Bigger scope or smaller pixels?

Hi all,

my question is inspired by this thread: How long will it take until we see the next generation of astrocams? - AstroBin and the fact that I have to carry my equipment to my stargazing location and therefore always watch out for low bulk and low weight.

Let's assume you want to image DSOs at an image scale of 1 arcsecond. You could do so with a camera with tiny 2 micrometer pixels (e.g. like the announced ASI 678) and a 420mm scope or with a camera with 3.76 micrometer pixels and an 800mm scope (give or take some mm of focal length in both cases).
The second setup would be much larger and much heavier - and you may need a larger and heavier mount in addition.

Would it be possible to achieve similar results with the "tiny pixel setup" (let's assume same focal ratio and same integration time) ? Or does such a "tiny pixel approach" not make any sense at all - if so, why not?

Clear skies
Wolfgang

No…it doesn’t quite work that way. Take a look at my presentation at AIC this year for an introduction on how to match cameras and how image sharpness works in terms of aperture, the atmosphere, and the sensor. You can find it here:

https://www.advancedimagingconference.com/articles/secrets-long-focal-length-imaging-john-hayes

John

Look up "imaging etendue" "concept" on CN ( https://www.cloudynights.com/topic/660553-etendue-calculator/ )

The thing is, light gathering capability of a pixel is affected by it's area - so for example, 3,76 um square pixels have an area of 3,76 x 3,76 = 14,1376, while 2,4 micron pixels have an area of 2,4x2,4 = 5,76. So 14,14/5,76 = 2,45 - so the larger pixels have much bigger (at same QE) photon collecting capacity. And it's hard to compensate for that factor with better f-ratio, you'd need to go from f/4 to f/2.8 to be able to "compete" with larger pixels. So the small 2,4-micron pixels are best for fast scopes, f/2.8, f/2 or f/1.4.

I have a f/4 setup with 2,4-micron pixels and it's painfully slow, need like 10-20 hours exposure. I plan to give up some resolution (I resample most of the time anyway), go to 3,76-micron pixels, stay at f/4 but increase aperture from 6 to 8" and expect to have 2,5x "faster" setup (imaging etendue calculation).

Peter

Not 100% sure what you're asking. I guess that, in theory your field of view would be similar… ish.. so a target will frame up quite similar in both setups.

Would I image DSO with it? - God no

1).. the 678 is a planetary camera. And not designed for DSO objects, so you're already limiting yourself there with such a setup up.
You lack the cooling and such, and also well depth and sensitivity.

The well depth of a 678 is almost 5 times less than a 533MC. A larger well-depth allows you to catch more photons in each image before a pixel gets saturated.. think of it as a bucket.. both with the same opening, but one is very tall, the other one is 5 times as short.. You start pouring photons into those buckets. One will fill up a lot sooner than the other. And any new photons are just being spilled out and not captured.

The low-ish well depth of the 678 is an indicator that it's more for fast exposures, like lucky imaging in planetary photography, because with long exposures you would fill that bucket and elements of your image can get saturated (mainly stars).. Obviously there is a lot more to this.. And a camera still needs to be matched up with the correct focal length..- So people with a better understanding of such things can probably explain it better.

The bucket of a 533 is also a lot wider. This allows it to capture a lot more photons per pixel before that bucket becomes full and saturated.

personally I run into this issue with my 183MC/MM, not as much the saturation, but definitely the light gathering power of small pixels.. They are great camera's. But it takes some work to gather the light in my F5 scope. - I see people with a 533 or 1600MM camera, posting great results after 3-6 hours of integration. And here I am, trying to collect 20+ hours before I'm happy enough with the results.

2) it highly depends on your area and seeing conditions. Aiming for 1 arc second can be great, but only if your atmosphere and seeing conditions allow for that to happen it.. if you're seeing conditions are worse than 1 arc second, you're not actually gaining anything by aiming for 1 arc second.

3) as for the weight. The weight difference between a 678 or a DSO camera is going to be minimal, depending on if you don't already have a powersupply on such a trip.

