Another Thread On CMOS and Darks or Bias
Posted: Sun Feb 25, 2024 6:45 am
I read another thread about Despeckling that was talking about the hot topic of CMOS and dark frames. I’m studying an online video series about PI done by a highly respected image processor. That series was covers this very subject, so I thought I might share what I’ve learned and a hypothesis.
FirstCMOS chips of late seem to have very low thermal noise. They do, in fact tend to have lower dark current, but they are also more sensitive. A more sensitive sensor means shorter integration time before saturating pixels. So my hypothesis (this is not an idea proposed in the video series) is that it seems like there is no dark current because exposures are shorter and the dark current is actually lower. There is still some dark current though.
The other thread I mention also posted sub frames of dark and bias frames. I propose a different method of evaluating the actual dark current. This is proposed in the video series, and I have yet to, but intend to, do this experiment with all of my cameras.
First, take a bias frame, then a series of darks, all at the temperature that light frames are typically captured at, of varying exposure times. Start with maybe 10 seconds, then 20, 30, 60, then 30 second intervals up to 600 seconds.
Next load the subs in PixInsight, then use PixelMath to subtract the bias signal from the dark signal. The resulting images are your true thermal signature at those exposure times. When you have that information you can reasonably determine the need or not for dark frames.
When I have time to run this experiment I’ll post the result here. Suggestions on what processes or methods are most appropriate to evaluate the data appreciated. I’m thinking the Statistics process should do.
First
The other thread I mention also posted sub frames of dark and bias frames. I propose a different method of evaluating the actual dark current. This is proposed in the video series, and I have yet to, but intend to, do this experiment with all of my cameras.
First, take a bias frame, then a series of darks, all at the temperature that light frames are typically captured at, of varying exposure times. Start with maybe 10 seconds, then 20, 30, 60, then 30 second intervals up to 600 seconds.
Next load the subs in PixInsight, then use PixelMath to subtract the bias signal from the dark signal. The resulting images are your true thermal signature at those exposure times. When you have that information you can reasonably determine the need or not for dark frames.
When I have time to run this experiment I’ll post the result here. Suggestions on what processes or methods are most appropriate to evaluate the data appreciated. I’m thinking the Statistics process should do.