No, in that way you’re mapping 20K levels to just 4096, so you’re not taking full advantage of the capabilities of the sensor. Or, quoting Jon:
Regarding your second objection:
Nobody here is underestimating the importance of dynamic range, for sure I’m not. But I think it is safe to say that dynamic range is much more important for LRGB than it is for narrow band imaging, as I mentioned above (“LRGB exposure in light polluted skys can actually benefit from low gain values”).
On the other hand, you can’t get rid of FPN noise by increasing the number of exposures, because by definition it is not a random noise. Dithering may help, as long as FPN is actually “fixed” (some sensors are affected by a sort of pattern noise that may shift slightly from an exposure to the next).
As the OP, this is actually the discussion I wanted here, how to properly expose these cooled cmos astro cameras, in hopes that we get more aids in that in SGP.
The latest versions of SharpCap Pro have some features that evaluate sensors and creates some of the graphs discussed here, and then goes on with a “brain” tool and a “smart histogram” to help you with proper exposure. (given what you think your total integration time will be) I’d like to see some version of that or other approaches here.
For those who doubt the importance of dynamic range I did a little experiment. I started with 12 bit Ha, SII, OIII stacked channels. I made a copy of each channel reducing them to 8 bits. I then processed each set the same way in Pixinsight - channel combination then a default STF stretch. I then used Photoshop to reduce the image size by 1/4. The choice is yours, but I’m sticking with 12 bits.
While not having the ability to understand all the discussion on here regarding gain and offset settings, I usually just use unity gain for all imaging/filters.
Could someone, in simple terms, tell me what length of exposures would it be best to use for each of the following settings: gain 75 offset 12, gain 139 offset 21 and gain 200 offset 50. And also which to use for LRGB and NB filters?
Based on these discussions, next time I image I’m going to try a gain of 60 offset 35. Why? It’s the sweet spot between most dynamic range and lower noise. With my setup, I expect to do 8 minute exposures. Try to use an exposure that gets the brightest stars almost to saturation (or perhaps a few saturated pixels) - that will give you the best dynamic range and the least noise from the camera. If you don’t get close to saturation, you are discarding dynamic range on the high end. Your exposure will depend on your aperture and focal length.
I think that exposures based on SQM are irrelevant. Changing the exposure doesn’t fix light pollution problems.
My takeaway is still that you should base exposure based on the low end, to make sure you above any of the noise inputs, so I think this table or similar is still relevant (asi1600)
“Median ADU shown in SGP:
Gain 0 Offset 10: 400 ADU
Gain 75 Offset 12: 550 ADU
Gain 139 Offset 21: 850 ADU
Gain 200 Offset 50: 1690 ADU
Gain 300 Offset 50 : 2650 ADU”
and then you either get away with no clipped stars (LRGB or OSC in my case) or they still clip (NB in my case) and you have to deal with it. Tone mapping and replace with RGB stars or separate short star exposures .
Re the sweet spot, I’m not sure exactly where it is as the DR graph isn’t in discrete steps. Here’s my asi1600mm-cooled sensor curve from ShapCap Pro:
The stuff in green are data points I tried to extrapolate.
In the graph Blue is Gain and Orange is DR.
So, is the “sweet spot” 60 or 75?
Right now I’m doing Gain 0 offset 10 for LRGB or OSC and gain 75 offset 12 for NB.
@alcol Do a Frame & Focus and check the image histogram. I would recommend doing a 10 second exposure with a bright light to find the max ADU you are getting, however, you have to be aware that there are likely some pixels that are stuck on and will mess up the max value in the statistics. If you look at the the non-stretched histogram you can get an idea of the max.
@dts350z Where are you getting the recommend offsets? When I tested my camera at 60 gain, I had to set the offset at 35 to get the minimum values off the floor. That is, at zero offset the minimum ADU was 16 and I had to get to 35 offset to see a minimum above 16.
I believe the chart I’ve posted in this thread, of Median ADU target vs. gain and offset, came from Jon Rista.
There are others, with slightly different math, that also came from cloudynights threads on asi1600 exposure.
What I would like to get to is the ability to use short exposures to predict the correct long exposure.
With DSLRs I teach people to do test exposures at the highest ISO the camera offers, say 12800, then multiply the good exposure results 2x for each iso stop. e.g. if you are going to do your subs at iso1600, multiply your good histogram iso 12800 exposure time by 8 (2x2x2).
That way you don’t spend 8 minutes finding out that you need a little longer exposure, and just guessing at how much longer.
Problem is, despite lots of effort and spread sheets, I can’t figure out how to do the same with an asi1600, I’ve haven’t tried since I’ve had the e-/ADU data, but still I think the different offsets maybe messing me up. Hitting >= to a median ADU seems a lot harder than just having a gap on the left of the histogram.
And, I have an FR in here for “auto exposure” because it also seems that median ADU for a given exposure changes with target altitude etc.
I hate to sound like a broken record but what you want is to have the brightest parts of the image just below saturation. Then dimmer parts of the image can take advantage of the lower bits. Ignore the median ADU - it will depend on the target. If you pay attention to the bright parts of the image then you should have more success scaling.
I’m saying keep the stars just below clipping (or some people advocate clipping a little bit). The image I posted earlier is a DSO and the brightest stars are a little below clipping so I could have exposed it a slight bit more. As you can see the DSO looks just fine. This nebulosity is not even as bright as a galaxy would be. If you over clip the stars they will start blooming.
By which I mean gain zero has twice the full well (a full stop) of the 60-79 gain “sweet spot”. or is every e- value in that range the same ADU, because of the 12bit DAC?
The fact that the full well is half at the sweet spot is just an indication of the doubled sensitivity at that point. The dynamic range is what counts and it almost equal at zero and the sweet spot.
For gain 0 through 61ish that is above 12, but is limited to 12 by the ADC being 12bits.
So what is actually driving the loss of Dynamic range as the gain goes up is the reduction in the full well values. So why does the full well value go down with increasing gain?
