Jupiter sampling C14 F11 vs F18 - 11 Nov 2022

[neil phillips]

13 minutes ago, geoflewis said:

Well here is an F18 RGB image captured immediately prior to the above sharp IR image, i.e. the timestamp is just 7 mins earlier. It's so much softer than the IR, yet still too noisy I think.

I don't know what to make of it, but maybe I'm just missing sharp focus operating at F18 on the RGB with the colour camera? 🤷‍♂️

Or seeing didn't support it. But the IR cut through better ?

[johnturley]

16 minutes ago, Magnum said:

I found this comparison I did on Sept 4th with my old 224C, colours is slightly different between the 2 as it was only a quick comparison, and I really cant be bothered to process it again from scratch.

Anyway the left image is at prime focus with my 12" LX200 but with extra spacing the disc measured 250 pixels which works out to be F13, the right image is using my 2.5x Barlow measuring 480 pixels and is bang on f25. I have enlarged the f13 image to match the f25 image size. They were taken about 30 mins apart and conditions were pretty consistent throughout the session. To my eyes the f25 image is way more refined.

Ive since upgraded to the 462C with smaller 2.9um pixels compared to 3.75um on the 224C so ive dropped my working FL down from f25 to f21 now to get me similar scale.

the ideal sampling maybe somewhere between the 2, but I would conclude same thing Neil said, its better to oversample than to undersample

Lee

Do you find the ASI 462 with its smaller pixel size a definite improvement over the 224.

John 

[geoflewis]

I'll round of my set of Jupiter images from last night, with my first attempt at Methane band (CH4). This is at F11 with the GRS just rotating into view and some other detail coming through too.

[Magnum]

7 minutes ago, johnturley said:

Do you find the ASI 462 with its smaller pixel size a definite improvement over the 224.

John 

I dont think the pixel size made any difference as I adjust the focal length with the barlow spacing to compensate. Well except at prime focus then the smaller pixels are an advantage, come to think of it, I wouldn't mind a 1.5um pixel camera then I could do away with the Barlow altogether and lose the extra glass from the equation 

The main reason I upgraded was my old Altair 224 was only USB2 and was limiting my frame rate, and as I was upgrading I thought I would go for the newer sensor which is more versatile as its  more sensitive in IR

Edited November 12, 2022 by Magnum

[Magnum]

4 minutes ago, geoflewis said:

I'll round of my set of Jupiter images from last night, with my first attempt at Methane band (CH4). This is at F11 with the GRS just rotating into view and some other detail coming through too.

Thats very good for your first methane band Geof, 

[geoflewis]

6 minutes ago, Magnum said:

Thats very good for your first methane band Geof, 

Thanks Lee, I've processed it about a dozen times. Tried drizzle x1.5 too and a bunch of different wavelet settings and think this is about as good as I can get from it. Must say it's far better than I anticipated, so your PM advice yesterday was a huge help, thanks. I also spent a few hours watch Chris Go on YT yesterday to get some pointers from him. All good learning I think. Might be interresting to see what I can do with the filter under better seeing conditions and by running a few shorter vids then de-rotating the resultant stacks as Chris Go recommends.. Also next year Jupiter will be much higher elevation, so more favourably placed....🤞

Edited November 12, 2022 by geoflewis

[vlaiv]

Here is what I find interesting in this comparison.

First two images are supposed to be F/11 vs F/18, right?

This means that Jupiter in one image should have 18/11 = x1.6363.... larger diameter.

I've measured following diameters in two images:

If smaller Jupiter has 470px diameter - larger should have 470 x 1.63636 .... = ~769px and not 524

Difference between two images is much less than F/11 and F/18.

Noise difference is very large indeed, and I wonder if images where manipulated in some other way?

If we check the native image, at F/11 it should have sampling rate of 0.153"/px and that would make 46.2" diameter Jupiter be ~301px

Maybe drizzle x1.5 was used for that image? It would be much better fit if drizzle x1.5 was used to F/11 and F/18 integrated normally (accounting for small FL change due to primary - secondary distance change when focusing).

 

[Space Cowboy]

Seeing conditions are the key. Considering the conditions were less than great the oversampling has done ok.

[vlaiv]

Here is a protocol that I would suggest for anyone wanting to compare two different samplings that will mostly make things equal.

It requires mono camera and some special processing.

We shoot single video - but we create two new videos out of it to be processed. Video is shot at exposure length suitable for smaller sampling rate but with higher sampling rate (use of barlow).

First video is created by binning each frame 2x2 and then taking every 4th frame (we discard other 3). This will give us lower sampling rate at exposure length of lower sampling rate with 1/4 of all frames equally spread over capture period.

