×

INDI Library v2.0.7 is Released (01 Apr 2024)

Bi-monthly release with minor bug fixes and improvements

Optimal Sub-Exposure Calculator help

  • Posts: 42
  • Thank you received: 11
Thank you all for responding to my request.

(I'm sorry for the delay in my reply; I was working on some performance improvements to the analyzer tool, and also trying to understand an anomaly I see when I process data from color sensors.)

But I've just created a folder on Google Drive that should be writable for anyone with this link.

drive.google.com/drive/folders/1nHlqDTHz...ofudq?usp=drive_link

I assume that Google drive will allow you to create sub-folders in this shared folder, so it would probably be best to create a "home" folder for yourself. And then upload zip files of your bias and flat images to that home folder.

If you plan to create new flats and bias images for this effort:

Remember that I will need two of each file type, (2 flats and 2 bias), for each gain value being plotted.

I use file naming format like this /%t/%T/%F/%t_%T_%F_Gain_%G_Offset_%O to help me distinguish the files. (But I can jut read the fits header if need be).

For flats I've been setting the calibration option for the ADU to 29000. But this is probably not critical, as long as the value you use is reasonable and consistent.

As I mentioned, I'm using a dimmable LED panel for my flats. But to reduce the brightness I ended up inserting about 30 sheets of printer paper between camera and the panel so that the exposure time for flats is around 1 second. But when I image at with higher gain values I've had to add a few more sheets of paper because the exposure time can fall rapidly at high gains.

I also found that my cameras leak light in quite badly. I've been creating my images in daytime, and find that I need to cover the camera with a dark cloth to avoid unwanted light in the exposure.

Thanks again for your help!
2 months 3 weeks ago #98272

Please Log in or Create an account to join the conversation.

  • Posts: 139
  • Thank you received: 16
I've loaded frames for the 4 cameras I mentioned. For the Moravian and FLI cameras they were taken with the camera in front of a lightbox, no scope. I was making photon transfer curves.

I've added the curves I obtained to the Moravian folder in case they are useful. Please take note of the page from the user manual I added as well, which points out that the "gain" is a command to the driver and is not a multiplicative gain, neither in linear or dB. The conversion from the 2 values I used to multiplicative gain is given there. I suspect this might be true for all Moravian CMOS cameras, by the way.

For the QSI I also made a photon transfer curve, so I must have the data (without scope) somewhere. As I can't find them I have instead used the oldest (and best quality) through-the-scope flats I could find. If this is no good I can fish out the camera and make some more. Same story for the SBIG ST2000XM, except I never made a photon transfer curve.

I will rework the Kepler and QSI photon transfer data into the same form as the Moravian plots (easier to interpret) if you find them useful.

Cheers,
Richard
The following user(s) said Thank You: Joseph McGee
Last edit: 2 months 3 weeks ago by Richard Francis. Reason: typos
2 months 3 weeks ago #98300

Please Log in or Create an account to join the conversation.

  • Posts: 206
  • Thank you received: 28
On the road for the next week but will upload files next week when I’m back at home base.
2 months 3 weeks ago #98331

Please Log in or Create an account to join the conversation.

  • Posts: 42
  • Thank you received: 11
Thanks Richard,

I was able to get the SBIG ST2000XM Data to process through my tool.
To be clear:
1) My tool only analyzes a 100x100 pixel region at the center of the sensor. (That's actually a recommend approach when performing this type of analysis).
2) The value shown for "Gain Scaled" has to do with adjusting the reported gain for the bit depth of the sensor vs the bit depth read from the fits file. So, for example, if the sensor were only 14 bits, then the Gain Scaled (e-/ADU) value would be 4 times the value reported as "Gain (e-/ADU)". (I'm working on a function that hope will detect the sensor bit depth by analyzing the intervals between pixel values, but that function is not yet active, so this output assumes the sensor is actually 16 bits).

Read Noise (ADU): 10.6161
Gain (e-/ADU): 0.6288
Gain Scaled (e-/ADU): 0.6288
Read Noise (e-): 6.6756
Full Well (e-): 41210.2823
Dynamic Range (Stop): 12.6351


After running this I looked up some specs on this camera, and the values my tool reports seem to be reasonable.

I will continue with testing on the other cameras.
2 months 3 weeks ago #98340

Please Log in or Create an account to join the conversation.

  • Posts: 17
  • Thank you received: 4
I've loaded data from ZWO ASI294MC Pro and ZWO ASI120mm-mini
The following user(s) said Thank You: Joseph McGee
2 months 3 weeks ago #98361

Please Log in or Create an account to join the conversation.

  • Posts: 42
  • Thank you received: 11
Hi Richard,

I've finished processing the data for your four cameras. I uploaded text files with results into each of your camera folders on the google share. Take a look if you are curious.

This had been a very productive exercise for me. I've made notes of a number of changes I plan to make to my calculation function.

For the most part my tool seems to be producing reasonable values. The only result I question is for the FLI Kepler, mainly because there was pattern noise in some input files, and the read noise that my tool computed was a bit higher than the value I found reported on the manufacturers site. (I actually ran a 2nd test for this camera where I selected different region to analyze, but the result was not substantially better.

