What ISO?

[jgs001]

Paul, I'm being thick here... I've captured the frames and ready to analyse. I can get the mean and stdev from the histogram. You mention about adding 2000 or 3000 adu to the difference... how do you do that ?

[narrowbandpaul]

Hi John

Firstly the subtraction (differencing) of two flats removes the fixed pattern noise.

Because the images are of very similar signal, doing a straight subtraction will leave lots of zeroes, clipping the histogram and destroying the std dev measurements

so we need to add 2000-3000 ADU to one of the frames (the one on top of the subtraction.....

open up the image math option in whatever programme you are using.

there should be an option to add an offset to one of the frames.

either that or in pixel math (which usually includes options like dividing/multiuplying pixel values by a constant) ADD a value of 2000-3000.

adding alters the average value of the differenced frame but perfectly preserves the std dev.....which is what we want (actually std dev^2)

hope that helps......

same for the read noise (differencing of bias frames)....remember the factor of root 2.....

let us know the results, and good luck with the analysis

paul

[jgs001]

Ah.. gotcha... I think I know what I need to do then. Thanks

[narrowbandpaul]

what camera are u using?

[jgs001]

Canon 450d. So very similar to the tests you ran. I do know the 450d noise handling is much better and there is no amp glow, even after 20 minutes.

[narrowbandpaul]

actually I am even more interested in your results.....

i have recently been using a modded 450D....

so I am keen to know your numbers...

saves me doing it! :-)

cheers

paul

[jgs001]

Damn... bad timing on my part

[narrowbandpaul]

on the contrary.......

[jgs001]

I'm messing something up in here I think... Here's my results...

Iso Gain Read Noise Full Well Dyn Range

100 286.64 4.05 40989.52 10111.63

200 274.60 5.83 39267.80 6741.08

400 268.09 20.85 38336.87 1838.48

800 268.17 60.68 38348.04 631.98

1600 271.30 109.35 38795.56 354.79

[narrowbandpaul]

did u remeber to square the standard deviation for the frame differenced images when calculating gain.....

ie.....

G= (F1+F2)-(B1+B2)/( (std_flat1-2)^2-(std_b1-2)^2

check to make sure you used the correct equations

[jgs001]

Paul, I've checked the formula a dozen times (before posting those up)... I'm using an Excel spreadsheet to do the number crunching for me.

The only thing I can think is I didn't take the frame properly...

Flats, I did what I do for all my imaging... I used my lappy screen running VBL maximised, camera set in aperture priority, only varied the ISO. I shot with a lens on, but set to infinity, manual focus, inside min focus distance, at f8 to eliminate risk of vignetting.

Bias, Shutter priority, highest shutter speed 1/4000 with the lens cap on.

Here's the values from the Bias Frames

ISO Mean St.Dev.

100b1 0

100b2 0 0.02

200b1 0

200b2 0 0.03

400b1 0

400b2 0 0.11

800b1 0.02

800b2 0.01 0.32

1600b1 0.04

1600b2 0.04 0.57

[narrowbandpaul]

send me two bias frames and two flats for each ISO

and I can get them analysed

why shoot with a lens? unlensed works well and helps assure an even illumination.

just send me your cropped 100x100 boxes....thats all i need.

paul

[jgs001]

I tried without the lens, and after trying for ages and not getting sensible results, decided to reshoot with the lens.

Pm me your email addy and I'll send you a zip

[narrowbandpaul]

pm'd

[jgs001]

email on the way

[narrowbandpaul]

excellent...i get to do some math now!

[narrowbandpaul]

john, I have emailed you the initial results!