Hello everybody,
After some attempts (5), it seems at last I succeeded in my first "bare-foot" DSLR
photometry. Sept 24,2012, starring: Alf Her, equipment: Canon EOS 1100D with 18-55 lens,
environment: light polluted but clean suburban sky, moon, 21°C, processing with IRIS 5.59
following the CitizenSky tutorial.
I said "it seems" because I actually don't master fully the math and physics besides
the "Reduction-Intermediate" spreadsheet, so I must trust what it produces.
Successful for me it means when I see numbers below 0.05 in "comparison stars fit error (max)"
and "check star mag error" cells, followed by "OK" in "Air mass plausibility check".
The values I got in these cells are 0.011 and 0.010 respectively.
I'd like to submit my data - I would like to upload a file in extended format - but I have some doubts about the
how to compile some fields:
-Mag Err: I would put the highest of the two error values I just said above.
-Filter : the tutorial says that the spreadsheet applies a correction to magnitude values
obtained from DSLR images (Green component) to match standard V filter values. So, what filter
should I select in this field? Johnson-V? or TG?
-Trans: I have no idea of what this means, so I would put "NO"
-Mtype: I would put "STD"
-CName, CMag: the spreadsheet requires some comparison stars, while here it seems I must select only
one of them. I would select the one with the smallest error.
thanks for any suggestion
Giuseppe
(BGMB)
Hi,
congratulations, the numbers look good.
As for the spreadsheet, it's no magic:it performs a linear fit of the data to find a color transformation coefficient that minimizes the squares of the residual errors of measured and catalog (V band ) magnitudes. So the magnitude that is computed for the variable star is in fact a "transformed" one and probably should be reported as such. The spreadsheet can also perform a correction for the differences in atmospheric extinction across the field, something that is usually only called for when the field is low in the sky. In that case it is just performing a two dimensional linear fit instead of a one dimensional when just doing a color correction fit. In other words: your comparison star measurements provide a set of data points and the spreadsheet is just trying to find transformation coefficients that result in the best agreement of your measurements and the catalog values of the comparison stars, taking into account that your camera's green filter is different from a standard Johnson V filter and (optionally) that some stars in the field are measured thru a longer column of air than others.
Clear Skies
Heinz-Bernd Eggenstein
Ok Heinz,
I'll enter "Transformed=YES" and "TG" filter in WebObs, these were the biggest doubts in my post. ù
About the good results, I wait the next clear sky night for another try (#7) and see if the good results of (#6) were "beginner's luck" or not :)
thanks
Massimo (BGMB)
This is
Congratulations!
This is very exciting. Once you get comfortable with the technique. You should make a nice write-up for the neofite and publish it in the DLSR astronomy groups... Perhaps even an article in the astronomy magazines. You could bring in hordes of new photometrist!
Juan Herrero
In fact, I wrote
Hi Juan,
In fact, I wrote about the mistakes I did in the previous 5 attempts. To teach something, a well reported error has the same value of a success.
Thank you for encouragement.
Massimo
BGMB
You've just
Giuseppe:
You've just become my new best friend! :-)
I got a Canon 1100D just two weeks ago with the aim of using it for bright star DSLR photometry as well. Very glad there is someone here who is also using that equipment.
----
Doc Kinne (KQR)
Did
you're welcome Richard.
Did you already make some tests?
Massimo (BGMB)
Richard:
Keep posting. I am in a position similar and I am finding your posts far more useful than some experienced posters who are too condescending toward green-as-grass FNGs like me to be of any use.
I only get one or two opportunities a month and I spend a lot of time on outreach for my Astronomy Club so progress is slow. I have not yet submitted any data as my check star error is still 0.1 mag or so. Much better than my initial attempts where the error exceeded my visual scatter but not yet clean enough to submit.
A lot of my problem is that the stock 50mm lens does not quite focus at infinity and no matter what I do results in doughnut star images-even when stopping down aperture. Also the image scale often results in insufficient separation between star images when I get to the Aperture Photometry command in IRIS. My results with the test data in the DSLR tutorials show consistent results so I am reasonably confident of my techniques there. For me it appears to be an issue of obtaining good images for IRIS to work with.
In spite of these I am having a lot of fun and have spent hours perusing my images for variables. All kinds of neat stuff is buried in those images.
Equipment-
Canon T3i with 50mm F1.8 lens mounted piggyback on a Meade LX-200 Classic with equatorial wedge.
Software- IRIS and the AAVSO spreadsheets and DSLR tutorials.
HI David,
Have you already tried to use a Bahtinov mask?
http://en.wikipedia.org/wiki/Bahtinov_mask
Clear skies,
Robert Fidrich (FRF)
Robert-
Thanks very much for the response.
I haven't tried a mask for the camera because until I get the techniques down I am only doing piggyback photometric images. A step at a time for me.
I had suspected from my other astrophotos an issue with this lens so I ran a series of test shots out at Calstar a few of weeks ago. This particular lens runs against the stop before reaching infinity.
Some of those lenses go past infinity for some reason. It was the case with our T1i. Try the live focus or a mask. Might help.
