Jupiter comparison, 70% v ignored level histogram.

[Pete Presland]

Just before i venture out again tonight, i would like to share a quick experiment from last night. 

As some of you might have seen, i have recently had a fascinating conversation with @vlaiv regarding Histogram levels, exposure gain....

So last night i tried ignoring the histogram level and just setting the exposure to around 6.5ms and the gain to 350. The Histogram was down at 28%, so i had to focus before reducing the gain.

I think the results are very interesting to be honest.  All SER files were batched stacked together in A/S3 and processed the same other than the curve adjustment was larger on the darker image.

There was clearly much less noise in the 28% Histogram image, which probably allowed for  slightly more aggressive sharpening. Although only a subtle amount more.

Usual settings 70% histogram, straight from A/S3

350 gain setting 28% histogram, straight from A/S3

Usual settings 70% histogram, after processing in CS2, curves, unsharp mask.

350 gain settings 28% histogram, after processing in CS2, curves, unsharp mask

I think it looks a better, more detailed image to me.  I am going to try it again tonight 🙂 

 

[neil phillips]

 Hi Pete Did you use the same sharpening on both images ? other wise the results are meaningless ?  Exposure and gain are the things that will influence the most. No need to have a impossibly low level. From my experiance gain noise is not always welcome either. Not just read noise.  Gain noise will require increasingly larger stacks to avoid a million dots.  How long was these captures ? Just tried to sharpen them. Both images are impossibly noisy compared to a 3 min stack i normally take

Edited September 23, 2021 by neil phillips

[Pete Presland]

The sharpening was almost identical, i felt i could sharpen the bottom image slightly more due less noise from a lower gain setting.  I thought that altering the gain, altered the read noise?

I am stacking 20% of approx 15000 frames.

[neil phillips]

50 minutes ago, Pete Presland said:

The sharpening was almost identical, i felt i could sharpen the bottom image slightly more due less noise from a lower gain setting.  I thought that altering the gain, altered the read noise?

I am stacking 20% of approx 15000 frames.

Of course Setting a gain setting too low will increase read noise. But too high a gain will also produce noise. But it will be gain noise. All i know is a 3 min stack produces around 22.500 frames at 125/secs exposure. With my 245mm mirror. Even a conservative 25% Will give 5750 frames. I use winjupos for considerably more frames. If your frame count is this low. Your images will be noisy as heck. Indeed putting both these images on reg 6.

The images are considerably more noisy.  Than i am getting Pete. 3 min stack 8ms exposure 60% histo 327gain ( a little low) 22.500 frames.  245mm mirror Works for me.

Edited September 23, 2021 by neil phillips

[vlaiv]

Three things here come into play.

1. Total imaging time

2. Level of read noise and number of stacked frames

3. Single exposure length vs seeing effects

These three are tied together in some ways so you can't optimize them independently.

Take total imaging time. Say you image for 3 minutes - that means that you gathered 180s of data, right? Not necessarily. Camera, computer and USB speed come into play here. If you used 5ms exposure - and your camera is capable of doing 150FPS and you imaged for 3 minutes - total imaging time will be 180 x 150 x 5ms = 135s - so 2 minutes and 15 seconds of recording instead of 3 minutes.

In the end - if you stack 5% of frames - you'll be stacking 5% of that 2:15 and not 3:00.

Level of read noise dictates difference between one single long exposure and stack of shorter exposures. If read noise was zero - non existent, then it would not matter if you stacked 1000 x 1ms or 100 x 10ms or 10 x 100ms or 1 x 1s - from SNR point of view. Thermal signal grows linearly with time, LP signal grows linearly with time, target signal grows linearly with time - which means that they all add up the same regardless of number of subs in total imaging time (and hence the same noise levels coming from them) - only read noise grows as number of stacked frames - each frame has one "dose" of read noise.

When read noise is non zero - more subs you stack (that add up to same total time) - worse resulting SNR will be.

In planetary imaging we are forced to use short subs and lots of them. That is the reason why we want as low read noise camera as possible. It is also the reason to set camera gain in such way as to minimize read noise.

Read noise comes in two parts - pre amp and post amp. Some of read noise is added before signal amplification and some after. Here lies the reason why higher gain results in less read noise.

Say we have A to be pre amp read noise and B to be post amp read noise. Resulting noise will be A * gain + B  (where gain is in e/ADU units and lower number means higher gain - you have less electrons per ADU - or more ADU units per electron - higher gain).

If gain is 1 e/ADU then total read noise will be A + B, but if gain is set to 0.5 e/ADU - then total read noise will be A * 0.5 + B - and that is lower value of read noise.

Three has to do with how seeing works and fact that we need to freeze it to avoid additional motion blur and just have wavefront aberration blur that we can asses for quality and accept / reject (AS!3 even does quality estimation around alignment points and accepts part of the sub that are good - if you turn that option on).

From all of the above, we have number of "rules" that we must take into account when choosing gain and exposure length.

