Advice on choice of Camera

[festoon]

I'm looking to upgrade the sensor on my grab and go imaging rig and wanted to seek some advice on here if posssible.

My rig is a Samyang 135mm f/2 lens mounted on an AZGti in Eq mode. Currently I'm using a ASI224MC-Cool CMOS sensor, but am quickly seeing the limitations of this 12 bit sensor and limited full well capacity. At unity gain the FWC is 4096 electrons. My imaging rig uses an intel M3 compute stick which I remote desktop to when imaging. I use sharpcap for polar alignment and NINA for my imaging runs to acquire data.

I'd like to stick to a sensor which has small pixels. I've tried 6.45um pixels on an ATIK414EX colour and was not happy with the undersampling, and much prefer the image with pixel size from the ASI224MC-cool at 3.75um.

The options I have been thinking about are

• QHY268C currently at £1950 in the UK. Advantages that I can see are still small pixels of 3.76um giving a resolution of 5.74 arcsec/pixel. 16 bit ADC so vastly superior full well capacity of 65,000 electrons at unity gain, and even at 0.4 e/ADU the FWC is 26,000 electrons. And a huge field of view due to the APS-C sensor size at 10.01° x 6.71°. Being a colour sensor its much easier to create full colour images, but one question I was wondering is would I get significant image resolution improvement moving to a mono and filter set up. The other question I had was file size and download over usb. The sensor is a whopping 26MPixel image. What do people think about the chances of the compute stick not coping with the usb file size or memory issues? Also the weight is 980g, so a significant weight addition onto my set up (compared to the 410g of the ASI224MC cool)
• QHY268M currently at £1800 in the UK. All of the above except I would also buy a filter slider (not a filter wheel - to keep costs and weight down) and filters. As mentioned above  - would I see significant resolution improvements over the colour sensor? Would the sensitivity (quantum efficiency) be significantly better without the bayer matrix? Disadvantage, to create colour images I'd have to change filters.
• Starlight Xpress Camera Trius PRO-814 colour currently £2000 in the UK. Pixel size is again 3.69um, so resolution is 5.63 arcsec/pixel. 16 bit ADC with full well capacity >15,000 with a gain of 0.3 e/ADU. Read noise will be a little worse than the CMOS sensors above at 3 electrons. Sensor size would give a field of view pretty large at 5.29° x 4.23°. No amp glow at all (CCD) - I imagine would probably get away with no darks. Weight is only 450g so similar to the camera I have now. I'd hope at 9MPixels my compute stick would still be able to cope.
• Starlight Xpress Camera Trius PRO-814 mono currently £2000 in the UK. As with the SX colour but needing a filter slider and filters into the equation - would I see significant resolution and sensitivity improvements over the colour sensor?
• ASI533MC currently at £900 in the UK. Pixel size is 3.76um giving a resolution of 5.74 arcsec/pixel. 14 bit ADC giving a full well capacity of 16,000 electrons at unity gain. Decent FOV at 4.79° x 4.79°. Weight only 470g so similar to my current set up. I'd hope at 9MPixels my compute stick would still be able to cope.

I'd really appreciate some insights, advice, and help on this. Many thanks in advance.

Edited January 25, 2021 by festoon

[festoon]

Not sure if this should be or is best in the quipment - cameras section. If so mods please feel free to move.

[vlaiv]

1 hour ago, festoon said:

My rig is a Samyang 135mm f/2 lens mounted on an AZGti in Eq mode. Currently I'm using a ASI224MC-Cool CMOS sensor, but am quickly seeing the limitations of this 12 bit sensor and limited full well capacity. At unity gain the FWC is 4096 electrons.

How exactly 12bit and full well capacity limits you?

[festoon]

At unity gain the fwc is 4096 electrons. For a given exposure I believe this is leading to bloated stars being over exposed in a moderate sky pollution even for exposures of 10s at f2 

[vlaiv]

3 minutes ago, festoon said:

At unity gain the fwc is 4096 electrons. For a given exposure I believe this is leading to bloated stars being over exposed in a moderate sky pollution even for exposures of 10s at f2 

Well, use exposures at 5s or 2s - you loose nothing.

