Why does changing the Offset on an OSC do this?

[vineyard]

Hello,

So I've been experimenting with my new (2nd hand) ASI294MCPro.

Here are 2 images of M31.  Both are 60 lights of 60s each. Both at -5C with gain 125.  Both calibrated w the identical dark, dark flat & flat files.  Both then just run through ABE & HT in PI.

The 2nd image shows the dust bunnies and the weird coloured corners (I actually encountered them on another image earlier & thought maybe it was dew!).

The only difference is that the 2nd image was taken w offset set to 64 (yes I know too high) whereas the first image w offset set at 0.

How does offset have such an impact on the image?  And does this mean that my offset64 lights are unusable & I should just trash them - or can they be salvaged somehow (I've tried taking flats etc w the same offset but it makes no difference).

Thank you!

Vin

(PS - I think I'm going back to offset=0 )

[MarkAR]

First image is good, second image is ART. Don't think it's worth the hassle of trying to recover them even if it's possible.

 

[vineyard]

22 minutes ago, MarkAR said:

First image is good, second image is ART. Don't think it's worth the hassle of trying to recover them even if it's possible.

 

Darn, I should have submitted it for that $10k prize!  Maybe next year 😂

[geordie85]

My guess is that it's something to do with calibration as I cannot see how changing the offset could result in this.

I may we'll be wrong though, I have been many times before 

[andrew s]

Can you post the histograms for the two images.

Did you used calibration frame with matched offset  to the two images or just one set for both?

Regards Andrew 

[ollypenrice]

46 minutes ago, MarkAR said:

First image is good, second image is ART. Don't think it's worth the hassle of trying to recover them even if it's possible.

 

Art? That's a bit hard on art! The second image is just wrong. Wrong does not equal art except in the parallel universe of that stupid astrophoto competition. 😁

In my understanding the offset in OSC imaging concerns the simple matter of which pixel has been filtered red, which green and which blue, across the chip. This is the Bayer Matrix offset. Not all chips have the same pattern so, for example, the top left pixel on your chip might be under a red filter, or a green, or a blue. Your stacking/calibrating/debayering software needs to know which pixel is filtered to give which colour or, obviously, it cannot produce a correct interpretation of the data.

If 'offset' has been given another and different meaning since I last calibrated OSC data I may be entirely wrong.

Olly

[tomato]

I think offset in relation to a CMOS camera is designed to shift the mean of the signal histogram by a fixed ADU value to the right.
 

Checkout this thread:

 

[vineyard]

@andrew s here are the histograms (offset 0 first and then offset64).  The offset has definitely boosted the left hand side a lot.

@tomato yes i had come across that thread when I was researching the ASI294MC before buying it.  Needed a cold towel on my head then (& on re-reading just now I still do!). It was that thread that prompted me to try the higher offset, in the hope that it would improve SNR, but it seems to have opened a can of worms!  (One thing I think I can confidently say is that it won't be AWB as I've never touched that & have only ever used it w EKOS).

@geordie85 & @ollypenrice I don't even know enough to feel comfortable guessing - it does seem v strange.  I'm quite curious to understand why this would be happening (there must be maths behind it), but I think come this Thursday's clear night I will be back at offset 0 - stick to what works sufficiently well for my purposes

[gorann]

@vlaiv please explain to us what offset on CMOS means. I think it has to do with pushing the histogram to the right, but you can probably explain it elegantly.

 

[andrew s]

Why only 8 bit as I think it's a 14 bit camera which converst its output to 16bit? Can you confirm you used calibration frames with 64 offset for the 64 offset image? 

My understanding of offset is that it is a bias added to the measured signal to prevent negative values. I.e. it moves the histogram to the right.

Regards Andrew 

 

[DaveS]

I found This on the QHY site, which you are welcome to rubbish as advertising puff.

[JamesF]

46 minutes ago, andrew s said:

My understanding of offset is that it is a bias added to the measured signal

Mine too.  Not that I have a CMOS camera that has such a control

James

[vineyard]

Thanks all. I can sort of intuitively follow the logic but it's puzzling that calibration doesn't take it away.  I thin its definitely the flats that are creating those weird effects b/c if I stack w/o them those don't happen (but I get dust bunnies).

@andrew sto your Q, that's just the histogram showing in 8 bit (I can change it to display more fine, the curve doesn't change but there's more precision to the curve).  On DSS the lights show up at 16bit Gray FITs.  I've double-checked and if I stack with gain 125 offset 64 flats as well, the same effects happen.  It's almost like the flats are too 'strong' (have been amplified in some way) and then when they get taken away from the lights there's too much signal being lost?  But I could be getting the logic all wrong there! 

[andrew s]

3 minutes ago, vineyard said:

Thanks all. I can sort of intuitively follow the logic but it's puzzling that calibration doesn't take it away.  I thin its definitely the flats that are creating those weird effects b/c if I stack w/o them those don't happen (but I get dust bunnies).

@andrew sto your Q, that's just the histogram showing in 8 bit (I can change it to display more fine, the curve doesn't change but there's more precision to the curve).  On DSS the lights show up at 16bit Gray FITs.  I've double-checked and if I stack with gain 125 offset 64 flats as well, the same effects happen.  It's almost like the flats are too 'strong' (have been amplified in some way) and then when they get taken away from the lights there's too much signal being lost?  But I could be getting the logic all wrong there! 

Yes, definitely looks like the flats. Sorry I am not familiar with the software you use so can't help. One test might be to use your master flats to treat a subset of the flats as if they were an image and see if that works.

Regards Andrew 

[vlaiv]

Offset does exactly what has been mentioned above - it "moves" histogram to the right / adds "constant" value to all pixels.

Issue with second image is not with flats - it is the fact that lights were taken with certain gain and offset and dark calibration was performed with different offset (possibly gain as well?)

This results in light not having only light signal in them but some of the offset as well (not properly removed by mismatched darks). When flat calibration is applied - it assumes only light signal is present. If this is not the case, flat correction "works" on offset as well and this creates mess since offset was not affected by light throughput in the first place as it was not part of light signal.

You should use higher offset - nothing wrong with that, but you need to match your darks to lights in offset and in gain. You can use flats that were shot at different offset/gain, but you must calibrate those with matching flat darks.

To reiterate:

Lights === darks

and

Flats === flat darks

but

Lights !== flats

=== is symbol for matching offset, gain, exposure length and other parameters. !== is opposite - no need for matched parameters.

 

[vineyard]

Thanks @vlaiv - I went back & checked and you're absolutely right, the darks I was using were offset 0 (gain was 125) - I had foolishly just assumed that since it was dark, as long as the temperature & exposure time were the same, that would be fine...

You're a genius (& a v helpful one at that too).

I will run new darks (gain 125, offset 64) tomorrow and report back!

Cheers all.

[MarkAR]

Cor Blimey, then it can be saved. Well done @vlaiv for clarifying things.

[vineyard]

It worked! Here's two images.

The first is the horrible one from up above, but this time done with darks set at the same offset.

The second is a stack of all the lights from the two different sessions. Same flats & dark flats used for all files.  But the offset0gain125 lights and darks were put together & the offset64gain125 lights and darks were put together.

All I've done is DSS, ABE & then HT - no other tinkering.

@vlaivyou're a star, thank you - I am going to revisit my Heart & Soul lights now

Cheers - thank you ALL for your engagement with this (incl @gorannfor tagging @vlaivinto the thread!).

Stay safe,

Vin