Photometry woes - can anyone advise

[billyharris72]

Hi all:

I've been looking to try some variable star photometry (hence deciding to post in this section) and was wondering if someone could offer some advice, as I seem to have hit a brick wall before I have even started.

I recently upgraded my camera from a Nikon D3300 to a ZWO ASI1600MM, which I finally got the chance to try last night on M57 (imaging with a 130P-DS), mainly with the idea of experimenting with creating an LRGB image. When I was processing in AstroArt (including flats and darks) I decided to try some photometry, picking stars in a range with decent SNR but not saturated (I think, though I've not ruled that out). When I looked at the results (on LRGB, luminance and green images) they were all over the place, with errors ranging from about 0.1 to 0.6 magnitudes.

I appreciate that I'm using a CMOS camera, not a CCD and that I wasn't using a V filter, but the AAVSO manual estimates a margin of error for DSLR photometry can be as good as 0.01 magnitudes, and these results were miles out compared to that.

I've attached the output text file below, which shows the magnitude estimates and the residuals (E_mag) for the reference stars (the stars with an error of "1.0" are the ones I measured afterwards and have similar or worse errors). The jpg shows the stars I selected. For anyone who fancies taking a look, the fit file itself is at: 

http://nova.astrometry.net/user_images/1719235#original

I'd appreciate any advice anyone has to offer, as I don't want to give up but don't see where to go with this. Thoughts that come to mind are:

1) The stars are saturated (seems unlikely as all have relatively low ADU for the camera, maybe about 10-20% of full well).

2) SNR is not high enough (AA advises >100, which they all pass easily).

3) Under / over sampling

4) For some reason the camera is not suitable (though my Nikon had the same issues)

5) Without a V band filter this level of error is normal (though seems to contradict AAVSO)

6) Something wrong with my flat (attached)

7) Something wrong with my annulus (ooh err). Was set at 3.5 / 1 / 5; messing about with it does not seem to make a huge difference.

8) Something wrong with the catalogue (GSC) that I am using. For some reason, AstroArt displays star magnitudes with an error, typically as high as +- 0.4 for stars of this magnitude in this part of the image. Am I using an inappropriate catalogue, or choosing bad reference stars?

9) I'm doing something else wrong that I don't yet know about.

If any kind photometrists out there are feeling in a charitable mood I'd be glad for any pearls of wisdom you might cast in my direction.

Cheers,

Billy.

starsoutput.txt

[Dave Smith]

Hi Billy

I'm not sure that I can be of much help but as no one else has replied here is my two p.

I only started this year having previously spent my time imaging nebulae. You haven't said what software you are using but I was advised to give Muniwin (which is free) a try and have found it relatively easy to use and it also gives an output suitable for the BAAVSS database and also the aavso database. It seems to automatically reject any readings that are wrong in some way, e.g. on Sunday night I took nearly 400 readings which were rejected due to saturation. I did check that before doing the run but clearly not well enough. It has a nice routine for selecting the best aperture. If you are a member of the BAA they do have a mentor scheme so that you automatically have an expert on hand should it be needed.

I hope that this may be of a little help.

Dave

[billyharris72]

Hi Dave:

Thanks for the suggestion - I'll check that out. To date I've been using Astro Art, which seems to be regarded as reasonable for amateur photometry, but I'll give Muniwin a try (the price is very attractive!).

I haven't had another clear night to shoot any more images, so have not been able to try for an improvement, but I've identified a couple of areas where I made quite critical mistakes. Specifically:

1) The flat looked okay, as did the histogram. Turns out it was hugely stretched to look okay and AstroArt always centers the histogram (unlike my DSLR). The averge pixel value was about 2k (it saturates above 50k), so horribly underexposed.

2) In the past I've never shot flat darks (AA uses separate darks for images and flats instead of a bias frame) as the thermal noise on my DSLR was low for those exposures the bias seemed low. Turns out the ASI1600 chip can show banding in the bias, and that this changes between reboots of the camera.