For example, you quote 533 with having 14bit ADC and 16000e at full well as camera that would suit you better.

Pixel size of each is roughly the same 3.75µm vs 3.76µm and QE is about the same at 80%.

Now take 4 exposures with ASI224 being 2.5s and one exposure of ASI533 of 10s - what difference will there be between the two?

If you add 4 exposures with ASI224 you will get 14bit and 16k FW capacity. You will get 2.4e read noise since read noise at unity gain is around 1.2e. ASI533 will have 14bit and 16k FW capacity and about 1.55e.

So, yes, ASI533 has slight edge in read noise - which is really not that important at F/2 if your subs are close to saturating - read noise will be drowned in LP noise and difference is minimal.

If bit count and full well capacity is your only concern - then you don't need to spend your money - you only need to change your approach - just simply take shorter exposures - for same total imaging time. Results will be 99.99% the same and you won't have saturated star cores.

Good reason to switch to new camera would be sensor size - you can capture much more of the sky in single go with larger sensor.

 

[festoon]

Thanks for the analysis @vlaiv

I’ve tried exposures at 1s and found walking noise to be an issue, but have not tried 2.5s for instance.

I can see the argument holding against the 533mc. But for the QHY268 surely there is an advantage there with a full well capacity of 65,000 at unity gain?

[happy-kat]

Doesn't dithering help with walking noise?

 

[vlaiv]

18 minutes ago, festoon said:

I can see the argument holding against the 533mc. But for the QHY268 surely there is an advantage there with a full well capacity of 65,000 at unity gain?

It wasn't argument against 533mc as much as argument for ASI224.

I don't mind you upgrading your camera - I personally value larger sensors as they are "faster" sensors (when paired with appropriate scope). My only concern is if you want to replace your camera for the right reasons.

With most cameras you will run into a problem of clipping the brightest stars. Only way around that is to use camera with 0 read noise - and such thing does not exist. Clipped stars are dealt with in different way. You take longer exposure for faint signal and in the end you do couple of short exposures for star cores.

You combine the two after stacking each in processing stage (best when still linear).

I don't believe that you are saturating your targets or that you saturate sensor due to LP. I'm also in rather heavy LP and I've just checked one of my subs that I've taken with ASI178mcc and Samyang 85mm F/1.4 lens (at F/2). Core of M31 has average ADU of about 2500 at gain higher than unity (this camera has lowest gain at about ~0.92) - that is about 2300e. This was with 60s exposure.

Granted, camera has 2.4µm pixel size - so ASI224 in these conditions would have 5500e - but that is 60s exposure and core of M31 - it saturates so easily.

[festoon]

My set up is unguided...have managed 90s with no issues

[festoon]

Thanks again @vlaiv. I really appreciate that you are asking relevant questions.

in general I am happy with the FOV, so not massively fussed about going to a bigger sensor area.

How are you combining exposures after stacking?

[festoon]

Also wanted to say @happy-kat..yes dithering would help. Trying to avoid if possible, but a guide camera may be a route I head down if it proves an advantage 

[vlaiv]

3 minutes ago, festoon said:

Also wanted to say @happy-kat..yes dithering would help. Trying to avoid if possible, but a guide camera may be a route I head down if it proves an advantage 

I got little guide scope for that purpose exactly - I can also get long enough exposure at F/2 with lens but I want to dither. Dither is rather important at randomizing noise stuff and helps with noise reduction.

7 minutes ago, festoon said:

How are you combining exposures after stacking?

That really depends on software that you are using, but simple procedure would be:

Stack both short and long exposures to their respective stacks. Align short stack with long stack (register them without stacking).

Use pixel math and replace all pixels in long stack that have value larger than 80% of max value with scaled value from short stack.

Scaled value from short stack simply means that if you took say 60s long subs and 2s short subs - you need to multiply short stack with 30 to get compatible values between stacks.

[festoon]

Here is an example image for 10s exposures. Yes the galaxy core is not overexposed (about 50,000 on a 16 bit scale) but the quite a few of the stars are overexposed and when stretching this leads to very bloated stars.

Also is an example of M42 with 3s exposures, and even then you can see stars that are over exposed

I guess if I exposed at 1s maybe I would have all the stars below the full well capacity. But I guess then I do have to worry about walking noise.