I plan to make some custom camera data files for you, in case you want to use the exposure calculator. But the Moravian camera might pose a problem for the exposure calculator. My design in the exposure calculator assumed that CCD cameras do not have an ability to select gain. But this camera seems to have discrete gain values that can be chosen. So, in a way it behaves rather like a DLSR that has discrete ISO settings. So I may need to make a change to the exposure calculator to handle this type of CCD camera.

Thanks again for providing the files.
2 months 2 weeks ago #98380

Please Log in or Create an account to join the conversation.

  • Posts: 139
  • Thank you received: 16
Actually the Moravian is a CMOS camera, not CCD. I know that INDI treats them all as CCD or DSLR, but it's quite common for CMOS ones to have selectable gain. The important point about it is that the "gain" is not actually gain but a look-up table entry from where the gain is read. That's why I attached the page from the user manual.

I agree the Kepler has a weird fixed pattern component. It can be calibrated out, but doesn't look good in the pre-calibration frames.

Did you find the photon transfer curves of any use? If so, I can present the data I already collected for the Kepler and QSI in the same form. These may be doing something similar to you, but over a wide range of flat exposures, and dark exposures too as a variant.

Cheers,
Richard
2 months 2 weeks ago #98381

Please Log in or Create an account to join the conversation.

  • Posts: 22
  • Thank you received: 0
Hi quick question why exposure time is in ms for me when i saw sec on several screens shared.
where i can select exposure time format
2 months 2 weeks ago #98405

Please Log in or Create an account to join the conversation.

  • Posts: 42
  • Thank you received: 11
Hi Vera,

The calculator should only display, exposures in seconds. But the calculation can produce extreme exposure times, both very short and very long, depending upon the inputs.
Can you share a screen shot of the calculator?
2 months 2 weeks ago #98417

Please Log in or Create an account to join the conversation.

  • Posts: 42
  • Thank you received: 11
Hi Richard,

For the Moravian, I made a small adjustment to the output from my tool to include the dynamic range value in both raw (no units), and in dB. So that it can be better compared to the information in the charts you provided.

My computed values are generally the same, as what I find in the charts. But there is a difference between my computation of Full Well, and that shown in the charts.The calculation for full well is gain (e-/adu) x max adu. But I use 65536 (on a 16-bit readout) as the max adu, but the charts seem to using 64000. Do you think that the reduction is to account for thermal noise, or maybe something else? (My tool does not yet consider thermal noise).
2 months 2 weeks ago #98420
Attachments:

Please Log in or Create an account to join the conversation.

  • Posts: 139
  • Thank you received: 16
But I use 65536 (on a 16-bit readout) as the max adu, but the charts seem to using 64000. Do you think that the reduction is to account for thermal noise, or maybe something else? (My tool does not yet consider thermal noise).

It's a bit subjective. The fixed pattern and shot noise fall off as full-well is approached (since there's no dynamic range left for the positive excursions of the noisy signal). So on the photon transfer plots you do see this fall-off. So you could assign the max value of the 16-bit ADC as the full-well capacity, or you could back it off a bit -- given the noise signal it's difficult to say exactly where it is.

I used 64000 in the plots but 65536 is equally valid. In either case the observed noise is significantly reduced, as you see in the plots.

Cheers,
Richard
The following user(s) said Thank You: Joseph McGee
Last edit: 2 months 2 weeks ago by Richard Francis.
2 months 2 weeks ago #98425

Please Log in or Create an account to join the conversation.

  • Posts: 42
  • Thank you received: 11
Hi Vladimir,

I ran my tool on the data from both your cameras, but I had to spend some extra time reviewing the results. I've uploaded some files to the Google share in case you are interested it seeing what my tool produced.

But here is a summary...

For the ASI-120MM:
1) The results from my tool seem reasonable, but these calculations do not match the ZWO Documentation for this camera very well.
My tool calculated higher read noise at lower gain settings than what ZWO graphs show. But my tool calculated lower read noise at the gains of 80 to 100, than than what ZWO graphs show.

2) The ZWO documents show several steps in the read noise values. My understanding is that these steps usually caused by timing changes in the camera read-out circuitry. But the steps are occurring between the gain values of the images that we used for the analysis. So the results from my tool cannot "see" exactly where these steps occur. (This is expected, and when I develop the process to automate the capturing of images to be analyzed, I will need to detect where these steps occur).


For the ASI-294MC Pro:
My results are very different from the ZWO Documentation. Most noticeable is the calculation that my tool produced from the images at Gain value 50, which are very different from ZWO documentation. The calculation for Gain (e-/ADU) is lower that I would expect. This result causes the calculation of the Read Noise (-e) to be much lower. (This is apparent in the charts from the spreadsheet I uploaded).

I've been working to analyze the source of this discrepancy, and found that in the calculation of the Gain (e-/ADU), part of numerator in calculation, (the mean value of the sum of the flat pixel data), had fallen slightly, while part of the denominator of the calculation, (the standard deviation of the difference of the flat pixel data), rose abruptly. This produced a much lower than expected Gain (e-/ADU) for the images at Gain value 50. I will continue to analyze this, to determine whether my code has some fault.

Thanks again for providing the data
2 months 2 weeks ago #98454

Please Log in or Create an account to join the conversation.

Time to create page: 1.248 seconds