And while it is important to consider your pixelscale, it might not be the most important thing. Seeing conditions, Aperture, Diameter of the scope all come into play. Oversampling is "fine" - it just takes longer to gather light. Undersampling can be fixed with drizzling..

2) it highly depends on your area and seeing conditions. Aiming for 1 arc second can be great, but only if your atmosphere and seeing conditions allow for that to happen it.. if you're seeing conditions are worse than 1 arc second, you're not actually gaining anything by aiming for 1 arc second.

He meant pixel scale, not (potential) resolution. To resolve detail of 1" (FWHM/PSF) you would need a sampling of about 0.33". Some argue for even less than that.

Thank you for your enlightening answers! As the saying goes: "There is no free lunch".

Clear skies
Wolfgang

I tested the 2.4 micron pixels of the 183 cameras, not all that great. The combination of larger pixels + larger aperture will increase the photon collection capability vastly.

With the 3.76u IMX455 and a larger CDK14 aperture (0.66x) to squeeze more light into the pixels get much more data vs. a smaller scope at the same scale.

I then down sample further, the scale of 0.46"/pix is just too high for the seeing here (2.2 to 2.5). This increases SNR which helps get an image in less time as clear skies are gold dust here (Vancouver).

Recent example with decent resolution is relatively short integration times (14hrs from Bortle 6)

My opinion is go for the largest aperture you can and avoid 2u pixels.

Link:
https://www.astrobin.com/full/vxsvh4/0/?mod=&real=

Looking at the pixel signal you can see how much stinger it is with a large scope at the SAME scale.

This is the "fast" Tak FSQ vs the CDKs with the sensors.

If you use smaller pixels on the Tak it will be even worse (a LOT lower pixel signal).

Look up "imaging etendue" "concept" on CN ( https://www.cloudynights.com/topic/660553-etendue-calculator/ )

Be careful with the concept of image etendue. The assumption there is one of an extended object - meaning increasing the FOV covers more of the object of interest. That is often not the case.

Consider two imaging setups, one with a “slow” but large aperture scope and one with a fast but small aperture scope. So long as the FOV of interest is covered by both systems, the large aperture slow scope will outperform the small aperture faster scope in SNR when both images are viewed at the same scale. The fact that small pixels have smaller area is largely irrelevant with sensors of the same technology since I can always upsize the pixels to a common size so long as the larger pixels give me the resolution I want. On a per square micron basis, the light collecting capability of sensors of the same generation tends to be very similar regardless of pixel size.

I would second the idea of going with the larger aperture scope so long as your system covers your desired FOV and you have no issues around portability.

So even if the focal length is too long for your seeing and you are oversampling, you can always use that extra aperture for more signal/faster data collection.

As john has shown in several presentations, the sampling rate can be complicated but you can work out general rules of thumb around 3x your seeing.

So with typical 2.5" seeing, you can go below 1.0"/pixel and get more resolution.

I capture in bin1 mode and see how the data was for that night (FWHM) and then decide how much to down sample.

Some nights/months is over 3.0 at times it can get to 1.7" so you can tune your down sampling rate accordingly.

So even if the focal length is too long for your seeing and you are oversampling, you can always use that extra aperture for more signal/faster data collection.

Rouz,
I'm sorry but I have to straighten out your comment. Unless you are talking about stars, more aperture has nothing to do with signal strength--or how fast you gather data! For extended objects, which includes everything except stars, the irradiance in the focal plane depends ONLY on the inverse square of the focal ratio. The signal is then given by the irradiance multiplied by the area of the pixel and by the responsivity (aka QE) of the sensor. The telescope aperture does not appear anywhere in that relationship. The relevant thing that does appear is the area of the pixel, which means that signal is proportional to the square of the pixel dimension. That's why small pixels always result in less signal--and (almost always) lower SNR. On the flip side, although huge pixels produce a higher signal, past a certain point, it's easy to under sample the image, which results in decreased image sharpness. That's why it is important to pick the right size sensor. You don't want it too small and you don't want it too big.

More aperture helps to image fainter stars and to generally produce bigger images (for a give focal ratio), but it doesn't help to produce a brighter image in the focal plane!

