Flats Panels & ADU

[Jkulin]

@angryowlposted sometime ago (Thanks) about these cheap flat panels, previously I had been using expensive EL Panels and I bought these to put aside for when my EL Panels finally went.

They were dirt cheap at the time and have risen back to their normal price, however I note that they now list a 10w version that might be even more useful: - http://www.beamled.com/biard-10w-panel-light-300-x-300mm-led.html

Before Christmas I ordered some Neutral density Lighting Gels (Sheets of Neutral density film used in the theatrical industry), specifically https://www.sblite.co.uk/neutral-density/265-rosco-299-12-neutral-density.html

In case I needed clearer I also ordered some 0.9 as well https://www.sblite.co.uk/neutral-density/264-rosco-211-9-neutral-density.html

So I did my first full session on NB over the 6th and 7th and last night I set everything up for my flats using these panels and gels.

Interesting APT has a CCD Flats Aid that will create flat plans according to what ADU you require the default suggestion being 20,000, this turned out to be really useful and enlightening because as a photographer of many years I had always worked on aiming to get the histogram into the middle or to the right as that gave you the best flexibility.

I was surprised at the results of the plan, I've attached images of the unstretched flats at 20K and 30K and the histograms for each using two sheets of the 1.2 ND Gels: -

This lead me to do some more research and the conclusions seems to be that 20K is the ideal ADU for flat panels for use in Astronomy, there was some suggestion that your flats should be representative of the dark sky, not sure about that, but it was furthermore suggested that using a higher ADU would create considerably more noise.

Now I am no scientist, just an average joe blogs who just so happens to like getting things right.

I wondered what your opinions were on here, does this seem correct.

[vlaiv]

Two things are important with flats (not entirely true, you need a bunch more important things for good flats, but none are related to ADU):

1. Linearity of sensor

2. Best SNR for flats that you can get.

For first item you need to know how linear your sensor is, and if you can find technical specs on it, then great, but with modern sensors you can almost be sure you will be in linear region if you steer of far ends of histogram, so avoid top and bottom 5% of histogram.

For second point, there is simple rule: get your histogram to far right without going into non linear region (if there is one), and also avoid clipping, and take as many flats as you can (is feasible in terms of time and processing).

[Jkulin]

Thanks Vlaiv, so not being anywhere upto your skills or knowledge, can you advise if using my Atik 383 is reporting the correct ADU, the specs are: -

Sensor Type: KAF-8300 17.6mm x 13.52m
Horizontal Resolution: 3354 pixels
Vertical Resolution: 2529 pixels
Pixel Size: 5.40µm x 5.40µm
ADC: 16-bit
Readout Noise: Typical 7e-
Gain Factor: 0.41e-/ ADU
Full Well: ~26,000 e-
Dark Current: ~0.1 electrons/second at -10°

I typically take 20 of each filter, I can do more, but am advised that this was the ideal.

When I take my OTA of the Pier I place it carefully on my dining room table until I am satisfied that the Flats are good, so there is no problem running it up to produce more if need be as my wife woudln't dare move it :-)

[michael.h.f.wilkinson]

11 minutes ago, vlaiv said:

Two things are important with flats (not entirely true, you need a bunch more important things for good flats, but none are related to ADU):

1. Linearity of sensor

2. Best SNR for flats that you can get.

For first item you need to know how linear your sensor is, and if you can find technical specs on it, then great, but with modern sensors you can almost be sure you will be in linear region if you steer of far ends of histogram, so avoid top and bottom 5% of histogram.

For second point, there is simple rule: get your histogram to far right without going into non linear region (if there is one), and also avoid clipping, and take as many flats as you can (is feasible in terms of time and processing).

Agreed, the histogram should be to the right, without entering into clipping. The reason for this is that you want a high photon (or electron) count does increase the noise, but the noise only rises as the square root of the number of photons (or electrons) captured, whereas the signal increases linearly. Thus the all-important signal-to-noise ratio increases as the square root of the number of photons captured. S/N also increases with the square root of the number of flats combined for essentially the same reason. use 100 flats and the S/N in the master flat will be ten times better than just using a single

[Jkulin]

So looking at the flats I showed at 20K & 30K ADU it would appear impossible to get the histogram to the right without producing 100K ADU, how can that be achieved with the Atik 383?

