Go Back   IceInSpace > Equipment > Astrophotography and Imaging Equipment and Discussions

Reply
 
Thread Tools Rating: Thread Rating: 10 votes, 5.00 average.
  #21  
Old 04-03-2013, 09:31 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by PRejto View Post
Hi Ray,

Thanks for this interesting post. Like everywhere this is discussed nobody can ever agree!! But, I think Greg is right in the sense that as these cameras are used more and more certain facts will become clearer no doubt. I'm in no position to argue the technical side of things.....

I'm wondering if you might share which camera you decided to buy (and perhaps why), and which scope you intend to test this with?

I wonder how it might perform with my TEC140 at f7 (.95 arcsec/pix) compared with my KAF8300 (1.14 arcsec/pix)? Not a significant improvement in resolution, but would I see much improvement in sensitivity for RGB and Ha? Would it be worth the decreased FOV? Of course, as reported by several, the possibility to not need darks or deep cooling, and lighter package are all appealing selling points.

Thanks,

Peter
Hi Peter. Put the two options into the model. Looks like your scope would be a good match for either chip for hi res imaging down to about 2 arc sec seeing.
  • The 8300 will give you a much larger field of view.
  • The system broadband sensitivity will be about the same with either camera (the bigger pixels of the 8300 help with sensitivity, but lose resolution).
  • The 694 will have much better SNR (maybe 2x) at low signal levels in Ha narrow band due to the low noise of the 694.
  • The 694 will give you significantly better resolution (if you have a target that covers 100 pixels on the 8300, the 694 will put 140 pixels on it). that would be a noticeable advantage in good seeing
  • The 694 has a somewhat better dynamic range than the 8300, and the very low read noise of the 694 gives you more flexibility in exposure strategy.

Not sure if you want an opinion, but if it was me, I would push the 8300 as far as I could before deciding if I needed a replacement - but that is entirely your call, based on what you want to do. And of course you also need to consider the ancillary mechanical issues.

hope this is helpful. regards Ray

Last edited by Shiraz; 05-03-2013 at 11:00 AM.
Reply With Quote
  #22  
Old 04-03-2013, 10:03 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by Peter Ward View Post
I'd honestly say my pedaling KAF based devices is a sideline...and I do try to keep any commercial biases out of any postings...that said..

I'm very much a user of Astro-imaging gear with KAF8300, KAF16803 and ICX694 senors (plus a number of Canon DSLR's) all being in my personal arsenal.

My KAF8300 does indeed saturate very quickly... but I'd not call 25% better performance a "little" improvement over the ICX694.

A boost in QE, for example, by 25%, is obvious when you look at the raw data.

I also think Sony's noise figures are a little rubbery...clever correlation filters that also scrub photon liberated electrons can easily make noise look good at the expense of subtle signal.

Also can you, or someone, point me to a SONY ICX694 data sheet with absolute QE?...I can only find 3rd party specs...which I frankly don't trust.

The only method I've read about...apart from deep cooling...that keeps thermal electrons in the background is "Skipper" CCD technology...sadly not commercially available as far as I am aware.

I suppose what I am still saying, when it comes to pixels, is: bigger = better.

The downside is, bigger (matching optical systems) often cost a whole lot more....
I guess we will just have to disagree Peter.

On the quantum efficiency issue, Clive's post points to an independent evaluation that compared the 694 directly with a 3200 - the measured QE was consistent with those published by such third parties as Apogee, Atik, Pt Grey etc. It really is very high.

I also think Sony's noise figures are a little rubbery...clever correlation filters that also scrub photon liberated electrons can easily make noise look good at the expense of subtle signal.
????? If Sony had found a way to get rid of noise and not signal, they would own the telecommunications world - they don't. What they are actually doing is more likely to be something like MPP, which is discussed some way through this: http://learn.hamamatsu.com/articles/ccdsnr.html

Note also the equation for dark current

D = 2.5 × 1015 × A × Id × T1.5 × e-Eg / (2kT) where A is the pixel area - which means that dark current is directly proportional to pixel area - big pixels are not necessarily better.



