Question for QHY268c users….

[Stuart1971]

I now have this camera and was out last night imaging the iris nebula using 2 min subs with just an Idas P2 light pollution filter from my bottle 6 backyard, now I have read lots of articles about the relationship between gain and offset, and for the most part I get it, but I used the lowest setting last night, the photographic mode and gain 0, but then when watching the histogram in NINA, after running test 2 min subs, I had to move the offset up to 100 to get a small gap to the left of the histogram curve, this seemed very high, as it only goes up to 255 so not much room for using with higher gain modes…

so was it too high or am I missing something, I am coming from a few years of CCD imaging, so this gain and offset stuff is all new to me….??

[teoria_del_big_bang]

Hi Stuart, Interesting and off the top of my head I can't think why you would need such an offset.
I must admit since I got my mono version of this camera I have had so few clear nights I do not have that much data to look back on and I hold my hands up to being a bit blase in that I have not really paid too much attention to the left side and where it cuts off. In essence I saw people were using offsets between 10 and 60 and in the early days I did look at this and settles for an offset of 25 (I normally use gain 56, offset 25 in mode 1).

I know its a mono version but essentially the same chip and it is colour when I put the filters in front so will look at my RGB data from last night ( normally would use NB filters but with the little moon last night and pretty clear skies early on took advantage to use my neglected RGB filters) so will take a look later today when I get chance.

Could it be anything to do with the IDAs filter ?  I doubt it and just pose the question without any science behind it. I used to use the same filter in my flattener permanently fitted but took it out at the start of the year to see if it was reducing the collected light too much as I do not have masses of LP with a view to doing some tests, as it is a 2" filter that fits permanently in the flattener it is not so easy to compare as unless I take the images on the same night with same PL and same moon presence then cannot really compare and again with so few clear nights just never got on with the test.

I will have a look at some of my data this year and get back to you if thats of any use.

And I make my new years resolution now to get back into this hobby properly and to try to make the best of any clear night, got rig out last night and was appalled (at the weather but also myself 🙂 ) that when I looked at previous data the last session had been early September 😞 

Steve

[teoria_del_big_bang]

Looking at my data from last night there looks plenty of flat line to left of curve.

So this is mode 1 offset 25. In all cases I have set that Vertical line to the last flat bit of the line where pixels read zero value. So in the RED image for example the histogram reads zero up to 0.0239 normalised (or  an actual value of around 1566) before any pixels actually have a value. 

Red

Green very similar at 0.0242 normalised.

And blue also similar at 0.0207

 

Whether this helps you or not I am not sure.
If you want to attach a raw image I can check in PI the same way if you want ?

Steve

[teoria_del_big_bang]

I also have some earlier Ha data on same target from Feb 2021 and I think I was using photographic mode 0, oddly enough the mode is not in the fits header for these images (it is in my more recent RGB images) but I am sure I started out with mode 0 then changed quite recently to mode 1.

What is in the header is gain and offset so Gain was 40 and offset 20 (not really sure why I chose those) and the area at the left is a bit less but its still there at 0.0065 normalised. So its pretty small but so long as it is there then that should be fine I think.

Steve

[david_taurus83]

Don't these have a low gain mode and a high gain mode? Where the likes of the ZWO version automatically switches to the higher mode at a certain gain value?

[Stuart1971]

10 minutes ago, david_taurus83 said:

Don't these have a low gain mode and a high gain mode? Where the likes of the ZWO version automatically switches to the higher mode at a certain gain value?

Yes, well they actually have 4 gain modes, 0,1,2,3,  and also different modes within each of those 4 too, but that was not my question, I was using mode 0 as stated, but wanting to know if an offset of 100 seemed high, for a mode 0 and gain 0 exposure…as that was what I had to use to get no pixels with a value of zero…it just seemed very high…

[david_taurus83]

Did NINA say they had a min value of 0 or just eyeing the histogram?

[teoria_del_big_bang]

Actually me comparing my 5 and 10 minute exposures that is also at a different mode is maybe not that helpful when your exposures were only 2 minutes.