Jim Roe
As for focus: for
Hi
As for focus: for photometry, i think a live view with zoom is more than sufficient: focus on a bright star and then defocus slightly to spread the star image across a few pixels. One of my lenses is giving me donut stars as well, but I haven't seen any significant negative effects of this on the photometry, so I would not worry too much about that.
clear Skies
hbe
Hi MDAV,
0
0
1
180
1029
ANSTO
8
2
1207
14.0
Normal
0
false
false
false
EN-US
JA
X-NONE
/* Style Definitions */
table.MsoNormalTable
{mso-style-name:"Table Normal";
mso-tstyle-rowband-size:0;
mso-tstyle-colband-size:0;
mso-style-noshow:yes;
mso-style-priority:99;
mso-style-parent:"";
mso-padding-alt:0cm 5.4pt 0cm 5.4pt;
mso-para-margin-top:0cm;
mso-para-margin-right:0cm;
mso-para-margin-bottom:10.0pt;
mso-para-margin-left:0cm;
mso-pagination:widow-orphan;
font-size:12.0pt;
font-family:Cambria;
mso-ascii-font-family:Cambria;
mso-ascii-theme-font:minor-latin;
mso-hansi-font-family:Cambria;
mso-hansi-theme-font:minor-latin;
mso-ansi-language:EN-US;
mso-fareast-language:JA;}
Bikeman is right, you don't want sharply focused star images for DSLR photometry because too few pixels of each colour will be exposed. The Bayer filter matrix is composed of an array of 2x2 pixels with one blue pixel, one red pixel and two green pixels. So when the software extracts, say, the green image it interpolates what the green intensity would have been on the red and blue pixels. It is better to defocus so FWHM is more than, say, 5 pixels. This will allow the software to interpolate intensities for each colour more accurately.
If you are using a recent electronic Canon lens it can be focused via USB cable using the BackyardEOS software. I have the same camera as you with a Canon EF 200mm f/2.8 L II USM Lens. BYE is excellent for focusing and capturing a series of images.
The image scale with your 50mm focal length lens may cause problems with nearby stars contaminating your measurement aperture when doing differential photometry. Even with a 200mm lens I find a lot of my targets have this problem. Not a great concern if all you want is time of minimum for an eclipsing binary. But a big issue if you need to compare magnitude measurements with other observers. Cheers,
Mark
Hi All,
I'm not sure where all the extra lines came from in my previous post. I composed it in Word before copying and pasting into the forum reply window. Cheers,
Mark
Thanks Giuseppe for starting this thread ! -- I find myself in nearly the same position as David (MDAV) -- short on photometry experience and also limited time :-( ...
I've worked through the DSLR tutorial and have got IRIS up and running.
I will contiue to follow this "Supplement" to the tutorial with great interest.
I appreciate all the helpful suggestions and sharing experience.
Walt (FWJA)
Hi all,
it's embarrassing to say after all the enthusiastic comments appeared here after my first post, but there was a mistake in my spreadsheet about the calculation of Air Mass for the check variable - no problems about the comparison stars and the variable, but now the data quality check reports an error of 0.135 mags for check variable and 0.011 max for comparison stars (the same as before) so I'll have to remove the measure from AAVSO database.
My spreadsheet derives from Tutorial's intermediate, but I inserted some more formulas to convert the RA coordinates from HHMMSS to angles and some cells where to store the aperture photometry circles diameters.
Doing this, I copy/pasted few cells with values only without the formulas.
The oct2 I took again images of Alf Her field, but this time I defocused too much and I didn't noticed that while capturing the images. The FWHM results about 18 and the spreadsheet says errors 0.084 on comp stars and 0.085 on check star.
But sincerely, I have doubts about what is better: the previous measurement had little error on comp stars and bigger on check star, and this difference raises the suspect that something is wrong in processing or reduction. I'll investigate again about that by retrying the measurement on IRIS with bigger circles.
In the last one instead are almost equal, and this gives me the impression that I took better images than in sept 24 (when the FWHM was 8).
Hi all,
I am a past CS DSLR team guy also involved in the IRIS tutorial. I have no good experience of the spreadsheet itself as I use another software.
Defocus is essential with DSLR, it is needed to reduce the undersampling effect of the Bayer cell structure as said by Heinz-Bernd. But it is also important to increase the electron storage capacity for each star. The "storage" capacity of a single pixel is only about 25000 e-, this is far to small to achieve a good SNR. It's usual to involve 100 sub-pixels to meet good instrumental conditions and avoid saturation. Then the blending with background stars is an issue, it shall be checked using the catalogs and proper software like SkyChart (free, using Tycho-2 and the GSC-ACT for fainter stars)
I concour with HB, the live view used at x10 on a bright enough star is excellent to calibrate the needed defocus, you can directly measure the spot diameter on the screen and ensure you use the optimal value (easily 5 to 10 pixels diameter). A 50 mm lens is a too small focal lenght from my experience and a reason for a severe blending. Several of us work with a 200 mm lens instead. At a given F# the telelens offers a much larger aperture surface and much more photons for a given star.
