Determining offset

[Shimonu]

Is there a methodology to determining what offset should be used with a camera? I understand it's somewhat individual but what kind of analysis can I do to find a good value or know that I have the wrong value?

[Seelive]

My first guess would be to look at the (linear) histogram, if you've got a spike at 0 ADU then then the offset needs to be increased?

[vlaiv]

Here is methodology that I would suggest:

- Set other parameters as per your preference (gain, temperature, whatever) - and start with some offset.

- determine what is minimum value your camera outputs (probably 0 but it can sometimes happen that floor is not 0)

- take set of bias subs. Maybe couple dozen of them - stack using minimum method

- inspect resulting stack for minimum pixel value - if it is equal to floor value of your camera (again, probably 0 value) - increase offset and go to step 3, if minimum pixel ADU is larger than floor - you are done.

[Shimonu]

6 minutes ago, vlaiv said:

Here is methodology that I would suggest:

- Set other parameters as per your preference (gain, temperature, whatever) - and start with some offset.

- determine what is minimum value your camera outputs (probably 0 but it can sometimes happen that floor is not 0)

- take set of bias subs. Maybe couple dozen of them - stack using minimum method

- inspect resulting stack for minimum pixel value - if it is equal to floor value of your camera (again, probably 0 value) - increase offset and go to step 3, if minimum pixel ADU is larger than floor - you are done.

Thanks! Just a couple of questions. How do I find the minimum my camera outputs, is this some publicly available info or something I need to find by testing? What do I use to inspect the output values? I'm guessing there is some tool that can give something like a min pixel value?

[vlaiv]

1 minute ago, Shimonu said:

Thanks! Just a couple of questions. How do I find the minimum my camera outputs, is this some publicly available info or something I need to find by testing? What do I use to inspect the output values? I'm guessing there is some tool that can give something like a min pixel value?

To me, simplest way would be to put offset to 0 and then do single bias sub.

Load that sub into any tool that will inspect things for you and get min value and histogram. With low offset - you should get clipping to the left of histogram - a bit looking like this:

bell curve will get clipped on left side. If this happens - you should look at min value and that is minimum pixel value that your camera outputs. In 99% cases it will be 0.

As for tool - I use ImageJ. It is free/open source tool for scientific image analysis. It opens FITS files - and can plot histogram, do analysis of image and so on.

We add offset precisely to avoid this sort of clipping to the left. If read value is 0 - you have no idea if actual pixel value was less than zero or zero and that is bad. For this reason we want bell shaped curve to be above zero in 99.99% of cases. If you shoot couple dozen of bias subs (or even say 8 of them) and look for minimum pixel value among all subs and all pixels - you are just decreasing probability that any pixel will be at or below floor (read noise has gaussian distribution and there is always chance some pixel will have lower value - point is decreasing that chance to acceptable minimum).

 

 

[bottletopburly]

i just asked astro bin users what offsets they used for the specific camera used .

[Shimonu]

On 24/07/2021 at 20:48, vlaiv said:

Here is methodology that I would suggest:

- Set other parameters as per your preference (gain, temperature, whatever) - and start with some offset.

- determine what is minimum value your camera outputs (probably 0 but it can sometimes happen that floor is not 0)

- take set of bias subs. Maybe couple dozen of them - stack using minimum method

- inspect resulting stack for minimum pixel value - if it is equal to floor value of your camera (again, probably 0 value) - increase offset and go to step 3, if minimum pixel ADU is larger than floor - you are done.

I figured this sounded easy enough but I hit a slight snag. Which tools allow you stack using minimum? I checked both AstroPixelProcessor and DeepSkyStacker but they don't have options for minimum. So is it something you do manually in an editor like Photoshop or GIMP?

EDIT: Even manually going through the bias frames in ImageJ I found an offset of 5 will give a minimum above 100. But from other comments I read that seems like a very low offset as people are saying even 8 which in some cases is default is very low. How temperature sensitive is this? I haven't exactly been running the cooling at max.

Edited July 25, 2021 by Shimonu

[vlaiv]

1 hour ago, Shimonu said:

I figured this sounded easy enough but I hit a slight snag. Which tools allow you stack using minimum? I checked both AstroPixelProcessor and DeepSkyStacker but they don't have options for minimum. So is it something you do manually in an editor like Photoshop or GIMP?

ImageJ will let you stack with minimum.

Just form a stack from your subs (which is just a "pile" or "collection" of images - in ImageJ lingo that is stack whether you stack it or not - you can simply do it by "File / import / image sequence") and select:

Image / Stack / Z Project and select minimum method:

1 hour ago, Shimonu said:

EDIT: Even manually going through the bias frames in ImageJ I found an offset of 5 will give a minimum above 100. But from other comments I read that seems like a very low offset as people are saying even 8 which in some cases is default is very low. How temperature sensitive is this? I haven't exactly been running the cooling at max.

If you shoot bias - which means minimum exposure time, temperature should not make much of a difference - but in principle - it is temperature dependent. As temperature causes dark current buildup and dark current will also raise signal level - it can skew results.

It will be gain dependent as well - that is why I mentioned that you should set your parameters the same way as when imaging - gain, temperature, etc ... and only vary offset to determine which value to use.

[Shimonu]

15 hours ago, vlaiv said:

ImageJ will let you stack with minimum.

Just form a stack from your subs (which is just a "pile" or "collection" of images - in ImageJ lingo that is stack whether you stack it or not - you can simply do it by "File / import / image sequence") and select:

Image / Stack / Z Project and select minimum method:

If you shoot bias - which means minimum exposure time, temperature should not make much of a difference - but in principle - it is temperature dependent. As temperature causes dark current buildup and dark current will also raise signal level - it can skew results.

It will be gain dependent as well - that is why I mentioned that you should set your parameters the same way as when imaging - gain, temperature, etc ... and only vary offset to determine which value to use.

I'll see if I can do it with more consistent and realistic temperatures.

Thank you very much for the help!