- Higher gain means lower read noise - good in every case

- Shorter exposure means higher impact of read noise, can cause "dropped" frames, or under utilization of total imaging time, but it is essential for freezing the seeing

- Using longer exposures can in fact be counter productive. Yes, you minimize impact of read noise and you ensure you have no "dropped" frames and you use whole imaging run for frame selection - but you severely reduce change chance of frame being sharp enough to be included in resulting stack because of atmosphere motion blur.

If seeing is exceptional - then use longer exposures

If seeing is not good - use shorter exposures

Always keep gain high as it reduces read noise

 

Edited September 23, 2021 by vlaiv
typo ...

[vlaiv]

2 minutes ago, neil phillips said:

Of course Setting a gain setting too low will increase read noise. But too high a gain will also produce noise. But it will be gain noise. All i know is a 3 min stack produces around 22.500 frames at 125/secs exposure. With my 245mm mirror. Even a conservative 25% Will give 5750 frames. I use winjupos for considerably more frames. If your frame count is this low. Your images will be noisy as heck. Indeed putting both these images on reg 6.

The images are considerably more noisy.  Than i am getting Pete. 3 min stack 8ms exposure 60% histo 327gain ( a little low) 22.500 frames.  245mm mirror Works for me.

Why do you derotate 3 minute video?

What do you use to stack your recordings? AS!3 can easily handle any planetary rotation that happens in 3 minutes (and even more) on 10" class aperture.

[neil phillips]

22 minutes ago, vlaiv said:

Why do you derotate 3 minute video?

What do you use to stack your recordings? AS!3 can easily handle any planetary rotation that happens in 3 minutes (and even more) on 10" class aperture.

I dont generally. Was just saying to Pete 3 mins as i often capture at that time untill i see seeing improve. Pete may not have wanted to use winjupos. So i mentioned the lower time scale that i capture (3 mins) that i dont bother de rotating then run off some  6 min captures that i do video de rotation on. Should probably have clarified

Edited September 23, 2021 by neil phillips

[Pete Presland]

These are my typical settings that use, when usually aiming for 70% Histogram level.

camera=ZWO ASI224MC
Diameter=47.25"
Focal Length=5500mm
Resolution=0.14"
Duration=180.003s
Frames captured=14985
ROI=696x598
FPS (avg.)=83
Shutter=5.907ms
Gain=425 (70%)
Gamma=50
Histogram=71%

[vlaiv]

12 minutes ago, Pete Presland said:

These are my typical settings that use, when usually aiming for 70% Histogram level.

camera=ZWO ASI224MC
Diameter=47.25"
Focal Length=5500mm
Resolution=0.14"
Duration=180.003s
Frames captured=14985
ROI=696x598
FPS (avg.)=83
Shutter=5.907ms
Gain=425 (70%)
Gamma=50
Histogram=71%

You are getting quite low FPS for some reason.

Even with 12bit capture, you should be able to pull 100FPS. With 8bit, you'll be limited with exposure length - 5.9µs is equivalent of ~169.5FPS, yet you are able to capture only half of that.

Don't think that you need to go with gain 425. Try setting it to 379 next time. With such high gain settings - read noise difference is small - and it's better to avoid quantization effects if possible and keep gain in form of 135 + N*61.

This is because unity gain is 135 and it doubles (or e/ADU halves) with every additional 61. This helps with rounding - in binary multiplication / division by 2 is simple shift of bits.

5.9µs is quite good exposure value.

 

[Pete Presland]

35 minutes ago, vlaiv said:

You are getting quite low FPS for some reason.

Even with 12bit capture, you should be able to pull 100FPS. With 8bit, you'll be limited with exposure length - 5.9µs is equivalent of ~169.5FPS, yet you are able to capture only half of that.

Don't think that you need to go with gain 425. Try setting it to 379 next time. With such high gain settings - read noise difference is small - and it's better to avoid quantization effects if possible and keep gain in form of 135 + N*61.

This is because unity gain is 135 and it doubles (or e/ADU halves) with every additional 61. This helps with rounding - in binary multiplication / division by 2 is simple shift of bits.

5.9µs is quite good exposure value.

 

That mystery is now solved. My new laptop has 3 USB ports, but only 1 is USB3. A little annoying as it is only 1 year old, i thought they would be all USB3. I have been plugging all my cameras in  USB2 ports, you have to laugh 🙂  A quick check and the frame rate is now above 100fps and 640x480 its 125fps

[vlaiv]

14 minutes ago, Pete Presland said:

That mystery is now solved. My new laptop has 3 USB ports, but only 1 is USB3. A little annoying as it is only 1 year old, i thought they would be all USB3. I have been plugging all my cameras in  USB2 ports, you have to laugh 🙂  A quick check and the frame rate is now above 100fps and 640x480 its 125fps

When using high gain and short exposures - you can use 8bit mode without any issues - it will give you even higher frame rates.

[neil phillips]

37 minutes ago, Pete Presland said:

That mystery is now solved. My new laptop has 3 USB ports, but only 1 is USB3. A little annoying as it is only 1 year old, i thought they would be all USB3. I have been plugging all my cameras in  USB2 ports, you have to laugh 🙂  A quick check and the frame rate is now above 100fps and 640x480 its 125fps

Well thats good definately need more frames. From those last captures