DSLR RAW Files - Black Point Correction

[Phil Leigh]

If I set DSLR_RAW prefs to black point correction enabled or not enabled it seems to make no difference... the image statistics are identical. A Bias frame for my 14-bit camera is ~512 regardless. Does that seem correct or would you expect to see any difference?
thanks
Phil (confused)

[bitli]

This is indeed the case for the Canon 350 (although the black point is at 16 AUD on 4096).
This depends on dcraw and the format of the DSLR raw file.  In practice all DSLR I have examined (just a few) have no real raw, they already do some calibration of the black point and certainly dead pixel correction.  It is better to ask for no black point correction to be on the safe side, but it does not seems to do much differences.

As a side effect you will see that the darks are also already partially calibrated.  Their median value will be about the same or lower than the bias.  So you should not create calibrated darks, unless you are ready to work with PEDESTAL. You do not need to calibrate the darks unless you want to 'optimize' (scale) the darks and this does not seem to work well (or at all) for DSLR (this may be different depending on the DSLR, this is not an issue of PixInsight, but of the pre-processing already done in most DLSR). For the few NIkon I have seen, this is even worse - most pixels are at zero on Bias and Dark. I am not even sure you want to bother with bias and darks on a Nikon, except maybe to help CosmeticCorrection to correct the hot pixels.

Naturally there may be some esoteric settings that I missed, tell me if you find them.

-- bitli

[Phil Leigh]

Hi Bitli - and thanks for confirming that setting the black point or not doesn't really do much.

I don't use darks at all (super cooled "DSLR's" - there's nothing in the darks that isn't removed by dithering) but I do use 200-frame SuperBias to calibrate flats and lights - seems to work OK... The Pi noise stats for flats and lights are definitely lower if I calibrate them with the SuperBias.

All the (14-bit) bias frames I've ever seen had a median close to 512 on the 14-bit scale in Image Statistics. However, a friend has just shown me his master bias from a 12-bit Canon and it has a median of 15.9 so I'm a bit confused... I was expecting his bias to be ~128.

[bitli]

Using the superbias is fine.  It is just subtracting the bias from the dark which does not work. I use the darks because I have the 'amplifier glow' (and sometime also 'battery heat glow'). Removing the gradients is much more critical than removing hot pixels, as having gradients from the sky, the 'ampligier glow' and may be less than perfect bias make it very difficult to remove.  Hot pixel will go away with dithering as you mentioned (or with CC).

The level of the Bias fixed by Canon is completely arbitrary (well, they want it low enough but not tool low taking into account manufacturing tolerance). So you do not see 128 on the 350 because it is not the same electronic as you model which would be scaled down. It is a different electronic, different amplifier, different AD converter, different temperature compensation and so on.  The 'pedestal' was chosen independently in the two type of DSLR.  Or maybe the engineer of the latest did not both looking at the value of the former :-)

May I ask which model you have?

-- bitli

[cdesselles]

On the subject of "super bias".  Exactly how does one go about producing only such a master bias to be able to save it for multiple uses on future calibration of flats and lights?