
2. Blank off the emitting area so that it is only slightly larger than the diameter of the OTA. For example, if like me, you have a 10" scope and a 4", and the panel is sized for the 10" you want to restrict the emitting radius when used with the 4".
That reduces the angles of rays into the tube. You actually end up with light entering the OTA over a smaller range of angles than with sky flats because the sky is a radiator over essentially a 180 degree angle and is not collimated like light from stars.
The most important things are that the light source be really flat, and have an appropriate continuous flux over the spectrum of interest. I am a bit concerned about sufficient illumination at the extreme ends of the spectrum, Ic in my case since I don't have a U filter, but I will soon find out since I just ordered one of these gizmos. I will also find out if I need one of the neutral density filters they mention. It wouldn't be good for an automated observatory if the intensity varies over the spectrum sufficiently that you need a neutral density filter for some passbands and not for others.
By the way, there is even Question about the need for continuous spectrum flux provided you are not using elemental filters, as long as emissions are present near the peak of each filter bandpass. See attached article.
Brad Walter, WBY
Great article, thanks. But, it would seem that the problem with the sky flat was not related to the lack of collimated light but the unavoidable gradient in sky illumination? In any event, the allure of the panels such as inquired about, for a robotic set up, is they provide a dust cover function. Based on this article, one advantage of these flip flap designs is it would seem to assure perpendicularity. Especially for a mirrored instrument where the pupil is necessarily somewhat removed from the panel, one wonders whether any error from imperfect collimation at the focal plane would outweight the potential advantages of the system and/or be material in the context of other inherent errors. A refractor system would seem more vulnerable but even there one wonders whether a well baffled system would not substantially deminish the effects. I opted for the wall mounted system out of fear for windage that the flap system would involve.
Oh absolutely. I wasn't trying to claim that the major problem with sky flats is the lack of collimation. I was only trying to point out that even sky flats are uncollimated and that this effect should not stop anyone from using flat panels.
I asked Gerd Neumann about the spectrum of his Aurora device and he sent me the attached with the caveat that outside 400-700 nm his test equipment lacked sensitivity. So althought it looks like it has essentially no output in the Ic band, which I expected, I simply will have to try ii through my Ic filter so see if it is useful in Ic. It looks pretty good in B, V and R and potentially has the same problem in U as in Ic. Also attached are a set of Astrodon UBVRI passband curves, for comparison.
I got interested in this when I discovered that a light box I had used with alacrity for several months was about as flat as Little Round Top and the lights had to be completely repositioned to make it acceptably flat (less than 1% variation).
Brad Walter
Hi Brad,
Yes, agree, I also use this solution. It provides a valuable improvement in case of instrument showing a serious problem but some errors remain (at a few percent level at edges). It also needs some optical consideration to determine the right size of the target, large enough to provide all rays normally reaching the sensor and tacking in account the strong defocus effect (as the instrument as to be set at infinity focus).
It is easy to apply to telescopes but not so to shorter focal length lenses.
Use of LEDs is also somewhat difficult as they have a non-smooth pattern. It needs a good additional diffuser to provide a uniform illumination of the target.
Clear Skies !
Roger (PROC)
If used in a light box I assumed they woule reflect of the back wall of the box and go through two diffusers similar to the common types of halogen or incandescent light boxes. In dome flats I assumed they would be on the outside of the scope baffled from the aperture and reflecting off the dome screen. Both ways seem to work.
Brad, if the illumination from LED is indirect yes, agree. But I have seen guys directly illuminating the flat target with LEDs !
Anyway it's difficult to ensure the target is uniformly illuminated. I usually check it with a photometer and I can say it's not straightforward to get it below 1%. Then the Lambertian condition of the target is also an issue (most are not at % level), including under "pupil" condition (for a telescope a couple of meters is in).
CS Roger
Folks,
I am very intrigued by this thread and by the realization of an EL panel that is full aperture for this particular purpose.
As far as I can tell, the issue is not whether it is perfect or not but whether it is a significant improvement over existing non-sky-based practices. While people can express their opinions and doubts, the proof of the pudding is in the eating. When someone puts it to the test and tells us the level the residuals are, we will know the answer.
I have to say that my expectation is that such an apparently-uniformly emissive, broadband surface sounds very promising relative to essentially any scattered light target or small number of illuminants target in the near field.
We shall see!
