Recommendations on my next imaging Camera after years with modded DSLRs?

[glend (Glen)]

Colin, Jon Rista over at CN, has developed a "Sparse Ditheting" formula which reduces time lost in dithering, ie you only dither after 'n' number of subs, worth a read. Its in the imaging forum.
And SGP now supports dithering every 'n' subs.

[RickS (Rick)]

Originally Posted by Shiraz

Rick, I think that you do get a signal-to-read-noise advantage from software binning. With 2x2, the signal goes up by 4x, but the read noise adds in quadrature and only goes up by 2x, so you get a 2x improvement in SNR - probably works out about the same as hardware binning with many chips, except that you also get the full 4x effective well depth (most hardware binning only gives about 2x).

Ray, I agree that you'll get a 2x improvement in overall SNR traded off for lower resolution. I was trying, albeit imprecisely, to make the point that you don't get the potential reduction in read noise that comes from HW binning.

The whole point of all this is that there is no perfect sensor or camera at present (perhaps ever.) The "best" choice depends on your requirements and the compromises you are willing to make. I like the big FOV of a KAF-16803 and I'm willing to do the long exposures required to make it work. If I wanted to target small galaxies in isolation then a small, low noise Sony sensor would be a better choice. The new CMOS sensors are interesting but don't appeal to me yet. In a year or two who knows?

Cheers,
Rick.

[Camelopardalis (Dunk)]

Given that many of the modern cooled astro cameras have very low dark current, the read noise is surely the biggest factor in determining the sky limiting exposure length. So, whether imaging LRGB or narrowband, a camera with lower read noise is advantageous unless it is completely outgunned by a camera with significantly better QE.

The quicker the subs are exposed adequately, the more subs you can rattle off in a night. The more subs you can rattle off in a sequence the better the SNR is going to be in the stacked image...to a point. There must be a crossover point where the read noise (additive) from more subs equals the read noise from a stack of fewer subs from a noisier camera. I'm sure one of you maths geniuses could show us the way

By my (possibly incorrect) mental arithmetic, with the "smaller" chip on the 1600 at roughly 1/4 the area of a full frame chip, a mosaic of 4 panels (and 64 megapixels!) works out at about 8e read noise. It'd be an interesting experiment to see how'd they'd compare.

I think this still means we need to deeply consider the best option to go with our kit and techniques...

[gregbradley]

Originally Posted by glend

Greg, the ASI1600MM-C is a mono camera.

My point in relation the advice i provide to the OP, is that he can probably get the ASI1600MM-C, plus the filter wheel, and a set of narrowband and broadband filters for less than $3K, and that equates to an entry level cheap much noisier KAF-8300 (Atik) alone. He is of course free to spend his money as he choses.

Oh I see. My error. Yes I agree that ASI1600 has the potential to outdo a KAF8300 but you'd have to be good with it as there are some amazing 8300 images out there which is a known and proven path.

Greg.

[gregbradley]

Originally Posted by glend

Greg, the ASI1600MM-C is a mono camera.

My point in relation the advice i provide to the OP, is that he can probably get the ASI1600MM-C, plus the filter wheel, and a set of narrowband and broadband filters for less than $3K, and that equates to an entry level cheap much noisier KAF-8300 (Atik) alone. He is of course free to spend his money as he choses.

Originally Posted by Atmos

This is only true if you don't overcome read noise. Take the 16803 which has a very similar QE to the ASI1600 but a RN of 8e- compared to the variable RN of the 1600. As long as they both overcome the read noise, their sensitivity is more or less identical. The 16803 will have to do longer subs to reach that point but it has MASSIVE wells to be able to accomodate for that.

Just my opinion but I feel the 16803 is still the best astro sensor out there despite it now being a bit old. Its a phenomenal sensor.

Greg.

[Atmos (Colin)]

Originally Posted by gregbradley

Just my opinion but I feel the 16803 is still the best astro sensor out there despite it now being a bit old. Its a phenomenal sensor.

Greg.

I agree with you Greg, read noise isn't the only limiting factor to consider for sub length. For instance, the ASI1600 running somewhere around unity gain has a read noise of about 2.3e- from memory. The FLI16803 runs at 8e-. The 16803 is noisier BUT has considerably larger pixels so on the same telescope becomes sky limited 2.15x slower which isn't that much when considering that is the difference between 60s and 130s! You get a much larger FOV but less resolution which can quite easily be recovered a bit by drizzling.

[lazjen (Chris)]

Well, the 16803 cameras are really a moot point for the budget proposed, particularly if you want to get a FLI. And especially when you add in the extras required. It would be a minimum 5 times the cost (as the ASI1600 setup).