I wonder about the use of Bathinov mask, how to use it to determine a calibrated defocus ? Any tutorial, instruction to do it ? I am also afraid the signal would be too low to be visible on the live view.
There are a couple of inaccurate points in the CS tutorial you probably noticed. I attach a revised and augmented version that has not been published.
Clear Skies ! (Here in France, the sky remains cloudy, a very bad year for observations ! )
Roger
Hi Roger
as in France, in Italy too the weather cloudy. So yesterday evening I surfed the net searching more docs about CCD/DSLR photometry with or without IRIS and I found these very interesting:
http://www.astrosurf.com/buil/iris/tutorial15/doc38_us.htm (Photometrie D'Ouverture - Buil)
http://www.astro.bas.bg/~petrov/papers/02romanishin-ccd_photometry.pdf
(i found in fact an italian version too)
With the infos found in Buil's article I found how to obtain the PSF / FWHM / shapes of stars in order to choose a correct size for photometry rings. The sizes I used until yesterday have been chosen in way to include the star within the first circle, but in this way some information got lost.
So I repeated again the photometry on my Alf Her images of Sept 24 and Oct 2 choosing a inner ring size a little bigger than the FWHM. (Romanishin says, in chapter 17.2, the best size is 1,4xFWHM)
More, I changed the check star choosing another one having an airmass more similar to the comparison stars ones and a b-v value similar to Alf Her.
In this way I got again decent values on my first image - sept24 (FWHM 8): 0,029 max error on comp.stars and 0,005 on check star.
The second image - oct 2(FWHM 18) still gives bad results: 0,148 on comp.stars despite the 0,010 on check star. My sensation - looking also on the PSF shape graph - is that the comp. stars light is spread on too many pixels so the ratio between the signal each pixel gets and the background sky noise is insufficient. In fact, I choose comp stars at about mag.6 while the check star is 5.030 on catalogue. So it gives better result that others, and the faintest comp star gives the biggest error 0,148 (mag. 6.740 on catalogue).
Roger, IMHO is important to add in your tutorial the part about PSF and shape graphs as shown in Buil's tutorial.
I have a question about the chapter 23 of Romanishin's : I choose my comp and check stars with "Cartes du Ciel" being careful about avoiding chosing known variable stars.
specially if I look for "bright" stars (mag < 7). So I read that exists a list of sample stars spread over all sky to be used for photometry calibrations. I'll try to get that list and see if I can be useful for my purposes.
But looking in the VSX, it seems that almost the stars are variable or suspected ones
Clear and dark skies
Massimo
(BGMB)
Hi Massimo
Yeah, finding good comp stars is always a big issue, as for DSLR photometry you'll want to have several in your field, and not just "somewhere" in the sky.
BTW, would you mind sharing the two imges in question? I think you have made many people interested in doing some hands-on experiments with your data, and one can discuss this much easier with the acutal exposures at hand.
My email is hbe[at]ehea-obs.org. If the stuff is too big to attach here and you lack webspace, I could put them online on my webspace with your consent, of course.
Clear Skies
HB
EDIT: correction of my email address :-)
Hi Heinz,
I loaded the original CRW and all the intermediate and final files produced by IRIS in my Google Drive. The public link is https://docs.google.com/folder/d/0B54Zm2R1qWqeQUJtLVBzQ3pld3M/edit.
There are both the images and spreadsheets of Sept24 and also Oct2 despite the bad quality of images - I think is not worth further effort on those, I prefer to wait another night and retry taking better pictures.
Clear Skies
Massimo (BGMB)
Excellent, thank you. I hope I'll have time to look into the data later today
CS
HB
Hi Massimo,
You got the good point from Buil and the curve would show you that 1.4 x FHWM is far too small ! This is also shown at the point 2.3 of the revised tutorial I attached. If you look at the curve under 2.3 you will see that the point where the electron count gets to 99% is about at a radius of 12 pixels ! This is very far from Romanishin. The FHWM was at about 4 pixels radius. Here to ensure 99% of the electrons are in, the factor is not 1.4 but at least 3. With a 12 pix radius you ensure the mag impact is at worst 0.01 mag. If you look at 1 mmag you need even a larger radius.
Romanishin uses a technique that's ok for the process used by professional but it doesn't work for us with DSLR. Due to the lens aberrations and defocus that both change depending the position of the star in the image we can't use that PSF fitting methode. The purpose of that fitting is to maximize the SNR but it's usable only when the star is at the center of the field of a pro telescope. With DSLR lenses the PSF "foot" varies a lot across our field. We use number of stars at various positions in the field, then this PSF fitting can't work. We have to set the circle on the star having the larger PSF "foot" corresponding to the targeted accuracy (like shown in 2.3) Then the same circle size shall be used for all stars involved.
Romanishin is a good author but his technique is for the pros and their specific equipment. Often it's not applicable to our specific DSLR "mapping" based technique. A couple of weeks ago we have seen that his way to evaluate the noise and the SNR is an approximation ok with the large flux of a large telescope and a deep cooled camera but not ok for us. He neglects the read noise and the size of the background circle that's ok with his technique. With our DSLR and small lenses it's unacceptable, the read noise can't be neglected, and our background is not measured the same way. This is two exemples that show refeering to the pro technique is not always good for us !
