Camera manufacturers are quite good about publishing data for Pixel Size and Full Well Depth. These are important but equally important are the sensor's response to UV. When shopping around I see QE graphs end at 400nm but happily publish IR response. Now when I see this I think to myself "UV must be a crime scene...move along nothing to see here."
A few months ago I purchased an Atik 314E at a good price. Upon receiving it I ran a battery of tests to determine Gain, Read Noise, Full Well Depth, and Dark Current. Everything was within spec. I captured an impressive 2-hour LRGB of M27. Great, so let's try photometry. M67 is out of reach this time of year but SA38 is well positioned. With just seven stars I got good scatter plots and trend lines. Tb and Tv aren't terrible but the ugly one is Tbv = 1.25. Undeterred I imaged BZ Cyg over three consecutive nights at or near minimum light. My V-mags weren't bad but my B-mags were horrible (see attachment).
I related this story to make the point that knowing one's QE is crucial. It would be nice to have the "full story" before making a purchase but who to go to if the manufacture decided that you don't need to see UV response?
Brian
Brian,
In the attached response curves for multiple CCD's commonly used for amateur photometry, the Sony ICX285 seems to fall about in the middle of the pack at 400nm and at the bottom of the pack at 350nm. I think these are the kinds of differences that color tranforms should be able to correct.
What filters are you using? Were your B magnitudes in the BZ Cyg light curve transformed?
Phil
Hi Phil,
It's from OptCorp: Astrodon J/C with checkmark on casing: "B***" but the filter cell is printed "Bu***" if I am not mistaken; I'd have to crack open the filter wheel to say for certain -- I may do that this afternoon.
Those are transformed magnitudes on the Phase Plot. I played around with comparison stars to try to get better readings. The Phase Plot shows the first try in which error was low but average is way off as you see. The second attempt got the average a lot closer to the light curve but the error was crazy high, like 0.4 mag.
Thank you for the QE link. Do you know which of those sensors are closest to standard, something that would give 0.9 < Tbv < 1.1 ??
Brian
Phil,
Here are my coefficients. (Depending on what version of OpenOffice you have, or Excel, the trendlines might not show.)
Thanks,
Brian
Phil,
Things are improving. I discovered that stacking helps a lot. Fortunately I captured 10 frames each per filter per session. Initially I selected the "best" B and V frames out of the set but I couldn't stomach the error. The biggest contributor to error in my opinion is that I am slightly under-sampling with 4.65um pixels at 418mm FL. Another factor is that I don't actively guide, just PEC, which is good but over the course of 4 minutes exposure I do experience some star elongation. So I found that stacking definitely helps.
There is another issue that I did not expect: atmospheric transparency can change quite significantly from night to night. The first night I was out was when I selected the exposure: 50s in V and 240s in B. On the following night I kept those exposures but noticed that the peak pixel value decreased significantly. Going forward I'll make some test exposures at the beginning of the session.
One other factor is that this camera has got a paltry 13ke- full well depth. The most I can get out of it is 2 mags of quality range. The range required for BZ Cyg is 1.875 mags. It's doable but I am not going to win any awards.
Brian
Brian,
Night to night variation in transparency: You should take the images in both B and V on the same night and, as much as possible, at the same air mass. If your first set of transforms used images from different nights you really should redo them. Things may look even better if you do that.
PEC, guiding, stacking: Here's something else you could try.
Do exposure tests to find out what is the longest exposure you can use which reliably gives you acceptable images most of the time (say 90 or 95% of the time). Use that exposure for your transform images (as long as that exposure time doesn't saturate any of your standards). Delete the few bad ones from the stacking list. Take as many images as you need to give a SNR of about 100 for the standard stars in the stacked image for each filter. I find that, for solar type stars, I need to stack 2 1/2 times more B images than V images to get about the same SNR (if I use the same exposure time for both filters).
Ultimately there can be a diminishing return when stacking lots of images, but I think usually the sky background is the limiting factor.
BTW, your transform plots looked fine. You've got the process for calculating the transforms down. The only problem was that the B-V range for your Landolt standards only goes to B-V of ~1.4 on the red end. This is also a problem for the M67 standards on the red end (at least it was the last time I looked). You can do better once NGC 7790 rolls around. I think there are several standard stars there with B-V = 1.7 or thereabouts. The problem with NGC 7790 is that the blue end is rather shallow.
In one of your previous posts you asked about what chips have color responses close to the standard system. I can't make a general statement, but I can tell you my own experience. Like you, I use Astrodon J-C filters. Both of my cameras give color transforms and magnitude transforms quite close to the standard system. One has the KAF 1603ME CCD, and the other uses the KAF 1001E.