[RickS (Rick)]

Originally Posted by glend

I believe people adapt their techniques to work to the strengths of their equipment. For example, with the ASI1600, owners are experimenting with short sub imaging and running large numbers of them, also using the varible gain feature to shoot short NB. With broadband subs of 30-60 secs not uncommon, but hundreds of them. This makes the mount and guiding much less of an issue. Deep wells are not really an issue for a camera with very low noise and short sub capability. Max depth on the 1600 is at Gain 70, but many of us shoot at Unity (Gain 139) and just shoot more subs, we gain our depth through stacking, not sub open shutter duration. As 50 subs at 300" is equal to 300 subs at 50 secs in terms of data acquired. Short subs also are less likely to be sky limited.

You seem to think that "sky limited" is a bad thing? It actually means that the limiting factor is the unavoidable shot noise from the target and sky glow... which is as good as it gets.

Originally Posted by Camelopardalis

The quicker the subs are exposed adequately, the more subs you can rattle off in a night. The more subs you can rattle off in a sequence the better the SNR is going to be in the stacked image...to a point.

Not even to a point, Dunk There are potential advantages of shorter subs but improved SNR is not one of them. For sky limited subs it is only the total integration time that matters. For read noise limited subs, more subs just means the same signal with more noise.

[Shiraz (Ray)]

Originally Posted by RickS

You seem to think that "sky limited" is a bad thing? It actually means that the limiting factor is the unavoidable shot noise from the target and sky glow... which is as good as it gets.



Not even to a point, Dunk There are potential advantages of shorter subs but improved SNR is not one of them. For sky limited subs it is only the total integration time that matters. For read noise limited subs, more subs just means the same signal with more noise.

Good points Rick. Had put together a summary that tries to say the same thing in a little more detail, so will post anyway - thanks to Dunk for raising the issue. Is in the usual pompous tone - please forgive . Matthew, if you feel this is inappropriate for your thread, will happily delete.

The number of photons detected is solely a function of the total exposure time. It doesn't matter if you use short or long subs, you get the same number of photons in total. Same applies to the total dark current, which is also summed over the whole integration time and also does not depend on sub length. Short subs do not give less total dark current and long subs do not give more total photons from faint targets - the total number of photons (signal) as well as the total dark current is completely independent of the sub length for a given total integration time.

The shot noise from the sky background after stacking is the square root of the number of (sky light electrons+dark electrons), so it is also solely determined by how long you image in total - it has nothing to do with the sub length.

The read noise is the only other source of noise and it depends only on how many subs you have - the total read noise in the full integration time is (the square root of the number of subs x read noise per sub), so the more subs you have, the more read noise you have. Low read noise cameras allow you to use more subs (ie shorter ones) before the total read noise becomes objectionable - which it will do if it gets up to a level where it becomes significant compared to the shot noise from the (sky + dark current). That is why the guys with high read noise cameras need to use fewer/longer subs than the folks with low read noise chips. However, with proper sub lengths, both will end up with the same result at the end of the integration. Of course, if there is no sky noise (eg is using NB with dark sky), everybody gets the best results using the longest possible subs (to give the lowest total read noise).

The total signal that you can gather from bright stars before they saturate is determined by the well depth applying to each sub multiplied by the number of subs. For example, 100x 1minute subs with well depth of 10,000e will allow the brightest brightest stars in the stack to reach 1,000,000e before they saturate. 10x 10 minute subs with a sensor having well depth of 100,000e, will allow the brightest stars to reach the same saturation level of 1,000,000e. Well depth by itself does not provide any advantage in terms of saturation or sensitivity - the thing that matters is the dynamic range (well depth/read noise) and there is no need for deep wells if you have low read noise. The other side of the coin is that, if you have high read noise, you must have deep wells.

the above discussion is based on additive stacking. Average stacking is generally used, but that is just additive stacking with everything scaled by the number of subs, so the conclusions are unaffected.

The main advantages of short subs are: increased resolution on the very extremes of lucky imaging; reduced requirement for high quality tracking and guiding; and the possibility that polar alignment can be a bit off, or even that altaz mounts could be used, without field rotation being a major issue. Low read noise does not provide any overall SNR advantage with well exposed broadband subs, but it can help when imaging dim narrowband targets under dark sky, where sky-limited exposures are not possible - if there is no issue with using long subs, you will get better results from a low read noise camera than from one with higher read noise.

[RickS (Rick)]

Originally Posted by Shiraz

Good points Rick. Had put together a summary that tries to say the same thing in a little more detail, so will post anyway.

