17 replies to this topic

[Tulloch]

There was a bit of discussion recently about framerates, and at my last outing I had a quick go at capturing at 4 different framerates on Jupiter. The seeing at the time was not particularly good, the focus was not as consistant as I normally like, with the planet going in and out of focus quite readily. Collimation was fine and I had the scope outside for a couple of hours before imaging, so I'm blaming the atmospherics. It got better later in the night, however this was a good test of the effect of frame rate during less than ideal conditions.

 

All images were taken from 3 minute videos run one after the other (after a quick focus check) with the the best 33% stacked from each video, processed identically in AS!3, Registax and Photoshop. Focus point was checked before each capture, the focus point was more difficult to confirm for the 200 fps capture which meant it took more time to find and so the gap between the frames in the animation below is longer.

 

There is a noticeable drop in quality for the 30 fps capture, the best is possibly the 100 fps capture but it may well just be the variable seeing at the time. The issues with the atmosphere not being stable when I did these captures may account for the drop in quality for the lower framerate. It's difficult to see the differences at 100% captured size, the link shows the images at 150% captured size (taken from a 50% reduction in size after 3x drizzle in AS!3).

 

Animated png at 100% captured size

 

 

Link to 150% capture size animation.

https://www.cloudyni...6445_128080.png

 

Celestron Evolution C9.25" SCT, 2.5x TV PowerMate and ASI224MC.

 

Andrew

Edited by Tulloch, 15 October 2021 - 05:59 PM.

[RedLionNJ]

An oft-reinforced point which you appear to be ignoring is that you cannot process all the samples identically and expect the results to form a valid comparison.

 

A slower frame rate is generally accompanied by lower gain and thus can be sharpened more aggressively without introducing quite so many noise-related artifacts.

 

I'm sure there are other differences to be expected in the processing regime too, but this is the one I see raised most often.

 

Sorry!

[dcaponeii]

I think in the end we all optimize for the imaging details that we become accustomed to.

[Ittaku]

Nice comparison. The 100fps seems to have more contrast and sharpening ringing rather than detail compared to the higher frame rates though. It's interesting to just compare the different rind effects as well, as rind tends to correlate with seeing if I recall correctly? But does that take into account the varying effect of different frame rates? My head's going to explode...

[rkinnett]

Are you stacking equal frame counts or equal percentages?  If using equal percentage, would you mind trying equal frame count?

 

Beautiful captures, by the way.

[Ittaku]

Are you stacking equal frame counts or equal percentages?  If using equal percentage, would you mind trying equal frame count?

That would be unfair as there is a lot more light captured per frame at lower frame rates.

[Tulloch]

An oft-reinforced point which you appear to be ignoring is that you cannot process all the samples identically and expect the results to form a valid comparison.

 

A slower frame rate is generally accompanied by lower gain and thus can be sharpened more aggressively without introducing quite so many noise-related artifacts.

 

I'm sure there are other differences to be expected in the processing regime too, but this is the one I see raised most often.

 

Sorry!

Thanks Grant, naively I thought it might the other way around - more numbers of frames stacked would lower the noise levels and so sharpening could be higher - but I seem to remember a square root relationship there also ... I tried to keep it as consistent as possible, but maybe that was a mistake...

 

Nice comparison. The 100fps seems to have more contrast and sharpening ringing rather than detail compared to the higher frame rates though. It's interesting to just compare the different rind effects as well, as rind tends to correlate with seeing if I recall correctly? But does that take into account the varying effect of different frame rates? My head's going to explode...

Yeah, I wondered about going through the whole processing and posting the results as they are a bit inconclusive... However, they don't really convince me to change my processing from 150fps ... yet.

 

Are you stacking equal frame counts or equal percentages?  If using equal percentage, would you mind trying equal frame count?

 

Beautiful captures, by the way.

I stacked the best 33% from each video, but I can stack equal frame numbers if you like. I'll also post the 33% raw stacks in full 32-bit format if anyone want to do the processing themelves

 

33% @ 150fps

https://www.cloudyni...445_7158030.png

 

33% @ 30 fps

https://www.cloudyni...445_7158963.png

 

33% @ 100 fps

https://www.cloudyni...445_7159035.png

 

33% @ 200 fps

https://www.cloudyni...445_7159099.png

 

Andrew

Edited by Tulloch, 15 October 2021 - 08:01 PM.

