Superbias Trouble

[bmhjr]

I am trying to make a superbias for my T3i.  I integrated 100 bias frames to make a masterbias.  The resulting superbias shows what looks to be some banding noise it may have picked up.  Is there any way to fix this superbias so it is usable?  The masterbias and superbias are shown below.

Bill

[TobiasLindemann]

Hi Bill,

what can help is, to take more biases for the masterbias, or to increase the the multiscale layers in the superbias script. Maybe 10 layers do ab better job in your case.

Tobias

[MikeOates]

Bill,

Bear in mind that the Superbias is stretched and may just look bad, it may only have a few ADU difference between the darkest and lightest areas. In which case, that may well be the true bias for your camera.

But as Tobias suggests, take more bias images, you really need more especially for a DSLR, 200 frames will make a big difference.

Mike

[dnault42]

Superbias is really designed for column or row oriented sensors like a CCD camera.  Most CMOS cameras have both row and column patterns giving you lumpy patterns like this due to the mixing of these patterns.

You can still get a good Superbias for a CMOS sensor but you have to go through a few more steps:
1) Run Superbias on your master bias image in Column mode
2) Bring up PixelMath and subtract the superbias from the bias with an offset: master_bias - superbias + 0.1
3) Run Superbias on this new image except switch to Row mode
4) Bring up PixelMath again and add the first and second superbias images subtracting the offset added in the first pass: superbias + superbias1 - 0.1

The resulting bias should have both column and row patterns.

Regards,
David

[bmhjr]

Thank you everyone for the advice.  I will take some more bias frames and then try the column/row steps as suggested.

Thanks,
Bill

[Carlos Milovic]

You may also try increasing the number of layers. This will reject more structures into the superbias.

[fnord123]

Superbias is really designed for column or row oriented sensors like a CCD camera.  Most CMOS cameras have both row and column patterns giving you lumpy patterns like this due to the mixing of these patterns.

You can still get a good Superbias for a CMOS sensor but you have to go through a few more steps:
1) Run Superbias on your master bias image in Column mode
2) Bring up PixelMath and subtract the superbias from the bias with an offset: master_bias - superbias + 0.1
3) Run Superbias on this new image except switch to Row mode
4) Bring up PixelMath again and add the first and second superbias images subtracting the offset added in the first pass: superbias + superbias1 - 0.1

The resulting bias should have both column and row patterns.

Regards,
David

Is this still the recommended way of doing superbias for a CMOS-based sensor (e.g. ASI1600mc)?  Why wouldn't one use the "Vertical and Horizontal" option in the Superbias process instead?  Is that for something else?

[dnault42]

Using the "Columns and rows" Orientation option doesn't work because both will contain the large scale gradients common on bias frames.  If added together (even if removing the median of one of the images to normalize it) those gradients will be amplified and cause over subtraction issues during calibration.

I used the process I described below just a few days ago on some DSLR data and it worked well.

By the way, this wasn't my original idea.  I think Juan described it first or maybe it was Carlos (I definitely saw it on this forum).

Regards,
David

[steven_usa]

Just wanted to ask -- was there a mix up in this thread?

Isn't T3i/600D a CMOS and ASI1600MC is a CCD?

So superbias is NOT recommended for CCDs for CMOS?   (for the reasons cited earlier)
EDIT: corrected.

[dnault42]

The T3i/600Da and the ASI1600MC are both CMOS cameras.  The ASI uses the Panasonic MN34230ALJ sensor (here's the datasheet for it).

Regards,
David