New sensors from On Semi

[gregbradley]

Originally Posted by Peter Ward

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.

I think you are downplaying the advantages of the extra QE. I noticed the 16803 being more sensitive in Ha straight away.

I understand that it may not seem like a lot but you'd have to add up how much exposure time you now need to take to get a clean and solid Ha image.

The difference with the Sony sensor with Ha and O111 is much more dramatic. Its obvious that you can get a solid Ha image in half the time than using the 16803.

For example I could see the main jet off NGC1097 from the Trius on the Honders in a very pushed 10 minute 1x1 image. That's the QE and the F3.8 at work.

To get a similar image using a KAI11002 would take probably twice as long perhaps even more.

It becomes obvious when imaging dimmer objects like dimmer galaxies not something usually done in the burbs but if you did then the difference is plain to see.

Greg.

[gregbradley]

FLI ML16 Availability is soon. Perhaps within a week or so from an email I got.

It could be a very good sensor. 39K wells is fine. The KAI11002 is not much more than that and I have never noticed an issue with wells with that.

Well depth is a bit of an overblown issue. Although in my personal experience 20K well depth is starting to become a problem with faster scopes.

What I see as the main advantage of well depth is stars are much more robust from a deep well camera compared to a small well camera. Stars can become easily damaged in processing from a small well camera and can lose colour more easily and get fluffy /furry edges/halos.

Greg.

[Atmos (Colin)]

Originally Posted by gregbradley

I think you are downplaying the advantages of the extra QE. I noticed the 16803 being more sensitive in Ha straight away.

I understand that it may not seem like a lot but you'd have to add up how much exposure time you now need to take to get a clean and solid Ha image.

The difference with the Sony sensor with Ha and O111 is much more dramatic. Its obvious that you can get a solid Ha image in half the time than using the 16803.

For example I could see the main jet off NGC1097 from the Trius on the Honders in a very pushed 10 minute 1x1 image. That's the QE and the F3.8 at work.

To get a similar image using a KAI11002 would take probably twice as long perhaps even more.

It becomes obvious when imaging dimmer objects like dimmer galaxies not something usually done in the burbs but if you did then the difference is plain to see.

Greg.

I am still considering the upgrade to a SX964 and possibly throwing in the AO in for kicks. Of course, all comes down to when funds are available

[Peter Ward]

Originally Posted by gregbradley

I think you are downplaying the advantages of the extra QE. I noticed the 16803 being more sensitive in Ha straight away.

I understand that it may not seem like a lot but you'd have to add up how much exposure time you now need to take to get a clean and solid Ha image.

.

I make no bones about it, I am, as having used a 95% QE sensor which gave terrible results for all sorts of reasons, I did not find the extra QE to be worth all the pain associated with that particular sensor.

But signal can also be gained in many other ways.

My AOX regularly delivers and extra 15-20% of detected flux....across the entire spectrum....and sure in an ideal world I'd have a 36mm x36mm Sony chip as well a large aperture AO.

But sadly it does not exist and is unlikely to do so anytime soon.

[Shiraz (Ray)]

Originally Posted by Peter Ward

My AOX regularly delivers and extra 15-20% of detected flux....across the entire spectrum.......and sure in an ideal world I'd have a 36mm x36mm Sony chip as well a large aperture AO.

no it doesn't Peter. Under some conditions, an AO may possibly increase the peak pixel flux on stars, but on things of actual interest, (extended objects like galaxies, nebulae etc) it will always slightly reduce the average signal. Sorry, but AO is not a substitute for QE

However, part of your ideal world is already here - the latest SX AO has a large 60mm aperture, which is plenty big enough for all Sony chips, the KAF16200 and anything else up to full frame 35mm. As for wanting big Sony chips, I don't give two hoots who makes my CCD, but I simply cannot understand why ONsemi is still producing their large pixel CCDs with such low QEs, when Sony and Aptina can routinely get above 70%, with full ABG, low thermal and read noise and all using small pixels - I don't get it.

[Somnium (Aidan)]

talking about the chips themselves ... what would be the recommended filter for the 16200, you would probably need 50 mm round or square filters. it would be a good candidate for for the FLI 2-7 or 5-7. or is there anything smaller you can get away with, i suspect not.

[RickS (Rick)]

Originally Posted by Somnium

talking about the chips themselves ... what would be the recommended filter for the 16200, you would probably need 50 mm round or square filters. it would be a good candidate for for the FLI 2-7 or 5-7. or is there anything smaller you can get away with, i suspect not.