This is why I thought maybe the better option was to go for a deeper well

I'd really be interested in trying to do as you suggest of combining images. In general I'm using Siril and GIMP as software. I'm not sure how I would do the pixel math as you suggest.

2021-01-22_18-47-56__-15.00_10.00s_0013.fits 2021-01-22_00-15-12__-15.00_3.00s_0292.fits

[festoon]

Also a bit off topic...but has anyone tried with NINA (taken from the NINA documentation)

Built-in Dithering

There are cases where there is no guiding equipment in use, and thus no PHD2, but where dithering is still desirable. Examples of configurations like this usually include small, portable setups that have a main camera and telescope or lens, but no guiding. In cases like this, N.I.N.A. can still effect dithering operations, but on its own through its Direct Guider facility. Once activated, dithering operations become available in the Sequence and are effected by N.I.N.A. directly.

[vlaiv]

7 minutes ago, festoon said:

Here is an example image for 10s exposures. Yes the galaxy core is not overexposed (about 50,000 on a 16 bit scale) but the quite a few of the stars are overexposed and when stretching this leads to very bloated stars.

Also is an example of M42 with 3s exposures, and even then you can see stars that are over exposed

I guess if I exposed at 1s maybe I would have all the stars below the full well capacity. But I guess then I do have to worry about walking noise.

This is why I thought maybe the better option was to go for a deeper well

I'd really be interested in trying to do as you suggest of combining images. In general I'm using Siril and GIMP as software. I'm not sure how I would do the pixel math as you suggest.

2021-01-22_18-47-56__-15.00_10.00s_0013.fits 2.43 MB · 0 downloads 2021-01-22_00-15-12__-15.00_3.00s_0292.fits 2.43 MB · 0 downloads

I think I now get why you want to change the camera - again, you don't

Did you add UV/IR cut filter between camera and the lens? Your stars are indeed bloated but that is due to IR and UV part of the spectrum and lens not being corrected for it.

By the way - this is really sensitive camera and that lens is really good.

I mean - look at this:

that is single 3 second exposure. You can already see running man starting to appear and almost whole "loop" of M42.

Similarly - look at 10 second exposure of M31:

Core is saturated because I did quick linear stretch - but at 10 seconds dust lanes already start to show and M110 is there ...

Stars do look bloated and this is because of above issue. Another thing to consider - this Samyang lens is good - but you want to use it at F/2.8 if you shoot with OSC camera. It will be sharper that way. F/2 is good for narrowband data and similar, but with small pixels and OSC camera - F/2.8 is better / sharper.

[vlaiv]

1 minute ago, festoon said:

Also a bit off topic...but has anyone tried with NINA (taken from the NINA documentation)

Built-in Dithering

There are cases where there is no guiding equipment in use, and thus no PHD2, but where dithering is still desirable. Examples of configurations like this usually include small, portable setups that have a main camera and telescope or lens, but no guiding. In cases like this, N.I.N.A. can still effect dithering operations, but on its own through its Direct Guider facility. Once activated, dithering operations become available in the Sequence and are effected by N.I.N.A. directly.

I did not know that - it it makes perfect sense, no reason why imaging software could not issue mount command between exposures and provide dither functionality.

[festoon]

3 minutes ago, vlaiv said:

Did you add UV/IR cut filter between camera and the lens? Your stars are indeed bloated but that is due to IR and UV part of the spectrum and lens not being corrected for it.

There is no filters added for these images. Are you saying the image would benefit from having a luminance filter and this would prevent a lot of the star bloating?

 

5 minutes ago, vlaiv said:

Stars do look bloated and this is because of above issue. Another thing to consider - this Samyang lens is good - but you want to use it at F/2.8 if you shoot with OSC camera. It will be sharper that way. F/2 is good for narrowband data and similar, but with small pixels and OSC camera - F/2.8 is better / sharper.

I have not given f/2.8 a go yet. I guess that would also mean a little less light gathering, and possibly reduce over exposure? Is it sharper as the lens is more forgiving to aberrations at f/2.8?

[vlaiv]

1 minute ago, festoon said:

There is no filters added for these images. Are you saying the image would benefit from having a luminance filter and this would prevent a lot of the star bloating?

