DSLR Photometry results

[7170]

I've been trying to do some photometry with my DSLR with a view to using it on variable stars as a quick "grab and go" solution when there is a gap in clouds etc.

Rather than looking at variables I decided to start off with comparing some fixed stars to identify how well I can estimate magnitudes with my DSLR, and the table above shows the results.

Using only two stacked frames (5s, ISO 800, F3.5 on my 5DMk2) with no darks, light polluted london sky, and just one comparison star I seem to be able to get to around 1% variance for the majority of the readings, with the worse being 2.54% out. Trying Chi Cas, against Upsilon1, Upsilon2 and omegaAnd comes up with M4.67 which is mag0.03 different (0.55%).

I'll be honest I am surprised at the results as it is not all that far off the 0.01-0.02 mag range often quoted for looking at exoplanet transits for example. Has anyone else tried this exercise as i'm interested to know how these results stack up - good or bad. The only thing I know for sure is I couldn't get it that close visually using my eyes!  

[billyharris72]

These results look good excet for a couple of oddballs. I tried this myself a while ago and really struggled. I was using a regression based approach to estimating magnitudes and was generally getting model residuals in the 0.01 - 0.03 range, which I thought was alright. What I struggled with was:

1. Occasional big outliers in the residuals (up to about 0.1 or so). I dealt with these by excluding them from the model.

2. When I compared my estimated values for the variables I was looking at with those in the BAA variable star database there wasn't that good a match up - in some cases I was up to 0.3 of a magnitude out, which is pretty drastic. While that was comparing mostly with visual estimates there are some very exprienced and capable people on there and I'd trust their data more than mine.

Not sure where I was going wrong, espcially as the models themselves generally looked pretty good. I'd be keen to know how you get on with this as it's still an area of interest.

Billy.

[Captain Scarlet]

Nice, interesting. Are you adjusting for the partial extinction "magnitude reduction" due to the increased air mass away from zenith the light has to travel through? I have found a formula for it and apply it in my "magnitude filtering" feature in my homemade star-map spreadsheet. It suggests an appreciable difference, extremely so as you get closer to the horizon. It might help reduce some of the differences.

Cheers, Magnus

[7170]

Well, last night there was a break in the clouds and I managed to get some more images of Cassiopeia. The table below is using the same settings as before but this time with a stack of 10 for lights, darks and bias.

With the exception of a few the magnitude variance is smaller. All of the ones with a large variance are in general not at the center of the FOV, are close to/at saturation or are not close together. The wide FOV means flats are probably required and this could account for some of the outliers.

@Captain Magenta, no I have made no adjustments for the position in the sky as Cassiopeia was directly above and all the star comparisons are using the one stacked frame. 

Another possible source of error could be a result of the colour of the star and the sensitivity of the CMOS to each colour. All I've done is take an average of the three colours and done comparisons on that, and a star with an uneven colour distribution could therefore be out. 

[Dave Smith]

This is an excellent piece of work. How are you deducing the magnitude?

I can strongly recommend the free software AstroImageJ

It is important to use flats otherwise any vignetting or dust bunnies will badly effect results. There is much help and advice on the BAA variable star section web page http://www.britastro.org/vss/ and also suggestions for target variable stars.

Dave

[billyharris72]

19 hours ago, 7170 said:

I have made no adjustments for the position in the sky as Cassiopeia was directly above and all the star comparisons are using the one stacked frame. 

Indeed, think you can probably get away without correcting provided the stars are close enough together. With differential photometry it's only the differences in extinction that matter and if the stars are close this should be small.

That said, I've never been sure just what size field is regarded as small enough for extinction not to be an issue.

[billyharris72]

19 hours ago, 7170 said:

Another possible source of error could be a result of the colour of the star and the sensitivity of the CMOS to each colour. All I've done is take an average of the three colours and done comparisons on that, and a star with an uneven colour distribution could therefore be out. 

Interestong that the values are still close despite this. Might be worth trying with just the green channel and seeing how the results compare?

Billy.