Luminous Landscape Forum

Equipment & Techniques => Medium Format / Film / Digital Backs – and Large Sensor Photography => Topic started by: marcmccalmont on January 25, 2011, 07:59:04 pm

Title: Why not a CMOS sensor in the Phase One IQ series?
Post by: marcmccalmont on January 25, 2011, 07:59:04 pm
I'm excited over the new Phase One backs because of the large screen and liveview since I have a Cambo wide RS and will not need the ground glass
But Pentax squeezed 14 stops of DR out of an APSc sized Nikon CMOS sensor so this begs the question why not develop CMOS medium format sensors?
Lower power, more DR, I would have liked to see a large leap in IQ this time not just small incremental steps
Marc
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: michael on January 25, 2011, 08:16:39 pm
I asked that question of Phase's head of technology and the answer simply is that CCDs still offer the highest image quality in large chip sizes, and that's what MF is ultimately all about.

Its coming, but possible another few years.

Michael
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: marcmccalmont on January 25, 2011, 09:11:24 pm
I'm glad you asked the question! If I were the emperor I would have:
Had Nikon develop a medium format sensor
Had Burr Brown develop a 24 bit a/d converter insuring 16 clean bits
Had Adobe implement their products with a full 16 bits
Had Pentax produce the DB for 10K!
Marc
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: Alan W George on January 25, 2011, 10:28:49 pm

Had Nikon develop a medium format sensor


As far as I can tell, Nikon is not in the sensor development business.  They use other's (e.g Sony) sensors.  Canon on the other hand.....
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: deejjjaaaa on January 26, 2011, 09:48:34 am
As far as I can tell, Nikon is not in the sensor development business. 

Nikon is - for example D700, D3100 have Nikon sensor... sensor D7000 was designed by Sony w/ certain Nikon participation (not sure how big Nikon's input was silicon-wise) and manufactured by Sony
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: aaron on January 26, 2011, 12:34:38 pm
As far as I can tell, Nikon is not in the sensor development business.  They use other's (e.g Sony) sensors.  Canon on the other hand.....

Not so sure about that, don't Nikon manufacture the Semiconductor Steppers that Sony and possibly Canon use to make their cmos sensors?   ;)

Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: BJL on January 26, 2011, 05:30:44 pm
I asked that question of Phase's head of technology and the answer simply is that CCDs still offer the highest image quality in large chip sizes, and that's what MF is ultimately all about.
I am sure that the MF people say that, but allow me a little skepticism, given that all the technical comparisons between CMOS and CCD photosites of the sme size do not come close to supporting that claim.

How about:
1. the market for sensors larger than 36x24mm is so much smaller than for smaller formats that, up till now, none of the companies that know how to make good active pixel CMOS sensors (Canon, Sony, Panasonic, and Nikon?) have seen adequate profit potential in developing those larger CMOS sensors in a form suitable for high end photography.
2. Kodak and Dalsa do not yet have active pixel CMOS sensor technology good enough to offer a significant performance advantage. Kodak in fact abandoned its CMOS efforts entirely.

Aside: Dalsa does makes some large CMOS sensors for X-rays and maybe telescopes, by "mosaicing" smaller sensor chips, but that adds visible join lines, unacceptable in high end MF photography. This mosaicing is NOT the same as "stitching", which produces a single large sensor chip. Canon has also designed some large sensors, but again suitable for uses like telescopes but not MF cameras.
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: BJL on January 26, 2011, 06:04:08 pm
This sensor design vs sensor fab. vs stepper maker question keeps coming up, so here is a summary as I understand it:

1. Nikon designs some of its own high end sensors, but out-sources their fabrication, and outsources sensors for its lower priced DSLR models and compacts from Sony, sometimes with Nikon input on design customizations.

2. Canon does similarly, but with more in-house: it designs and manufacture all its own DSLR sensors, but at the lower end of the product range, Canon outsources the sensors for most of its compacts, again from Sony.

3. Outsourcing sensor fab. is no big deal: there are multiple foundries that can make sensors once one gives them the design, so there is no reason to think that Nikon is tied to Sony for fab. of its sensors. In fact, I would bet that Nikon can get the fab. done cheaper in Korea or Taiwan or China.

4. Nikon and Canon both make steppers, so quite likely Canon uses its own steppers, though it seems to be struggling a bit in the stepper business, in which ASML and Nikon are the top two (ASML well ahead as #1.)