Quote:
Originally Posted by clive milne View Post
Ray,
you might find this interesting:

http://www.astrosurf.com/buil/isis/noise/result.htm

Christian Buil presents a noise and electronic gain evaluation of a sample of astronomical CCD cameras including; KAF8300, KAF1603, KAF16803, KAF3200, KAF402, KAF11000, KAI2020, ICX694, ICX285.
Thanks for that Clive - a very useful resource and points out how good your 3200 is.

Quote:
Originally Posted by gregbradley View Post
Hi Rally,

An interesting thread. I agree sampling is definitely an important factor and matching pixel size to focal ratio is a well known datum. 1 arc second per pixel is one approximation that AP recommends and others use .66 arc seconds per pixel as being more specific to using the Nyquist sampling theory in application (3 times sample is better than the minimum 2X).

As far as gain goes in 8300 chip I don't know anything about that so I take your work for it.

As far as small pixels will have low well depth and thus likely to show white stars - I have seen examples of better looking stars in shorter exposures in fast scopes. Longer scopes probably less of an issue but then too long a focal length and the small pixels are no longer giving you proper sampling and as you point out you lose sensitivity. I have seen that with my own eyes and this is true.

As far as small pixels not having large dynamic range I think that has some assumptions. It is assuming the same read noise and as the post linking Christian Buils work on measuring some of these common chips you can see the Sony chip is in totally another league with lower read noise whilst higher QE. As dynamic range is a function of read noise (too much read noise and you lose differentiation between some levels of shadow or brightness).

As I thought Sony is totally out in front with clear air from other sensor manufacturers and are leaving them all in the dust. Canon is several generations behind Sony. Sony is also just going from strength to strength. They just last week signed a cross sharing of imaging patents with Aptina who has a huge number of imaging patents. Aptina often make excellent sensors for Nikon cameras for one.

So that means Sony will have access to Aotina's technology to actually increase well depth on any sensor of any pixel size. There is a video of Aptina President describing the technology.

A few years of development at this rate and we'll all be using some sort of hitech CMOS sensor unless True Sense releases some new advanced chips. Probably a while yet but I already see the signs.

My prediction is this 694 chip will be very very hard to beat and will see a lot of people switching their 8300 cameras for one or at least new buyers going to the Sony Darkside.


Greg.
thanks for the summary Greg. Interesting that Aptina is tying up with Sony - they have some incredible small pixel technology - did you read the future directions white paper on their website? If only some of these guys would make some of their excellent DSLR chips without the Bayer filters - one would think it would not be that hard to do. Problem is probably that nobody high up in Sony is even aware that there is a tiny amateur astronomy market or that some of their chips find their way into it.

regards Ray

Last edited by Shiraz; 04-03-2013 at 10:51 PM.
Reply With Quote
  #23  
Old 04-03-2013, 10:48 PM
clive milne
Registered User

clive milne is offline
 
Join Date: May 2011
Location: Freo WA
Posts: 1,443
Quote:
Originally Posted by Shiraz View Post
Thanks for that Clive - a very useful resource and points out how good your 3200 is.
Your welcome.

A couple of points though;
I am not sure that Christian Buil actually states that he measured the QE of the Sony chip relative to the Kodaks, he may have been relying on a third party. Be that as it may, some things really stand out; the low noise of the Sony chip enables it to achieve the highest ratings in the group despite its relatively shallow well depth. Even if the published QE figures were somewhat in error, it would still be amongst the best in its class. This probably needs some qualifiers, but for the application you are going to use it for this is unlikely going to be an issue.
Also, there appears to be some variation in the performance of cameras fitted with ICX694 chips and whether you bin them or not. (which implicitly negates the well depth limitations of the chip at the expense of over(?) sampling)

As for my 3200... ahh well, I find myself spoilt for choice.
I now have an FLI IMG6303 to play with so the ST10 is somewhat redundant.
Reply With Quote
  #24  
Old 04-03-2013, 11:18 PM
Peter Ward's Avatar
Peter Ward
Galaxy hitchhiking guide

Peter Ward is offline
 
Join Date: Dec 2007
Location: The Shire
Posts: 8,088
Quote:
Originally Posted by Shiraz View Post
I guess we will just have to disagree Peter.

On the quantum efficiency issue, Clive's post points to an independent evaluation that compared the 694 directly with a 3200.......