John

The signal is then given by the irradiance multiplied by the area of the pixel and by the responsivity (aka QE) of the sensor. The telescope aperture does not appear anywhere in that relationship.

But isn't that increased signal created completely by focusing a larger angular area of the object of interest per sensor area? And neglecting the effect of pixel read noise, wouldn't the SNR on a per angular area of the object of interest (which is dictated primarily by the number of photons gathered, i.e., shot noise) be greater with the larger aperture scope regardless of F ratio? I bring up this point because it is commonly thought that a faster scope is better. Yes, a faster scope will produce a "brighter" image on the sensor, but that advantage would be lost when a common viewing scale is used, wouldn't it?

The signal is then given by the irradiance multiplied by the area of the pixel and by the responsivity (aka QE) of the sensor. The telescope aperture does not appear anywhere in that relationship.

But isn't that increased signal created completely by focusing a larger angular area of the object of interest per sensor area? And neglecting the effect of pixel read noise, wouldn't the SNR on a per angular area of the object of interest be greater with the larger aperture scope regardless of F ratio? I bring up this point because it is commonly thought that a faster scope is better. Yes, a faster scope will produce a "brighter" image on the sensor, but that advantage would be lost when a common viewing scale is used, wouldn't it?

Personally, I don’t like thinking in focal ratio for the simple reason:
Focal length and pixel size determines image scale in object space. For my conditions, for a seeing limited scope, a pixel scale of 1“ is probably adequate. Then: give me as much aperture as possible and what my mount can carry. And that’s where the practical issues come into play: we cannot have arbitrary focal ratios.

If you keep the image scale constant in object space, a larger aperture collects more signal.
That‘s what John has written.
Signal is proportional to pixel size ( p ) squared and inverse focal ratio ( fr ) squared: s = p^2/fr^2
(please read = in the sense of „proportional to“ as I don’t have all math characters here)

If your resolve fr as focal length ( fl ) over aperture ( a ), then you have the image scale ( res ) in your equation:
s = a^2/fl^2 * p^2 = a^2 * res^2

Therefore, if res is constant, and aperture increases, you get more signal.

Björn

PS/EDIT: to clarify. if you buy a larger scope and don’t take care about your sensor, you don’t gain anything. You either start binning or you buy a sensor that matches your scope.

John,

I probably didn't word it right.

I'm looking at pixel signal, not irradiance.
The way I think of it, when we look at an image, I see pixels not irradiance. Higher SNR is a cleaner image that allows me to pull out more detail.

If we keep the same scale of say 1.0", I can use a larger aperture slow scope and manipulate (bin) pixels that will have far greater SNR than small bin1 pixels used with a small scope with a fast focal ratio.

Here is an example of extended objects, both at the same image scale, both with the same camera, filters, and location.
Both resampled to the same scale of 2.29" / pixel

A 76mm Triplet @ f/4.5 with 140 mins (28 x 300s) of integration
318 CDK @ f/5.3 with 15 minutes of data (3 x 300s)
Interestingly the 3 subs from the big CDK have more SNR.
APO SNR = 35.02 db with 140mins @ f/4.5
CDK SNR = 39.62 db with 15 mins @f/5.25

At the same scale - 15mins with the larger scope has more SNR than 140mins with the "fast" smaller scope

Björn Arnold:

If you keep the image scale constant in object space, a larger aperture collects more signal.
That‘s what John has written.

Therefore, if res is constant, and aperture increases, you get more signal.

Björn

PS/EDIT: to clarify. if you buy a larger scope and don’t take care about your sensor, you don’t gain anything. You either start binning or you buy a sensor that matches your scope.

That's what I mean.

At scale = x
Get the largest aperture you can
Find or create pixels to achieve your target scale of x

For me, X is somewhere between 2.5~3.5x seeing.

Focal ratio isn't key as pixel size can now be manipulated so easily.
The reason is that we are looking at images produced by the combination of telescope + sensor.

A fast focal ratio by itself meaningless, my phone's camera lens is f/1.6

What you can not compensate for is aperture.

Rouz

Focal ratio isn't key as pixel size can now be manipulated so easily.
The reason is that we are looking at images produced by the combination of telescope + sensor.