5. To those who insist that it is always best to do everything in-house, never outsourcing, I have one word: "Apple". The turn-around of Apple from its dark days was driven in part by Tim Cook's initiative to get Apple out of the manufacturing business entirely, outsourcing to the most cost-effective suppliers and assembly plants. Even fab. of its own new A4 chip design for iPhone/iPad/iPod is outsourced ... ironically from Apple's smart phone and tablet rival, Samsung. And Apple seems be surviving!
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: ErikKaffehr on January 27, 2011, 12:20:17 am
Hi,

You happen to have links to CMOS and CCD comparisons? I have the impression that CMOS is in advantage but I have not seen any real data.

Best regards
Erik


I am sure that the MF people say that, but allow me a little skepticism, given that all the technical comparisons between CMOS and CCD photosites of the sme size do not come close to supporting that claim.


Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: mediumcool on January 27, 2011, 02:41:01 am
... Pentax squeezed 14 stops of DR out of an APSc sized Nikon CMOS sensor ...

My understanding is that Pentax has been using Sony sensors since ending the romance with Samsung.
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: BenNorton on January 27, 2011, 04:04:17 am
Not so sure about that, don't Nikon manufacture the Semiconductor Steppers that Sony and possibly Canon use to make their cmos sensors?   ;)



Yes they do. I worked for Nikon in Kingston Upon Thames for a good few years and the Stepper market is probably the largest sole area of profit for them.
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: BernardLanguillier on January 27, 2011, 05:40:19 am
Yes they do. I worked for Nikon in Kingston Upon Thames for a good few years and the Stepper market is probably the largest sole area of profit for them.

I believe that Imaging represents 60% of their revenue nowadays.

Regards,
Bernard
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: ondebanks on January 27, 2011, 07:34:01 am
I asked that question of Phase's head of technology and the answer simply is that CCDs still offer the highest image quality in large chip sizes, and that's what MF is ultimately all about.


"highest image quality " - maybe, but in their current best implementation, this is true only at base ISO, and only taking short exposures. It is these limitations which apparently forced Nikon to abandon CCD for CMOS in their better DSLRs. You'll often see it said that CCDs are inherently capable of superior performance to CMOS, but as of 2011 and in recent years past, CCD developers have failed to capitalise on this capability, while CMOS developers have greatly surpassed predictions such as these ones from 2001 (http://www.siliconimaging.com/ARTICLES/sensors.htm); some of its statements such as "Many CMOS imagers don’t perform at the same level as CCD imagers. Most notably, CMOS imagers can have high fixed-pattern noise, low sensitivity to light, high dark current, focal plane shutter effects, and some difficulty scaling to smaller pixel sizes" are now demonstrably wrong on just about every point.

"in large chip sizes" - AFAIK there are no comparable (large format, small pixel) commercial CMOS chips, so isn't this like Phase's head saying that large CCDs are the best for MF simply because there are currently no alternatives? In a race where I am the only entrant, I am the fastest runner! (and you should see me trying to run  :-[ )

Ray
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: yaya on January 27, 2011, 07:57:31 am
"in large chip sizes" - AFAIK there are no comparable (large format, small pixel) commercial CMOS chips, so isn't this like Phase's head saying that large CCDs are the best for MF simply because there are currently no alternatives? In a race where I am the only entrant, I am the fastest runner! (and you should see me trying to run  :-[ )
Ray

The biggest challenge for CMOS developers is to be able to make them at sizes larger than 24X36 and to maintain a reasonable (read financially viable) yield. This, for the time being, makes this debate irrelevant...

Yair
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: PierreVandevenne on January 27, 2011, 08:33:06 am
CMOS have evolved up to the point where you can say that what used to be the full calibration of a CCD frame and then some is now performed at each pixel level (that's clear from the Sony sensor technology pages and docs). The signal is cooked in a way that is what the photographer needs. The last remaining advantage of the CCD is, imho, its straightforwardness which allows the end user to know exactly what has been done to the signal because he does it himself (or at least is free to design a clean system and control all aspects of the post-processing) and larger sensel sizes when a _faithful_ large dynamic range is required: it is clear to everyone that we aren't going to extract a real 16 bits of DR with a well capacity of 50.000 e.  Building large CCD sensors make sense because there are a lot of scientific and military applications. Building large CMOS doesn't make economic sense yet I guess.

By and large I agree that Phase is marketing that aspect heavily as a deliberate choice while in fact they don't have any choice as of today. This being said, the "rawness" of the signal also gives them (or their provider) the opportunity to work directly on it, with their own recipes, which could potentially be better than their competitors.

The dalsa sensors for example are kind of smooth compared to the Kodak ones. When you look at pure RAW data from Kodak sensors, you can clearly the typical "sandy" aspect of the sensor's noise. You don't get that with the Dalsa sensors, which essentially means that their raw data has already been processed, if only in terms of pixel response uniformity.