Excellent summary, Ray. Hopefully, everybody will read it and we'll never have to have this discussion again

[g__day (Matthew)]

Folks,

I am really appreciating all this information. I hadn't thought of the ASI1600MM-Cool - but it does look like a really performer. From one of the CloudyNight reviews - it looks like with higher gain it had pretty bad AMP noise - that could be processed out. The idea of Binning 2x2 to get a 80K well hadn't occured to me either. Also pleasing to see Bintel stock it - as I find them 100% reliable to deal with over a very long time!

Be interested in knowing if anyone here has one and is staisfied with it. While the 16803 chip from FLI or S-Big that Greg mentioned looks - the ant's pants - with filters and focusers it probably North of $10K so I would work my way towards that I think - one I am sure of my skills and the set up matches.

I am finding this a very, very interesting thread. The Cloudy Night reviews mentioned the author was going from uncooled DSLR to the ASI1600MM-Cool - wish he had posted some best DSLR versus best ASI pictures for comparision on his edge 8" SCT.

Many thanks all, keep it coming,

Matthew

[Atmos (Colin)]

That is a very good summary Ray, a lot more succinct than I'd have probably been able to so as I do waffle on a bit at times

Getting an FLI PL-16803 with Centreline CFW, Astrodon LRGB and 3nm narrowband filters will cost you about $35,000 AUS at the current USD/AUD rate. So, yeah, it's pretty expensive My only concern with the 1600 with your SCT Matthew is that the pixels on the ASI1600 are a bit on the small side. It'll be fine for your refractor BUT at 0.33"/pixel on your C9.25 the sampling is not quite ideal.

The 16200 sensor with 6 micron pixels would be a better all rounder.

[Shiraz (Ray)]

Matthew, the amp noise is a minor issue. I did an apples-to-apples noise comparison with my (excellent) H694 camera. The noise from an "amp glow" section of the 1600 is shown in the top panel and that from the corner of the 694 in the bottom panel of http://www.astrobin.com/full/253351/...d=Noneattached . The amp glow+read noise in the worst area of the 1600 is lower than the dark current+read noise in the best region of the H694. These are 5 minute darks with bias subtracted, so what you see is the dark current+read noise. The two images have been scaled so that the degree of stretching extends over the same range in photo-electrons. Although the amp glow in the 1600 looks fairly bad by itself, it is actually miniscule when compared to other very good cameras and easily handled with flat calibration.

As to whether it would suit you, maybe as a 4mp @7.6micron chip with software binning on your 9.25 it would be a good match - take Colin's advice though, the pixels are too small without binning. The chip is just small enough to use 1.25 filters as well, which is a bonus. It is a bit different to use though - more parameters than you need to set on other cameras and the USB implementation is not as bulletproof as it might be. Overall though, it is an impressive camera, regardless of price.

[gregbradley]

Well depth by itself does not provide any advantage in terms of saturation or sensitivity - the thing that matters is the dynamic range (well depth/read noise) and there is no need for deep wells if you have low read noise. The other side of the coin is that, if you have high read noise, you must have deep wells.


I don't think that is quite true Ray. Deep wells allow you to capture bright areas in longer exposures before blowing out the highlights. Deep wells allow you to do longer exposures and retain star details and colour.

I have imaged with a few shallow well cameras. It may be the fact of large fast aperture but I found that shallow wells require shorter exposures otherwise your stars will be a mess to process later.

Also images taken with a deep well camera (like the 16803) the stars are very robust and take processing much more than the shallow well cameras.

So there is a very distinct advantage to saturation there. Shallow wells means you really have to do short exposures or use narrowband filters to reduce the light and get longer exposures there but even then perhaps not as long as a deep well camera. Deep wells are a plus not something to be ignored. Personally I am done with shallow welled cameras. Too hard to process well.

The KAF 16200 is at the minimum I would want - around 39K much like the 11002 at around 50K.

Greg.

[g__day (Matthew)]

A slight aside - I see Software Bisque's the SkyX Camera add on manages ZWO cameras directly - does that mean I could manage one of these Camera's from TSX (very useful for Closed Loop Slewing) or do people find its better to control these cameras directly through vendor specific tools?

I could still use it thru the Sky to do the closed loop slew - cause the pointing accuracy that gives me tends to be arc seconds precise.

[Shiraz (Ray)]

Originally Posted by gregbradley

Deep wells allow you to capture bright areas in longer exposures before blowing out the highlights. Deep wells allow you to do longer exposures and retain star details and colour.

I have imaged with a few shallow well cameras. It may be the fact of large fast aperture but I found that shallow wells require shorter exposures otherwise your stars will be a mess to process later.


Greg.

Well that is what I said Greg - shallow well cameras Must Have short subs to avoid saturation. If you don't use short subs with a shallow well camera, you can be guaranteed to have burned out data. But the point is that you can get good results from short subs with low well depth cameras because they also generally have low read noise. That means that short-sub/shallow-well cameras can go just as deep as long-sub/deep-well cameras - provided you take the same total exposure (ie lots of subs for the shallow well cameras).