So basically, I shot a flat frame that did nothing but pump a whole load of bias noise and banding into the image.

So my next step is to work on the flat frames by getting the exposure right and shooting flat darks; I'm also building a flat box (based on an A3 tracing light box) which should make the whole process easier and more reliable.

Fingers crossed...

Billy.

[dokeeffe]

Billy,

Another thing to check.. What is your image scale in arc seconds per pixel? The star's fwhm should be about 2 or 3 pixels. 

Derek

[robin_astro]

Hi Billy,

I do not do photometry regularly but I know some of the basics.

How did you chose your reference stars, where did you find their brightnesses and how are the values described (eg which filter) ?   Normally these are chosen from charts produced specifically for the star you are trying to measure eg BAA AAVSO etc where the reference stars have been checked for variability, the brightness measured to high accuracy and chosen to be similar in brightness and colour to the target, particularly important for unfiltered measurements, which can be very different from V mag values depending on the colour of the star.

Note 0.01mag absolute accuracy in magnitude  (~1%) is tough to achieve. Are you sure the 0.01 figure quoted for DSLR was not precision or uncertainty ie how repeatable the measurement is, rather than the absolute accuracy ?

Cheers

Robin

 

[billyharris72]

Thanks for the replies all, which have been very helpful. I've done a few things to try to improve the results, including checking that the FWHM is sensible (I'm over sampling a bit but nothing major) and using the AAVSO database to make sure I have the right check stars. In addition, I've built a flat box (using a tracing light box and translucent plastic sheet) and invested in a V band filter. I've also started manually checking Astro Art's (UCAC 4) map values against the AAVSO values manually (they don't always agree on the V-band values and my own modelling suggests that where they don't the AAVSO ones are the right ones). I'm now getting errors down to about 0.02 or 0.03 magnitudes, so a pretty big step in the reight direction, and probably plenty good for now.

Billy.

[dokeeffe]

Awesome! well done

[Darryl]

There are a multitude of variables that can affect photometric measurements. In particular, the colour response of your system compared with the catalogue system and air-mass effects can easily add a few percent uncertainty.

Also, the point spread function of your stars is likely to vary across the field of view, meaning that stars near the edge will measure slightly faint compared with those close to the centre for a given measurement aperture diameter. You can combat this by using large measurement apertures but that then risks adding uncertainty from background sources.

Systematics casued by imperfect flat-fielding can also play havoc, a problem i've had most significantly when using a 2.5m class professional telescope.

An excellent primer on photometry is "A Practical Guide to Lightcurve Photometry and Analysis" by Brian Warner: http://www.springer.com/us/book/9780387333915. Well worth the $20 investment.

Cheers,

Darryl

 

[Darryl]

I've just had a look at the plate fit for your first post, 2.65 arcsec per pixel is pretty large to get decent photometry. I'm not sure what the seeing was at the time but i'd always aim for at least 2 pixels per FWHM. Less than this and pixel to pixel sensitivity variations/ flat field errors are really problematic especially as moving a filter wheel or any other element in the optical train (or a bit of flex in the system as you change elevation) between taking flats and lights will cause small offsets in the flat.

I've done a fair bit of photometry both with amateur and professional scopes so have some experience with chasing down the 'last few percent'. It can be both fun and frustrating all at the same time! 

It's also worth remembering that a significant fraction of stars do genuinely vary. Variables would have been weeded out if comparing your photometry with a 'standard' such as a Landolt or Henden field, however out in the 'wild' i'd guess that perhaps 5-10% of 'field' stars might either vary intrinsically or due to binarity or spots. This fraction would increase further in 'young' star fields such as in Orion or Taurus.

Cheers,

Darryl

[billyharris72]

Thanks for the input Darryl; these are good points. To be honest my problem when I posted that was unrealistic expectations as to what was required for photometry. I basically tried to use a (binned, hence image scale) test image of m57 to "try out" photometry (selecting unsaturated stars at least). Since learned it's a bit more complicated than  that...

Funnily enough I actually have recently bought that book and looks pretty good.

Billy.