[vlaiv]

5 minutes ago, Jkulin said:

Thanks Vlaiv, so not being anywhere upto your skills or knowledge, can you advise if using my Atik 383 is reporting the correct ADU, the specs are: -

Sensor Type: KAF-8300 17.6mm x 13.52m
Horizontal Resolution: 3354 pixels
Vertical Resolution: 2529 pixels
Pixel Size: 5.40µm x 5.40µm
ADC: 16-bit
Readout Noise: Typical 7e-
Gain Factor: 0.41e-/ ADU
Full Well: ~26,000 e-
Dark Current: ~0.1 electrons/second at -10°

I typically take 20 of each filter, I can do more, but am advised that this was the ideal.

When I take my OTA of the Pier I place it carefully on my dining room table until I am satisfied that the Flats are good, so there is no problem running it up to produce more if need be as my wife woudln't dare move it :-)

Well quick search on internet gives this graph for linearity of Atik 383:

So it is pretty linear over whole range. If you pay attention to this graph you will notice that there is slight deviation from ideal line in central region so probably best to avoid this.

On RHS there is pixel value, so 20000 is in nice linear but bottom region. I would advise you to aim for either 60000 ADU count for mean/median ADU value of flat, or keep it on 20000 if you have trouble with flats at 60000.

Just for reference I do take a lot of flats, mainly because camera I use has 12bit so my ADU count for flats is in region of 3500-4000. I use 256 flats - that improves SNR over single flat 16 times.

[michael.h.f.wilkinson]

1 minute ago, Jkulin said:

So looking at the flats I showed at 20K & 30K ADU it would appear impossible to get the histogram to the right without producing 100K ADU, how can that be achieved with the Atik 383?

Not sure: the zero point of the two graphs is different. The maximum is at 64K. It might be that the software is scaling the histogram between minimum and maximum, and that there are a few hot pixels yielding 64K

[vlaiv]

3 minutes ago, Jkulin said:

So looking at the flats I showed at 20K & 30K ADU it would appear impossible to get the histogram to the right without producing 100K ADU, how can that be achieved with the Atik 383?

You can't really get to 100K ADU mark with that camera. Top value that you can have is somewhere around 65000ADU (65535 to be precise if we just look at 16 bits that can be used to record ADU value, and also Full Well / Gain factor will get you - 26000 / 0.41 = ~63414)

You can control ADU level of flat in two ways - either have adjustable brightness flat panel, or more simply change exposure length of flat frame

[Jkulin]

Thanks Michael and Vlaiv, the 100K ADU was an exaggeration :-)

Indeed I can remove one of the Neutral Density Gels and that would give me a considerably higher ADU, but wouldn't that defeat the object? I'll run some off tonight with one of the sheets removed and post the results tomorrow.

[vlaiv]

Just now, Jkulin said:

Thanks Michael and Vlaiv, the 100K ADU was an exaggeration :-)

Indeed I can remove one of the Neutral Density Gels and that would give me a considerably higher ADU, but wouldn't that defeat the object? I'll run some off tonight with one of the sheets removed and post the results tomorrow.

No need to do that if it is too much trouble. Flats taken at 20000 ADU will work, and will work well provided you take plenty of them.

Difference in SNR in single flat frame between 20000 and 60000 ADU flat is something like x1.7 - just take x60 of 20000 ADU flats instead of x20 and it will be almost the same as if you've taken x20 of 60000 ADU flats. Or take x128 and it will be better than both of those two combinations

[Jkulin]

Thanks Vlaiv that makes sense as theoretically the way I do my flats can be done at anytime as long as the focus and rotation isn't changed, so will try and capture 60 of each NB filter this evening and then see what the results are like.

[ollypenrice]

Be careful on a shutter camera to make your flats long enough to avoid the wipe of the shutter introducing a gradient. Also don't forget to use a master bias as a dark-for-flats. (You can do dedicated darks for flats but it is pointless, statistically.

I've tried taking twenty flats and taking eighty. I could see no real world difference.

Olly

[Jkulin]

Thanks Olly, yep the 383 can catch you out with the shutter.

Master and Super Biases were created when I got the camera in August and work fine, will update them every six months.

I use PI for my processing.