once again, a collaboration with @borodog (Mike) has helped forge a new SharpCap feature
Dark frame scaling is designed to help in situations where the dark frames in use don't exactly match the conditions for the light frames. The most obvious reason this might happen is if you are using an uncooled camera, so you have no real control over the sensor temperature. The same code can also _try_ to correct for changes in exposure and/or gain, but I would advise capturing matching darks if at all possible in those situations.
Anyway, in more detail...
What does Dark Scaing do?
When dark scaling is active, SharpCap will try to calculate a correlation between key pixels in the dark frame and the light frame to work out a scaling factor that needs to be applied to the dark frame to give best correction when the dark frame is applied. For instance, if the light frames are being taken at a colder temperature than the dark frame, it's likely that the scaling factor will be less than one, meaning that the dark frame will be scaled down before subtraction to get best results.
This calculation is done on a frame-by-frame basis, so if the camera temperature changes during a session, SharpCap's dark frame scaling will be constantly adjusting to the changes
How do I activate it?
If all the right conditions are in place to use Dark Frame Scaling (see below) then you will see a control appear two below the dark frame selection control, which you can use to turn dark frame scaling on or off
When is Dark frame scaling available?
Dark frame scaling is available (and the on/off control will show to enable/disable it) if all of the following conditions are met
- The camera is in a high bit depth (12/14/16 bit) mode
- Camera sensor analysis data is available for that camera in the selected mode
- A dark frame has been selected in the 'Subtract Dark' control
Dark frame scaling may also be available when loading previously saved FITS files via the Virtual Folder Monitor camera (providing the FITS files have been saved with a recent version of SharpCap 4.1, which adds FITS headers related to the e/ADU and bias levels of the camera when the images were captured).
Why don't I need a bias frame?
Dark scaling techniques traditionally use a combination of a dark frame and a bias frame to calculate the new, scaled, dark frame data. Essentially you subtract the bias away from the dark frame, scale the result of that calculation then add the bias back on.
Right now, SharpCap isn't using a bias frame, instead it's using the measured bias level from camera sensor analysis data - using a single constant value across the whole frame.
In order to get benefit from using a true bias frame, you need to capture a very large number of individual bias frames to make a master bias (probably several hundred - with lower numbers random noise dominates over the pattern noise you hope to find in the master bias). Even if you do that, the variation of a master bias is low with modern CMOS cameras (I measured a standard deviation of 0.2e), meaning the effects are small and are reduced further if the dark scaling factor is close to one (small corrections made by dark scaling)
How do I find out if it is doing anything?
You can look in the SharpCap log, where the scaling applied to each frame will be recorded along with a measure of how strong the correlation that SharpCap found in the data was ...
Code: Select all
Info 15:22:24.123311 #44 Dark frame scaling of 0.6287x applied (scaling explains 86.0% of warm pixel variability)
Alternatively, just hover your mouse over the 'Dark Scaling (Experimental)' control to see similar information in a tooltip...
Note... having left the camera used for the test above running for several minutes while writing this article, the sensor has warmed up, and now the results being seen are more like this...
Code: Select all
Info 15:28:20.773099 #44 Dark frame scaling of 0.8841x applied (scaling explains 89.8% of warm pixel variability)
What are the limitations of dark scaling?
If a scaling factor outside the range of 0.1 to 10.0 is calculated, it will be ignored and no scaling will be applied. Additionally, if the correlation calculated by SharpCap explains less than 10% of the variability in pixel brightnesses in the light frame, no scaling will be applied.[*]
What improvements can I expect?
Subtle reductions in noise in the darker parts of the image where dark frame correction is more accurate, better handling of warm/cool pixels, reducing their impact on the final image.
Why is it experimental?
Basically because of limited testing so far - while it's working as tested so far, there may be situations where it should work and it doesn't, or where it calculates the wrong scale factor for some reason and makes a mess of things. Please use with care and keep an eye on progress/results as you are capturing to ensure that everything is working as expected. Please do report any bugs/problems/poor results (ideally with captured images, darks, etc so the problem can be reproduced in testing).
cheers,
Robin