Dual rig with different focal lengths (?)

[WolfieGlos]

Hi all,

With some apparent clear skies forecast this week (fingers crossed....), I'm thinking about possibly dual rigging to obviously capture more data, and to try and finish a target I started in November. It's suffering from a measly 5 hours so far, after culling the worst subs.

To date, it has been imaged with a Starfield 102 + 0.8 reducer with a Canon 800D. Given the equipment I have to hand, I can mount a Canon EF 70-300L zoom lens on top of the tube rings and connect a 585mc. If I set the zoom to 240mm I get basically the same FOV as shown below.

However, the resolution will obviously be different.

The SF102 will be imaging at 1.34" p/p whilst the zoom lens would be at 2.49" p/p (and circa f/4.5 instead of 5.6). Will this have a detrimental impact on the final image? The main core of the nebula is already looking great, but it needs more data to reduce noise - and then hopefully pull out the dust for a more satisfactory image (currently it's very mottled).

I know zoom lenses are not ideal, I'm just trying to use what I have to hand and to understand the possible implications of doing this - and before I decide to make a more permanent dual setup using something like a FMA180 instead of the zoom lens. If this is not a good idea that's fine.

Thanks!

[Elp]

Should work fine as long as your "main" scope data is providing the meat of the resolution for the detail and you're doing minimal to no scaling on this main data. I mix and match optics and camera data all the time, though I try to keep specific bandpass sessions using the same equipment (eg all HA is done with one setup, RGB with a other etc etc).

[tomato]

You probably shouldn’t but I have combined data from a F2 RASA8, F7 Esprit 150 and an F8 RC on M31, APP takes this in it’s stride.

[WolfieGlos]

Thanks both. All of the data will be RGB (no narrowband).

It’s M78, and it isn’t visible to me until it reaches SSE (circa 150 degrees) so I’ll have a few hours to get it setup.

Allegedly we have some clear skies tonight so perhaps I’ll give it a go and see. I suppose worst case scenario is that I’ll have a load of unusable data from the 585mc - or two versions of the same target with different camera/scope combos.

Edit; I just remembered, I did combine 72ed and 102ed data for M31 before Christmas now I’m thinking about it, and it seemed OK. If I get some data maybe I’ll post the individual and combined results.

 

Edited January 7 by WolfieGlos

[ollypenrice]

Do it-.More data beat less data. You may not really reach a true resolution of I.3"PP anyway, and 2.49 is perfectly respectable. There are some very fine little details in M78 but the bulk of the image is about faint stuff, so signal is the key.

Olly

Edit: when you co-register the data, register the lens to the scope - but I'm sure you would anyway.

Edited January 7 by ollypenrice

[WolfieGlos]

2 hours ago, ollypenrice said:

Do it-.More data beat less data. You may not really reach a true resolution of I.3"PP anyway, and 2.49 is perfectly respectable. There are some very fine little details in M78 but the bulk of the image is about faint stuff, so signal is the key.

Olly

Edit: when you co-register the data, register the lens to the scope - but I'm sure you would anyway.

Thanks Olly, and that was my thinking for the faint stuff.

I would have stacked the images from the two setups separately (with their own calibrations) and then “stacked the stacks”. Assume this is what you mean by co-registering? Using Siril I think this is the only way I could do it.

However. Right now, with 4 different forecasts showing it as clear until midnight, of course….its cloudy.

So I might not get to try it tonight anyway 🙄  Hopefully the forecast is correct for later this week.

[ollypenrice]

11 hours ago, WolfieGlos said:

Thanks Olly, and that was my thinking for the faint stuff.

I would have stacked the images from the two setups separately (with their own calibrations) and then “stacked the stacks”. Assume this is what you mean by co-registering? Using Siril I think this is the only way I could do it.

However. Right now, with 4 different forecasts showing it as clear until midnight, of course….its cloudy.

So I might not get to try it tonight anyway 🙄  Hopefully the forecast is correct for later this week.

