The AAVSO Guide to CCD/CMOS Photometry with Monochrome Cameras states the following about CMOS camera gain settings:
"When multiple gain settings provide the same (or similar) dynamic range, you should choose the lower gain setting."
This statement applies to cameras in which the user is able to select a HCG (high conversion gain) setting. The Guide says nothing else (at least, nothing that I have found) about CMOS gain settings. I use a ZWO ASI1600MM camera which has 12 bit ADC and no HCG setting. My attempts to select settings for optimal precision (for time series of EW eclipsing binaries) have been based on trial and error, but I've recently done a more systematic trial. Images were taken through a V filter and a Canon 200m f/2.8 prime tele lens.
Exposures of 30, 60, 90 and 120 seconds were taken at each of the following gain settings:
Gain 139, system gain 1e/ADU
Gain 79, system gain 2e/ADU
Gain 49, system gain 3e/ADU
Gain 19, system gain 4e/ADU
Gain 0, system gain 5e/ADU
For each combination of exposure length and gain setting, 15 frames were taken.
For 10th mag stars best precision and SNR were seen at an exposure of 120 seconds at system gains of 2 or 3 e/ADU. These settings, for in-focus images, allowed substantial ADU counts with no saturated pixels, whereas many pixels were saturated at 120 seconds and a system gain of 1e/ADU. At system gain settings or 4 and 5e/ADU, ADU counts are low, precision is not as good, and SNR values were lower, even at exposures of 120 seconds.
For brighter stars the strategy adopted is to use lower gain (higher values of system gain) and the longest exposure possible. For 'very bright' stars (6th mag for the setup used in this trial), images must be defocussed to avoid saturated pixels.
Using low gain settings (higher system gain) for higher dynamic range is sometimes mentioned as a desirable aim in discussion of CMOS photometry. My experience leads me to ignore dynamic range and optimise precision by trying to select the 'best' combinations of gain and exposure, as above.
I'd be interested to know if other observers have found the above principles to be useful, or if anyone has quite different ideas.
Dear observers, I've got ZWO ASI432MM planetary uncooled mono camera for variables. I understand, that idea is strange or even stupid, but it is my limit. I did not make startest yet and seen just light and darkness on my smartphone screen (I plan to use camera with Asicap application). But I'm at a loss: there are not only exposure, gain but offset too!
Could you please recommend optimal values of these three parameters for vars stacking? I hope to reach some seconds exposure on my old black HEQ-5 without autoguiding, but gain??... and offset?!?! Or show me please the shortest way for finding of optimal exposure, gain and ohhh offset.
Hope for your kind help!
Optimal exposures for your system can only be determined by you. Maximum exposure (unguided) will be limited by the tracking accuracy of your mount. Up to that limit, optimal exposures within the range of linearity of your camera will depend on the brightness of your target and comp stars. Systematic trials of exposures of various length (from, say, 15 or 30 seconds up to a few minutes) for non-variable stars of various magnitudes will soon give you the data you need to decide on what exposures to use. Take several images at each exposure setting to make sure you image through an entire cycle of your mount’s worm gear. Personally, I would take at least 10 images at each exposure setting, so that you can see how much variation in measured magnitude of non-variable stars there is from frame to frame. I like to plot the results and calculate standard deviations.
The AAVSO Web Site has resources to help you with other aspects of imaging for photometry. If you haven’t done so already, you should read the AAVSO Guide to CCD/CMOS Photometry with Monochrome Cameras. On the AAVSO Home Page click Education CHOICE Courses Manuals Videos > Observing Manuals > Guide to CCD/CMOS Photometry …
Note section 3.2.2, Image Scale on page 31. Your camera has 9 micron pixels, which are unusually large for a CMOS camera, so this section is particularly important for you. Note also section 3.2.3, Seeing and Sampling.
Offset is in section 220.127.116.11 on pages 42 and 43. I don’t know what the settings are for offset with your camera. Mine (ZWO ASI1600MM Pro) range, I think, from 0 to 100. The offset adds the same count to every pixel in an image. I use an offset of 10, which adds 10 ADUs to each pixel. An offset of 20 adds 20 ADUs, and so on. Note that in 12 bit ZWO cameras, these native offset counts (and the counts in all images) are scaled up by multiplying by 16. The purpose of the offset is described well in the Guide. If you take bias frames at various offsets and look at the image histograms you will soon see what the offset does.
The section on Gain (18.104.22.168) starts on page 44. I quoted from this in my previous post. The instruction there is specifically for cameras with HCG (High Conversion Gain), which your camera has at gains of 140 and above, and can be seen in the discontinuities in the manufacturer’s graphs of gain versus DR (dynamic range) and gain versus read noise. My camera does not have this.