With a 34.6mm diagonal on the sensor 36mm filters aren't going to cut it. 50mm round will be the cheapest option.

[Peter Ward]

Originally Posted by Shiraz

no it doesn't Peter. Under some conditions, an AO may possibly increase the peak pixel flux on stars, but on things of actual interest, (extended objects like galaxies, nebulae etc) it will always slightly reduce the average signal. Sorry, but AO is not a substitute for QE

I have never found an AO to make things worse...and agree it's not the same as QE....nor is tracking or focusing the telescope accurately ( to which AO is a much closer analogy)....

I was thinking more of the overall system.

QE is detected vs actual fux.

I really don't give a toss if I have a 100% QE sensor attached to a system that simply disperses light across the entire field without actually bringing it to focus....sure the light is being detected by the sensor...but the stars are simply not seen above the background soup in that absurd conclusion...hence making the system's QE effectively zero.

I am however far more interested in an imaging system that concentrates the flux to tight + intense footprints.

I've been using AO's of one sort or another for over a decade and know (despite your statements to to contrary) they help do just that, time and time again, and have the data to prove it.

Higher pixel values by better QE or AO? Sure, not the same, but the higher intensity result is a rose by any other name here.

Sure I'd like a higher QE sensor +AO (as I said earlier, not available), but getting higher values in lieu of a more efficient CCD, via an AO, works for me.

[Shiraz (Ray)]

Originally Posted by ericwbenson

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

interesting point Eric - basic question then is, how big does the sensor need to be? - guess that depends on the targets of interest. I have never felt a need for anything bigger than 6 mp (ie less than 1 degree^2 at optimal sampling in local conditions), but then, I have no interest in extensive nebulas and am not doing survey work. The new chips generally up the ante by providing more pixels in their market segments - as you point out, for other users, a larger pixel count could be worthwhile, even if the pixel design is a bit behind the times.

Another possibility is that more pixels make more sense in really good seeing, so maybe the bigger chips will have more payoff on mountaintops, rather than in Australia?

edit; been thinking about it and, for interest, at equivalent sampling, a 694 could pretty much match the sky area of an 11002 using just 2 panels, so in time/etendue terms, the systems would be almost equal.

[Slawomir (Suavi)]

Originally Posted by Shiraz

edit; been thinking about it and, for interest, at equivalent sampling, a 694 could pretty much match the sky area of an 11002 using just 2 panels, so in etendue terms, the systems would be almost equal.

Good point. I would also like to point out that ICX834 would do just that with one panel and with 1.25" filters. I know pixels are small, but so is the read noise (about 3e- with FLI and QSI) and QE is on pair with ICX694. I have a very positive experience with ICX814 (also tiny pixels) and narrowband imaging and I feel that ICX834 is usually easily overlooked and also quite underestimated sensor by amateur astrophotographers.

[gregbradley]

Originally Posted by Peter Ward

I make no bones about it, I am, as having used a 95% QE sensor which gave terrible results for all sorts of reasons, I did not find the extra QE to be worth all the pain associated with that particular sensor.

But signal can also be gained in many other ways.

My AOX regularly delivers and extra 15-20% of detected flux....across the entire spectrum....and sure in an ideal world I'd have a 36mm x36mm Sony chip as well a large aperture AO.

But sadly it does not exist and is unlikely to do so anytime soon.

I agree. I have always been able to tell if an AO unit was used in acquiring an image.
The images have more punch.
Much like a non AO system on night of good seeing.
Greg

[ericwbenson (Eric)]

Originally Posted by Shiraz

interesting point Eric - basic question then is, how big does the sensor need to be? - guess that depends on the targets of interest

It would seem most professional installations are heading towards multi-multi-megapixel cameras..it make sense if you've already invested buckets of money on the telescope/site/support/buildings and all you have to do is change the electronics to get more throughput.

Originally Posted by Shiraz

edit; been thinking about it and, for interest, at equivalent sampling, a 694 could pretty much match the sky area of an 11002 using just 2 panels, so in time/etendue terms, the systems would be almost equal.

2 panels? For the same telescope the KAI11002 has ~7x times the sky coverage, the KAF16803 10x the coverage (with less read noise, lower dark current, better QE, deeper wells and higher cost than the interline chip).

Regards,
EB

[gregbradley]

Not sure what Ray meant by sky coverage as the 694 sensor is much less than 1/2 the size of the 11002.