Although, one should stay within reasonable limits. If the ideal sampling would be realized through 10um and your sensor has 2.5um, you have 16 read events and each adds noise (obviously I am talking about CMOS). If you can avoid that, you should.

If one does the math, the SNR is proportional to the pixel size and thus for the above example, the SNR of the good match would be 4 times the SNR of the highly oversampled sensor.

(Again: true for CMOS but not CCD)

Björn

A fast focal ratio by itself meaningless, my phone's camera lens is f/1.6

What you can not compensate for is aperture.

Exactly this. Blindly applying formulas for etendue as suggested in that CN post would suggest that my 35mm f/1.4 camera lens will outperform my 92mm aperture triplet, or even a 14" aperture CDK! Obviously that's ridiculous. It will certainly put down a brighter image on the sensor but do this purely by covering a vastly larger area of sky. It will collect far, far fewer photons from every square arc second of the sky than the two larger scopes. Take the image produced by the lens and view it at the image scale of the 14" CDK and compare it to an image taken using the CDK and there really will be no comparison.

As for SNR, yes, on a pixel level, SNR depends on pixel size, but in object space, SNR is governed purely by the aperture of the scope. The light entering your scope does not know or care what the focal length of the scope is. This is why, viewed at a common image scale, large aperture scopes will outperform small aperture ones for the same integration time, as demonstrated by Rouz in his comparison.

A fast focal ratio by itself meaningless, my phone's camera lens is f/1.6

What you can not compensate for is aperture.

Rouz

I take this as the summary of the discussion. Thank you all so much for sharing your amazing knowledge with a beginner. This discussion helped a lot to put some things straight in my head.

Looks like I have to start excercising to take heavier gear into the field. Maybe replacing weight of the astro photographer by weight of AP gear is a healthy strategy ...

Clear skies
Wolfgang

Blindly applying formulas for etendue as suggested in that CN post would suggest that my 35mm f/1.4 camera lens will outperform my 92mm aperture triplet, or even a 14" aperture CDK!

In terms of "how much total exposure do I need to produce a noise-controlled deep sky image", it probably will.
In reality, you of course decide what kind of system you are looking for = resolution in "/px and (minimum) FOV. Then you can compare the speed via imaging etendue.

My 150mm f/4 newton with QHY183C has etendue of 12833, which is super slow. I need like 10-20 hours to get a decent image at 0,83"/px.
My old Borg 71FL f/3.6 + QHY8L setup (huge pixels) had etendue of 112545! 2-3 hours were enough for decent image, of course at 6,1"/px.
Best match for 183 sensor would be RASA11 at f/2.2 - now that is the exact same FL as my current 150 f/4. RASA11 would have etendue of 42422, so 3,5x faster. But here you will notice we are all taking about the same thing - so I kept the FL, pixel size and sensor size (and thus FOV and resolution) and increased the speed by ... going to faster f-ratio which is accomplished by increasing the aperture. So, larger aperture wins because it automatically means faster f-ratio at fixed focal length.

As for the OP's question - I decided to abandon my small pixel 0,83"/px setup (etendue 12833) and settled for slightly "worse" resolution at 0,97"/px with 200mm f/4 and IMX571 camera with larger pixels - etendue of 30000. I do not really care about FOV as long it's practical, I have no issues with cropping. It is heavier, but at fixed location, so not an issue. For portable setup with resolution around 1" (+- 0,3), I would consider RASA8+183C (1,24"/px, etendue 51300), Sharpstar 130 f/2.8 + 183C (1,36"/px, etendue 26200), Sharpstar 150 f/2.8 + 183C (1,18"/px, etendue 26200) or 150 f/4 + Nexus 0.75x corr/red = 150mm f/3 + 183C (1,1"/px, etendue 22800). Of course, the downside of all these systems is tilt/collimation, which will be a nightmare due to f-ratios and need to do it often.