My next point is the use of the "intermediate spreadsheet". It's not that simple in fact. The choice of the comp stars is critical. If the color of the stars is not well balanced between the high and the low extinction areas of the field an error occurs in the determination of the color transformation coefficient and the extinction coefficient. The single equation system that computes the two coefficients can't properly separate them if that color "balance" (or color/extinction gradients decoupling) is not satisfied. This is somewhat a limit of this technique. Depending the case the error is negligible or unacceptable. One way is to use a massive set of comps (several tens) spreaded into the field, but that needs a specific automated software. I use another technique that separates both operations, I did describe it in a paper submitted to JAAVSO but I have no news about it at time being.
Variable stars: It's true the var catalogs do not consider most microvariable stars, and all stars are more or less variable (to see the Kepler's findings !) But ok this is not at a point we would be in trouble. The Tycho-2, Hip prep and Hipparcos catalogs have various flags that show the stars variability, you could also find a "scatter" of their observations that's a good indication of the stars stability. Some stars got the flags "C" and "S" that says they have been seen constant or are considered standard stars (the best). The CDS references are: I/259, I/196, I/239. Beside this there are standard stars fields ( Landolt, Cousin..) but most those stars are too faint for us or are too south to be accessible.
Clear Skies ! (Tonight it's ok here, I will make a couple of epsilon series)
Roger
Hi Roger,
The screenshot shows the shape and growth curves I obtained from Alf Her the Sept24 - the boxed star (Oct 2 images are not considered here due to garbage quality). My Google Drive link to these images is in my answer to Heinz.
The FWHM is shown in the display window (8.42 X 8.02 Y), so choosing the conservative Romanishin's formula, I should use an inner circle of at least 8.42 x 1.4 = 11.8 while the graphs are showing me that above 10 pixels of radius, the curves becomes flat, so what's the use of taking bigger circles sizes? Even multiplying by 1.4 it seems too much!
In other words: the example in your tutorial says you have FWHM 4 and the growth curve becomes flat above 12 pixel radius: why this difference? did I defocus too much?
Thanks again
Massimo
(BGMB)
Hi Massimo,
FWHM is a diameter, the growth and shape curves are in radius. Looking at your curves I see 0.01 mag would be stabilized at about 11 pixels that's a 22 pix diameter or 2.7 times the FWHM. Not exactly the 1.4x from Romanishin !
Maybe your defocus is a little bit large, I see a peak on the bottom due to some aberrations. I usually prefer to defocus less to get a well round profile and then expose longer to get a trail in addition. The trail provides a much more uniform spread of the photons across the pixels and the pixel capacity is better used. But ok it's true IRIS has only circle, I use an elongated aperture that fit the trail.
Clear Skies,
Roger
Hi Roger,
FWHM is a diameter.... naaaah
ok now I understood.
By the way, I was in doubt that capturing a trail could confuse IRIS in stacking phase so I preferred to not exceed 5 secs, but the next time I'll take 2..3 series with a narrower focus and increasing times. My fear is also to catch too much noise too, since I take photos from a light polluted place. Let's see. Thank you
I have another question about IRIS but it's more related to software development than astronomy.
I tried to automatize the calibration-registering-stacking procedure. I've seen that IRIS has very limited scripting capabilities - for instance is not possible to catch a value returned by a command and pipe it to following one and neither save it into a variable to be used later. So I canned a semi-automatic procedure through a python script that generates a sequence of pgm files to be launched manually from IRIS console.
The problem I found is that is not so straightforward to find the right command or sequence of commands that produces the same results of the menu commands explained in the tutorial so I'm not able actually to perform the Preprocessing step.
The PR command in fact produces a sequence of greenish images that looks very different from the images produced by the "Preprocessing" menu command and don't know what's wrong.
Being not open source, I have no way to compare what PR and "Preprocessing" menu command really do.
Clear Skies
Massimo
(BGMB)
Hi Massimo,
Thanks for the images, I had a look at a couple of them, img_10.pic, dark_2.pic. The first thing I noticed is the "system" black level at 2048 instead the classical 1024. Apparently Canon did push it due to the fact they increased the ISO range to 6400. Not ideal for us, but ok, it's workable.
Then the pixel max level of img_10 is about 3037. That means the signal/pix is only about 900 (3037-2124). The sky background level being about 76. All this is very low. You need to expose much more to get a good SNR. I have no idea at which ISO those images have been made ? Given the noise level (if dark_2 is a single dark shot ? ) your ISO should be about 200~400 and you are probably near a 1-to-1 e-/ADU ratio (this is the good ISO setup when an high flux is not available). Then you can push the exposure 10 folds to 10000 ADUs without problem and increase a lot your SNR.
With the 50 mm lens you could push 4x to 20 sec, the trail should be small enough to be ok with IRIS. Then you could reduce the defocus by 0.6x . But this is true the stacking softwares don't like stars that are not pin-point ! They don't like the defocus either... As an exemple, DSS doesn't work at all with our defocused stars ! This is one of the reasons for which I do not stack and measure each individual image (using my own automated software). It's possible with IRIS if you use only 5~6 stars and 5~10 images depending the exposure. It doesn't take much more time than stacking (IRIS being slow stacking !) This is very interesting at beginning as you can calculate the average AND the sigma of the series. The sigma shows you the level of uncertainty of your observations and helps you to improve it.