There is one point perhaps taken into account with your sensitivity formula and that is a larger sensor collects more light than a smaller one.

Well known in digital cameras. A full frame (35mm sized sensor) DSLR always performs better than a smaller APSc or smaller sensor for nightscapes etc. Its just simply the larger collecting areas. Even more true for CCDs where there aren't really large spaces between pixels like CMOS does.

Greg.

[Slawomir (Suavi)]

Originally Posted by gregbradley

Not sure what Ray meant by sky coverage as the 694 sensor is much less than 1/2 the size of the 11002.

Ray clearly stated "at equivalent sampling", so to my understanding his statement is correct, and that's also why I mentioned that ICX834 with equivalent sampling is an interesting (and cheaper) but overlooked alternative to KAI11002.

[gregbradley]

I can't see the 814 being in any way equivalent to the 11002.The KAI11002 is a full frame sensor meaning a massive wide field of view versus a tiny one with the 814. I am not sure what you meant, perhaps you meant that with a short enough focal length they can get a similar field of view as the 11002 on a 3X longer focal length?

But for the 814 to match the 11002 in field of view you would probably have to go down to 300mm lenses on an 814 to match the FOV of the 11002 at 900mm. The 11002 is like 3X the width and height of the 814.

36 x 26mm versus 12mm x 10mm or so for the 814. 50,000 well depth versus 16,000 or so for the 814. Lower QE for the 11002 and higher read noise but 9 micron pixel suits lots of scopes and full frame means more total light collected.

Mind you I like the tiny FOV of the 694 as I use it for my close up, zoomed in camera. The 16803 and 11002 give the wide expansive widefield images that an 814/694 can't unless you do a mosaic which opens the door to all sorts of mismatches and tricky processing.

The allure of these large sensors is the large wide field of view in a single image with a shorter focal length scope (that has a large enough corrected field which is now in the realm of higher end scopes). Or on a longer focal length scope good for galaxies like your RCs, CDKs etc. This is one of the problems of the larger sensors - everything needs to be upgraded. Focusers need to be at least 3.5 inches, corrected field on the scope needs to be around 46mm or larger and no flex, everything nice and square. Reducers no longer work to the corners, you need a flattener on most scopes. Cost goes way up. Everything becomes heavier, the need for a higher capacity mount becomes critical. Stiffer adapters become more important, flex is more likely to be an issue.

Greg

[Shiraz (Ray)]

Originally Posted by ericwbenson

It would seem most professional installations are heading towards multi-multi-megapixel cameras..it make sense if you've already invested buckets of money on the telescope/site/support/buildings and all you have to do is change the electronics to get more throughput.

This only applies to those who are doing survey work Eric. It all depends on the target and if you are doing a whole sky survey I agree that large etendue is paramount. For high res investigations it isn't and for example, it seems that the Keck2 has a 1kx1k array as its high res NIR sensor. That is enough to cover the isoplanatic patch, so that's all that is needed. In contrast, the ESO has gone to a lot of trouble on the VST survey scope, which has a 256mpix array. At entirely the other end of the scale, I am primarily interested in imaging galaxies and smaller objects at high resolution, so anything much over 0.5 degrees and about 6mpix would be a waste of time and money in my seeing. I don't want images of a galaxy surrounded by an acre of stars, but for someone who wants to image large nebulas, more etendue would be a good thing - but it isn't a goal for all astronomy.

Originally Posted by ericwbenson

2 panels? For the same telescope the KAI11002 has ~7x times the sky coverage, the KAF16803 10x the coverage (with less read noise, lower dark current, better QE, deeper wells and higher cost than the interline chip).

Regards,
EB

but you wouldn't put a 694 on a 2+m telescope - it would be a heavily oversampled system. If you put it on a 1m scope (ie at the same sampling, as was specified - thanks S), it will cover the same sky as an 11002 on a 2m scope using just 2 panels. eg an 11002 on a 10inch f8 will have the same resolution, sampling, etendue etc as a 694 on a 10 inch f4. However, the 10 inch f4 with a 694 will have twice the Ha sensitivity on targets within it's field of view, due to the higher QE.

