New sensors from On Semi

[Atmos (Colin)]

The application of a CCD does need to be taken into consideration along with everything else. Some of the Apogee CCDs are fantastic for scientific research because of their HUGE pixels, enormous well depth and dynamic range but typically tend to fit within the 0.25-4 megapixel category. Some of those sensors have pixels 3x the size of the KAF-8300, 96% peak QE and don't have ABG. At this point you're tailoring your exposure time to reflect what it is that you're wanting to detect. A 0.5mp 24 micron CCD is designed for a purpose and astrophotography isn't one of them To put it another way, the KAF-16803 is on my future wish list but I certainly wouldn't be doing planetary imaging with it.

[rally]

Read noise is an issue silently waiting here.

Extremely low read noise cameras allows us to capture a very wide dynamic range and to capture faint detail (low photon count data) without corruption

I do wonder if award winning Ha data is more a case of being able to resolve the extremely faint detail (aka very few photons) as opposed to resolving just the bright stuff where we have bucket loads of photons to capture - after all - displaying that which isnt always seen, resolved or captured produces the Ooh Aahs.

So the art (both the practice and the finished result !) becomes more about capturing and processing the low photon count regions rather than the bright regions.
Some of that is equipment related and much is skill related.

The chip specs will help of course, but so will everything else help to resolve such fine faint detail. Noise is our enemy and it comes in many disguises - seeing, camera noise, dynamic range of chip and of data collection technique, image processing, optics, light pollution, gradients, atmospheric effects (transparency, extinction etc), focus, tracking, guiding, PE, vibration . . . - its a big list - in fact its almost everything we work towards.

So equating hours of total exposure time to results doesnt necessarily tell us anything about the data or the noise inherent in the process, but then neither does Qe in isolation - there is lots at play.

The ideal camera of course for pretty pictures ignoring physics and finance is a huge well depth with an appropriately small image scale, negligible read noise (less than -1e), Qe as close to 100% across the full imaging spectum UV to IR
Its currently a bit of a dream for amateurs but who knows how much further sCMOS will come.
sCMOS cameras with such low read noise have become affordable for mere mortals.

[Peter Ward]

Originally Posted by Shiraz

none of this has the slightest bit to do with CCD QE Peter. ..........It would be foolhardy to ignore it.

I'm not ignoring it, I'm simply stating ( as Rally has done quite eloquently) with so many other factors, the hope that a 15-20% increase in absolute QE will result in imaging sessions made in Valhalla is forlorn one in my experience.

I made this very step in going from a Kaf11002 to a 16803 sensor some years ago. There was no signal level revelation, hence despite the extra horses under the bonnet, I found myself working on other aspects of my imaging system much as a motor racing team does eg, better weight, suspension, steering, braking all help to improve lap times, to better take advantage of the extra HP.

[RickS (Rick)]

Originally Posted by rally

Read noise is an issue silently waiting here.

You're preaching to the converted there, Rally, and read noise was discussed earlier in the thread. The reason it got distilled down to a discussion on QE was that we were comparing two sensors that both had pretty awful read noise specs compared to newer Sony sensors.

Originally Posted by Peter Ward

I'm simply stating ( as Rally has done quite eloquently) with so many other factors, the hope that a 15-20% increase in absolute QE will result in imaging sessions made in Valhalla is forlorn one in my experience.

You're welcome to state that, Peter, but I didn't see anybody claim that QE was magic pixie dust or the only thing that mattered, only that it was a significant factor.

Cheers,
Rick.

[gregbradley]

What would be awesome but unlikely to occur is a larger Sony sensor with the clean read noise, high QE and deep wells.

Since Sony is putting all its eggs in the CMOS market then it may have to come from an advanced CMOS sensor.

Greg.

[RickS (Rick)]

Originally Posted by gregbradley

What would be awesome but unlikely to occur is a larger Sony sensor with the clean read noise, high QE and deep wells.

Since Sony is putting all its eggs in the CMOS market then it may have to come from an advanced CMOS sensor.

Greg.

Yes, I'll have one of those, thanks Greg

[Peter Ward]

Originally Posted by RickS

.......
You're welcome to state that, Peter, but I didn't see anybody claim that QE was magic pixie dust or the only thing that mattered, only that it was a significant factor.