If Sony had found a way to get rid of noise and not signal, they would own the telecommunications world - they don't. What they are actually doing is more likely to be something like MPP, which is discussed some way through this: http://learn.hamamatsu.com/articles/ccdsnr.html


regards Ray
Clive's post is indeed interesting...and I have to agree presents a great case for high absolute QE with the 694. Remarkable really.

MPP has been around for a while. Most CCD's including Kodak use the technique. Yet, while Sony chips are indeed cleaner looking, I suspect you really have made my case: getting rid of noise without any signal loss is probably not happening here.

The Hamamatsu link is pretty much a condensed version of my copy of Jansick's substantial tome on the subject. Both authors point to cooling as THE method to reduce dark current....which is indeed higher with larger pixels....yet while you can chill a large pixel a tad more, it's very much harder to expand small pixel to improve its well depth. On chip binning you say? Well, yes, but doesn't that defeat the purpose of those tiny pixels?
Reply With Quote
  #25  
Old 05-03-2013, 08:26 AM
rally
Registered User

rally is offline
 
Join Date: Sep 2007
Location: Australia
Posts: 896
I wonder if we are missing the fact that the Dynamic Range of the chip whilst important its only just one parameter of the many needed.

I'll give an example
For the argument lets take a hypothetical (extreme) example of a Skipper Chip that is 'available' to amateurs to illustrate the issue.
This hypothetical chip has the advantage of extreme low read noise per the published Skipper CCD design potential of 0.2e-
These chips have actually been built !

Its Dynamic Range is therefore 88db or 25,000 using the standard calculation

If we compare that to a few cameras we know and use . . .
STL11000m - 73db or 4545
Sony ICX294 - 77db or 4000
DSIPro ICX254 - 67db or 2261
KAF8300 (std) - 70db or 3188
ST10 KAF3200 - 79db or 8750

Then on first inspection 88db looks extremely impressive doesnt it.
Well ahead of the pack

But if I then reveal that the well depth in this hypothetical example is only 5000e- what would we say ?

The reality is that it probably wouldn't be much good for our purposes
Yet 5000 is actually 12.3 bits, so in theory it ought to be quite useful.

When Sony release a 15mp, 9um chip then we'll have something !

Rally
Reply With Quote
  #26  
Old 05-03-2013, 03:39 PM
PRejto's Avatar
PRejto (Peter)
Registered User

PRejto is offline
 
Join Date: Jan 2011
Location: Rylstone, NSW, Australia
Posts: 1,397
Quote:
Originally Posted by Shiraz View Post
Hi Peter. Put the two options into the model. Looks like your scope would be a good match for either chip for hi res imaging down to about 2 arc sec seeing.
  • The 8300 will give you a much larger field of view.
  • The system broadband sensitivity will be about the same with either camera (the bigger pixels of the 8300 help with sensitivity, but lose resolution).
  • The 694 will have much better SNR (maybe 2x) at low signal levels in Ha narrow band due to the low noise of the 694.
  • The 694 will give you significantly better resolution (if you have a target that covers 100 pixels on the 8300, the 694 will put 140 pixels on it). that would be a noticeable advantage in good seeing
  • The 694 has a somewhat better dynamic range than the 8300, and the very low read noise of the 694 gives you more flexibility in exposure strategy.
Not sure if you want an opinion, but if it was me, I would push the 8300 as far as I could before deciding if I needed a replacement - but that is entirely your call, based on what you want to do. And of course you also need to consider the ancillary mechanical issues.

hope this is helpful. regards Ray
Thanks Ray, and, yes, I always value an opinion!

May I ask another? What about well depth vs light pollution? Is it true that light pollution will have a greater effect on a chip with a smaller well depth? For example if I compared the KAI4022 to the ICX694 the KAI chip well is 2x the well size ICX694. If LP in a given subframe took up say 5000 that would be 25% of the ICX694 but only 12.5% of the KAI. Would that then give a better dynamic range to the KAI chip? If so, imaging in a light polluted environment, would it be better to trade higher QE for greater dynamic range?

I ask because I struggle with background noise imaging as I do from Sydney, and I'm wondering if what I really need is greater dynamic range rather than another camera with a small well. Or, have I got this wrong and the greater SNR of the ICX694 would make a big difference?

