The Need for a Large Number of DSLR BIAS Subs

[jimwaters]

I have seen several long posts on Cloudy Nights about people taking 200 plus BIAS subs and making a Master BIAS.  There’s even an article on DSLR Astrophotography saying that the more the better – 150 to 250.  I think some of these is based on BIAS noise for the Nikon D5100 camera.  IMHO this sounds like overkill and should be considered based on the camera model.  I use a Canon 6D and I stretched one BIAS sub almost 100% and can see very little background noise.  My 7DMkII is even better.  I did the same for my Ha modified T3i and yes there was noise.  I created a Master BIAS for the 6D and T3i cameras.  For the 6D I used 49 subs and 99 for the T3i.  For me Darks and Flats are more important especially Darks for the T3i.

What am I getting wrong here?  BTW - I have also seen posts saying that an 'odd number' of subs are better when stacking.  Any idea why?

[pfile]

you are probably right about dependence on camera model - some canon models have really bad banding patterns in the bias which need lots of subs to resolve. others don't.

dunno about odd numbers but maybe people are thinking that for median stacking modes you need a single frame which is the median value itself. after all with even numbers of subs there's no sub exactly in the middle. still i can't see that it matters, for sets with even numbers of elements the median is still defined - the mean of the two central values. and as the documentation for ImageIntegration clearly shows, median stacking methods are only 80% efficient compared to averaging stacking methods, and so there's no good reason to use median stacking, given the plethora of pixel rejection methods that PI provides for average stacking methods.

rob

[jkmorse]

I like lots of subs for any camera or CCD.  I typically shoot 100 for a very clean 6303E chip.  The key is that you only need to shoot new bias frames once every few months, not with each image set.  I tend to shoot a new dark and bias library every quarter.  Since you are only shooting bias frames every few months and they take very little time to shoot, why not go for more rather than less?

For what it's worth,

Jim

[msmythers]

I shoot a lot of darks, flats and bias,75,200,200. I have a different situation then most since I shoot from a very hot, humid climate with very bad LP. The night before last at 5AM it was 73 degrees. The average December/January night temperature is 65, not much of a difference. I shoot with an old Nex-5 and have to shoot short exposures, 2 minutes and under. My thought was the more cal files the less noise introduced from those cal files. Through testing different numbers of calibration files I found my images were a slight bit less noisy with more files. Is 200 a magic number for me, no. It's just an easy number to count vs say 215 or 175. That's just my setup. If I was capturing with cooler temps with better SNR then I would hope to reduce those numbers.

With my camera I can do 200 flats in 5 or 6 minutes. Bias in under 3 minutes. Darks a little more then an hour to 2 and half hours depending on exposure time.

On the good side for me is I can take darks inside my house during the day since the inside and outside temps are so close. I also don't have to take them all at the same time, one night is the same as the next for 85 percent of the year. The bad thing is I can take darks during the day since my temps are so close.




Mike

[mschuster]

Just playing around:

If sky limited, noise in the sky is three times read noise (or more), so in quadrature read noise adds 5% (at most) to total noise in a light. This penalty remains the same in the integration.

Suppose you use the same number of calibration frames as light frames (eg. 10 lights, 10 bias, and 10 darks). The plot below shows the quadrature combination of read noise in the integration plus noise in the calibration frames adds about 15% to total noise. Or about 11% if you do a dark only calibration.

Doing 4x the number of lights (eg. 10 lights, 40 bias, and 40 darks), brings the penalty down to 8% (7% dark only). Which is only a couple percent larger than the unavoidable 5% read noise penalty.

So it may not help much to use more than 4x. Improvement will be small.

Key point: the more lights integrated, the more cal frames needed to keep cal noise in check.

Thanks,
Mike

[jkmorse]

Mike,

I like the 4x guideline.  It makes a lot of sense but that still puts folks in the hundreds of cal subs if they are shooting the typical 30+ lights. 

I would be curious to get your thoughts on Flats, however, since those are definitely limited by available sky for those of us shooting sky flats.  I have excellent results from only 5 flat subs to build my masters.

Best,

Jim

[mschuster]

Jim,

Yes, if you capture just at sky limit. One option to reduce # cals is to expose longer so read and cal noise are even further buried by the larger sky noise.

On flats, I am not sure. At least two issues, their "flatness" and their SNR.

My panel flats are not truly "flat" (spec is 1%, that might be optimistic), but on my Ha neb fields it is not really possible to model a gradient, so I don't DBE.

Flat SNR shouldn't limit target SNR, but I think target SNR's more than say a couple hundred are not really necessary. So pick flat SNR say 400. That requires 160,000 e-. My camera full-well is 25,000 e-, expose at ~50%, say 10,000 e- per flat frame. So 16 frames are needed.

Thanks,
Mike

[jkmorse]

Mike,

Again, good analysis and it gives me a lot of comfort.  My 6303E has a full well capacity of 100k, I expose about 40% and shoot five sky flats which puts me around 200,000 e- which should be plenty to get me where I want.

Thanks,

Jim