Yes, I think making 2 images is the only way. I would then co-register the larger FOV to the smaller in Registar and crop it to fit. As for how to combine them, I think you'd want to give each one a test stretch of its own to assess noise and resolution. I'd probably give each a basic stretch, not too hard, put one over the other in Photoshop layers, set the opacity to 50% at first and then move the opacity slider each way to see where I got the best result. I'd flatten and then stretch harder. There are lots of ways to kill this particular cat, depending on software.

Olly

[tomato]

In APP I would stack each scope/camera combination and then combine the stacks, you can do it all from the raw sub frames, but my PC gets quite hot and bothered in the process.

[ONIKKINEN]

Siril will happily stack both datasets to one image if you want to do that. I think you could try that first before trying the stack the stacks method as there is a decent chance you get a better result integrating both sets to one stack.

To force registration to one of the scopes you need to have the first input image as the one you want the registration frame to be.

[WolfieGlos]

8 hours ago, ONIKKINEN said:

Siril will happily stack both datasets to one image if you want to do that. I think you could try that first before trying the stack the stacks method as there is a decent chance you get a better result integrating both sets to one stack.

To force registration to one of the scopes you need to have the first input image as the one you want the registration frame to be.

Thanks Oni, I’ll give that a go - but how do you stack both datasets to 1 image? Is that using Sirilic? I know how to do it deep sky stacker, but that’s all.

I have generally been using scripts up to this point (and then reviewing the graph to cull the worst subs and then restacking), but I’ve decided to learn to do it manually and this is the first image I’m trying this on, and doing a comparison I achieved a better result. So far, after culling, I have 162 x 120s spread across SIX sessions such has been the poor weather - the initial session I retained 90 frames, whilst some of those sessions have resulted in about 8 frames retained. So I already have the “stack the stack” going on - which is what I’ve always done to date for multiple sessions in Siril.

Out of interest, how is stacking both datasets to one image any better? Isn’t it taking the same data to produce a final stacked image? 

[ONIKKINEN]

On 09/01/2024 at 01:20, WolfieGlos said:

Thanks Oni, I’ll give that a go - but how do you stack both datasets to 1 image? Is that using Sirilic? I know how to do it deep sky stacker, but that’s all.

I have generally been using scripts up to this point (and then reviewing the graph to cull the worst subs and then restacking), but I’ve decided to learn to do it manually and this is the first image I’m trying this on, and doing a comparison I achieved a better result. So far, after culling, I have 162 x 120s spread across SIX sessions such has been the poor weather - the initial session I retained 90 frames, whilst some of those sessions have resulted in about 8 frames retained. So I already have the “stack the stack” going on - which is what I’ve always done to date for multiple sessions in Siril.

Out of interest, how is stacking both datasets to one image any better? Isn’t it taking the same data to produce a final stacked image? 

I think you could use Sirilic for that, but its been a few years since i used it so not so sure on exactly how to do that, might get convoluted with how many nights you have going on here.

What i would do is first calibrate both sets from all the different nights first and save them to their own folders somewhere and then stack with Siril manually (also, i feel your pain, i have 3 hours on the horsehead from 3 nights. Sometimes i wonder why even bother with the hobby...). Since everything is calibrated its now as simple as importing all the data to a sequence using the Conversion tab in Siril (just drag and drop there) and then registering it all using global star alignment, inspecting the plot and making rejection choices if you wish, and finally stacking it all. The only choice you have to make here is what dataset you want to register to, and that is decided by the first image imported in to the sequence. So as an example have the first image be from the bigger scope - then when you do global star alignment in the registration tab all the following images in the sequence are transformed to fit that and get its pixel scale. Its important to not do the 2-pass method here, as that can choose some other image as the registration frame so only global alignment will work if you want to force Siril to stack to a specific frame.

When stacking it all to one image you get a better and more accurate average of all the input subs - so in theory better signal to noise ratio. Stacking set 1 first and set 2 second and averaging those 2 will be less accurate than just forming one stack out of it all. But since you have rather significantly different scopes and data sets here, im not quite sure what is the best way forward in practice rather than theory. Try the one stack method and compare to the 2 stacks averaged and see if there is a difference. Might not be too noticeable, in which case you could do it either way but my money is on the 1 stack method providing a better result if only by a slight margin.