But it is becoming harder and harder to have a clear picture of what is going on. There's no dispute than the D7000 K5 sensor is very impressive and that it seems to provide a significant improvement in terms of DR for photographic purposes. But some things don't fit too nicely: the sensor appears to be almost too good to be true in terms of DR vs well capacity and it is a bit hard (at least for me, which doesn't mean much ;-)) to model a simple virtual sensor replicating its documented behavior. When you see the impact a mere sensor bias change has for the "photographic quality" of deep blacks...




Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: John R Smith on January 27, 2011, 09:15:02 am
Regardless of sensor type or manufacture, what about the effect that the signal A/D converters have on the result? We never seem to see this discussed or measured anywhere, but in audio (which is another of my passions) the quality of A/D converters when recording music is quite crucial to the end result.

John
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: PierreVandevenne on January 27, 2011, 09:38:44 am
Of course, manufacturers will always say that they are using the "highest quality component". But these days, in CMOS, it isn't like selecting a component and plugging or smting to a board. See for example

http://www.sony.net/SonyInfo/technology/technology/theme/cmos_02.html#page03

where something like a CCD A/D converter used to look like this

http://www.astrosurf.com/audine/images2/can.gif
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: eleanorbrown on January 27, 2011, 10:04:44 am
I'll add a question regarding sensors for 35mm vs medium format...why would leica choose a ccd over CMOS ?  I have a phase back...dalsa ccd and the leica m9 ...kodak...both with ccd sensors. Eleanor

I am sure that the MF people say that, but allow me a little skepticism, given that all the technical comparisons between CMOS and CCD photosites of the sme size do not come close to supporting that claim.

How about:
1. the market for sensors larger than 36x24mm is so much smaller than for smaller formats that, up till now, none of the companies that know how to make good active pixel CMOS sensors (Canon, Sony, Panasonic, and Nikon?) have seen adequate profit potential in developing those larger CMOS sensors in a form suitable for high end photography.
2. Kodak and Dalsa do not yet have active pixel CMOS sensor technology good enough to offer a significant performance advantage. Kodak in fact abandoned its CMOS efforts entirely.

Aside: Dalsa does makes some large CMOS sensors for X-rays and maybe telescopes, by "mosaicing" smaller sensor chips, but that adds visible join lines, unacceptable in high end MF photography. This mosaicing is NOT the same as "stitching", which produces a single large sensor chip. Canon has also designed some large sensors, but again suitable for uses like telescopes but not MF cameras.
Title: FF CCD better for highly off-perpendicular light with microlenses?
Post by: BJL on January 27, 2011, 10:48:23 am
I'll add a question regarding sensors for 35mm vs medium format...why would leica choose a ccd over CMOS ?
In the M8 and M9, Leica had the special challenge that some of its lenses deliver light to the sensor at a highly off-perpendicular angle (far more so that with any SLR lens), which is difficult to do when microlenses are also wanted for the sake of better low light performance. This was solved with Kodak's innovation of off-set microlenses. I believe that [active pixel] CMOS designs also inherently require the microlenses to be further from the wells than with the simpler full frame CCD design because of the extra circuitry on top of the well, and the microlenses also have to be stronger to get the light into a target region that is a smaller fraction of the photosite, both of which makes the "off-perpendicular" problem worse for CMOS than for FF CCD.

So there are two possible reason why Leica went with Kodak FF CCD's for it s rangefinders:
1. the inherent advantages of CCD in respect of handling highly off-perpendicular light when microlenses are wanted, and
2. Kodak's willingness to work on a custom design for a low volume, high price product.

Kodak and Dalsa seem far more oriented to the more specialized, low-volume end of the sensor business, while Canon, Sony, Panasonic etc. are more oriented to higher volume products. And when it comes to volume comparisons, even the original 1Ds, at 24,000 a year, was a high volume product compared to the Leica M8, or M9, or any MF sensor.

The "custom design" argument probably applies to the R2 sensor, where Kodak was willing to  design and produce that 45x30mm sensor exclusively for one very low volume product. Maybe this is also why Olympus originally chose Kodak for its 4/3 sensors, before Panasonic saw the potential for a far higher volume through the transition to Micro Four Thirds, which I suspect was in Panasonic's plans from the beginning.
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: BJL on January 27, 2011, 11:04:03 am
I believe that Imaging represents 60% of their revenue nowadays.
That is what I recall reading too, though a few years ago Nikon Precision (the stepper/scanner division) was half or more of Nikon. The change is in part is for the bad reason that NP has been losing a lot of ground to the ever more dominant ASML, which now has about 80% market share.