Software: Yes, IRIS is very limited in scripting, it misses a kind of "for...next" to be usable for series. I have no experience with python, for long time I develop all my software under an APL system in APL language (Dylog APL, not free). C/C++_like languages would take me far to much time.
Clear Skies !
Roger
Hi Roger,
I see Google Drive refused to load all the images I put in my shared directory, so some of them remained here. I will delete the intermediate files and keep only the raw, the final and the spreadhseet online.
The images were 13, plus 5 darks and 5 flats.
Images and darks have been taken at 800ISO 5sec f5.6. Flats have been taken in full auto.
"With the 50 mm lens you could push 4x to 20 sec, the trail should be small enough to be ok with IRIS. Then you could reduce the defocus by 0.6x . But this is true the stacking softwares don't like stars that are not pin-point ! They don't like the defocus either... As an exemple, DSS doesn't work at all with our defocused stars ! This is one of the reasons for which I do not stack and measure each individual image (using my own automated software). It's possible with IRIS if you use only 5~6 stars and 5~10 images depending the exposure."
Here you mean the "Automatic Photometry" I guess. IRIS allows to record up to 5 stars including the variable itself and outputs the magnitudes of each of these stars in a sequence of images. The spreadsheet should be a little arranged, because the magnitudes to be put in lightblue cells won't come directly from IRIS but from an average of the magnitudes read from each picture.
Clear Skies
Massimo
Massimo,
I was considering only the manual solution. It doesn't take a lot of time to peak 5~6 stars in a picture. I had no good experience with the automated solution of IRIS. The auto placement of the apertures seems not accurate enough.
Roger
If you are looking for comp star you can try to use Seqplot:
http://www.aavso.org/Seqplot
https://sites.google.com/site/aavsosequenceteam/Home/how-to-use-seq-plot
But the AAVSO Sequence Team members can also help finding appropriate comp stars for you.
http://www.aavso.org/request-comparison-stars-variable-star-charts
Hi all,
here is a DSLR photometry for eps aur from pictures taken in oct 23, with narrower FWHM, max error 0,025
https://docs.google.com/folder/d/0B54Zm2R1qWqeNlNicjhFenVyMms/edit
I'd like to take shots of Mira, the next time.
clear skies
Massimo (BGMB)
Hi Massimo,
Good job !
I think we had similar weather, poor many days, but I also got an observation the same night ! It was not perfect, the extinction was unstable. As usual, I made 5 series of 10 images with the 450D, a 200mm Nikkor at F4, ISO 100, 10 sec. each image.
Oct. 23rd, 2012 22h7m UTC, the 5 series results (sigma is such of each series mean)
eps AUR V = 3.024 (0.005), 3.028 (0.003), 3.005 (0.003), 3.015 (0.005), 3.023 (0.005)
HR1558 V = 6.080 (0.004), 6.088 (0.004), 6.081 (0.005), 6.083 (0.004), 6.090 (0.004)
HR1550 V = 5.710 (0.003), 5.703 (0.004), 5.709 (0.003), 5.707 (0.005), 5.701 (0.009)
HR1644 V = 6.200 (0.005), 6.193 (0.006), 6.199 (0.010), 6.197 (0.006), 6.191 (0.008)
I don't use the CS methode but what I call "VSF" This is a Johnson's V like Synthetic Filter made by combining the R and B channels of the DSLR with the G channel. It's totaly independent of the B-V catalog values when calibrated. It also takes in account the atmospheric reddening. Then the neutral extinction gradient is independently computed through image mapping.
The issue with the CS technique (or similar) is the dependance between the color correction and the extinction correction. If the stars color is not well balanced along the extinction gradient it results in significant error. By the way the choice of the comp stars is very difficult for this technique. A too large distance between stars makes the issue more critical. This is the reason why I prefer to use comp stars near the target and as much as possible "bracketing" it (no extrapolation).
The three HR1550, 1558, 1644 are very interesting for eps AUR, they are near, reasonnably bracketing, and have colors not too far eps. You should have HR1644 as you see 1550 and 1558. I would recommand to test those three comp. stars. Your ensemble is very excentered from eps and could be a problem. I am very interested to see at what level of accuracy we can reconciliate DSLR measurments when using the same comparizon ensemble, eliminating bracketing issues.
The values you have for eta seems not the standard (3.172 -0.18) ?
Very interesting indeed !
Cheers,
Roger (PROC)
Hi Massimo, Hi Roger.
Interesting, I managed to get a series of exposures under reasonably good conditions early in the morning of Oct 20th, when eps Aur was almost in the zenith but with some haze.