[Ittaku]

What did the as3 graphs look like, if I may ask?

[Tulloch]

What did the as3 graphs look like, if I may ask?

Sure, here they are ...

 

150 fps

 

 

30 fps

 

 

100 fps

 

 

200 fps

 

Edited by Tulloch, 15 October 2021 - 09:11 PM.

[Ittaku]

Sure, here they are ...

 

150 fps

 

30 fps

 

100 fps

 

200 fps

Thanks, assuming the absolute quality ratings can be compared (and I'm not really sure they can but whatever) they do seem similar except for the 150fps scoring a lower "best" score, but since its curve goes down slower it probably balances out over the 33% of frames.

[Ittaku]

Since my imaging is done at 30fps and I guess my processing is "optimised" around 30fps, I've done the basic processing I would normally have done. I had to tone it down slightly on the 100fps to avoid excessive contrast. Perhaps it's all in the processing, as I'd struggle to tell them apart, apart from the more front on view of the GRS.

 

 

Animated

 

Sorry too lazy to label them neatly, they're in increasing frame rate order.

[Tulloch]

Yep, at 100% captured size it's difficult to see a difference - zooming in a little does show some differences, I look at the white ovals in the south polar region, some of the smaller brown dots, the extra contrast in the poles. But it's a close run thing ...

[Foc]

Interesting to see the Zwo fps comparison in more typical conditions. The 100fps image looks a bit more contrasty but not sure that is purely a fps effect. I have been a bit of an agnostic on fast versus very fast fps in my inland location, but have not tried down around 30fps because the turbulent and cloudy affected seeing over the last 18 months suggested that changes for the better might occur very quickly and transiently.  Obviously even huge changes in speed in either direction will not make constantly poor seeing produce images like good seeing, so the impacts may be modest for most of us. 

 

I wonder if 30fps works noticeably better or is more cost effective disc-spacewise when coastal or other conditions provide steadier seeing of whatever quality.

 

I suppose even 1/30th of a second is a somewhat short time interval and I remember Andrew, your past comparisons with the 'slow' Canon apcs providing competitive images when matched against the fast 224mc, still I seem to remember you eventually found that more unsteady conditions often favored the fast fps 224mc.

[rkinnett]

That would be unfair as there is a lot more light captured per frame at lower frame rates.

Not true if he used the same exposure across the board.  And it’s still a worthwhile experiment in any case.

 

The idea is that at some point, it’s more beneficial to cash in on high frame rate by being more selective rather than stacking more frames.  At 100 fps, stacking 33% of frames from 3 mins is 6k frames.  I suspect that’s past the point of diminishing returns, hence no improvement at 200 fps.  If, in the 200 fps set, you stack the best 6k frames, rather than keeping the same percentage and including the next best 6k (12k total), then you’re being twice as selective, and I think that should yield slightly more detail (~15% “quality” improvement per graphs).  That’s my hypothesis anyway.

[Ittaku]

Not true if he used the same exposure across the board.

I'm not sure what you're getting at. That's impossible by nature. The gain needs to be inversely proportional to the frame duration. Higher gain doesn't net you the same amount of photons hitting the sensor in a 5ms frame as it does in a 33ms frame? 200fps gives you a 5ms frame, and 30fps gives you a 33ms frame.

Edited by Ittaku, 16 October 2021 - 03:43 AM.

[Tulloch]

Interesting to see the Zwo fps comparison in more typical conditions. The 100fps image looks a bit more contrasty but not sure that is purely a fps effect. I have been a bit of an agnostic on fast versus very fast fps in my inland location, but have not tried down around 30fps because the turbulent and cloudy affected seeing over the last 18 months suggested that changes for the better might occur very quickly and transiently.  Obviously even huge changes in speed in either direction will not make constantly poor seeing produce images like good seeing, so the impacts may be modest for most of us. 

 

I wonder if 30fps works noticeably better or is more cost effective disc-spacewise when coastal or other conditions provide steadier seeing of whatever quality.

 

I suppose even 1/30th of a second is a somewhat short time interval and I remember Andrew, your past comparisons with the 'slow' Canon apcs providing competitive images when matched against the fast 224mc, still I seem to remember you eventually found that more unsteady conditions often favored the fast fps 224mc.

