Robin Glover on gain, exposure time, e/ADU and read noise curves

[SkyHiker]

One of the most frequently asked questions is, what exposure time and gain to choose. Robin Glover (the author of SharpCap) gave two excellent presentations on this topic:





To me this is the most definitive answer that I know of. Enjoy the presentations!

[SkyHiker]

I did an analysis of the gain and dynamic range curves of my ASI2600MC. It has a dual conversion sensor that has a pre-amp that has a low conversion gain of 1 below 10 dB and a high conversion gain of about 3.3 above 10 dB. I derived a model for this and fit the parameters to the curves in the ASI200 manual. You can see a nice jump down in the read noise at 10 dB (and a jump up in the dynamic range).

Then I applied Glover's formula to find the optimum minimal exposure time for a few cases, at 0, 9.99 and 10.01 dB, for Bortle 5.8 (where I live) and 1.0 (Grand Canyon or so).

What Glover does not cover is when the stars get blown out. I did some math for that approximating the photon flux for stars as about 30 times that as the light pollution where I'm at, Bortle 5.6, based on a histogram of a HorseHead Autosave.tif that was lying around. So, a bit of handwaving with the numbers but not total nonsense. I calculated if the stars got blown out for the same 6 cases as the exposure length calculation, accounting for the different well depth at those gains. There was one case where the stars get blown out.

At any rate, for preventing star blowout it's best to take subs at 0 and 10.01 dB (0 and 100.1 in ZWO's 0.1 dB units). That is also best for minimizing the sub exposure length.

It's a bit nerdy but I got a better understanding of how my camera works and what settings should be best. I attached a little PPT converted to PDF about what I described that I just presented to my local astro club Tuesday night Zoom gathering. Enjoy and don't get a headache.

exposure.pdf

(1022.01 KiB) Downloaded 34 times

[Juno16]

Thanks Henk.

Interesting stuff. I really need to give this another try. I looked at it awhile back and found for my sky brightness level, my exposure times would be very short. Couldn't do it!
I like to save all of my lights and dealing with tons of subs and tons of data to store (not to mention processing times in PI WBPP) didn't sound like what I wanted to do. Granted, I do have lots of room to improve on my images!

Of course, I very possibly didn't run the calculations correctly and that is why my exposure times were so short. Have you calculated your optimal exposure time yet? Sorry if I missed it in your post.

[SkyHiker]

Of course, I very possibly didn't run the calculations correctly and that is why my exposure times were so short. Have you calculated your optimal exposure time yet? Sorry if I missed it in your post.

No problem. To me the benefit was to get a better understanding of my camera (and the performance graphs in the ZWO manual), and the whole picture of stacking in function of gain and exposure time. Making the model of the ASI2600 gave me a good understanding of the camera and what the ZWO graphs mean.

Glover solves the minimum exposure choice. For the Bortle levels that most of us have, the minimum exposures are remarkably short. Nobody would pick those short times because you would have to stack a lot more subs than you like. So, in practice, that lower limit is more or less irrelevant. It is nice to know that taking shorter subs is perfectly OK, and that the need for 10-, 20- to 30-minute-long subs for the best results is a myth.

A bigger problem is saturation of the stars, which can easily happen with multiple-minute exposures especially if the gain is on the high side. This is nothing new. The math that I did should give some guidelines. Or you can just take a sub of that length and check if the stars are saturated so you can forego the math altogether.

The math that I did for some examples was for RGB. If you use filters the minimum exposure lengths scale up by a factor discussed by Glover. The saturation times scale up accordingly as well, so the numbers become larger, but staying clear from the lower and upper limits should not be harder.

[Juno16]

Of course, I very possibly didn't run the calculations correctly and that is why my exposure times were so short. Have you calculated your optimal exposure time yet? Sorry if I missed it in your post.

No problem. To me the benefit was to get a better understanding of my camera (and the performance graphs in the ZWO manual), and the whole picture of stacking in function of gain and exposure time. Making the model of the ASI2600 gave me a good understanding of the camera and what the ZWO graphs mean.

Glover solves the minimum exposure choice. For the Bortle levels that most of us have, the minimum exposures are remarkably short. Nobody would pick those short times because you would have to stack a lot more subs than you like. So, in practice, that lower limit is more or less irrelevant. It is nice to know that taking shorter subs is perfectly OK, and that the need for 10-, 20- to 30-minute-long subs for the best results is a myth.