Edited January 12 by ONIKKINEN
Typo

[WolfieGlos]

That makes sense, thanks for the detailed explanation. I deleted my previous process/registration so I'll have to rerun those, but that's not an issue.

I think I'll try two versions; I'll take the images I capture tonight (fingers crossed) and then do my usual "stack the stacks", and then run a separate stack registering them all together and compare the difference before doing any image processing. So far my 6 nights are all from 1 scope/camera combination, it will only be tonight when I add the second that I will have a different (but additional) scope/camera to add to the mix.

3 hours on the horsehead in 3 nights? Wow. These last few months have been terrible, but I found last year to be pretty decent up until mid October. Hope your skies clear soon.

[WolfieGlos]

That was a mammoth test, but finally I got there and I tried the different methods of stacking - and back on topic, with different scope/camera combinations.

So first of all, using just the DSLR + 102ED, I added two more sessions this week and using the "stack the stacks" method, the readings from the final stack (using noise estimation in Siril), gives the following:

Background noise value (channel: #0): 3.359 (5.126e-05)
Background noise value (channel: #1): 3.339 (5.096e-05)
Background noise value (channel: #2): 3.376 (5.151e-05)

However, registering all of the calibrated frames and stacking as one (as described by @ONIKKINEN above), the noise estimation was as follows:

Background noise value (channel: #0): 2.865 (4.371e-05)
Background noise value (channel: #1): 2.866 (4.373e-05)
Background noise value (channel: #2): 2.850 (4.348e-05)

This is the result from 312 x 120s subs at ISO-400.

Both final stacks have the same reference frame, so they have the exact same FOV, and these values are before any cropping or processing.

Both of the above tests included the same number of dark frames for each session (sensor temperature varied by no more than 3 degrees in each session, as measured in ASTAP) and flat/bias calibration, all run manually in Siril - no scripts.

So there is a definite difference and benefit to doing it that way. Which now makes me think I haven't extracted all that I could have done from previous images.

Those are the numbers, but as they say a picture says a thousand words, and interrogating the images by eye, I see very small differences in the stacked (and unprocessed image) but in this particular case it is most noticable in the corner where I have a part of Barnards Loop. All images are unprocessed and set to "AutoStretch View" in Siril (applying a histrogram stretch for viewing only). Below examples:

Full image - Stack the Stacks:

Full image - One Stack:

Reflection Nebulosity - Stack the Stacks:

Reflection Nebulosity - One Stack:

Barnards Loop Corner - Stack the Stacks:

Barnards Loop Corner - One Stack:

 

For me, the clear winner is stacking as one, not "stacking the stacks".

 

So I then tried adding the data from the ASI 585MC + 70-300 lens. This was using the "One Stack Method", adding an additional 80x60s subs at Gain 252 with darks/flats/darkflats, and unfortunately the rotations didn't quite align - see bottom right corner. The lens also suffered from tilt/distortion/bad star shapes, so the stars don't align in the final stack giving a processing headache. However, with additional data it does provide a lower noise estimation:

Background noise value (channel: #0): 2.405 (3.669e-05)
Background noise value (channel: #1): 2.386 (3.641e-05)
Background noise value (channel: #2): 2.394 (3.654e-05)

I'm dubious about those values anyway (it is "estimated" afterall), and given that this was only an additional 01:20 hours of frames, that seems to have dropped by quite a lot. The 585 will certainly produce less noise than a DSLR but still.

Anyway, pictures:

Full Image - One stack with additional data from another camera/scope combo - note star misalignment, particularly in the top left:

Reflection Nebulosity - One stack with additional data from another camera/scope combo - note star misalignment:

Barnards Loop Corner - One stack with additional data from another camera/scope combo - note star misalignment:

 

So, as a two part test, my conclusions are:

1) Different scope/camera combinations can be combined successfully.

2) Registering and stacking as one for multiple sessions, is better than stacking the stacks.