When I was imaging with my DSLR, the shutter speed was fixed at 1/30 sec. I found that whenever there was more than just a breath of wind, my imaging suffered and the results were not as good. With my one-armed alt/az mount not being the most stable of platforms, moving to the higher framerate camera certainly made a big difference.

 

I did a comparison of noise vs frame numbers for the ASI224MC and Canon 700D a while ago, linked below (if you haven't seen it before ).

https://www.cloudyni...sct-test-three/

 

Not true if he used the same exposure across the board.  And it’s still a worthwhile experiment in any case.

 

The idea is that at some point, it’s more beneficial to cash in on high frame rate by being more selective rather than stacking more frames.  At 100 fps, stacking 33% of frames from 3 mins is 6k frames.  I suspect that’s past the point of diminishing returns, hence no improvement at 200 fps.  If, in the 200 fps set, you stack the best 6k frames, rather than keeping the same percentage and including the next best 6k (12k total), then you’re being twice as selective, and I think that should yield slightly more detail (~15% “quality” improvement per graphs).  That’s my hypothesis anyway.

The framerate was inversely proportional to the shutter speed. The capture details were:

 

Capture 1: shutter speed 6.666ms, gain 280, 150 fps

Capture 2: shutter speed 33.33ms, gain 150, 30 fps

Capture 3: shutter speed 10.00ms, gain 250, 100 fps

Capture 4: shutter speed 5.000ms, gain 305, 200 fps

 

I've found that for my longer captures (10 minute derotated videos @ 150 fps), stacking 10,000/90,000 frames allows me to sharpen more than lower stacks without introducing more noise (there's a thread here on derotation which touches on this if you are interested).

 

I haven't got around to stacking a certain number of frames equally across the 4 videos, and I don't really think it's fair to do it this way, as you need a certain number at a high framerate to reduce the noise level, and stacking too many at a low framerate means using too many lower quality frames. I think stacking a certain percentage is fairer all around.

Edited by Tulloch, 16 October 2021 - 03:31 AM.

[rkinnett]

Ittaku, you’re right that frame rate is limited by shutter speed, but you do in fact have control over frame rate within that limit.  It’s a setting in FC.  Not sure there’s any good reason to use that capability, but it does exist.

 

Andrew, I agree, it’s not useful to re-stack the latter 3 data sets to match the frame count of the 30 fps stack.  Instead, please consider stacking just the 200 fps data set at the same frame count (6k) as the 100 fps set.  That would enable 2 interesting comparisons to the stacks you already produced:  200 fps 6k vs 12k frames, and 100 fps vs 200 fps at 6k frames.

 

I hear what you guys are saying about using equal percentage as a “fair” comparison in the sense of minimizing variables.  That’s totally valid.  It’s sound experiment design, but it’s an incomplete experiment if the overarching goal is to evaluate the relative advantage of higher frame rate.  There is a logical explanation for why we’re not seeing a difference in the 100 fps and 200 fps stack comparison: 12k stacked frames is well past the knee in the curve above which adding more frames no longer helps to reduce noise or add detail.  The equal-percentages approach does not maximally leverage the advantage of higher frame rate.  Without exploring further we can’t conclude that 200 fps is not beneficial over 100 fps.  Does that make sense?

 

Selectivity is the essence of lucky imaging.

 

By the way, forgive me for being so insistent.  Your data sets are much higher quality than mine on my best nights and are better for systematic comparison.  I appreciate you posting these.

[Tulloch]

OK, I think I understand now.

 

Here are the 6000 frame stacks, processed identically to the last set. As excepted, the noise levels are higher for the faster framerates since the intrinsic noise levels are higher due to the increased gain.

 

Animated png showing 150, 100 and 200 fps with 6000 frames stacked at 150% captured size.

https://www.cloudyni...6445_186967.png

 

Here are the raw stacks in 32 bit format, if you want to play with the post processing yourself.

6000 frames at 150 fps

https://www.cloudyni...445_7235501.png

 

6000 frames at 200 fps

https://www.cloudyni...445_7235612.png

 

Andrew

Edited by Tulloch, 16 October 2021 - 05:16 PM.