Thanks,

Peter
Reply With Quote
  #27  
Old 06-03-2013, 06:14 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by rally View Post
I wonder if we are missing the fact that the Dynamic Range of the chip whilst important its only just one parameter of the many needed.

I'll give an example
For the argument lets take a hypothetical (extreme) example of a Skipper Chip that is 'available' to amateurs to illustrate the issue.
This hypothetical chip has the advantage of extreme low read noise per the published Skipper CCD design potential of 0.2e-
These chips have actually been built !

Its Dynamic Range is therefore 88db or 25,000 using the standard calculation

If we compare that to a few cameras we know and use . . .
STL11000m - 73db or 4545
Sony ICX294 - 77db or 4000
DSIPro ICX254 - 67db or 2261
KAF8300 (std) - 70db or 3188
ST10 KAF3200 - 79db or 8750

Then on first inspection 88db looks extremely impressive doesnt it.
Well ahead of the pack

But if I then reveal that the well depth in this hypothetical example is only 5000e- what would we say ?

The reality is that it probably wouldn't be much good for our purposes
Yet 5000 is actually 12.3 bits, so in theory it ought to be quite useful.

When Sony release a 15mp, 9um chip then we'll have something !

Rally
Well depth by itself does not tell you very much. Anyone who gets a chip like the one you describe into our hobby market will make all our current CCDs and high performance tracking mounts pretty much obsolete. Being able to use hundreds of short exposures without the noise going up much would open up the possibility of "lucky imaging" for DSO. It would also mean that there would be no need for mounts that could track perfectly for long periods of time. Provided there are no export restrictions, you may be able to currently buy EMCCD cameras with with less than 1 electron read noise (eg Andor, E2V) - if you have very deep pockets.

Quote:
Originally Posted by PRejto View Post
Thanks Ray, and, yes, I always value an opinion!

May I ask another? What about well depth vs light pollution? Is it true that light pollution will have a greater effect on a chip with a smaller well depth? For example if I compared the KAI4022 to the ICX694 the KAI chip well is 2x the well size ICX694. If LP in a given subframe took up say 5000 that would be 25% of the ICX694 but only 12.5% of the KAI. Would that then give a better dynamic range to the KAI chip? If so, imaging in a light polluted environment, would it be better to trade higher QE for greater dynamic range?

I ask because I struggle with background noise imaging as I do from Sydney, and I'm wondering if what I really need is greater dynamic range rather than another camera with a small well. Or, have I got this wrong and the greater SNR of the ICX694 would make a big difference?

Thanks,

Peter
The QE advantage of the 694 is offset by the larger number of photons intercepted by the bigger 8300 pixels - at the expense of reduced resolution though - both chips have about the same broadband sensitivity with your scope, just different resolution.

I don't think well depth makes any difference in heavy light pollution. The sky signal comes with a whole bunch of Poisson noise and if the sky is bright, system SNR is completely determined as the sqrt of the number of photons detected. Thus, you could get almost the exact same result (in both SNR and headroom) from any camera by choosing an appropriate number of exposures and exposure durations - eg exposing 8 frames with an 8300 or 10 shorter frames with a 694 would give ~the same result (same target signal, same SNR, same headroom). The low noise of the 694 will not help at all in this situation, since read and thermal noise will be swamped by sky noise in almost any camera - unless you do a huge number of short exposures. If you are having star saturation problems due to the sky offset, you might try more numerous short exposures to reduce both sky and star signals - you should be able to take a lot of them before you will start to see any read noise.

If you decide to do narrow band imaging, there will be very little residual sky noise and the low read noise of the 694 would give it a significant advantage, even if you trade some of it off to give more headroom.

I clearly need to expand my model to incorporate various exposure strategies in different sky conditions - thanks for the inspiration to do this.

regards Ray
Reply With Quote
  #28  
Old 06-03-2013, 08:25 AM
rally
Registered User

rally is offline
 
Join Date: Sep 2007
Location: Australia
Posts: 896
Ray,

I agree it will have an interesting place, but I don't think it would open it up to "lucky imaging" on DSOs !
Thats the whole point of my argument about well depth - it wont be catching enough photons to record any of the faint nebulosity !

A short exposure will simply get zero photons from such a weak signal source or that the shot noise will the same or similar to the signal noise.
The only way to capture this sort of detail is a very long exposure.