As for the eta Aur magnitude, there had been much discussion on it during the CitizenSky campaign IIRC, I dug out a message from Arne Henden with the quoted magnitude here:
http://www.citizensky.org/forum/mini-observing-campaign-secondary-eclip…
I'm basically using the method as described in the CS spreadsheet, but lately I automated it in a way so that I can throw in some stacked images and the automated pipeline does all the photometry by itself automagically. It is using the tools SourceExtractor and Scamp as distributed in the "Scisoft" package by the European Southern Observatory (ESO). It selects all stars brighter than a given magnitude as comp stars as long as the stars are not flagged as variable in the reference catalog data (I use the ASCC 2.5 catalog) and they are not blended with other stars (but I overruled some variabily flags manually (e.g. for the eclipsing zeta Aur) and used the comparison star magnitudes from the CS spreadsheets for the stars given there).
I did 5 x 10 shots, so 10 exposures are stacked for each series. Each exposure was 3.2 seconds with a legacy f=50 mm Pentax lens, stopped to f/2, on a rather noisy but lightweight and compact Olympus E420 DSLR (w/ FourThirds sensor).
I got these values for eps Aur:
3.08073, 3.01841,2.98495,3.01395,3.0065
The mean of these 5 values is 3.02091 and the standard deviation is 0.03. So ithe mean is in the same ballpark as Rogers measurements but he gets more consistent results using better equipment and longer exposures (my E420 only has 12 bits)
Here are my values for the comparison stars that Roger mentioned:
HR1558 (TYC 2906-2736-1): 6.09904 , 6.02023, 6.04984, 6.06661, 6.01945 (mean: 6.05103) The ASCC2.5 catalog value is 6.058
HR1550 (TYC 2902-3380-1): 5.71099, 5.61613, 5.62586,5.67262,5.61487 (mean: 5.64809) catalog value used was 5.682
HR1644 (TYC 2907-780-1): 6.26893, 6.24391, 6.2029, 6.20929, 6.19942 (mean: 6.22489) catalog value: 6.203
At ca 3 mags below the target variable, those stars are already a bit on the faint side for my 12 bit chip I guess. Beyond 6th mag I get less than satisfactory photometry for the exposure setup I used.
I will try to put your image from the google docs folder thru my pipeline later tomorrow.
CS
HB(EHEA)
Thank you Roger,
this week I got 3 consecutive clean sky evenings, but "clean" is a big word here...
Anyway, my spreadsheet of Eps Aur has been produced after many attempts with different comp stars. At the beginning I took the same used in the original CitizenSky tutorial, but I got error > 0,2 on check star (ome aur) so I replaced it with a fainter one, HR1558 that gives a much better result. Then I replaced Lam Aur (error > 0,05) with HR1550 and the result is what you've seen.
After reading your message, I made other 2 spreadsheets:
In the one ending with "-4" I replaced 58 Per (the farest from eps) with HR1644. The results are not improved, the max error on comp stars is a little increased. In the "-5" sheet, I replaced eta Aur with TYC 3345-2184-1 obtaining a worser result. I'll try other stars again but I doubt I could improve significantly the accuracy. This time I tuned better the FWHM (i learned how to use the "live view" function!) the next time I'll try like you doing a serie of images at lower ISO and higher times.
About eta Aur, usually I choose the comp. stars with Skychart first, then I check on VSX to ensure they're not variables and finally I take the values from SIMBAD.
I tried to use the photometry tables from VSP but I found too complicated to handle AUIDs.
Cheers
Massimo (BGMB)
Hi Massimo, HB,
HB, yes, Tycho-2 is good we both know. But the mag you provided look Tycho VT mag (?) that would be in a different photometric system than Johnson. VT can't be used for Johnson without a transformation that's not always obvious as we have been talking about number of times.
The past discussion at CS was really unclear to me ! Later Arne said a number of times Mermillod is the good reference and this is the one that has been adopted by Brian for the tutorial (eta / 3.172) The values I use are eta=3.172, HR1550=5.710, HR1558=6.080, HR1644=6.200. Zeta is more a problem as very red, its Mermillod value is 3.755 but I see it about 3.735, in fact its color seems unstable. Mermillod gives clear info on the accuracy, number of observations... this is why it's important. But it's very incomplete... Only some stars are well documented.
In my mapping fitting with the hereabove values, the standard deviation in a good sky is near zero mmag, and most of the time below 5 mmag. Deviations of that fitting at 10 mmag generally means the extinction had a problem, like stratification... I think those values are the right ones, at least they are well coherent.
The catalog references remain a problem for us when applying our "ensemble" mapping technique. Tycho-2 seems to me the best approach as the most coherent. Next, its VT system is very similar to the DSLR G response and nearly no color transform is needed between them ! (high in the sky) I think the best is to work first with Tycho-2 for all stars and then transform the end result to Johnson when needed. Mike Bessell has developed an interesting table for such transformation.
Cheers,
Roger (PROC)
[quote=Roger Pieri]
HB, yes, Tycho-2 is good we both know. But the mag you provided look Tycho VT mag (?) that would be in a different photometric system than Johnson. [/quote]
Hi,
The mags are all in V system (reference mags are from ASCC2.5 and measured mags are transformed to V) , I just added the Tycho identifies for reference as my pipeline uses them to index the stars it detects.