Don't get me wrong, I am not advocating that anyone with a large scope change over to the Sony chips - they just came up as real world examples of what was out there in the discussion on whether a change from 30% QE to 60 % QE actually resulted in a doubling of sensitivity, rather than a 30% change - and why Adaptive Optics cannot possibly give 15-20% more detected flux on extended targets. Somehow that whole discussion went down some odd rabbit holes, but that is where the mention of Sony chips came from - they have higher QE than the new ONsemi ones and the question that is pertinent to the thread is - why can't ONsemi achieve similar QE/noise specs on their big pixels?

Greg, the equation for sensitivity is pixel-based - the size of the field of view is not considered.

regards Ray

[Slawomir (Suavi)]

Originally Posted by gregbradley

I can't see the 814 being in any way equivalent to the 11002.The KAI11002 is a full frame sensor meaning a massive wide field of view versus a tiny one with the 814. I am not sure what you meant, perhaps you meant that with a short enough focal length they can get a similar field of view as the 11002 on a 3X longer focal length?

But for the 814 to match the 11002 in field of view you would probably have to go down to 300mm lenses on an 814 to match the FOV of the 11002 at 900mm. The 11002 is like 3X the width and height of the 814.

36 x 26mm versus 12mm x 10mm or so for the 814. 50,000 well depth versus 16,000 or so for the 814. Lower QE for the 11002 and higher read noise but 9 micron pixel suits lots of scopes and full frame means more total light collected.

Mind you I like the tiny FOV of the 694 as I use it for my close up, zoomed in camera. The 16803 and 11002 give the wide expansive widefield images that an 814/694 can't unless you do a mosaic which opens the door to all sorts of mismatches and tricky processing.

The allure of these large sensors is the large wide field of view in a single image with a shorter focal length scope (that has a large enough corrected field which is now in the realm of higher end scopes). Or on a longer focal length scope good for galaxies like your RCs, CDKs etc. This is one of the problems of the larger sensors - everything needs to be upgraded. Focusers need to be at least 3.5 inches, corrected field on the scope needs to be around 46mm or larger and no flex, everything nice and square. Reducers no longer work to the corners, you need a flattener on most scopes. Cost goes way up. Everything becomes heavier, the need for a higher capacity mount becomes critical. Stiffer adapters become more important, flex is more likely to be an issue.

Greg

Hi Greg,

I was suggesting that ICX834 (not 814) could be a cheaper alternative to 11002.

And you are right, ICX834 would need to be put at the end of a much smaller telescope to give the same FOV as 11002, but that what I meant - maybe I did not explain it correctly.

Agreed, ICX834 has tiny pixels but read noise in only 2e- with FLI MLx834 - I really think when combined with FSQ106 (1.21"/px) it would be a very capable relatively fast narrowband imaging setup.

KAI11002 would need to be combined with 1500mm fl to give the same resolution, and both setups would have about the same FOV.
Yes, KAI11002 has much deeper wells but also significantly higher read noise and significantly lower QE making it, in theory, less suitable for narrowband imaging, in spite of being more expensive (as you pointed out- larger mount, larger filters etc).

So if someone's on a budget (relatively speaking) and has a smaller telescope, I feel that ICX834 presents an interesting and in theory quite capable option.

[gregbradley]

I totally agree the 834 is an amazing sensor and I have been very impressed with the quality of your images with a modest refractor and your QSI690.

It does emulate a larger sensor when used with a short focal length refractor and reducer and I think the Sony sensors and the KAF8300 have lowered the cost of imaging a lot as you can get superb results from a smallish and modest refractor with these sensors instead of the larger and more expensive APOs and required expensive mounts.

The full frame sensors as I mentioned really add to the cost and complexity of the system and not many scopes can cope with them really. None of them are cheap either. Not many reducers work with them as well without coma in the corners.


Greg.

[gregbradley]

Onsemi bought from True Sense who bought Kodak's sensor division.

Kodak went down an engineering path. My understanding is that these sensor plants are worth billions to setup. So once they go down a path that is it for some time. Canon was/is in that predicament where their sensor plant was (not sure what the current state of play is) behind the times compared to Sony. I read once Canon's plant was running on something like 500 micron size and Sony's at 15 microns.

Sony is the world's largest sensor maker. It also acquires companies that have technology they want for their sensors. That's how the Sony Exmor sensor got such good low light low noise performance - by buying the tech from someone else in CMOS.

Sony also has a technology sharing agreement with Aptina who holds a lot of technology patents for sensors.

I can't see OnSemi changing these sensors. The money involved probably does not make sense to do so.

So its either a future Sony CMOS sensor that is ubeaut or perhaps some new Chinese player (have a look at QHY 42 for example) to bring something new to the table.