Cheers,
Rick.

Fair 'nuff. Perhaps I've read too much into Ray's comment:

"QE is of fundamental importance and there is no way round that"

as my experience has been a 10-15% fundamentally wasn't that important

[Shiraz (Ray)]

Originally Posted by Peter Ward

Fair 'nuff. Perhaps I've read too much into Ray's comment:

"QE is of fundamental importance and there is no way round that"

as my experience has been a 10-15% fundamentally wasn't that important

Well I stand by it, QE is fundamentally important - suggesting that it isn't "that important" is a novel approach....

The QE difference between an 11002 and a 694 translates into a difference in imaging time of 2x at Ha. That is to say, you can get the same result in half the time at 60% QE (eg 694) compared to 30% QE (eg 11002).

I don't know what your 10-15% figure refers to, but the sensitivity change on going from 30% QE to 60% QE is:

DOUBLE,
TWICE,
TWO TIMES,
2x

- not really sure how else to put it.

Also, you can get this performance advantage even after doing all the optimisations that Rally noted. This is nothing to do with "Valhalla", it's nothing magic - it just seems to be a tiny bit of commonsense. Why would you not want to double the number of targets that you could image in any given time?

And then there is read noise .....

[rally]

Originally Posted by Shiraz

And then there is read noise .....

Good one Ray,

Yes - I know I am preaching to the converted

But its still possible my point is missed - especially when we are talking about faint detail in low photn count areas of our images

If we have for simplicity -10e read noise and the levels of faint Ha signal is only from very low (say 1e) through 10e and say up to -20e
Then the signal to noise ratio goes from 1:10 (meaning no useful data) through 1: up to 2:1 - add shot noise, dark noise, and all the other possible noise its still a fairly sad situation for useful data . . . - but this is using up the first 4 to 5 bits of our 10-14 bits of data !

Now if we have a -1e read noise camera (and all other things being equal) then the situation becomes 1:1 through to 10:1 and up to 20:1
This is vast improvement improvement in signal quality and will thus produce an image that is (at least for the faint data) not just 20% better but more likely 1000% better

So who cares about an increase in Quantum efficiency of 20 or even 40% when these sorts of gains can arise by low read noise cameras.

Anyone care to analyse this better than my rough numbers

[ericwbenson (Eric)]

Originally Posted by Shiraz

Well I stand by it, QE is fundamentally important - suggesting that it isn't "that important" is a novel approach....

The QE difference between an 11002 and a 694 translates into a difference in imaging time of 2x at Ha. That is to say, you can get the same result in half the time at 60% QE (eg 694) compared to 30% QE (eg 11002).

I think most amateur astronomers (and some professionals too) are not only interested in just QE. I think they are looking for, and perhaps they don't consciously know it, is what the pros sometimes call 'etendue', the product of system throughput (collected photons) times sky area. Lower QE can be tolerated in a big chip because you don't need four panels to cover the object of interest, Hence the total time spent imaging a big object is actually half for the KAI11002 in your above example.

Regards,
EB

[Peter Ward]

Originally Posted by Shiraz

I don't know what your 10-15% figure refers to, but the sensitivity change on going from 30% QE to 60%......
.....

It refers to the change from my KAF11002 based camera to my current KAF16803..... about 30% vs 45% QE in Ha.

But it really made stuff-all difference in my data...the faint stuff was simply a little less noisy.

Sadly Sony don't make overly large sensors, so despite some impressive QE numbers, it would give my RC16 a field of view similar to a drinking straw....hence the 16803 indeed saves me some time in other ways.

[Atmos (Colin)]

There are a considerable amount of things that need to be considered. It was only about three months ago that I was looking into what camera to get for my first imaging setup. In the end it came down to either a KAF-8300 or the ICX964 and ultimately went with the KAF because it ended up at about half the price.

As some theoretical comparisons however, they have pixel sizes of 4.54 and 5.4 micron. When calculating their absolute efficiency against one another their total area has to be taken into consideration. The KAF has ~41% more surface area per pixel. Comparing peak QE gives 77% vs 56% which makes the ICX 37.5% better at peak (green) but 50% better at Ha. At their peaks the KAF is ~3% better but the ICX is ~41% better in Ha.