Doug
Doug,
I'll let everyone know in about a month. I figure it will be about that long before it arrives and I have chance to test it. Mew Mewlon with a 6303 CCD chip camera is a good testing ground since it vignettes rather drastically. I am particularly interested to find out if it is useful in the Ic band. I don't have high expectations, but we'll find out.
Brad Walter
Brad, did you ever get your EL panel in and tested?
Hi Doug,
When I am saying there are issues it is not just opinion or doubt, this comes from experiments and measurement results. I am a guy from the visualization technologies R&D and I think I got some experience in optics experiment and design after many years in the field.
There are problems with flats !
Electroluminescent panels are only one case, all of the techniques being used for flat have problems on various optical aspects, the most critical being the vignetting compensation (that strongly affect our photometry).
Attached are three curves. It relates to a classical 80 mm refractor. The yellow curve is the flat profile on the horizontal axis of the sensor made with a collimated light flux (like the flux of a star, similar to what is shown in the paper distributed by Brad). The blue is the same H profile of a classical flat made with a strong, near Lambertian, transmissive diffuser near the pupil of the refractor (like an electrolum panel). The green curve is also a classical flat but made with a white target limited to the field of view of the refractor and the sensor. The target is surrounded with a black material to eliminate as much as possible the unwanted diffuse light. The target is set at two meters from the refractor pupil. The targets uniformity has been checked with a photometer.
No problem ?
As a visu tech guy I would also mention NO screen: projection, LCD, plasma, CRT... are Lambertian (one of our usual problem) They are all VERY FAR to be Lambertian.
I didn't mention the limited target solution at start as I consider it's not perfect and needs more work.
Clear Skies !
Roger (PROC)
Thanks Roger. the curves are informative. I expected some variation but not as much as between the yellow line and the Blue curve. So It looks like the best thing for the EL screen is to move it away as far as you can within the confines of your set up and still have it cover the FOV and mask off any excell illuminated area with non reflecting black material. In my little observatory that is only half a meter, but it will have to do.
I have found that black matte foam sheet material made for kids arts and crafts (It goes by a bunch of names like funny foam, crazy foam or just foam sheet at Walmart) works very well as a mask. I used it to cover the outside of my light box made of white foam board so that it wouldn't disturb others if used it at a star party and I have used it to mask around flat field reflecting surface. It is very black and it reflects very little light. It is flexible, cuts easily with scissors or a knife, glues well (but certinly not to the EL screen) and does not absorb water. So dew is no problem and it keeps dew out my light box foam board.
Brad Walter
Hello Doug;
I was hoping someone was going to post their raster results, comparing panels. No one has so far. As Doug points out, its not an opinion of which flats are better, there is an objective test that measures how good one's flat field process is. It is the Raster test as described in the paper that Brad posted.
I have attached my raster test from last week. I made at least 2 mistakes. I was trying to make the test short, so as to have uniform transparancy throughout the test. I used 1 second exposures, and then realized that scintillation was probably interferring.
The second, was that I had to use month old flats. I will revise this.
Posted below is the result. The repeatability of 4 raster runs was .014 mag one sigma. Throw in the lack of flatness, brings this to .04 mag one sigma. Looks by eye that 3 sigma is about 0.1 mag. I plan to use 20 second exposures and fresh flats and will repost.
So lets see your Rasters.
Gary
Gary, Are your 5 series 5 different X positions at 5 different Y positions, in otherwords is this a 5x5 matrix, or the same 5 positions repeated 5 times? Do you do this every evening or just just periodically, like transformation coefficients, and whenever you change anything? I guess if you are doing sky flats you would do it every night. I would think dome flats or light box flats would be very repeatable, as long as you hadn't changed anything in the image train, the light source or the relative positioning of the two.
Brad Walter, WBY
Hello Brad
The first plot was a 5x5 raster, ie 25 positions on the chip. I repeated the test 4 times, so there are 100 data points total. I calculated a std dev of the 4 data points at each position. Typical std dev was 0.014 mags.
I do sky flats every clear night. Rarely do the raster test. Mostly do Raster to track down stray light, etc. I don;t change things very often, most frequent change is about 6 months to change filters in the filter wheel.
As for dome flats, light box flats, and ELU panels, been there, done that and have settled with Twilight flats as the best solution. I can use twilight flats for a week or so with no problem, at 4 weeks, they show signs of some dust donut leak thru--ie the dust donuts are not completely eliminated. Sometimes weather forces to use old flats. I the weather is good, I take then everytime I observer.