CS
HB(EHEA)
Hi
I changed in my sheet the reference for mags to Mermillod / GCPD and changed a couple of stars to get better result. I'll publish now on Google Drive, the latest one - Eps Aur 2012-10-23-8-(Mermillod) is the best, Eps Aur=3.045 with error 0,002 on check star and max 0,019 on the worst comp star.
I think that to get better results at this point I can only take better and more shots the next time.
Cheers
Massimo (BGMB)
Hi Massimo
I've now managed to put your image from the google docs folder thru my pipeline. Even tho the image itself looks much less noisy than mine for example, the photometry shows some unexpected scatter.
From the looks of the calibration graph (catalog magnitude vs. instrumental magnitude) I would actually suspect that there might be something wrong with the de-bayerization. Can you tell what procedure you are using to just extract the green channel(s)? Just a hunch, to make sure there isn't a problem.
CS
HB (EHEA)
Hi HB,
I partially automatized the calibration procedure described in the tutorial. Since IRIS has some limited scripting capabilities, I wrote a python script which generates each time 3 IRIS scripts, then I call the scripts in sequence.
So, the finalG.pic has not been produced from menu command "Digital Photo / RGB separation" but i call the command SPLIT_RGB on my final.pic, that generates finalR finalG and finalB.
I believe the results should be the same, but I found that - specially for preprocessing - there is not always a correspondance menu command=script command in IRIS, so there could be errors in my scripts, it's still experimental work.
I published the composite pic and the R-G-B splitten pictures on my Google Drive so you can take a look.
by the way,
I'd like to start using my linux box for astronomical purposes.
I read about IRAF in Romanishin, but the tutorial ignores it, maybe because installation requires some UNIX experience, or because is a cumbersome toolkit or because not suited for DSLR, I don't know. Do somebody know?
Hi!
I do something very similar, I use an awk script to generate an IRIS script.
IRAF: Someone on the CitizenSky Workshop II mentioned that there is a running joke among astronomy students about the "axis of evil": "Iran, Iraq, IRAF" ...
Anyway, you should take a look at the Scisoft Collection from ESO: http://www.eso.org/sci/software/scisoft/
It has many astronomy related professional tools for Linux, including IRAF, but also SourceExtractor and Scamp, two tools that I found very useful.
The downside of Scisoft is that it is maintained only for a very ancient Linux distribution (Fedora 11 for which it it guaranteed to run, it might run on others as well ...). To get Scisoft running reliably without much hassle, I found it most conviniert to run it inside a virtual machine (I use VirtualBox from Oracle) even on Linux. So you will first install Fedora 11 in the VM, and then add Scisoft from a software repository.
CS
HB (EHEA)
I'm happy you mentioned software sunning in Linux environment. One of my friends run IRAF in Ubuntu. Later I also want to try, together with DS9, but unfortunately it needs an older GTK :(
I wanted also to try PIXY, in order to search for new variables, transient objects on my DSLR images, but PIXY is also quite an old software and it seems being not compatible with my Ubuntu 10.04 or 12.04 :( It seems there is some Java incompatibility. So I still need to use IRIS to blink my images, but I'd love to have something that makes this automated. Anyone managed to use PIXY and DS9 + IRAF in Ubuntu LTS?
Hi Massimo, HB,
I did some analysis of your final images, I agree with HB there is a problem of color mix. The color ratio of each star is flat instead reflecting their color - with ADUs normallized for HR 1558 (B-V~0) :
eta eps zet HR1550 HR1558 HR1644
R= 0.945 0.979 1.006 0.957 1 1.024
G= 1 1 1 1 1 1
B= 1.007 1.003 0.973 1.011 1 0.993
The common values are:
0.952 1.296 1.772 1.078 1 1.247
1 1 1 1 1 1
1.062 0.747 0.536 0.93 1 0.799
I don't think it's a question of de-Bayer (it would be less flat) This is probably a problem of registration and/or stacking using an IRIS command designed for monochrome images instead a CFA. Then the color are mixed up. I already saw such issue with IRIS, the scripting offers a lot of different command and options, sometime confusing...
The next possibility would be a wrong de-Bayer but I don't think it's the case. In fact there are two de-Bayer chains in IRIS: one for the DSLR (Digital Photo...) and one for the astro cameras. The RGGB sequence for the DSLR is provided by dcraw.dll, if the camera is well identified no problem. Next the RGGB sequence for the astro-cam is set in the "init" file of IRIS. Some commands use the first, some others use the second... also a possible confusion.
Cheers,
Roger (PROC)
Hi all,
There is a serious probability that my scripts use one or more commands not suited for this task. I can repeat the registering-->stacking--> conversion-->splitting manually through the menu commands as described in tutorial, I will publish then results on Gdrive as usual.
What command did you use to get the color ratio of each star?
About SciSoft and other linux software: I've seen there is a year of difference between Fedora 11 release date and my OpenSuse 11.3 so I'm almost sure there will be problems in installation of some packages.