Scientific CMOS is one possible future. Sony seems to have left the CCD business to concentrate on CMOS so I take it they believe that is where the demand is (Smartphone camera sensors, digital cameras, industrial vision).

Now we have backside illuminated CMOS as a standard Sony CMOS camera sensor so perhaps just a few more development iterations and CMOS may in fact be more advanced than CCD?

One advantage CCD had over CMOS I thought was less space around each pixel needed for circuitry. CMOS is around 40%. I don't know what it is for Kodak CCDs. I assume very little if any.

Greg.

[ericwbenson (Eric)]

Originally Posted by Shiraz

I am primarily interested in imaging galaxies and smaller objects at high resolution, so anything much over 0.5 degrees and about 6mpix would be a waste of time and money in my seeing. I don't want images of a galaxy surrounded by an acre of stars, but for someone who wants to image large nebulas, more etendue would be a good thing - but it isn't a goal for all astronomy.

Hi Ray,

Well funny enough my main interest has also always been small faint galaxies and hence why I was content with an ST7 then an ST8 for a long time. However when it came time to upgrade for various reasons, such as better cooling that I needed in Australia vs Canada, remote operation hence sealed CCD chamber (no desiccant swaps), better image transfer system etc, I also started looking at various CCDs. The good old KAF3200ME had a lot going for it except for it's tendency to bloom easily (high QE, 6.8um pixels, no ABG = PITA for processing), and it was still fairly costly. KAF6303M looks good except again no ABG and lower QE (due to no microlenses) than my ST8XME. So the obvious choice was the KAF16803 which has all the good stuff and decent QE, plus allowed me to use my scope for more than just galaxies, I could actually image nebulas larger than the Horsehead! I don't think the latest Sony chips were available then, they would have been contenders except for the pixel size mismatch.

Originally Posted by Shiraz

but you wouldn't put a 694 on a 2+m telescope - it would be a heavily oversampled system. If you put it on a 1m scope (ie at the same sampling, as was specified - thanks S), it will cover the same sky as an 11002 on a 2m scope using just 2 panels. eg an 11002 on a 10inch f8 will have the same resolution, sampling, etendue etc as a 694 on a 10 inch f4. However, the 10 inch f4 with a 694 will have twice the Ha sensitivity on targets within it's field of view, due to the higher QE.

Changing the focal length to get the same sampling muddies the waters since for constant aperture (hence sky flux) the fratio must decrease for the smaller pixel CCD, and the problems associated with fast optics rear their head - no free lunch.
At constant aperture a bigger CCD is 'usually' easier to use than a small one. Yes it implies longer focal length, but that does not mean longer tube (enter catadioptrics), nor different requirements on tracking (where pixel scale is the only factor), it means slower optics which are generally more tolerant in the making and holding in place, plus longer focus zone.

Originally Posted by Shiraz

Don't get me wrong, I am not advocating that anyone with a large scope change over to the Sony chips

And I am not certainly advocating everybody go out and buy a KAF16803. I think we actually see things in much the same way, though with slightly different experiences colouring our perception. Frank discussion is always interesting.

Originally Posted by Shiraz

and why Adaptive Optics cannot possibly give 15-20% more detected flux on extended targets.

This is an aside that could be it's own thread IMHO after using an AO8 for 5+ years, amateur AO doesn't provide much boost for big CCDs if the mount is good enough to start with (most PMEs that are not overloaded, almost all Astrophysics mounts if the legend serves to be correct, and any Renishaw encoder equipped mount qualifies). Maybe one day I will use AO again and change my mind, but for the cost+complexity to add to my system it will be awhile
However, that being said, from empirical observation the signal from extended sources such as faint jets or spiral arms can be greatly affected by the seeing conditions. Perhaps it is the faint structures that get washed out diminishing the contrast, or simply the feature being smeared out against the background, not sure, but a single low FWHM subs is always worth many crappy high FWHM subs in a stack.

Originally Posted by Shiraz

why can't ONsemi achieve similar QE/noise specs on their big pixels?

Perhaps the answer is simple, they don't have the tech, can't afford to develop the tech, and Sony just doesn't give a hoot about the large CCD market segment. Amateur, and to some extent pro astronomy, have always been at the mercy of other commercial endeavours (e.g. medical imaging, surveillance) for CCD development, I can't see how this will change in the forseable future.

Best regards,
EB