However, the only downside I have, is using Siril, the amount of HD space required to generate the one, registered stack. Fortunately it informs you if you don't have enough space for a stack, and I had to clear out 25gb of space just to do this (it's only a 350gb drive!!!).

All of the calibrated files are circa 285mb each, so 125gb. Ditto for the registered files, so that's another 125gb so a grand total of 250gb of HD space required - ignoring the initial light, dark and flat frame files, final stacked file - and anything else on the HD! This computer also took 3 hours to register the frames, an hour for me to cull the worst ones, and then another 5.5 hours overnight to stack the 312x120 + 80x60s frames, so there's another consideration here and that's the time spent to do this.

Usually I shoot longer subs (so fewer files and storage) but the local LP in my area affects the eastern and southern parts of my sky - bad for anything in Orion - especially in broadband.

Next step is to process the results and see if my conclusion still stands. We're away for the most of the weekend so that'll have to wait for now.

[Elp]

Siril is a swine for needing disk space, I'm constantly doing the stacking process then removing the files it generates as I'm going through it. I think there's a option to link to the files instead rather than creating new files or sequences but I've never used this. Alternatively I work from a USB SSD drive which is fast when I don't have the space on the computer.

[ONIKKINEN]

8 hours ago, Elp said:

Siril is a swine for needing disk space, I'm constantly doing the stacking process then removing the files it generates as I'm going through it. I think there's a option to link to the files instead rather than creating new files or sequences but I've never used this. Alternatively I work from a USB SSD drive which is fast when I don't have the space on the computer.

There is something like this - symbolic links. They create a "virtual" file that takes no space on disk and can in fact exist on another disk so you could save some space, but this step only applies for the initial sequence you create. Registration will still always write new files since the files themselves are modified to perform the registration, so the minimum space required to stack a dataset is:

1.  The calibrated subs, on any disk on the PC
2.  The same subs registered, in the working directory of Siril.

So the minimum space one needs on a single hard drive is just the initial calibrated dataset, or twice that if there is only one drive. If symbolic links are not used then you do need 3x the initial calibrated subs worth of space (calibrated subs, the same subs but created to a sequence in Siril, the registered data).

Edited January 14 by ONIKKINEN

[Adam J]

16 hours ago, WolfieGlos said:

2) Registering and stacking as one for multiple sessions, is better than stacking the stacks.

Not surprising you are improving your ability to reject outliers by having a larger sample size if you stack the stacks then each stack will incorporate more outliers as the lower number of frames means that they will incorporate additional outlining pixel values into the average value for each sub stack. At that point you would need a very large number of sub stacks to enable outlier rejection to the operate on the stack of stacks (25 or more usually) and you don't have that many sum stacks. 

Adam

[WolfieGlos]

22 hours ago, ONIKKINEN said:

There is something like this - symbolic links. They create a "virtual" file that takes no space on disk and can in fact exist on another disk so you could save some space, but this step only applies for the initial sequence you create. 

Yes, those symbolic links were a life saver for disk space. A dedicated USB SSD drive for astro work is definitely on the cards, even before @Elp suggested it as that’s really all I can do for such a large amount of data at this stage. I forgot to say that as these are DSLR raw files, of course they start at 25mb each, each converted to fits files (50mb each) so that’s another load of storage - before any calibrating. So after calibrating them in siril, the raw files were moved onto an external HD and the fits deleted, before any registering could take place. No other way I could do it.

Stack the stacks is a lot easier for HD storage, and quicker to stack each session, but losing out on data and having higher noise is the trade off which I hadn’t been appreciating - somewhat obviously with hindsight.

I’m glad I made this post and tried both adding different data and registering all data to one stack, it’s been a very good learning exercise. I will try processing the results this week and see if there is any noticeable difference. Thanks for the help all.

[Elp]

I tend to stack the stacks, makes it more manageable in my mind and you can be certain you're checking each session for issues by doing this. The way I see it, you'll generally have noise to post NR anyway. The one time I've been surprised by a noise free image (or a smoother result) was when I was imaging in infra red, it gives a very smooth image but whether it's actually of any use...