Gary
Yes targets are a problem. The company in Canada that made the lambertian screens is no longer making them. About the best you can do is use a projector screen if you are doing dome flats. I have tried to find intensity vs. angle specs for them but have had no success. The only specs you seem to get are size and weight and for some of the better ones viewing angle. Some have viewing angles as wide as 160 degrees ( = 80 degrees on each side of the normal) so it seems they would be relatively constant over a small angle of say 15 degrees from the normal . Spectral resonse info is even more scarce.
Brad Walter
As it is already said in this thread, for (ultimate?) scattered light correction in the flatfield frames, one can use stars. Those stars doesn't need to have well defined standard magnitudes, just constant brightness. For such schemes see e.g. Manfroid 1995 (http://adsabs.harvard.edu/abs/1995A%26AS..113..587M) or more recent Ubercalibration data processing scheme (http://arxiv.org/pdf/1111.2058v2.pdf). Both of them include "rastering". I'd also suggest to read Manfroid's paper just for background - well written and besides "complex-looking-but-easy-math", has numerous important points.
It is known, that quite a good substance to use for lamp-illuminated flat field panels is barium sulfate. It's white and it's spectral reflectance is very flat across whole visible (and near-infrared?) spectrum:
From: http://www-kpno.kpno.noao.edu/glaspey/KP/FFLamps.html
---------------------------------------------------------------------
Here is the recipe for the high reflectance white paint, as taken from the paper by Wu et al. (1972, AAS Photobulletin 1, 9).
"To make one pint, heat 50ml of distilled water to 52-66 C (125-150 F) and slowly add 2.25 grams of elvanol polyvinyl alcohol (Grade 72-60), mixing until all the alcohol dissolves.� To this mixture add 150 ml of distilled water and 200 ml of 200 proof anhydrous ethyl alcohol.� Then slowly add (stirring continuously) 227 grams of USP grade barium sulfate to the mixture.
In practice it seems best to mix this paint just before use. A coating of Krylon 1502 flat white paint was sprayed as a primer.
-------------------------------------------------
And of course, from AAVSO site: http://www.aavso.org/files/ch8_0.pdf
Hello Tonisee
Thanks for pointing me to these papers. Much to absorb. However,
"Adopting the global linear least-squares (\ubercal") approach developed for the Sloan Digital Sky Survey (SDSS), we derive corrections for all observing runs, which indicate that the original sky flats were nonuniform by up to 0.13 mag peak to valley in z band, and by up to half that amount in BVR." from the Ubecal paper.
I guess my Raster is not so bad. If Sloan can be off by 0.13 mags. Of course, these papers describe how to get much better.
Gary
Hi Tonisee:
This is an interesting document. Can you tell us what it is, is there any more of it, is it current (out of date, under construction?), and where can we get more of it?
Thanks
Richard
I believe, you pointed to "AAVSO document". I don't have slightest idea! But I suspect that it is written by Arne. Hopefully it was not a wrong decision to link it here. After all - I found that link using Google.
That chapter is an early release copy from the CCD book, and was used for the CHOICE course on image calibration. A later version spins on my disk, but you won't find it through Google.
My understanding of the flip flat is that you turn it on and then use the camera shutter to adjust the intensity. In that case, any nonuniform deviation for short exposures is caused by the shutter, not the EL panel. This was being discussed on the CBA list.
Flat-fielding is an art. There are many ways of doing it; most work; you just need to understand what is going on and be prepared to test your assumptions.
Arne
Hello
Just a minor correction here. The flip flat does, according to their web page and op manual, have a means of adjusting the intensity of the panel to accomodate narrowband, normal band and wide band filters.
I also have an early one that I picked up on Amart, of unknown manufacturer, that has a 10 setting output from the power supply to do the same thing. I suspect that most do this.
This relieves one of the need to adjust the exposure as the only means of setting the level of the flats.
Gary
PS: I am still waiting for someone to post their raster results with an ELU or sky or dome flats.
Yes I got it. It appears very flat using a spot light meter. You need the neutral density filter for BV&Rc unless you want to use the built in "flasher" that blinks the panel on and off. Blinking potentially adds variables so I don't use it. There are several different density filters available with it. I got the medium one. I am using exposures of 10 sec for B, 7 sec for V and 6 sec for R. to get about 21,000 counts maximum in each. That tells you the light output of the screen is falling off toward the red and it falls of dramatically for Ic.There is not enough light output in Ic to use the neutral density filter. Integrations were several minutes each. However, without the filter in place The Ic integrations are about 36 seconds for my set up. Further, the intensity remains the same throughout 20 sets of BVR flats. That was not true for my light box that used halogen lamps. The output varied 10% or more over the set. I don't know how much of that was the SmartHome electronic dimmer I was using to control the lamps and how much was due to temp changes of the lamps.