Yes, you need to use UV/IR cut filter with refractive optics and this camera - otherwise you'll have significant CA and bloating of stars.

This camera has only AR coated window but is very sensitive in IR part of spectrum, as it is originally meant to serve as surveillance camera (or rather sensor - not camera itself). It was purposely left with only AR window to enable astronomy use at IR wavelengths - like methane bands for planetary imaging.

You can use standard luminance filter or even LPS filter if you have one - that will help with light pollution.

3 minutes ago, festoon said:

I have not given f/2.8 a go yet. I guess that would also mean a little less light gathering, and possibly reduce over exposure? Is it sharper as the lens is more forgiving to aberrations at f/2.8?

Not only it will be sharper but will also further reduce any residual chromatic aberration. These lens are sharp and fast as far lens go, but are not up to telescope sharpness standards.

This means that they have some residual chromatic aberration when wide open.

Look at image of M31 I took with Samyang 85 F/1.4 stopped at F/2.0:

Look at that bright star - it still shows halo of blue. Other stars show red halo. Here is what different color channels look compared next to each other for that image:

That is R, G and B. Green is sharpest and looks good. Red is bloated and blue really has that halo issue.

In the end, I took artificial star and tested different filters and apertures to find good balance.

Here is comparison without any filtration I believe:

F/1.4, F/2.0, F/2.8 and F/4

You can see that at F/2.8 things start to look good and there is not much to be gained by going to F/4

[festoon]

Thanks @vlaiv...I'll give f/2.8 a go. Also looking through my kit I do have a 1.25" UV/IR filter that came with my old Revolution imager. I'll also give that a go and compare.

[festoon]

So here is a 3 second exposure at f/2.8 with the UV/IR filter in place. Still you can see many stars that are overexposed (including the M42 core). 

2021-01-25_23-05-02__-15.10_3.00s_0104.fits

[festoon]

Just wanted to illustrate the point about wanting to do exposures greater than 3s, I guess preferentially 10s or above.

Image 1 30mins of 3 sec exposures. Stacked in DSS, processed in Siril, levels adjusted in GIMP. I haven't applied a flat to this image yet. Whilst detail is good, I have not applied any noise removal, but clearly walking noise is present.

28 x 10 second subs (only 4 mins and 40 seconds...all I had before the trees obscured the image). I've applied a flat to this image. For sure this could do with more data, but at this exposure the walking noise is much less. However the raw subs show many stars that are overexposed (as attached)

2021-01-26_00-18-17__-15.00_10.00s_0022.fits

[vlaiv]

11 hours ago, festoon said:

Just wanted to illustrate the point about wanting to do exposures greater than 3s, I guess preferentially 10s or above.

I hope you understand what you have there in couple of minutes I takes hours to make something like that with different equipment.

In any case, why not experiment with setting gain to 0. Yes, it will increase read noise considerably, but it will also increase FWC to 19K - which should provide you with at least x4 exposure time.

Another option is to combine very short exposures with long exposures to fill in over exposed areas like we discussed. Stars seem much better controlled now as far as bloating goes, so at least that is fixed.

[festoon]

Thanks @vlaiv. I guess I never appreciated how sensitive the setup is. If I went for another sensor like the ones I put above would the sensitivity be reduced compared to what I have now?

I'm not sure about the physics behind fixed pattern walking noise. I guess the effect is more prominant at lower exposure times as the SNR is lower. If I decrease the gain setting on my setup, will walking noise be the improved, get worse, or be the same for a given exposure. My thoughts are it would get worse. I think overall I'm limited by the dymanic range of the sensor to detect faint and strong signals at the same time. If I decrease the gain to zero, the dynamic range will increase slightly, the full well increases to 19k (x4 improvement) and the noise only increases by x2.5. So I'd have an overall dynamic range increase of x1.7.

[vlaiv]

25 minutes ago, festoon said:

If I went for another sensor like the ones I put above would the sensitivity be reduced compared to what I have now?

That is not easy question to answer as so many things are involved.

It really depends on the way you use it and it's not straight apples to apples comparison.