Still, Nikon stands out at the only "camera maker" for whom camera equipment is the parent company's main line of business. (Before you say "what about Hasselblad", its parent company is Shriro (http://www.shriro.com). )
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: MrSmith on January 28, 2011, 06:20:09 am
The biggest challenge for CMOS developers is to be able to make them at sizes larger than 24X36 and to maintain a reasonable (read financially viable) yield. This, for the time being, makes this debate irrelevant...
Yair

and when this happens i'll be at the front of the queue.
a 42-50 mpixel back/camera with 48x36 chip, usable up to 1200 asa and live view that works.
Canon have already shown that they can design/make lenses with bigger image circles (24ts-eII) it's well within their capabilities to design a 35, 50, 80, 120macro and 150 to cover this size chip.

unlike small scale medium format manufacturers that have launched then folded or just died a slow death a Canon or Nikon (or Pentax) could produce such a camera and lens system and develop it further.
the price tag would be high but certainly not higher that the current players, the rental houses would not shy away from investing either (never seen a contax or rollei HY6 camera for hire in the U.K.)

no idea if this will happen but i'm happy to wait as i'm still waiting for a digital back with the same noise at 800asa as a canon/nikon
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: mediumcool on January 28, 2011, 07:05:07 am
My dream would be a Foveon sensor — S2 dimensions would work.

A Foveon with [high-enough] sensitivity and a [large-enough] yield of a [large-enough] sensor would be a unique beast (told you it was a dream).

Could even be marketed as the Bayer Basher!
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: yaya on January 28, 2011, 07:35:36 am
and when this happens i'll be at the front of the queue.
a 42-50 mpixel back/camera with 48x72 chip, usable up to 1200 asa and live view that works.
Canon have already shown that they can design/make lenses with bigger image circles (24ts-eII) it's well within their capabilities to design a 35, 50, 80, 120macro and 150 to cover this size chip.
I don't think that neither Canon nor Nikon has an interest in marketing a camera larger than 35mm
Quote
unlike small scale medium format manufacturers that have launched then folded or just died a slow death a Canon or Nikon (or Pentax) could produce such a camera and lens system and develop it further.
Just as a reference...first Leaf DCB (36x36, 4MP) came out in 1992
First Canon Powershot 600 (0.6MP, chip size of a baby's fingernail) came out in 1996
First Nikon D1 (2.7MP, a fairly small CCD) came out in 1999
All 3 companies still make digital cameras/ backs successfully and 80MP at 54X31mm the Aptus-II 12 is still more than twice as large and approx X4 the resolution...
Quote
the price tag would be high but certainly not higher that the current players, the rental houses would not shy away from investing either (never seen a contax or rollei HY6 camera for hire in the U.K.)
In the early 2,000's Contax was probably the most popular MF rental platform, at least in London, alongside the RZ and 500 series Blads
Re Hy6 you're right but the I believe AFi is still available in rental through Peartree (http://peartreephoto.com/peartree-equipment-rentals.htm)
Quote
no idea if this will happen but i'm happy to wait as i'm still waiting for a digital back with the same noise at 800asa as a canon/nikon
At what size print?
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: MrSmith on January 28, 2011, 09:09:04 am
I don't think that neither Canon nor Nikon has an interest in marketing a camera larger than 35mmJust as a reference...first Leaf DCB (36x36, 4MP) came out in 1992
First Canon Powershot 600 (0.6MP, chip size of a baby's fingernail) came out in 1996
First Nikon D1 (2.7MP, a fairly small CCD) came out in 1999
All 3 companies still make digital cameras/ backs successfully and 80MP at 54X31mm the Aptus-II 12 is still more than twice as large and approx X4 the resolution...In the early 2,000's Contax was probably the most popular MF rental platform, at least in London, alongside the RZ and 500 series Blads
Re Hy6 you're right but the I believe AFi is still available in rental through Peartree (http://peartreephoto.com/peartree-equipment-rentals.htm)At what size print?

rental i'm talking about the big three. pro-center/flash center/calumet(and not forgetting the excellent Teamwork). peartree i have heard of but they certainly were not around when i started using rental accounts, Direct lighting were the only other rental house that offered a range of kit that an account was worth opening.
i don't remember seeing a contax on the shelves just blad and mamiya RZ. i find the idea of contax being the most popular rental platform hard to believe. i was assisting in the early 2000's in London and never once touched a contax camera only Hblad/mamiya/sinar/linhoff
even now the couple of dozen photographers i know 3 have H/blad the rest use phase backs

i'm fully aware of the digital history and the big jumps in technology/affordability/usability that occurred
(for me personally they were 1ds2, 5d, 5dII, 22mpixel (h/blad/phase) and 39mpixel (when all the 5x4 users i know jumped ship to digital)

maybe canon and nikon are not interested in making a back/camera system but if they did and the chips performed the same or better than the current 35mm offerings then i wouldn't entertain a phamiyaleafblad. and until a usable 800iso appears i'm still not going to.