Second video is created by summing each group of 4 frames to make output frame (stacking without alignment). This will give us higher sampling rate at  exposure length of higher sampling rate with 1/4 of all frames equally spread over capture period.

We have thus managed to create videos - each at appropriate sampling rate, each with realistic exposure length to reach same signal level, each with the same number of subs in same seeing / same conditions.

Drawbacks - recording will have x2 read noise it would usually have and we are limited to x2 difference in sampling rate (we could not do above x1.636363 example).

Next part has to deal with same processing. This is tricky as wavelets are applied based on feature size, so they must be set differently.

Thus we must create two images to be processed like this:

We stack both videos with same stacking parameters (same percentage of subs, same number of alignment points and relative size of alignment points being the same - which means that smaller sampling  rate will use half the size alignment point) to get two planetary images.

We then create two images for processing by taking smaller planet image, enlarging it x2 and putting it side by side with larger planet to form single image. We similarly take larger planet and reduce its size to half and put it next to small planet.

This way we can test processing at both sizes.

We can even create third image where we simply put them side by side without resizing and process that. My only concern is that there could be bias in processing and parameters selected so that one of the two looks better.

 

 

[Magnum]

25 minutes ago, vlaiv said:

Here is what I find interesting in this comparison.

First two images are supposed to be F/11 vs F/18, right?

This means that Jupiter in one image should have 18/11 = x1.6363.... larger diameter.

I've measured following diameters in two images:

If smaller Jupiter has 470px diameter - larger should have 470 x 1.63636 .... = ~769px and not 524

Difference between two images is much less than F/11 and F/18.

Noise difference is very large indeed, and I wonder if images where manipulated in some other way?

If we check the native image, at F/11 it should have sampling rate of 0.153"/px and that would make 46.2" diameter Jupiter be ~301px

Maybe drizzle x1.5 was used for that image? It would be much better fit if drizzle x1.5 was used to F/11 and F/18 integrated normally (accounting for small FL change due to primary - secondary distance change when focusing).

 

If the f11 one has not been resized then yes its coming out at about f16.5 for me rather than f11

[geoflewis]

35 minutes ago, vlaiv said:

Here is what I find interesting in this comparison.

First two images are supposed to be F/11 vs F/18, right?

This means that Jupiter in one image should have 18/11 = x1.6363.... larger diameter.

I've measured following diameters in two images:

If smaller Jupiter has 470px diameter - larger should have 470 x 1.63636 .... = ~769px and not 524

Difference between two images is much less than F/11 and F/18.

Noise difference is very large indeed, and I wonder if images where manipulated in some other way?

If we check the native image, at F/11 it should have sampling rate of 0.153"/px and that would make 46.2" diameter Jupiter be ~301px

Maybe drizzle x1.5 was used for that image? It would be much better fit if drizzle x1.5 was used to F/11 and F/18 integrated normally (accounting for small FL change due to primary - secondary distance change when focusing).

 

Sorry Vlaiv, I tried and failed to resize the F11 Jupiter image to the F18 image. All I achieved was to make both images including the border the same size, but clearly there was a larger ROI for the F11 image.. Here is the F11 image before I resized it.

Does the maths work now. I get F11 v F18 as that is what FireCapture reports.

[geoflewis]

22 minutes ago, vlaiv said:

Here is a protocol that I would suggest for anyone wanting to compare two different samplings that will mostly make things equal.

It requires mono camera and some special processing.

We shoot single video - but we create two new videos out of it to be processed. Video is shot at exposure length suitable for smaller sampling rate but with higher sampling rate (use of barlow).

First video is created by binning each frame 2x2 and then taking every 4th frame (we discard other 3). This will give us lower sampling rate at exposure length of lower sampling rate with 1/4 of all frames equally spread over capture period.

Second video is created by summing each group of 4 frames to make output frame (stacking without alignment). This will give us higher sampling rate at  exposure length of higher sampling rate with 1/4 of all frames equally spread over capture period.

We have thus managed to create videos - each at appropriate sampling rate, each with realistic exposure length to reach same signal level, each with the same number of subs in same seeing / same conditions.

Drawbacks - recording will have x2 read noise it would usually have and we are limited to x2 difference in sampling rate (we could not do above x1.636363 example).

Next part has to deal with same processing. This is tricky as wavelets are applied based on feature size, so they must be set differently.

Thus we must create two images to be processed like this:

We stack both videos with same stacking parameters (same percentage of subs, same number of alignment points and relative size of alignment points being the same - which means that smaller sampling  rate will use half the size alignment point) to get two planetary images.

