Auto Exposure


I was just reading the FR post on Sky Flats, and it reminded me of something I was thinking about the other night; “Auto Exposure”.

Some background:

I used to live at a darker location and imaged with a DSLR. exposures were always 8 min and always gave good histograms.

Similar with a debayered or monochrome DSLR and narrowband, in that exposure times were consistent over the night and between targets (however I acknowledge that a small gap to the left of the histogram was the only exposure criteria).

Now I have cooled cmos astro cameras, both mono and OSC ZWO ASI1600s. I got the mono camera just before moving from the fairly dark site and was super happy with early Nebulae results at “minimum read noise” gain and 300 second exposure (I was clipping stars however).

Then I started reading the endless threads on short exposures with that camera and started messing with gain, offset, and exposure settings to match various guidelines posted, and, to quote from the above mentioned Sky Flats post, “it’s like trying to pin jelly to a wall”, especially at my new light polluted location.

Yes now I am aware of both a minimum or median ADU target AND the desire not to clip or saturate stars, but the needed exposure seems to be very inconsistent over the night and/or between targets, moon age, etc. Even to the point where I can decide on a setting and then find it is too fast between finishing all the per filter test exposures and running the sequence!

So, the FR, similar to auto focus, could we not have auto exposure? Input criteria would be median ADU and maximum ADU. Figure it out per filter ahead of the run, and adjust periodically (like focus) as the sequence runs.

And, hey, it would be handy for Sky Flats too (median ADU could set at 45000 for example).

Bonus points a way to transfer the auto exposure times to a darks sequence.



Surprised no one has commented on this.

SharpCap 3.1 has something along this line (although weather has prevented full testing on my part).

There is a manual routine to evaluate your camera at different gain and offset per exposure (light and dark), Then there is a “smart histogram” that tells you where to expose, along with a “brain” wizard for setting exposure.

requires the native, vs. ascom driver, and it seems offset is missing in newer ZWO native drivers (but I digress)?


I use CCD cameras at the moment but I am looking seriously at CMOS ones for a second rig. I can imagine that with high gain, to achieve low read noise, you are at the mercy of changing conditions. Not sure how clipped stars work in the overall scheme of things but it seems that there is too much emphasis on read noise and not enough on dynamic range. The trouble with autoexposure as a concept is what is considered a good exposure? Bright stars will always clip - so if you had an algorithm to avoid clipping, you might end up with 10s exposures. This kind of judgement is not something I would want to leave to a computer. Where is the fun in that?


I have asi1600’s both mono and color. So far I can find settings where the low end is satisfied, according to recomendations for median ADU (as shown in SGP) for a given gain and offset, AND the stars are not clipped, so I would think that would be a “good exposure”, but yes, that seems to be non-deterministic and changes with altitude, etc. as I described in the top post, so, I would appreciate a little computer help (especially while I sleep) as I need hours and hours of data due to light pollution.

I’ve recently learned that part of the problem (exposure prediction) is the camera response, at least in part, is non-linear. I think that’s part of what SharpCap is doing in evaluating the sensor, so it can use and actual plot of response, vs. a single number for gain vs. response.

Here is a cut and paste from that SharpCap sensor evaluation (this is for my zwo asi1600mc-cool)

Gain Value e/ADU Read Noise (e) Full Well (e) Relative Gain Rel. Gain (db) Dynamic Range (Stops)
0 4.96856316964722 3.45519613081167 20351.234742875 1 0 12
60 2.64874235344584 2.59538382394716 10849.2486797142 1.87581973127112 5.46382199561611 12
61 2.61638761309931 2.58246741100859 10716.7236632548 1.89901647017873 5.57057462782852 12
100 1.68388746691009 2.14275057985103 6897.20306446371 2.95065036546918 9.39835502467633 11.6523317998627
200 0.52601883179064 1.75454856053239 2154.57313501446 9.44559941463228 19.5045904663471 10.2620864772308
300 0.163351614932051 1.59244359197703 669.088214761681 30.416370059849 29.6621476676846 8.7148103529752
400 0.0520305586294723 1.52358458420632 213.117168146319 95.4931736372482 39.5994465400233 7.12803341146884
500 0.0162164889326697 1.08361219082992 66.4227386682153 306.389576083732 49.7254797148821 5.93775677322606

and here is a table of recommended median adu (as shown in SGP) for various gain and offset pairings.

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

For this image:

I used gain 0 offset 10 and set the exposure so the median value was above 400 ADU AND the stars weren’t clipped (at least at the start of imaging each night).

That turned out to be 3 min subs (F5.6 12" RC with focal reducer). Using data like that from the sensor chart above, it would be nice to determine that exposure with higher gain, shorter exposures vs. waiting 2 and a half minutes to discover that I 150 sec exposure isn’t quite long enough (As I do with DSLRs and iso settings).


Lots of discussion of exposure, gain, and offset (cooled cmos cameras) here:

Which I think reinforces the need for an “aid” or “tool” to set exposure parameters.


Choosing the right exposure requires individual judgement. That applies to all cameras and technologies since they were invented. The CMOS cameras throw a few more variables into the mix, but so did gain/binning with CCDs to some extent.

In a previous life, I designed and patented a darkroom meter. I approached Ilford Ltd. “Thank goodness it’s not automatic” was their technical director’s reply. This meter provided the tools for the photographer to assess the image on the baseboard before they made a print. It did not remove any control from them but gave them tonality information with which to make an informed decision.

The tools do exist already in SGP in the form of histogram, picker etc. One person’s logic on determining a ‘good exposure’ will be very different to another’s. I’m just approaching CMOS cameras and it reminds me of choosing a developer. Too many combinations. Three is probably plenty. I plan to get to know them and use them well. I can see my sequences potentially exposing each filter with 2 settings, though, if too many stars are clipping, that is a problem (for my style).

The read noise issue is made worse by the bit depth of the sensor’s ADC, 12-bit is a compromise. Give it a few years and 14-bit and maybe even 16-bit may be more common, at the same price. The quantisation noise will reduce then and the read noise/gain debate will fizzle out.


I think there is a pretty good census in the digital photography world that the best exposures are close to saturation with the brightest parts of the image. To do otherwise, you are throwing away dynamic range at the top end.


Absolutely - very similar to the transparency film days.
In monochrome however, there was an additional gain control, effected by the amount of film development, that introduced the fun and games of the Zone system. In that, the exposure and development were inseparable, just as they are here with gain/offset and exposure time. Ansel Adams didn’t make his monumental images with his camera on auto.