I would say that 294mc pro is significantly more sensitive. On paper that does not make much sense since ASI224 is very sensitive as is - but don't look at raw numbers when comparing two cameras - look at how you are going to use them.

For example, look at this:

This is ASI224 paired with Samyang 135mm lens vs ASI294 paired with 500mm FL telescope.

You get same field of view and you can bin ASI294 to get the same resolution as ASI224. But you'll be using ASI224 with 48mm of aperture while you can use ASI294 with 100mm of aperture (you can find 100mm F/5 telescope or combination of aperture and reducer/flattener).

Now we have same FOV, same resolution - but ASI294 is using x4 more light collecting surface - hence it is x4 faster, regardless of the fact that Samyang is at F/2.8. However, you can't mount ASI294 + 4" scope on AzGti mount and camera + scope will cost something like x4 as much as ASI224 + Samyang lens (you really need 100mm triplet to get good performance and field flattener / reducer can cost as much as only Samyang lens).

That would be apples / apples comparison - which is still not comparing the same things - one setup is vastly more expensive not taking into account that you also need x3-x4 more expensive mount to be able to image with it - and portability is not going to be nearly as good.

If we on the other hand compare apples to oranges, then you'll get this:

You'll get vastly larger field of view with ASI294 and Samyang 135 and more sensitivity because ASI294 has larger pixels - but that will also mean less resolution. Less resolution is not necessarily bad thing with this lens as 3.75µm pixels are still too small for this lens (lens operates the best at 30 lines per mm - without loss of sharpness - and that is equivalent of 15µm pixel size - x4 smaller pixels are going to produce softer image).

43 minutes ago, festoon said:

I'm not sure about the physics behind fixed pattern walking noise. I guess the effect is more prominant at lower exposure times as the SNR is lower. If I decrease the gain setting on my setup, will walking noise be the improved, get worse, or be the same for a given exposure.

I still don't really understand how walking noise forms. I do understand that it is related to read noise / noise floor and that it forms when there is uniform drift in one direction  - which often happens when not guiding, but I wonder if it has something to do with the way data is calibrated / stacked?

What stacking software are you using?  Yes, you mentioned - DSS. Walking noise exclusively happens in DSS as far as I'm aware and for that reason I believe it could be related to issues with calibration or registration.

Could you maybe post 3s subs - whole lot that you used to produce first image with walking noise - so I can try to process it in ImageJ to see if there will be walking noise present as well. I'm suspecting bilinear interpolation being primary culprit and also 16bit precision when doing calibration.

47 minutes ago, festoon said:

My thoughts are it would get worse. I think overall I'm limited by the dymanic range of the sensor to detect faint and strong signals at the same time. If I decrease the gain to zero, the dynamic range will increase slightly, the full well increases to 19k (x4 improvement) and the noise only increases by x2.5. So I'd have an overall dynamic range increase of x1.7.

Dynamic range is really important when doing single exposure photography. It really does not matter much when doing stacking.

Stacking even 2 subs is increasing dynamic range by additional 1 bit. If you stack say 128 subs - you'll be adding 7 bits of dynamic range to base sub.

Having cameras with deep wells helps a bit - but does not solve the problem. Say you have camera with 14 bits of dynamic range. That is ratio of about 16000:1, or about 10 magnitudes of difference. Can you see 16 mag star in your image? Well, magnitude 6 star is going to saturate in single exposure - no way around it - even with 14 bits of dynamic range.

Well - there is a way around it - combine exposure lengths like I mentioned.

For this reason - dynamic range is really not important quantity for imaging. What is important is read noise - as it determines how long your regular exposure should be (and rule here is simple - make read noise small in comparison to any other noise source - often that is LP noise). Full well capacity again is not as important - if you have saturation - you'll need to take shorter / filler exposures.  These can be needed even if you have very deep wells.

In fact - even faint targets like galaxies can have more than 10 magnitudes of difference in brightness - cores are often at mag 17-18 while faint outer regions can be mag 28 - mag 30. It is not just stars that can saturate. Even in your example - we have M42 core saturating - that can happen on any camera.

[festoon]

Looking at the data...the walking noise is much better after 1 hour of 3 second exposures data compared to just 30 mins

Lots of detail in there so quite happy with the result.