Pedestal for flats

Wesley Mitchell

Hi,

Is there anyway to apply an input pedestal to my flats using WBPP? I see that I can do it if I run flat calibration as a separate step using ImageCalibration.

Thanks

Wes

BlockHead

You could use Pixel Math and an image container... but no, not in WBPP.

But... is this a one-off issue? Why do you need to do this?

-the Blockhead

Wesley Mitchell

This is the only way I've made progress at all on my over-correcting flats. I found that by adding a pedestal (output, not input) in image calibration, I could get a master flat that doesn't over correct.

BlockHead

Hmm... this just points to a calibration issue.

I assume you are calibrating the flats.

You need to look at the value of the bias/dark you are using.

It is too large (and this means you are subtracting too much).

Why is it too large?? That is what needs to be discovered by testing the camera and acquisition software.

Is your camera subtracting (calibrating) the flats for you (and you don't know it). All kinds of goofiness is possible. 

Whether application of a pedestal is done through a file or by hand- doesn't matter really But the file you generate contains pixel-to-pixel correlations/information that your addition of a value will not capture.

My solution still stands. You could run the image container on the flats before using WBPP... but I think this is an easy thing to figure out. It is a single subtracted value coming from your fle.

-the Blockhead

Wesley Mitchell

OK, this gives me something to look at. I have of course examined the values of the bias and dark frames. They are pretty uniform, right around .0024-.0025. 

The Ha/SII subs are really dark, background around .005 ± .0003. The defects in the subs are very faint indeed, not really visible.

I took my flats around the ADU midpoint, could they be too bright?

The camera is a player one poseidon-m, IMX 571 chip. I have the gain set at 125, which is right past the point where the low-read-noise mode kicks in. Is there something I can glean from the performance curves?

I have the offset when taking darks set to 20. Should I try setting it to zero? I set it to 20 because I noticed I was getting some zeroes in the dark frame when the offset was zero.

How do I go about testing the camera? What should I be looking for?

Does it boil down to how I am taking the darks? This is at a remote site, the dust cap is on, I point to a high altitude some time after astronomical dark to take the darks. Should I ask them to also put a cloth over the scope to further reduce light?

Thank you. These issues are very frustrating to me.

Wes

BlockHead

You mention the offset you apply to the darks... but did you have the same offset and gain for the Ha/SII subs?

The offset and gain of your calibration frames need to match your subs. That is an instant no-go if not.

So...here is the question. Your Ha/SII subs are effectively dark frames right? 

Unless you are imaging under bright skies (and long exposures)- maybe the darks and lights should have a background that is closer to each other? It could be 0.005 is OK...but I expect almost like a dark frame.

So your offset of 20 (or whatever) not giving you zeros in darks is good. There is no really issue with making it larger. 

Flat field errors are almost never about the flats being too bright or dim- other than noise considerations. It isn't this.

I still think something isn't matching...and that is the reason for the issue(s).

-the Blockhead

Wesley Mitchell

Interesting comment about the ha/sii subs effectively being dark frames. I never imagined a difference of 0.0025 would be significant ... but .... I subtracted off .002 from the sub and calibrated with the master flat and master dark using PixelMath, and got a near perfect calibration...

The offset  for the subs was 40, not 20. That's a difference of 20/65536, which is only 0.0003, but I'll try to get some exposures with offset 20 just to double check. Gain is 125 on all frames.

The issue is that the expression med(flat)/flat is completely non-linear in both the values from the raw flat and the subtracted values from the bias, so it's really hard for me to understand how the fraction behaves when either quantity is varied. For pixels where flat is below the median, the expression gets larger (more correction is applied) as you subtract a constant from each pixel. That's why I was chasing adding a pedestal to the flat, since that would make the correction smaller.

BlockHead

The offset and and gain really must match to be certain you get the intended results between calibration data and subs. There isn't wiggle room.

When the camera applies an offset of "20" and I not convinced this equals "20" in 16-bit sense. The offset is a hardware applied voltage or ADC counting scheme..not just a straight up number in the sense I think you mean it. My gut says I am on the right track here.

-the Blockhead 

Wesley Mitchell

Well, congratulate your gut appropriately. Once I gathered some new data with the correct offset of 20 to match everything else, it all fell in line and gave me nicely calibrated masters.

Thanks

Wes