#apertus IRC Channel Logs

off to bed now ... have a good one everyone!

good day

I captured some series ouf ~50 darkframes from ~.9ms to (1000/24)ms exposure time

the general goal is to find a way to use a single / a few darkframes for the darkframe calibration and not need a specific darkframe for each exposure time

so using two reference frames, one (`b`) that is basically the extrapolation of a darkframe at 0ms exposure time and a frame (`m`) that represents the exposure dependent part, so that the final darkframes for a exposure time of `n` ms is calculated as `n * m + b`

b and m are just determined by a per pixel linear regression over the 51 averaged darkframes I captured

(for the darkframes `real_df` with exposure time `n` ms, this is just `real_df - m * n - b`)

which is just the histogram of these residuals for each of the darkframes I captured

(so the actual variance of the residuals is visually vastly overstated)

a few examples that show the residuals for some specific darkframes: https://files.niemo.de/darkframe_residual_examples.pdf

any ideas for better models?, ill try some high degree polynomials next, but going higher than maybe second or third degree seems like arbitrary overfitting very quickly

any opinions on how flatfield correction will interact with this? maybe we want to "flatfield correct the `m * n` term"?

so this is for 4x gain: https://files.niemo.de/darkframe_4.0x_residual_linear_model.pdf

interestingly, the residuals do not match up completely

currently we cannot expose as long as the frame period

yeah, but I expect that this should some day be the case, right?

very short exposures like 1/8000s might be interesting

well they don't work currently :P

what is the low end currently?

ok

but that sounds like an issue that might be gone with the sequences/generator decoupling that Bertl is currently implementing

oh another question: how often do we expect longer exposure times to be interesting? (100ms to tens of seconds), are we fine with expecting people to capture specific darkframes for these ranges?

if the process is somehow guided through in the camera itself

not yet, I just though about that aswell, but forgot to include capturing the temperature while capturing the darkframes

its in the sensor registers if you captured raw12s as well

nope I did not

one thing I would also find an interesting check is if the fpn can be assumed constant for a particular combination of settings

so if you capture the same darkframe set one day and the same set again a week later

will the pattern match up perfectly

but also with temperature yes

but these should hopefully be somewhat temperature independent, if they use a "good" voltage reference

so it might be interesting to capture those aswell

can you capture the voltage information that is visible during the lvds training process (or was it another command) with the darkframes as metadata?

that should be trivial pretty much

great

morning folks!

off for now ... bbl

https://twitter.com/NASA/status/1512449836511121415?ref_src=twsrc%5Etfw%7Ctwcamp%5Etweetembed%7Ctwterm%5E1512449836511121415%7Ctwgr%5E%7Ctwcon%5Es1_&ref_url=https%3A%2F%2Fwww.derstandard.de%2Fstory%2F2000134800475%2Fstart-geglueckt-erste-private-crew-ist-unterwegs-zur-iss

https://twitter.com/NASA/status/1512449836511121415

@Axiom_Space

yay! *G*