So in short the ICX is basically at par or better with ~19% better native resolution. The biggest difference however comes from the read noise where the ICX has about half, 4.5 vs 8-12 (depending on readout speed) and manufacturer). Earlier this week I took my telescope to my dark site for the first time, managed to get a 60s test exposure before the clouds came over :/ From that I got ~ 150 ADU as background. With a gain of 0.389 in my QHY9 that works out to ~ 58e- per minute or ~ 25.67e-/arcsec/minute and this equates to ~ 41 e-/minute or ~ 137 ADU/min on the ICX. The ICX has a QE 37.5% higher so that then becomes ~188 ADU/min.

A while back Ray (Shiraz) wrote an equation for optimal exposure length being (RN*10)*(RN/gain). From this I know my QHY9 needs a background of ~1860 (RN of 8.5e-) ADU but the ICX only needs ~675 ADU. KAF-8300 hits sky limited in 742s where as the ICX964 does it in ~ 215s. Of course on the other hand it appears like the KAF-16803 only needs ~140s. That’s the benefit of massive wells depths (gain of ~ 1.6e-/ADU) even with a read noise of ~10e- and having large pixels. On the other hand the 11002 would require ~ 338s

Pretty sure all that maths stuff is correct, haven’’t double checked but it feels like it is in the right direction. Everything was calculated at peak QE as my initial test was done in luminance.

[RickS (Rick)]

Here's the KAF-16200 data sheet:
http://www.onsemi.com/pub_link/Colla...AF-16200-D.PDF

It's a shame they don't have any data on how read noise varies with clock rate.

Cheers,
Rick.

[Somnium (Aidan)]

so, when are some of the big players such as SBIG, FLI etc going to jump onto this. seems like the 16200 would fit well in the STXL body, making use of the self guiding FW and AO-X

[RickS (Rick)]

Originally Posted by Somnium

so, when are some of the big players such as SBIG, FLI etc going to jump onto this. seems like the 16200 would fit well in the STXL body, making use of the self guiding FW and AO-X

FLI is doing one: http://www.flicamera.com/16.html

[Slawomir (Suavi)]

Thank you Rick for the specs.

Read noise is stated at 14e-, so in an astro camera I would not expect anything near the read noise of the latest sensors from Sony. Nonetheless it should be an interesting option for astrophotographers.

[RickS (Rick)]

Originally Posted by Slawomir

Thank you Rick for the specs.

Read noise is stated at 14e-, so in an astro camera I would not expect anything near the read noise of the latest sensors from Sony. Nonetheless it should be an interesting option for astrophotographers.

My pleasure, Suavi!

I'd expect at a lower clock rate the read noise will be less. Probably not as low as 4e- or 5e- but hopefully down into the single digits at least.

Cheers,
Rick.

[Slawomir (Suavi)]

If the read noise will be on pair with KAF8300, then it could potentially become a quite popular sensor among astrophotographers. It is over twice the size and twice the resolution of the venerable 8300 (but hopefully the cost won't double!) also with good pixel size.

[Somnium (Aidan)]

Originally Posted by RickS

FLI is doing one: http://www.flicamera.com/16.html

oooo any word on pricing or availability before i shoot off an email to them ?

anyone know if SBIG is looking into it? i am considering a new CCD and this fits my spec needs perfectly but i want the SBIG AO option

it looks like the well depth is not stellar ... 39k at 6 microns

[RickS (Rick)]

Updated FLI specs here: http://www.flicamera.com/spec_sheets/ML16200.pdf

Read noise 7e- @ 3MHz and 11e- @ 12MHz.

Originally Posted by Somnium

oooo any word on pricing or availability before i shoot off an email to them ?

$6495 USD for grade 2 sensor, $7495 USD for grade 1. Dunno about availability.

Originally Posted by Somnium

anyone know if SBIG is looking into it?

Question for Peter Ward?

Originally Posted by Somnium

it looks like the well depth is not stellar ... 39k at 6 microns

About what I'd expect. The area of a pixel is about half that of the KAF-16803 and the well depth is scaled similarly.

Cheers,
Rick.