Weird sensor analysis results with Touptek ATR585M

[nurbson]

Thanks for the hint, but drivers are the same on both systems.

[admin]

OK, there seem to be a couple of possibilities left

1) Something to do with hardware differences USB controllers or something - I think this is pretty unlikely to be honest - it's a perfectly good reason why a camera might not work at all, but I've not seen that sort of issue leave the camera basically working but with subtle issues.

2) Some setting saved on the problem computer - either a SharpCap one or something that the ToupTek drivers/SDK are using. You could try making a new user profile on the computer, then run SharpCap in the new user profile and see if that makes any difference. Pretty much all SharpCap stored settings/data are stored at the user level rather, so a new user profile will be basically clean from the SharpCap point of view. I can't be so sure about how any ToupTek settings might be stored though.

cheers,

Robin

[nurbson]

Thanks for your patience with me
I finally figured it out: it was the size of the measure rectangle. It simply was to big or it included a dead pixel, either black or white. These are my final results

Screenshot 2025-05-26 194255.jpg (91.52 KiB) Viewed 3332 times

May I ask one final thing: I asume you are one of the developers of SharpCap. Would it be possible to implement an additional option to only measure certain gains? Something like SharpCap.SelectedCamera.FixedGainsForSensorAnalysis([100,150,200,250]). The idea behind this is that after one runs an initial analysis, they are only interested in a specific interval and that option would speed up the process a lot.

[admin]

Hi,

thanks for reporting back with the cause - actually I'm a bit surprised that hot/dead pixels would have so much effect! The analysis is not supposed to begin unless you have selected an area of at least 10000 pixels, so you would have thought that the effect of any outliers would be fairly small. Maybe it is worth changing the code to discard outlier values, or to use the median rather than mean value to address this. I will have to test that out.

I don't think that limiting the gain values would have a big impact on the speed - there are still a good deal of steps that need to run anyway (like the sensor depth measurement and the linearity check). The best way to limit the extra time spent measuring different gains is to put the camera into an ROI (say 800x600) before doing the analysis. That means that the camera can run at really high frame rates during the dark measurements, speeding that up considerably.

cheers,

Robin

[nurbson]

I was supprised, too.

How about adding an "additional measurement" button after the first analysis has finished. One can enter some gains and the results from the first run get reused.