Peter

My 150mm f/4 newton with QHY183C has etendue of 12833, which is super slow. I need like 10-20 hours to get a decent image at 0,83"/px.
My old Borg 71FL f/3.6 + QHY8L setup (huge pixels) had etendue of 112545! 2-3 hours were enough for decent image, of course at 6,1"/px.
Best match for 183 sensor would be RASA11 at f/2.2 - now that is the exact same FL as my current 150 f/4. RASA11 would have etendue of 42422, so 3,5x faster. But here you will notice we are all taking about the same thing - so I kept the FL, pixel size and sensor size (and thus FOV and resolution) and increased the speed by ... going to faster f-ratio which is accomplished by increasing the aperture. So, larger aperture wins because it automatically means faster f-ratio at fixed focal length.

I suspect we are just coming at this different ways. Yes, your 150mm f/4 Newtown+QHY183C would be slower than your 71mm Borg+QHY8L, but you could improve its etendue by binning pixels so long as the FOV was covered. Binning pixels is easy, but increasing aperture isn't! The best way to illustrate this is to image an object that can be covered by both systems (say a planetary nebulae or small galaxy), then present it at a common image scale. The 150mm f/4 will outperform the Borg in that case. Most of us intuitively know this and would never use a small refractor to image a PN or small galaxy in preference to a larger aperture scope. The point is, what you want to image matters. There isn't a single formula that covers every case, but rather a deeper understanding of the subject.

The way I see it, apart from the fact that it should be obvious that aperture matters and matters a lot per se, is that resolution is king and to achieve resolution you need aperture and the more the better.

Thank you Peter, I already had the Sharpstars in mind and even had contacted some of the owners who seem to be quite happy with them. However I still refrain from "collimation hell". RASA or the Sharpstars may not be a good choice as the first Newtonian although according to some of the users the Sharpstars seem to keep their collimation well. Maybe one of those 150mm f5 Newtonians is a good first step into Newtonian world.

Ok, will head over to the treadmill now …

Clear skies
Wolfgang

I did not read all answers, I must confess, but I wanted to add my zwo pennies worth out of my experience:

1. I own both ASI183 versions, the C and the M and yes, it is not really easy and not really a huge pleasure to use them at my 1250mm/f4 Newton. Why - because none of the subs seem to be really "sharp" and in focus and the HFR is at its best at around 3.1-3.3 if seeing permits. The reason is the dramatic effect of too small pixels compared to the resolving power of the scope. Downsampling helps, but not with a OSC and I prefer therefore to use one of my other cameras with pixel dimension of e.g.3,76micron.

2. If you think about a shorter focal length with a larger aperture - keep in mind, that the depth of field on the camera side is dramatically decreasing. I know what I am talking about from my RASA11 - it is not too easy a job to get a APS format camera properly aligned - tilt is one problem, the other is a correct focus plane. I know, I am far from perfect often and just give up to use the time for collecting photons instead of spending hours with tweaking my equipment.

3. Transportation of a wide-open - like f2 or so - equipment is much more a risk of loosing its alignment than with an f4 or f6 equipment.

I know, the ideal scope would be f1 with a depth of field of an f8 and the simplicity of handling like a f8 refractor…but this does not exist

CS
Georg

Björn Arnold:
If the ideal sampling would be realized through 10um and your sensor has 2.5um

As of today, we as amateurs are really limited to the 3.76u pixels of the IMX family (571, 455, 461, 411 etc).

With the numbers of these sensors I see little point messing about with others like the 183, 294, KAF16803 or even Gsese4040.

If you look at the FWC they can hold per sqr. micron of area, nothing comes close to these for now. Excellent QE, no amp glow, and excellent price.

As for CMOS binning, yes not as good as CCD, but keep in mind the read noise is extremely low and so the dark current.

So for now, its best to consider pixel size = 3.76 and bin accordingly.

Rouz

Etendue is of little use unless you are after FOV.

Compare 2 identical telescopes
A: With the APS-C QHY268 / ASI2600
B: With the full fame QHY600/ASi6200

B will have a LOT more etendue. When you image an object like M51, crescent nebula, etc. The object will look EXACTLY the same.

Sure you will capture more surrounding area but that's it.

To increase the "quality" you need more SNR, so its best to pursue pixel signal instead of etendue.

Rouz

Related discussions