Of course a much shorter exposure will have the histogram shifted much further to the left, with the same settings but if you were to then take a 10 minute exposure the whole histogram would move to the right.

Also won't using a zero gain also cause the whole histogram to move to the left  because you now need a lot more photons to reach fullwell ?

And using gain zero will also cause there to be quite noisy at low exposures.

So maybe the issue is just the short exposure (as it is a test shot) and using gain zero that means you have to use a high offset and if you use a longer exposure the whole histogram will move to the right and also a higher gain value would also move it to the right.

Now I think this is correct but I admit some of this I do not fully understand which is why I took an interest in your thread hoping somebody with much more experience could explain the issue.
It is my lack of knowledge why when using mode 0 I chose a gain to the right of that big drop at gain 30 to reduce the readout noise.

Does any of that make sense or am I barking up the wrong tree here ?
Maybe needs somebody better at the technical side of all this like @vlaiv ?

Steve

 

[vlaiv]

In order to figure out best offset setting for given gain - you only need bias exposures.

Procedure is rather simple:

1. set some offset

2. shoot a number of bias subs with camera (say 16)

3. stack those subs using minimum method

4. examine statistics of resulting stack. If minimum pixel value is larger than 0 (or whatever minimum value for camera is) - you are done, if not, raise offset and go to step 1.

Visually, histogram can be "glued" to left side - but that is just visual thing, you really need statistics in order to know values. If you have 0-65555 levels in pixel value and histogram on screen is say 300-400px wide, how can you tell if histogram is at pixel value 56 or 72 or is at 0? Visual histogram does not have enough resolution to provide feedback for that.

Btw, setting offset a bit "too high" is really not a big deal - it won't affect your image at all in the end. It will reduce your full well capacity a bit - but that is not big deal, you'll need to take short exposures anyway if you want to capture star cores that are usually blown in regular exposures.

[Stuart1971]

40 minutes ago, vlaiv said:

In order to figure out best offset setting for given gain - you only need bias exposures.

Procedure is rather simple:

1. set some offset

2. shoot a number of bias subs with camera (say 16)

3. stack those subs using minimum method

4. examine statistics of resulting stack. If minimum pixel value is larger than 0 (or whatever minimum value for camera is) - you are done, if not, raise offset and go to step 1.

Visually, histogram can be "glued" to left side - but that is just visual thing, you really need statistics in order to know values. If you have 0-65555 levels in pixel value and histogram on screen is say 300-400px wide, how can you tell if histogram is at pixel value 56 or 72 or is at 0? Visual histogram does not have enough resolution to provide feedback for that.

Btw, setting offset a bit "too high" is really not a big deal - it won't affect your image at all in the end. It will reduce your full well capacity a bit - but that is not big deal, you'll need to take short exposures anyway if you want to capture star cores that are usually blown in regular exposures.

Yes, this is what I have now learnt, to set the offset from a bias stack, and that it makes no difference what the exposer length is, so if I do the procedure you mention, with a bias stack and let’s say I use a Gain of 0, and an offset of 30,  and this works well, then whether I take a 2 min sub or a 10 min sub at gain 0, the same offset of 30 will apply…correct…?

[Stuart1971]

2 hours ago, david_taurus83 said:

Did NINA say they had a min value of 0 or just eyeing the histogram?

Yes, just eying the histogram, I have now learnt about the reading the statistics instead….this gain and offset lark is all me to me after years of using a CCD and not having to know about it…

[Stuart1971]

I think because I was eying the histogram in NINA, in a small window, this was my issue, the curve looked as if it was hard against the left edge, so I kept raising the offset to move it to the right, but now I put the graph full screen I see there is a big gap to the left and the lowest pixel value with 100 offfset was actually 1580…so it was way to big, but no harm done….

[david_taurus83]

2 minutes ago, Stuart1971 said:

Yes, just eying the histogram, I have now learnt about the reading the statistics instead….this gain and offset lark is all me to me after years of using a CCD and not having to know about it…

Yes, keep an eye on the numbers in the statistics. It will tell you the mean reading as well as the max, 65k on a star for example, and the min, your bias number. 