The red herring seems to be that short subs can look very thin, so the temptation is to make 'em look more gutsy by hitting them with longer exposure. That is a fatal mistake that will burn the data. Short subs with low read noise cameras look thin because they individually carry only a very small fraction of the final signal and so they should look thin and crappy- it is only when you stack that you can see how good the data is and how much dynamic range you have. It is quite scary to use the ZWO1600 with 1minute subs - the target in each sub can look like a light frosting of brighter noise with slight structure, but really deep stuff just pops out when you stack 300-400 of them. I normally set the well depth to around 4000e and rarely have anything saturated in the field - I am quite capable of saturating stuff in processing, but rarely start out that way.

[Atmos (Colin)]

Originally Posted by g__day

A slight aside - I see Software Bisque's the SkyX Camera add on manages ZWO cameras directly - does that mean I could manage one of these Camera's from TSX (very useful for Closed Loop Slewing) or do people find its better to control these cameras directly through vendor specific tools?

I could still use it thru the Sky to do the closed loop slew - cause the pointing accuracy that gives me tends to be arc seconds precise.

TheSkyX allows ASCOM camera control so it doesn't matter either way. Right now I use TSX which uses the camera via MaximDL. This allows me to have several programs at a time using the the camera I can take a photo in MaximDL, FocusMax, TSX and Sequence (ASA mount control) all without having to reconnect the camera.

[Camelopardalis (Dunk)]

Originally Posted by RickS

Not even to a point, Dunk There are potential advantages of shorter subs but improved SNR is not one of them. For sky limited subs it is only the total integration time that matters. For read noise limited subs, more subs just means the same signal with more noise.

Originally Posted by Shiraz

The number of photons detected is solely a function of the total exposure time. It doesn't matter if you use short or long subs, you get the same number of photons in total...

Even better

Having read all this good stuff makes me think I'm going to be (eventually) trying the 1600 on my SCT (with reducer) too I originally jumped in with both feet thinking it would let me continue imaging through the QLD summer and maybe dabble with some narrowband from the inner city...

[Peter Ward]

Originally Posted by Shiraz

The number of photons detected is solely a function of the total exposure time. It doesn't matter if you use short or long subs, you get the same number of photons in total.

I don't agree...as the argument falls apart with very low flux emitters, where you can't ignore shot noise.

If you are only capturing say...one photon every five seconds...then a 1 second sub will likely fail to capture anything 4 or so times out of five.
This is assuming you have zero read noise and perfect QE....which is simply not available at present....the reality is you won't capture anything but noise.

But if we go deeper...a five second exposure will capture that photon. A ten second exposure will capture two etc.

Yes, you need exposure time, but reduced to is absurd conclusion, a million millisecond exposures is unlikely to produce anything useful.

[Shiraz (Ray)]

Originally Posted by Peter Ward

I don't agree...as the argument falls apart with very low flux emitters, where you can't ignore shot noise.

If you are only capturing say...one photon every five seconds...then a 1 second sub will likely fail to capture anything 4 or so times out of five.
This is assuming you have zero read noise and perfect QE....which is simply not available at present....the reality is you won't capture anything but noise.

But if we go deeper...a five second exposure will capture that photon. A ten second exposure will capture two etc.

Yes, you need exposure time, but reduced to is absurd conclusion, a million millisecond exposures is unlikely to produce anything useful.

That isn't how it works Peter - it is perfectly sensible to count subs that have no target photons. The signal is the total of the photoelectrons and the noise is the combination of shot noise and (read noise*sqrt(numbersubs)). For deep narrowband, the shot noise may be almost non-existent compared to the read noise, so the SNR = totalphotoelectrons/RN*SQRT(numbersubs). This equation applies to any sub length - there is no requirement for every sub to have signal in it.

To prove the point, have a look at M&Ts wonderful deep Helix. http://www.iceinspace.com.au/forum/s...d.php?t=127240
The outermost regions are identified to be equivalent to 9 photoelectrons/pixel - detected in 16x 1 hour subs ie, about half of the subs had no target electrons at all. But there's the signal - really beautiful and very faint shock fronts.

The 16 subs with the 16803 would have produced about 40e rms noise total. A camera with 2e read noise would produce the same 40e total read noise using 400 subs over 16 hours, so you could get the same result using 400x 2.4 minute subs - and you would still only have target photons in 9 of the subs. ie, all else being equal, identical results to M&Ts could have been produced by a hypothetical low noise camera using vastly more short subs. Of course, with broadband the situation is dominated by sky noise, but but the two systems would be equivalent in this environment as well.