We then create two images for processing by taking smaller planet image, enlarging it x2 and putting it side by side with larger planet to form single image. We similarly take larger planet and reduce its size to half and put it next to small planet.

This way we can test processing at both sizes.

We can even create third image where we simply put them side by side without resizing and process that. My only concern is that there could be bias in processing and parameters selected so that one of the two looks better.

I think I'll pass on that......😖

[vlaiv]

3 minutes ago, geoflewis said:

I tried and failed to resize the F11 Jupiter image to the F18 image.

Ah, ok, I did read that you tried to do it - but I figured out that you left images as is after not being able to make them the same size.

[geoflewis]

25 minutes ago, Magnum said:

If the f11 one has not been resized then yes its coming out at about f16.5 for me rather than f11

The entire image was resized from 416x416 to 600x600, but of course Jupiter was in a larger relative ROI in the smaller image. Just my hopeless attempt to make the images look similar sizes.....🙄

Edited November 12, 2022 by geoflewis

[vlaiv]

2 minutes ago, geoflewis said:

I think I'll pass on that......😖

I would not mind doing it if someone shoots the movie in mono with 14" scope at twice critical sampling rate at 2ms exposure length

[vlaiv]

Just now, geoflewis said:

The entire image was resized from 416x416 to 600x600, but of course Jupiter was in a larger relative ROI in the smaller image. Just my hopeless attempt to make the images look similar sizes.....🙄

In order to get same images - here is what you do.

You resize smaller image to x1.636363 size or 163.63% and then crop to the central 600px.

This will give you both same planet size and same image size (in pixels).

[geoflewis]

5 minutes ago, vlaiv said:

In order to get same images - here is what you do.

You resize smaller image to x1.636363 size or 163.63% and then crop to the central 600px.

This will give you both same planet size and same image size (in pixels).

Thnaks Vlaiv, yes, if exactly F18 to F11, which I am not sure, but the princple is good and something for me to consider in future, if/when I want to do a comparison.

[geoflewis]

8 minutes ago, vlaiv said:

I would not mind doing it if someone shoots the movie in mono with 14" scope at twice critical sampling rate at 2ms exposure length

I have some old Mars videos (AVIs) from 2020 shot at F24 (according to FireCapture). They are mono 180s duration, but captured at 13ms. Equipment was ASI290MM (2.9mn px), C14, x2 TV PM, ADC. Any use?

[vlaiv]

35 minutes ago, geoflewis said:

I have some old Mars videos (AVIs) from 2020 shot at F24 (according to FireCapture). They are mono 180s duration, but captured at 13ms. Equipment was ASI290MM (2.9mn px), C14, x2 TV PM, ADC. Any use?

Won't work well I'm afraid.

13ms is too long and would defeat the purpose - but it does show that one needs longer exposure when using F/24.

[geoflewis]

21 minutes ago, vlaiv said:

Won't work well I'm afraid.

13ms is too long and would defeat the purpose - but it does show that one needs longer exposure when using F/24.

Ok , no worries Vlaiv. I'm not sure about that, but maybe. I was using 13ms based on advice from a friend in Australia. Others suggested faster and slow fps, but I just went with that..... I'd not really thought it about it previously.

[vlaiv]

50 minutes ago, geoflewis said:

Ok , no worries Vlaiv. I'm not sure about that, but maybe. I was using 13ms based on advice from a friend in Australia. Others suggested faster and slow fps, but I just went with that..... I'd not really thought it about it previously.

13ms is too long.

There is something called coherence time related to the seeing - it is period in which atmospheric distortion is almost non changing / frozen.

Look at this paper:

https://articles.adsabs.harvard.edu//full/1996PASP..108..456D/0000456.000.html

This is important graph:

It shows how "similar" (correlated) are wave fronts at time 0 and time t0 (being - atmospheric coherence time).

In another words - if coherence time is 0.5 - and you sample at 10ms - you get 0.1 similarity between start and end wave front - substantially changed

But if coherence time is 10ms and you sample at say 5ms - you get correlation loss factor close to 1 (good correlation).

And this is interesting part:

Atmosphere changes on very short time scales and if you want to avoid motion blur - you really need short exposures - like 5ms or less.

[geoflewis]

I decided to process the earlier 3x6 min RGBs captured at F18, then de-rotate and stack the 3 images. NB I did not de-rotate the individual SERs, so some loss of fidelity will have occured due to rotation over each 6 min period.

 

Edited November 13, 2022 by geoflewis
Updated image

[Magnum]

17 hours ago, vlaiv said:

13ms is too long.

There is something called coherence time related to the seeing - it is period in which atmospheric distortion is almost non changing / frozen.