I can check what libraries are missing and install them afterwards separately. In the worst case I can still create a Fedora 11 vbox af hoc.
cheers
Massimo
Massimo
(BGMB)
[quote=ret45]I can check what libraries are missing and install them afterwards separately. In the worst case I can still create a Fedora 11 vbox af hoc.[/quote]
From what I've read from ohers trying to install Scisoft on anything else but Fedora 11 or Scientific Linux 5 (also old), it's pretty hopeless and I didn't even try. I went straight for the vbox option which worked like a charm.
I'm curious to see your results after a modified stacking that takes care of the right color channels being stacked on each other. A quick way would be to split the calibrated images with the IRIS command "split_cfa" into 4 images each, one per channel (2 greens of course). And then stack the green images. The split images will have half the resolution of the original frames, of course, which might require to adjust the photometry aperture used.
I bet the result will be much better photometry!
Cheers and CS
HB (EHEA)
If I remember well, split_CFA is one of those astro-cam commands, the RGGB sequence is defined in the "init" file. The existing definition should be for a Kodak chip, and I think the sequence is GRBG, anyhow not the Canon's one (I got the problem in a test in the past). Needs checking.
CS
Roger (PROC)
Hi!
I would never trust the documentation / init settings anyway: split_cfa will produce 4 images and those two images out of those four which give almost identiacl photometry results for the bright (but not yet saturated) stars in the field will be the two green ones.
CS
HB
Sure HB !
You can determine it by various testing, even the B and R... Some of us uses the three colors, I think you know. Anyhow it's well explained by C.Buil, in general I belive him, maybe not exactly 100% ? What is important is to know that sequence is not automatically defined and checking it by any way is needed.
Cheers,
Roger (PROC)
I tried a photometry of eta aur on a uncalibrated pic, just to guess the color order of split_cfa for my Canon EOS 1100-D.
The result are:
guessing
img_c1 eta aur = -10.671 G1
img_c2 eta aur = -10.498 B
img_c3 eta aur = -9.546 R
img_c4 eta aur = -10.610 G2
so the order should be GBRG since the first and last value are the closest.
cheers
Massimo
Massimo,
As promized here-attached is an exemple of B-V extraction from the B and G DSLR channels. The results are not perfectly in agreement with the catalog values but well within their usual uncertainty.
I did check my notes from the review of IRIS I made beginning 2010, it says the split_CFA command parameters order is well GBRG for several Canon cameras and GRBG for Nikon cameras. We are in agreement.
Cheers,
Roger
Hi Bikeman,
I tried to install scisoft 7.7.0 on a Fedora 11 vbox downloaded from virtualboxes.org but there are many dependences missing. I should need the DVD ISO but I could not find it yet.
Massimo
Massimo,
I don't use IRIS, I know it only by making reviews, benchmarks between solutions... I work with my own software under an APL System (Dyalog APL). Dyalog is an high level pro tool used by specialists only. It's very expensive to distribute applications as full Windows runtimes as this requires a commercial license and fees for each runtime distributed. But it's very easy, instant, to make calculation under APL. This is just like writting the equation on a paper (the APL's work sheet), APL does all the next for you.
But, ok, the APL source code (directly executable, more or less like Java) is an ISO standard created by IBM, by the way free. Then Dyalog offers low cost licenses for restricted "personal" use of the system. Maybe a possibility in next future... If I have time enough ! I should also say there is a good, free, but simple, APL. It doesn't provide the graphics, GUI and many possibilities Dyalog offers. Up to recently it had no communication path to Windows or any other sysytem, but that has been recently implemented. I should explore if it's possible to run a simplified version of my software in it... same, if I find time for !
Making the color ratios is very simple, just get the "intensity" from the IRIS report .dat after measuring the stars of the R,G and B images and compute B/G, R/G... Then it's often possible to get an acceptable B-V just applying a classical transformation to -2.5xlog(B/G) ! Easy to implement it in your spreadsheet.
Getting a Johnson intensity (not mag) from R,G and B with my VSF is also simple V=G+axR-bxB, then the mag VJm=-2.5xlog V. It delivers you a color corrected mag, 100% independent, then you can make a very good extinction mapping from that color corrected mags. No more risk of color / extinction cross-contamination. With a 450D, high enough in the sky: a=0.284 ; b=0.224 (AM<1.5 , extinction coef=0.23 at 550 nm, more or less the mean in most urban places)
Cheers,
Roger (PROC)
Hi Roger,
"Making the color ratios is very simple, just get the "intensity" from the IRIS report .dat after measuring the stars of the R,G and B images and compute B/G, R/G... Then it's often possible to get an acceptable B-V just applying a classical transformation to -2.5xlog(B/G) ! Easy to implement it in your spreadsheet."
In the 10th spreadsheet, available under my gdrive/"manual as in tutorial" I added these calculations. including the B-V values.
The input data have been obtained through a fully manual procedure like the one described in manual. I've seen my scripts contain more than one error even in calibration, so I have to work on them until I can get the same final image as in manual procedure.
This time, I took the intensities instead of mags, leaving the spreadsheet to calculate the mags. The errors on magnitudes now are within 0.05
For color ratios, my calculations are on the raw values read directly from IRIS, without any transformation or correction but I wonder if it's the right procdure because the B-V values look completely wrong.
About Fedora11, I found the ISO at last.