But I agree a more holistic approach is going to be better than isolating one and trying to deal with that on its own - one that considers all the parameters and then apply them to the huge number of different imaging applications and variations in telescopes and local environments.

Its likely to be difficult to display since its a multidimensional puzzle, and once embedded in a hidden formula or buried in a bunch of tables less likely to be visually meaningful.

I wonder how much you could display in a 3D cloud plot - at least you would have three dimensions

Don't misunderstand my approach - I am not trying to hammer any particular chip in favour of another, on the contrary there is a need for them all !

Rally
Reply With Quote
  #29  
Old 06-03-2013, 07:14 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by rally View Post
Ray,

A short exposure will simply get zero photons from such a weak signal source or that the shot noise will the same or similar to the signal noise.
The only way to capture this sort of detail is a very long exposure.


Rally
Hi Rally.

thanks for raising this interesting offshoot from the original post. Did some basic calculations on a standard CCD (10e read noise) and the one you describe (.5 e read) with broadband imaging of NGC253 as the target under dark skies with a 12 inch scope at 1 arcsec scale.

The result are very interesting and I am a bit hesitant to publish because they will be ridiculed - but here goes anyway:
  • standard CCD: 30:1 SNR at roughly 13x5 minute exposures - headroom maybe 600,000:1
  • low read noise CCD: 30:1 SNR at roughly 4000 exposures of 1 second - headroom maybe 20,000,000:1 - could comfortably work faster than 1 Hz

At 1 second or less exposure, much of the seeing and most tracking errors would be defeated if a stacking program like Registax was used. Resolution would potentially be determined more by the scope than the seeing. The mount would not need to be anything special at all to keep a scope on target for 1 second periods.

In anticipation that the idea of detecting a few photons in each sub is not accepted, try the lucky imaging websites - a quote from http://www.ast.cam.ac.uk/research/in...maging.methods is: "Basden et al6 have further demonstrated that it is possible to achieve close to full quantum efficiency and photon counting operation even at signal levels of a few photons per pixel per frame, something that is actually quite a high signal rate in terms of photons per pixel per second, given that these devices can be operated at many tens or hundreds of frames per second' They are using scopes with 100x the aperture, but they are also running at 100x the speed I propose.

I am looking at a little under 1 photon per pixel per second. And you don't have to be worried about detecting part photons either - when you get down to this signal level, it becomes a statistical problem - some pixels have exactly one photon, some have 2 and many have exactly none. Add 4000 such exposures together and you have the equivalent of a long exposure result, provided read noise is small enough. I have not looked at narrow band imaging yet.

There are already examples of high framerate imaging on the planetary imaging forum - some imagers use up to 200 frames per second to cut through seeing with bright targets - read noise is the killer though. It will be a major upheaval in how thing are done if very low read noise chips become available to let the same thing be done for deep sky imaging.

Bit of a diversion from the 694, but still interesting. We are accustomed to thinking that we need really long exposures to go deep, but this is only because of read noise polluting low level parts of the image. If you can get rid of thermal noise (most chips largely do) and also read noise, there is absolutely nothing left but signal and you can then use many short exposures - or whatever else you would like to do. However, I suspect it will be a fair while before we have access to this type of technology. You can buy EMCCDs from Andor, E2V and others, but they cost a heap.

I understand that there will always be some hidden gotchas in new technology, but even if the eventual performance gets to 1/10 of what the simple calculations suggest, it will be game changing. Thanks for bringing it up.

regards Ray

Last edited by Shiraz; 06-03-2013 at 08:36 PM.
Reply With Quote
  #30  
Old 06-03-2013, 11:04 PM
rally
Registered User

rally is offline
 
Join Date: Sep 2007
Location: Australia
Posts: 896
Ray,

Thank Peter for that - he raised it first !
I just used it as an example for a thought experiment since it seemed relevant.

I'll look forward to seeing your results.

Shot noise on a 1e- signal is 1e- ! then add in the rest of the noise.
It might be hard, but then I guess it just becomes a question of just how deep you can go before the noise defeats those particular signals.

Rally
Reply With Quote
Reply

Bookmarks

Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +10. The time is now 02:45 PM.

Powered by vBulletin Version 3.8.7 | Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Advertisement
Bintel
Advertisement
Testar
Advertisement