The neutral density filter is a thin plastic film inside the flat panel sandwich.Therefore it is not something that you insert and remove on a nightly basis. I removed it and sandwiched it between two sheets of non reflective plastic. This plastic is clear but appears to have a matt finish not the pebble finish you often see for light diffusers. The filter sandwich may have an affect on flatness. I am experimenting with edgings for it and will let you know after I run a grid test with a star. I should have that done next weekend, weather permitting. I haven't had many nights without clouds since the beginning of spring and even when the nights start out clear, low clouds form starting at about 01:30 as the atmosphere drops below the dew point.
My intent was to hang the flat panel on the wall of my telescope shed like a flat screen. I haven't been able to do that yet. I have to build a mounting frame that will allow me to easily add and remove the filter "sandwich." Right now I am just parking the telescope at the zenith and manually placing the flat panel and filter on top. That doesn't work if there is wind since I can't close the roof with the telescope in that position.
I'll let you know the results of the star grid test.
Brad Walter WBY
I built a flat field panel using an 11x11" EL panel. It is sandwitched behind a diffuser panel and foam board with light attenuated further with white artist paper. Two pieces of foam board are in front with aperture holes for my 8" and 6" OTAs so they form fit over the end of the OTA. Just put the panel on the scope and shoot flats. Turn off the light and shoot flat darks. Kind of like a flat light box. The V flats appear very flat, I will be testing B and I soon. Light intensity cannot be controlled and the El panel emits very little light in the near IR, so exposures vary: about 1 second in V, 5 seconds in B and 120 seconds in I.
Ed
It looks like the EL panel gives a very flat field.
I had to make the grid twice.I was lazy the first time and just tried to use a simple dark subtract for the data frames before flat correcting. That didn't work at all. I ended up with big radial over the FOV the background int the center was much brighter. So I did it a second time using all of the same calibrations I would normally use for a real observing run: 20 darks each at corresponding time and temp for data frames, dark frames and flat frames, 30 Bias frames and 20 flat frames at around 23,000 max counts for each.the chip is a KAF 6303 in an SBIG STXL camera. I did a 3 x 3 grid for each frame. the target star positions were varied horizontall at ~ center and offset +/- ~1200 pixels from the central position (about 15 arc minutes) and vertically at ~ center and offset +/- ~ 650 pixels. It turns out I didn't get the vertical offset equal. The top row was offset about 680 pixels on average, but the bottom was offset only around 618 pixels.
I used Landolt nothern Standard star SA 38-358, a 9.9 magnitude star with B=V at ~1.1. CCD error equation values for All magnitude measurement gave the same stochastic error at about 0.0027 magnitudes in all positions.Airmass was over a narrow range of 1.045 to 1.051.
All measuremnts except the bottom left deviated less than 1% in net flux from the star at the central position. The bottom left had a significantly larger deviation of 1.5%. When the star was in this postion it also had a significantly larger FWHM than at all other positions including in the imediately previous and subsequent images. I only took one image per position. There is very poor corrolation between variations in net counts with radial distance from central star position, or with X / Y position. In other words there doesn't seem to be a gradient pattern. There is also very poor correlation between FWHM and radial separtion from the central star postion. In fitting a trend lines of Net counts and FWHM vs. radial separation I forced the trendline to be zero at zero separation. That is probably not a valid constraint since there is only one observation at the central position. The trend lines are substanitally flatter if this constraint is relaxed. To do a thorough job I should have gone through position sequence about 5 times and averaged them. Because of the variation in FWHM, I made measurements using a rang aperture radii from 6 pixels to 12. The data is for the middle of the range. average difference of measurements form the central location is slightly larger at 12 pixels and the left bottom location still had the largest variation. With a 6.0 pixel aperture the average flux variation from the middle position was 0.1% larger as well but the maximum deviation was 0.1% smaller and the star having the maximum deviation shifted from the left bottom to the center top which varied in net countsby 1.4% compared to the central position. The variation of the left bottom star net counts decreased to 0.7% which supports the randomness of the variation in measurements.