A bigger problem is saturation of the stars, which can easily happen with multiple-minute exposures especially if the gain is on the high side. This is nothing new. The math that I did should give some guidelines. Or you can just take a sub of that length and check if the stars are saturated so you can forego the math altogether.

The math that I did for some examples was for RGB. If you use filters the minimum exposure lengths scale up by a factor discussed by Glover. The saturation times scale up accordingly as well, so the numbers become larger, but staying clear from the lower and upper limits should not be harder.

Got you there Henk.

Absolutely the stars are the biggest issue. I definitely could use improvement there.

Are you going to try using his guidelines for exposure time? What time will you try for your location?
My sky is bright, but I really don't want to go below 30 second exposure times. Actually, I haven't gone below 60 seconds.

Very much looking forward to your results!

[SkyHiker]

Absolutely the stars are the biggest issue. I definitely could use improvement there.

Are you going to try using his guidelines for exposure time? What time will you try for your location?
My sky is bright, but I really don't want to go below 30 second exposure times. Actually, I haven't gone below 60 seconds.

If we agree that the star saturation is the biggest problem, then we focus on the gains where we have the most headroom.
Your 533 sensor is actually very similar to my 2600 in terms of read noise and dynamic range, it is just 14 bits instead of 16.
So, the e-/ADU gain graph is a factor 4 higher with the 533 because of those 2 bits.

The places where we have the most headroom is at 0 dB and right after 10 dB (100 in ZWO units).
Looking at the last slide for Bortle 5.8, the ADC value relative to the full well at 0 dB is 3467/52428=0.066, whereas at 10.01 dB it is 587/16580=0.035. That means that there is more headroom at 10.01 dB.

Keep in mind that these numbers are based on a ratio estimate of stars vs. LP that is based on one of my images. That ratio depends entirely on how many pixels are covered by a star. My rig has a larger focal length, the collimation is flopping around like crazy, the G11 is struggling with the 65 lbs., so I bet your stars are much smaller and that ratio will be higher in your case. So, adjust it based on what you see in the histogram.

Anyway, continuing the line of thought, we stay at either 0 dB or 10.01 dB and increase the sub exposure to the point where the stars start clipping. In our case, the 10.01 dB point looks best to start with.

That's what I can come up with so far, I hope it makes sense because I am making it up as I go.

[Juno16]

Absolutely the stars are the biggest issue. I definitely could use improvement there.

Are you going to try using his guidelines for exposure time? What time will you try for your location?
My sky is bright, but I really don't want to go below 30 second exposure times. Actually, I haven't gone below 60 seconds.

If we agree that the star saturation is the biggest problem, then we focus on the gains where we have the most headroom.
Your 533 sensor is actually very similar to my 2600 in terms of read noise and dynamic range, it is just 14 bits instead of 16.
So, the e-/ADU gain graph is a factor 4 higher with the 533 because of those 2 bits.

The places where we have the most headroom is at 0 dB and right after 10 dB (100 in ZWO units).
Looking at the last slide for Bortle 5.8, the ADC value relative to the full well at 0 dB is 3467/52428=0.066, whereas at 10.01 dB it is 587/16580=0.035. That means that there is more headroom at 10.01 dB.

Keep in mind that these numbers are based on a ratio estimate of stars vs. LP that is based on one of my images. That ratio depends entirely on how many pixels are covered by a star. My rig has a larger focal length, the collimation is flopping around like crazy, the G11 is struggling with the 65 lbs., so I bet your stars are much smaller and that ratio will be higher in your case. So, adjust it based on what you see in the histogram.

Anyway, continuing the line of thought, we stay at either 0 dB or 10.01 dB and increase the sub exposure to the point where the stars start clipping. In our case, the 10.01 dB point looks best to start with.

That's what I can come up with so far, I hope it makes sense because I am making it up as I go.

Making sense Henk.

I always image at a gain of 104. I guess that I will try reducing exposure time (for broadband) to 30 seconds the next time I'm out. I really don't want to process that many files (or store them), but my file size is much smaller than yours due to my smaller sensor.
Wouldn't hurt to try it out once anyway, just to see.

I just looked in PI at my Owl Cluster image stars. The brightest stars are still >1 (looking at the linear masterlight with the preview readout). They are close, but it wouldn't hurt to get the cores in the 0.7-0.8 range.

Thanks for the idea Henk!