Please explain to me this noise

[AstroScience]

Hi all,
I'm open and debayering 3 Light frames which were recorded with DSLR at 28 C temperature with ISO1600.
Noise evaluation for those 3 files reports:

Gaussian noise estimates:
s0 = 1.331e-003, n0 = 0.212 (MRS)
s1 = 1.282e-003, n1 = 0.283 (MRS)
s2 = 1.319e-003, n2 = 0.216 (MRS)

Gaussian noise estimates:
s0 = 1.359e-003, n0 = 0.210 (MRS)
s1 = 1.308e-003, n1 = 0.280 (MRS)
s2 = 1.349e-003, n2 = 0.214 (MRS)

Gaussian noise estimates:
s0 = 1.329e-003, n0 = 0.210 (MRS)
s1 = 1.280e-003, n1 = 0.280 (MRS)
s2 = 1.319e-003, n2 = 0.215 (MRS)

Now, to calibrate those I'm using Master Bias which created from 250 Bias frames at ISO1600 and Master Dark which created from 30 Dark frames at ISO1600 with exactly same 28 C temps.
After calibration with Master Bias and Master Dark with noise scaling and debayering I'm getting noise evaluations accordingly:

Dark scaling factors:
k0 = 1.491

VNG debayering: done
Noise evaluation: done
Gaussian noise estimates:
s0 = 1.013e-003, n0 = 0.556 (MRS)
s1 = 9.803e-004, n1 = 0.700 (MRS)
s2 = 8.908e-004, n2 = 0.544 (MRS)

----------------------------------------------

Dark scaling factors:
k0 = 1.500

VNG debayering: done
Noise evaluation: done
Gaussian noise estimates:
s0 = 1.024e-003, n0 = 0.555 (MRS)
s1 = 9.927e-004, n1 = 0.700 (MRS)
s2 = 8.972e-004, n2 = 0.541 (MRS)

-------------------------------------------------

Dark scaling factors:
k0 = 1.544

VNG debayering: done
Noise evaluation: done
Gaussian noise estimates:
s0 = 1.007e-003, n0 = 0.553 (MRS)
s1 = 9.766e-004, n1 = 0.699 (MRS)
s2 = 8.818e-004, n2 = 0.541 (MRS)

After calibration it shows that only channels R got reduced in noise in all of the three files and channels G and B got injected with noise.
Can someone explain what's going on here?

I have also run calibration on those files without dark scaling and got those noise evaluations accordingly:

s0 = 1.042e-003, n0 = 0.417 (MRS)
s1 = 1.001e-003, n1 = 0.539 (MRS)
s2 = 9.275e-004, n2 = 0.388 (MRS)

s0 = 1.059e-003, n0 = 0.404 (MRS)
s1 = 1.019e-003, n1 = 0.525 (MRS)
s2 = 9.407e-004, n2 = 0.372 (MRS)

s0 = 1.037e-003, n0 = 0.412 (MRS)
s1 = 9.977e-004, n1 = 0.534 (MRS)
s2 = 9.198e-004, n2 = 0.381 (MRS)



 

[cs_pixinsight]

Sergio, it looks like the noise dropped for all subs in all channels.  The exponent on the noise estimates went from e-003 to e-004, so the decimal point shifted after calibration.

I am curious why the dark scaling factor is so high given the matching dark.

Craig

[AstroScience]

Hi Craig,
silly me, I'm totally ignored those numbers, thank you for pointing this out to me.

As for the high factor, all I know is that Darks were 420 sec long while subs were 540 sec long, all same temps.
That must be it?

Also, do I have to optimize my calibration files to look for a factor of 1 or less?

[Geoff]

As for the high factor, all I know is that Darks were 420 sec long while subs were 540 sec long, all same temps.

Slightly off topic, but your darks should always have an exposure length that is no less than your subframe exposure.
Geoff

[AstroScience]

Hi Geoff,
not off topic at all and I'm glad you joined conversation as I'm filling the empty spots of information in my head.

I were totally relying on the Vicent's paper on Master Calibration Frames:

"IC is optimized to work with calibration libraries: don't worry about differing temperatures and exposure times between dark and light frames. IC will always rescale the dark noise to match every light frame. "

What actually happens if darks are shorter than the subs, scaling cannot be properly calculated?

[Geoff]

Scaling assumes that the darks are linear, so that if I expose twice as long I get twice the dark current and if I expose half as long I get half the dark current. If this were really the case it wouldn't matter what length your darks were.  Unfortunately, this is only a reasonable approximation, so when you scale down, errors in this assumption are reduced, but when you scale up, errors are magnified.
Geoff

[AstroScience]

Thank you, Geoff.