Sensor Analysis with Canon EOS 6D (DSLR)

[chris_stargazer]

I have purchased SharpCap to analyse my DSLR sensor (astromodified Canon EOS 6D) after watching Robins talks on YouTube.
My aim is to use measured read noise data to calculate minimum exposure length for my location and equipment (https://tools.sharpcap.co.uk/).
I installed ASCOM 7 from https://www.ascom-standards.org/ and ASCOM DSLR driver from https://github.com/vtorkalo/ASCOM.DSLR .
ASCOM DSLR Setup: Connection method = CanonSdk, Image mode = RGGB, Store in the Location (i set a custom folder), ISO 1600

My camera is in front of a dimmed flat panel with 21mm lens at f/16 (to reduce overall brightness).
First run this morning technically worked:

Screenshot 2024-10-18 104431.png (163.55 KiB) Viewed 1538 times

1. Full Well Capacity
ISO 100 is 98345.45 e- but I other sources (e.g. https://photonstophotos.net/Charts/Sens ... istics.htm) indicate 74.261 e-.
-> Is this correct or a misconfiguration?

2. Very low read noise
Analysis shows 0.65 e- read noise at ISO 1600.
https://photonstophotos.net/Charts/RN_e.htm shows shows 2.445 e-
-> Where does this come from? Again a misconfiguration?

3. ISO 3200 and ISO 6400 not included
Statistics for ISO 3200 + ISO 6400 was not included.
I found viewtopic.php?t=6450 and added
SharpCap.SelectedCamera.SetGainsOfInterestForSensorAnalysis([1,4,7,10,13,16,19])

4. now ist stopped working
I have also changed "Preferred Still Format" in SharepCap to "FITS".

The entire analysis now runs a while (measures gains, measures read noise, asks my to uncover the camera) and than stops with an error message on ISO 10.000 or ISO 16.000. Seems it cannot determine the right exposure time (over saturated green pixels).

Screenshot 2024-10-18 192059.png (930.21 KiB) Viewed 1538 times

Any ideas?

[admin]

Hi,

sensor data on DSLR cameras is always a little hit and miss, because there is more processing of the image going on between the sensor and SharpCap (and we don't know what sort of processing is happening). Astro cameras tend to be designed to just read out the raw pixel value and let the capture software/processing software deal with everything from there on, but DSLRs just don't work that way. You will particularly hit this as you test in regions where the DSLR might be doing more processing (ie noise reduction at higher gains, etc).

Looking at the individual points, I can see that the low read noise for ISO 1600 (and 1250) is probably related to the relative gain of those ISO values being measured incorrectly - the relative gain at ISO 1600 should be roughly double the reading at ISO 800, but in the table it rises from 12.67 to 71.36! Essentially this means that something is going wrong during the measurement of relative gains at different ISO values, leading to the incorrect 'relative gain' measurements, which feed on into incorrect e/ADU and read noise measurements.

I think that the same thing is stopping SharpCap from analyzing above ISO1600 - in an early stage of the analysis, SharpCap does a rough 'gain range' measurement to determine what gain values (ISOs) it will measure at. The limit is that nothing more than about 300 times brighter than minimum gain will get measured. That would normally be about ISO 32000 or so, but with the relative gain being measured too high, SharpCap is thinking that the limit is being hit much sooner and is stopping at ISO1600. Unfortunately tricks like setting the gains of interest will not help - those are still filtered by the 300x limit as described above.

Actually, looking down the 'relative gain' column in the data, we see that it is always too high - it should be roughly ISO/100, but at 400 ISO it is 4.97, 800 ISO it's 12.67, etc. The error seems to get worse at higher ISOs. I can't quite think what might be causing that sort of consistent error in the measurements, since the measurement technique is fairly simple (adjust from gain 100 to 200, then adjust exposure to bring the image brightness back to what we saw with 100 - the ratio of exposures matches the ration of gain100 to 200 in the opposite direction). I do wonder if some sort of non-linear adjustment is being added to the data in the camera that might account for this?

I doubt that you will be able to do much to improve the measurements, since the analysis isn't really designed with DSLR data in mind (probably not worth the wear and tear on the shutter). You do however see enough realistic data to understand why low ISO is usually a problem with DSLR astro imaging (very high effective read noise values).

sorry not to be much help!

Robin

[chris_stargazer]

Hi Robin,
thank you for your prompt response.
Maybe the flat panel is one root cause. I recall that my Canon EOS produces strange flats with exposures faster than 1/200 seconds.
My measurement attempts costed ~3.000 shutter counts already, I will therefore not try it again with DSLR.

I assume the sharpcap smart histogram feature will not work for my Canon EOS 6D, for the same reason correct?
Can you give me a hint how to enhance your formula for minimum exposure time s = C * R^2 / P to include thermal noise?

Regards,
Christian

[admin]

Hi Christian,

yes, the shutter count required to do the analysis is a problem It's not really optimal for DSLR cameras because it is designed for astro camera that are typically running in a constant capture video mode, so it's necessary to be sure that changes to settings have been applied for long enough to actually appear in the frames arriving from the camera.

I think you are right on the smart histogram not working - at least the measurement will fail to give the right sky background value. What will happen is that SharpCap will turn up to high ISO to try to use shorter exposures to get the background measurement. That high ISO data is dodgy, so the background measurements will be dodgy, even if used to calculate for ISO values where the analysis is reasonable. If you can get the sky background light pollution rate from some other source then enter it manually into the smart histogram, it should give reasonable suggestions for the range of ISO values where the data is good. You can potentially get the light pollution rate in e/pix/s using the SharpCap histogram, since that will show the equivalent number of electrons for an ADU value based on ISO and the sensor data (again, use a 'good' ISO level for this). Then divide that electrons/pixel number by the exposure length for e/pix/s.

cheers,

Robin