I'm puzzled with Noise in Light frames/Integration - Flat frames/Integration

[Sedat]

Hi,

I have started using PI for calibrating, aligning and stacking very recently. Instead of using the BPP, I'd like to use the dedicated tools to have a better control on the whole process. Thanks to the Noise Evaluation script, I was able to track the noise levels after each step. The following data belongs to Luminance images shot by Atik 460 EX Mono camera:

Uncalibrated sub
m106_1_600sec_1x1_L_frame30
Calculating noise standard deviation...

* Channel #0
?K = 2.304e-03, N = 157934 (77.90%), J = 4

Calibrated sub - with Bias and Flat Masters
m106_1_600sec_1x1_L_frame30_c
Calculating noise standard deviation...

* Channel #0
?K = 2.297e-03, N = 4647476 (76.88%), J = 4

CosmeticCorrection applied sub
m106_1_600sec_1x1_L_frame30_c_cc
Calculating noise standard deviation...

* Channel #0
?K = 2.295e-03, N = 4786605 (79.18%), J = 4

StarAligned sub
m106_1_600sec_1x1_L_frame30_c_cc_r
Calculating noise standard deviation...

* Channel #0
?K = 2.030e-03, N = 4484671 (74.19%), J = 4

Stack of 30 frames
integration
Calculating noise standard deviation...

* Channel #0
?K = 4.611e-04, N = 3256902 (53.88%), J = 4

So, I'm getting more than 2x more noise in the final stack.

On the other hand, the situation with flat frames is like the following:

Uncalibrated flat sub
m106_1_0_76sec_1x1_L_frame1
Calculating noise standard deviation...

* Channel #0
?K = 5.214e-03, N = 5910648 (97.78%), J = 4

Calibrated Flat sub - with Bias Master
m106_1_0_76sec_1x1_L_frame1_c
Calculating noise standard deviation...

* Channel #0
?K = 5.214e-03, N = 5910693 (97.78%), J = 4

MasterFlat - Stack of 40 frames
MasterFlat_L
Calculating noise standard deviation...

* Channel #0
?K = 1.400e-03, N = 5864821 (97.02%), J = 4

With the flats I'm getting almost 1/4th of the noise of the individual frames.

The BiasMaster used in the calibration of these light and flat frames - Stack of 200 frames
BiasMaster_1x1_0C_20130729_200x
Calculating noise standard deviation...

* Channel #0
?K = 2.791e-05, N = 5170396 (85.53%), J = 4

The SuperBias produced from the above BiasMaster
SuperBiasMaster_1x1_0C_20130729_200x
Calculating noise standard deviation...

* Channel #0
?K = 9.400e-06, N = 2611685 (43.20%), J = 4

The subs at the beginning and end of the bias capture
Target_Set_1_0_001sec_1x1__frame1
Calculating noise standard deviation...

* Channel #0
?K = 3.089e-04, N = 5915885 (97.86%), J = 4

Target_Set_1_0_001sec_1x1__frame200
Calculating noise standard deviation...

* Channel #0
?K = 3.092e-04, N = 5913929 (97.83%), J = 4

Can somebody please explain me what's going on here ? Also, why is the SuperBias more noisy ?

I've started reading the relevant chapter in The Handbook of Astronomical Image Processing but it will take some time to go through all the details there. I thought I could get some quicker response from the gurus here.

Many thanks

Sedat

[pfile]

in the stack, you have 4x10^-4 and in the sub you have 2x10^-3.

that's 1/5 the noise, not 2x

rob

[Don]

Hi Sedat,

Pay attention to the exponents in those noise standard deviation figures.  1e-03 is ten times higher than 1e-04.  So the noise standard deviation of your stack of 40 lights is less than 1/4 of that in the uncalibrated sub, and the superbias has nearly 1/3 of the noise standard deviation value of the master bias.  This is as it should be ...

Cheers,

Don

[Sedat]

Phewww! I knew I was making a stupid mistake somewhere... never looked at the exponential figures

Thank you both!

[Sedat]

Any guesses as to why the less than 1 sec flat subs are noisier than the 600 sec light subs ?