Look at this paper:

https://articles.adsabs.harvard.edu//full/1996PASP..108..456D/0000456.000.html

This is important graph:

It shows how "similar" (correlated) are wave fronts at time 0 and time t0 (being - atmospheric coherence time).

In another words - if coherence time is 0.5 - and you sample at 10ms - you get 0.1 similarity between start and end wave front - substantially changed

But if coherence time is 10ms and you sample at say 5ms - you get correlation loss factor close to 1 (good correlation).

And this is interesting part:

Atmosphere changes on very short time scales and if you want to avoid motion blur - you really need short exposures - like 5ms or less.

That an interesting and good to know, luckily by chance I have been using 5ms as my default in colour images of Jupiter and 2-3ms on Mars. Can get get that speed at any f ratio ive tried upto f25 with the 12" scope, I remember previously I had a C9.25 but always felt limited by the amount of light that could put onto the sensor so had to use longer exposures and higher gain. Since moving to the 12" scope I don't have to worry about using slow shutter speeds except for Uranus and Neptune or imaging in deep IR and Methane.

Lee

[geoflewis]

5 minutes ago, Magnum said:

That an interesting and good to know, luckily by chance I have been using 5ms as my default in colour images of Jupiter and 2-3ms on Mars. Can get get that speed at any f ratio ive tried upto f25 with the 12" scope, I remember previously I had a C9.25 but always felt limited by the amount of light that could put onto the sensor so had to use longer exposures and higher gain. Since moving to the 12" scope I don't have to worry about using slow shutter speeds except for Uranus and Neptune or imaging in deep IR and Methane.

Lee

Hi Lee, I have say, this is another area where theory and reality do not align. I've had a few discussions with Anthony Wesley (Bird) and Niall MacNeil, both imaging from Australia and they do not capture at anything like those high rates. Both have done a lot of their own analysis and Niall suggests that optimum is 13ms / 76fps  (hence why I used that) and Bird even slower at 65 fps. Here's is what Bird wrote about it in a discussion with Niall, which he shared with me, so I hope it's ok for me to repeat here.

Well it all depends on the seeing of course, there are so many variables that effect the seeing. Personally I run at about 65fps, that seems good enough for me, but I could also believe that your test of 75fps makes sense in the seeing that you had. At 65fps I can reach max hardware gain in the camera quite easily and still have a relatively low noise image. I've tried speeds up to about 90fps in the past, but 65 seems just fine for me on most nights. I've had nights in the past where fog reduces my speed to about 10fps but the seeing was excellent and I didn't have any trouble with blurring.

Something to remember is that the data gets noisier as you shorten the exposure - AS!3 somehow has to lock onto and align features on each of these raw frames and noise in the frames will directly affect its ability to get an accurate alignment, so increasing noise (ie shorter exposures) will likely make it harder to get the best alignment. It's a tradeoff to find the right value, short enough to remove the blurring and not so short that AS! can't get a good alignment. There's no magic bullet, if the data is too noisy then you'll get penalised by AS! not being able to accurately lock onto the features.

You could broadly speaking divide the "seeing" into two components - ie the stuff that can be corrected in post-processing by AS!, and the stuff that can't (ie global jetstream blurring). No amount of high speed fps can remove the nasty jetstream blur, that really needs the sort of adaptive optics that you only find on large professional telescopes. Post-processing software like AS! can remove the simpler lateral-motion modes that come from seeing cells large enough not to destroy the high frequency component, these cells are mostly located in air masses close to the scope, at most a few hundred metres above, and sometimes a lot closer, and so they change all the time and vary from night to night. 65-75fps should be fast enough to handle that when you consider how little motion this allows for between successive frames.

There is something to watch out for - and something that I've experienced in the past - where short exposures can give you the illusion of more contrasty images, due to the lower number of shades in each frame. This is due to data loss, and is roughly the same effect as if you apply a lot of gamma darkening while capturing to make the image look "better", but then you pay the penalty later in processing. This is also one of the reasons I always use 16 bit mode on the camera so I can have more headroom during processing if I need it.

The high quality of their images speak for themselves. Take a look at what Niall is posting on AstroBin.....https://www.astrobin.com/users/macnenia/ I was going to pick out some examples, but there are so many that you can take your pick....

[Space Cowboy]

The conditions in Australia might be more suited to slower shutter speeds 🙂 Certainly worth noting how noise can effect stacking alignment though. I do remember getting better results with my old DFK at 30fps compared to 60 because of the noise issue.  I might try slowing down shutter speed and using less gain but when the sky is like a foaming ocean its a pig,