Phil
Brian,
it sounds that your setup has moderate aperture (up to 80 mm?). If your exposure times are relatively short (is 240 sec in B actually 10x24 sec?) and airmass somewhat significant, you will experience scincillation effects. They hit hardest in blue filter(s). For "white light", where effect is smaller than in B filter, see e.g.: http://astro.corlan.net/gcx/scint.txt
Best wishes,
Tõnis
Good guess, actually 71mm.
1x240s in B.
So far I've tried to avoid stacking by taking different exposures in B and V such that the brightest comp star comes in just under saturation. For BZ Cyg that means 50s in V and 240s in B. Unfortunately I get quite a lot of star elongation at 240s due to tracking errors. Phil's recommendation is to use 50s for both B and V and then just stack more B-frames to raise the SNR. That is what I will do.
It is interesting that Phil should mention the ratio of 2-2.5 to 1 when stacking more B than V for Sun-like stars. I came up with the ratio of 2.01 to 1 a few days ago when I white-balanced my V and B filters against a G2V star. (I like to dabble in astrophotography but I only have a 5-filter wheel so I swapped out the wideband Green and Blue filters and replaced it with the photometric V and B filters. So instead of doing LRGB, I'll do LRVB! That is how I derived the 2.01 to 1 ratio of B to V.)
When doing AP I use different exposures for LRGB instead of a single exposure. I find that I can get a higher SNR in less time that way. Generally for my Bortle 5 skies my longest exposure is 61s in wideband Blue with 2x2 binning. That exposure is short enough that I get a high percentage of round stars with just PEC.
My modified plan is to image BZ Cyg in B and V using 50s exposure each and then stack twice the number of B's.
Also I will redo my transforms by imaging Landolt SA38 at 80s.
Thanks,
Brian
As someone whose workflow requires imaging of 4-6 Landolt standard fields in 3 filters each every session, I can offer a good word for Landolt fields SA38 (unusually bright), as well as SA32, SA35, SA41, and SA107, and SA108 all currently available (June) at least for a few hours each night from the Northern Hemisphere. Also, SA20 offers an especially rich field of ~ 20 bright stars; it is now available afer ~ 2 AM.
If you can observe that late, NGC 7790 (fainter) is now available after about 1 AM in the Northern mid-latitudes at present (June); it will rise ~ 2 hours earlier each month.
Hi Eric,
Please excuse my naivety but do you image Landolt fields for the purpose of calculating new transforms each session?
Thank you,
Brian
Hi Brian,
Good question, but no, I only get transforms every 6-12 months; they change very slowly, especially in cool and very dry climates like mine, which are kind to filter integrity.
I run Landolts every night (1) to determine my nightly zero-point magnitudes (relating my instrumental magnitudes to standard magnitudes), and (2) to check extinction in each filter/passband against typical local, seasonal values. Possibly overkill, but for several Northern Landolt fields the stars are pretty bright and quick to image, so the cost of getting this confidence is low.
Hi Phil,
A major improvement after following your suggestion! As you recall I do not actively guide so I rely on the accuracy of my polar alignment and PEC. You suggested that I keep the exposure time low in order to increase the probability of good frames. I followed that advice for Landolt SA38 for new transforms and for BZ Cyg which was 6 hours after maximum light.
I began the process with the V filter. With assistance from AstroImageJ I placed an aperture over the brightest comparison star in the field in order to plot the peak pixel value in realtime. I adjusted the exposure such that the majority of points fell in the 45,000 ADU to 55,000 ADU range. The exposure that worked best for SA38 was 60s and for BZ Cyg, it was 45s. For the B filter I used the same exposure. (Now with 20/20 hindsight I might want to experiment with the exposure and increase it by 25%. As I discovered later the SNR of the faintest comparison stars were under 100 even after stacking.)
I captured more frames than I needed for stacking since I knew that I needed to throw some away. In a spreadsheet I calculated the mean and standard deviation of the peak pixel values and then chose to throw away those that were more than one standard deviation from the mean. With respect to stacking I used 10 frames in V and 22 frames in B.
For the seven comparison stars in SA38 that were within the capability of my scope and camera the range of SNR was 128 to 255 in the V-stack and 66 to 132 in the B-stack.
For BZ Cyg the SNR range was 128 to 266 in the V-stack and 96 to 143 in the B-stack.
Finally I transformed the B and V stacks for BZ Cyg. Please see attachment. I want to shrink the error bars around the B-mag. If I understand the math properly the way to do that is to increase the SNR of the stars in the B-stack. The V-stack is fine in my opinion.
One closing note: my transform coefficients for SA38 using the method I described is a bit alarming: Tbv = 1.4046, Tb_bv = 0.2187, and Tv_bv = -0.0694. It will be interesting as I fill out more of the light curve.
I know now that I should invest in a new mount capable of active guiding. As I watched the realtime plot in AstroImageJ, every so often there would be a high quality frame -- the peak pixel value would shoot up. With active guiding those would be the majority!
Thanks,
Brian