My conclusion is that the EL panel even with the neutral density filter used in an external assembly comprised of the filter sandwiched between two non-reflective rigid plastic sheets (with white tape around the edges) gives a very flat field for a 15" EL panel used with a 10" aperture telescope. I noticed that the EL illumination has a sharp roll-off near the edge of the panel when used with the external neutral density filter assembly. I do not notice this in the panel itself without the external assembly. I did not notice whether the roll-off occurs when the neutral density filter is installed inside the EL panel assembly. I took the filter out before using it to make flats. I suspect the roll off is cause primarily by my external filter assembly but am not certain.
Attached are some 200 Row and a 200 Column wide averaged profile plots from Mira Pro that span the Image. The 3 Row profiles are centered at row 300, image center and row 1883. The column profile is at image center. These are made using a couple of the SA 38-358 images used to create the grid in my previous e-mail. I just picked one of the images that didn't have a bright star located in the profile area.
In this case pdf in the file name stands for Pretty Darned Flat.
Brad Walter,WBY
Hello Brad
I cannot open xlsx files. Can you save them as xls files?
Gary
Hi Gary,
Try this Windows converter it is free :
http://microsoft-office-file-converter-pack.en.softonic.com
It will converts all office data from xlsx to xlx and docx to doc.
Regards,
Velimir
I was very impressed with your knowledge of making flats. I dread taking flats.
What have you found that works best or is fine for taking indoor flats. I am embarrassed to have to spend 379.00 for my C14. I am looking at the spike for taking flats. I am very much interested in your opinion.
Thanks in advance from Russ Garrett in Birmingham, Alabama!
russ@russgarrett.com
205-276-9651
Attached are versions saved as .xls files. the standard deviation calculations may vary from xlsx.
Hello Brad
Thanks for posting these files in xls format. I was able to open and study them. Very nice work. I have one question, what type of magnitudes are those listed? Are they Instrumental, Maxim, STD, Other?
It will be interesting to see what happens when you run the raster test with multiple images at each location. This may help beat down some of the noise.
Gary
Gary,
The magnitudes in the magnitude column (immediately to the right of the Name Column) are raw instrumental magnitudes - -2.5*LOG(NETCOUNTS *GAIN/EXPTIME) The Landolt standard V mag is given (manually entered) in the "Mag Std" column to the right of S/N and the difference between standard V mag and the instrumental mag is given in the residual column. The range of airmasses spanned was really small and near the meridian, so for the purposes of this this exercise that column can be taken as a zero offset. (not really a zero point offset since extinction and transformation are lumped in). I know you know Mira Pro. These measurements were done without selecting a comparison star since I only care about changes in magnitude with position. I don't care what the actual magnitude is.
We have had relatively poor quality skies at my central Texas location this year. I wanted to get through a set of grid positions quickly between intermittent bands of cirrus clouds. On top of that my observing time is usually short. I generally run out of sky before 1:30 or 2:00 a.m.. when the atmosphere falls below the dew point a couple of thousand feet up and low clouds form within 10 or 15 minutes over the entire sky. It doesn't get dripping wet at ground level but a couple of thousand feet up the sky acts like some hit the smart glass control to give the sky some privacy from humans.
Brad Walter, WBY
Hello Brad
Thanks for the info that these were Instrumental Mags. Sometimes they are differential and that obviates the whole experiment. These are good results. As you have seen, we are pretty much dependent upon a very good night to get the true instumental mags and the flatness of the equipment.
I have posted a graph of the latest data I have. Its about 4 months old. I have not been able to get telescope time here at MMO while the REU students are doing their projects. I have made a few modifications to the scope, but have not had an opportunity to do the raster test after the mods. Its getting better. The plot shows +/- .006 mags. Its a 6x6 Raster repeated 4 times, for 144 total data points.
Gary,
Did you look at the PDFs in my follow up e-mail? Its another approach but they also indicate that the flat source is flat. Otherwise the background level would show some indication of gradient in at least one of the traces.
Brad Walter, WBY
I am confused by the raster graph. The Vertical axis is labeled DEC position S-N and the horizontal Axis the RA position. How can I see the variation in intensity with position? Is the Dec axis scale actually Delta magnitudes from the the value at center of the Image or perhaps from the average for the star in all positions in all images?
Brad Walter, WBY
Hello Brad
Sorry for the confusion, these are delta magnitudes from the average after removing extinction. There are 6 positions along Dec and 6 Positions along RA. Same thing as you plotted I believe.
The lines merely help distinguish which data is which.
I have added to the label on the Y-axis.
Gary