[Stuart1971]

1 minute ago, david_taurus83 said:

Yes, keep an eye on the numbers in the statistics. It will tell you the mean reading as well as the max, 65k on a star for example, and the min, your bias number. 

Yes, so the Min figure shown is the one I need to alway be above 0….correct…

[david_taurus83]

Just now, Stuart1971 said:

Yes, so the Min figure shown is the one I need to alway be above 0….correct…

Yes, it will be, or should be always. When you take a light you should never have a min value of 0. You'll have your bias signal in there as well. The offset value is added to this so even dead pixels will have a value above 0. So when you subtract a flat dark or dark you should never create 0 value pixels in your image because the offset keeps it above the threshold.

[vlaiv]

34 minutes ago, Stuart1971 said:

Yes, this is what I have now learnt, to set the offset from a bias stack, and that it makes no difference what the exposer length is, so if I do the procedure you mention, with a bias stack and let’s say I use a Gain of 0, and an offset of 30,  and this works well, then whether I take a 2 min sub or a 10 min sub at gain 0, the same offset of 30 will apply…correct…?

Yes, offset is related to readout of signal - it does not matter how long it was exposed. As long as you have it above zero with bias - anything else will just add signal and raise it more above 0 - either dark current in darks or target / LP signal in lights.

[davidmalanick]

On 06/01/2022 at 15:17, vlaiv said:

In order to figure out best offset setting for given gain - you only need bias exposures.

Procedure is rather simple:

1. set some offset

2. shoot a number of bias subs with camera (say 16)

3. stack those subs using minimum method

4. examine statistics of resulting stack. If minimum pixel value is larger than 0 (or whatever minimum value for camera is) - you are done, if not, raise offset and go to step 1.

Visually, histogram can be "glued" to left side - but that is just visual thing, you really need statistics in order to know values. If you have 0-65555 levels in pixel value and histogram on screen is say 300-400px wide, how can you tell if histogram is at pixel value 56 or 72 or is at 0? Visual histogram does not have enough resolution to provide feedback for that.

Btw, setting offset a bit "too high" is really not a big deal - it won't affect your image at all in the end. It will reduce your full well capacity a bit - but that is not big deal, you'll need to take short exposures anyway if you want to capture star cores that are usually blown in regular exposures.

This seems like the easiest method of computing offset I have ever seen. Could you please go into more detail on program usage?  I assume you are using Pixinsight. That is what I used for step 3. My question is on #4, where do you find minimum pixel value?  What process, what field is it?

Thanks!

 

[vlaiv]

7 hours ago, davidmalanick said:

This seems like the easiest method of computing offset I have ever seen. Could you please go into more detail on program usage?  I assume you are using Pixinsight. That is what I used for step 3. My question is on #4, where do you find minimum pixel value?  What process, what field is it?

Thanks!

 

Hi and welcome to SGL.

I don't use PixInsight, but rather use open source software called ImageJ.

For above procedure you only need two commands:

- image / stacks / z-project (with min as projection method)

- analyze / measure

(these are menu items).

I'm sure PixInsight also allows for above to be done easily, just look up how to check image statistics (average, median, min, max, standard deviation and such - and use min), and since it can stack - I'm sure it can use minimum as stacking method.

[Richard_]

On 09/07/2022 at 10:02, vlaiv said:

Hi and welcome to SGL.

I don't use PixInsight, but rather use open source software called ImageJ.

For above procedure you only need two commands:

- image / stacks / z-project (with min as projection method)

- analyze / measure

(these are menu items).

I'm sure PixInsight also allows for above to be done easily, just look up how to check image statistics (average, median, min, max, standard deviation and such - and use min), and since it can stack - I'm sure it can use minimum as stacking method.

That's correct, you can go to Process > All Processes > Statistics. From the drop down menu you'll need to select the sub in question. 

There's another drop down menu underneath image name where you can choose bit depth (most people use 16-bit to represent 0-65,535) and whether to normalise the values or not (in other words, show as 0-65,535 or replace with a scale of 0-1 I believe). 

In the below example which I pulled from Google images, the sub is called "Saturated_Image2", with a 16-bit depth. The minimum here is shown as 14,190.0.