What gain should your use? I think you should experiment. I am aware that some experienced photometrists in the AAVSO recommend the lowest gain setting (zero gain) for photometry, because dynamic range is then maximum, which is beneficial when there are both bright and faint stars to be measured in the same field. For me there are two points about this. First, there may be no necessity to measure both faint and bright stars at the same time, UNLESS the choice of comp stars is limited. Second, benefitting from higher dynamic range may be possible for 16 bit cameras, but tests with my own 12 bit camera (see previous post) do not show that the gain with the highest dynamic range yields the best precision, UNLESS very bright stars are targeted, in which case the lowest gain may be necessary to avoid saturation. In this situation, I also need to defocus to avoid saturation.
A final point is that, at zero gain, your camera has very high system gain (about 24 e/ADU), which is why the manufacturer can graph a full well of 97K electrons at this gain setting. This means, of course, that 24 electrons will generate only 1 ADU. Quantization errors could be significant. I think all you can do is to experiment with various combinations of gain and exposure, and determine the precision of your measurements. In general though, for the faintest stars you are able to measure, my personal view is that you could start at a gain of 272 (unity gain, system gain is 1e/ADU), and try using lower gain settings for brighter stars. This is what I do.
A lot of thanks, dear Roy, for your explanations and advices! Indeed, I have to study manual and train before real photometry.
I have the same camera as you Ray but I've only started using it since the autumn as a change from my old DSLR. I was recommended to use the 139 gain by someone who usually does imaging and I've stuck with that. I adjust the exposure time to give a high pixel value without saturating. This normally means I'm limited to stars greater than mag 10. During processing I always check that the target star, reference star and check star are not saturated.
But I have one question for you. Do you have to make separate dark and flat frames for each gain setting?
I use the same gain setting for darks and flats that I use for the corresponding science images, which is recommended by the AAVSO Guide to CCD/CMOS Photometry.
Mine is a monochrome camera. A set of flats must also be taken for each filter used. The same principle applies to colour CMOS astronomical cameras: a different set of flats is needed for each colour channel.
Concerning gain settings for photometry, I have expanded the analysis of precision to time series taken during the past several months. The results differ from those of the trial (of images taken during one night) I reported in this thread on November 20 last year. In the various time series, 'good' precision (standard deviation of the check star mag of 0.006 or 0.007) could be achieved at any gain setting from 139 (system gain 1e-/ADU) to zero gain (system gain 5e-/ADU). Focus varied, with images most nights being slightly defocussed (I image through a camera lens and use manual focus, so it's tricky to get perfect focus). FWHM (in pixels) varied, but there was no correlation between FWHM and photometric precision. The 'best' correlation was between higher precision and longer exposure (varied between 30 and 120 seconds, no autoguiding). No exposure shorter than 60 seconds achieved 'good' precision. Images were binned 1x1 or 2x2.
In practice now, I find it 'best' to use system gains of 3 to 5 e-/ADU (depending on the brightness of the target and comp stars) then set the exposure on each night by trial and error to get a 'good' signal without saturation.
For those who remember the (too many) posts from me on this forum in 2020, my strategy has now completely changed. Back then I deliberately defocussed and used a system gain of 1e-/ADU. Now, images are close to focus (except for very bright stars) and the system gain is almost always > 1e-/ADU.
You stated: "The same principle applies to colour CMOS astronomical cameras: a different set of flats is needed for each colour channel."
I assume what this means is that you take one set of flat images with your one-shot colour camera but you then separate them into individual colour channel flat images before application to the separate colour channels of the science images?
In fact, the DSLR Guide states the following:
"Color separation may be performed before or after image calibration. It does not matter which order you
choose so long as all data (calibration frames and science frames) are treated identically."
"Note: Depending on which software you use, color channels may need to be extracted prior to calibration.
It is very important not to mix the calibration frames for different color planes."
The only OSC photometry I have done was with a DSLR camera, and that was a few years ago. I found that if I took only one set of images for flats, then separated the colour channels and looked at the stats of the images from each channel, the average ADU varied among the channels.
Perhaps I was being too careful, but after trialling various exposures for the flats, I took three sets at different exposures. One set yielded the R master flat, the second the B master and the last the G master. I tried to get the average ADUs about the same for each channel. Of course they were not identical, but were fairly close.
I did not actually do a study to see if the extra work made any improvement to photometric precision over the alternative: flats for the three channels from a single set of exposures.