print size? any size the client might end up wanting to use the image and not complain "what is all that gritty stuff everywhere"

Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: Doug Peterson on January 28, 2011, 09:11:59 am
no idea if this will happen but i'm happy to wait as i'm still waiting for a digital back with the same noise at 800asa as a canon/nikon

http://www.captureintegration.com/2010/03/30/phase-one-p40-versus-canon-5dmkii-iso-800/

An IQ180 will go to ISO3200 at 20mp (ISO800 at 80mp).

Doug Peterson (e-mail Me) (doug@captureintegration.com)
__________________

Head of Technical Services, Capture Integration
Phase One Partner of the Year
Leaf, Leica, Cambo, Arca Swiss, Canon, Apple, Profoto, Broncolor, Eizo & More

National: 877.217.9870  |  Cell: 740.707.2183
Newsletter (http://"http://www.captureintegration.com/our-company/newsletters/") | RSS Feed (http://"http://www.captureintegration.com/2008/08/11/rss-feeds/")
Buy Capture One 6 at 10% off (http://"http://www.captureintegration.com/phase-one/buy-capture-one/")
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: madmanchan on January 28, 2011, 09:36:02 am
I'm glad you asked the question! If I were the emperor I would have:
Had Nikon develop a medium format sensor
Had Burr Brown develop a 24 bit a/d converter insuring 16 clean bits
Had Adobe implement their products with a full 16 bits
Had Pentax produce the DB for 10K!
Marc

FWIW, the Adobe part you mention already exists for Camera Raw and Lightroom (>= 16 bits).
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: yaya on January 28, 2011, 09:45:25 am
Calumet had LOADS of Contax kit in London at the time and it was THE digital platform until the H1 came along

Oh and there are more than 24 photographers in London, trust me :-)

At one point (early 2006), the yellow pages had nearly 16,000 (!) registered within 2 miles from Old St. Roundabout...I don't know ALL of them but I know quite a few...
Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: MrSmith on January 28, 2011, 09:59:07 am
and it took the kodak rep a day to get from one end of great sutton st to the other ::)

Yaya, i only hired once from calumet, the lens was so worn all the distance markings had rubbed off and the front element was wobbly, at least the pro-center replaced every hire item after 2-3 years depending on use.

i'm fully aware of what sensor+ can do but i want it to do that at 40mpixel not 15 especially if i'm paying 16K for it, until then i'll hire or borrow.
i look forward to sensor manufactures increasing the sensitivity and noise  characteristics of their chips over mpixels because to me anything over 40-50 doesn't make my life easier, higher iso usability does.
 i'll sit tight until that day comes maybe canikon will have changed their minds by then :-)
Title: Re: FF CCD better for highly off-perpendicular light with microlenses?
Post by: eleanorbrown on January 28, 2011, 11:04:04 am
thanks for taking the time for this lengthly answer! eleanor

In the M8 and M9, Leica had the special challenge that some of its lenses deliver light to the sensor at a highly off-perpendicular angle (far more so that with any SLR lens), which is difficult to do when microlenses are also wanted for the sake of better low light performance. This was solved with Kodak's innovation of off-set microlenses. I believe that [active pixel] CMOS designs also inherently require the microlenses to be further from the wells than with the simpler full frame CCD design because of the extra circuitry on top of the well, and the microlenses also have to be stronger to get the light into a target region that is a smaller fraction of the photosite, both of which makes the "off-perpendicular" problem worse for CMOS than for FF CCD.

So there are two possible reason why Leica went with Kodak FF CCD's for it s rangefinders:
1. the inherent advantages of CCD in respect of handling highly off-perpendicular light when microlenses are wanted, and
2. Kodak's willingness to work on a custom design for a low volume, high price product.

Kodak and Dalsa seem far more oriented to the more specialized, low-volume end of the sensor business, while Canon, Sony, Panasonic etc. are more oriented to higher volume products. And when it comes to volume comparisons, even the original 1Ds, at 24,000 a year, was a high volume product compared to the Leica M8, or M9, or any MF sensor.

The "custom design" argument probably applies to the R2 sensor, where Kodak was willing to  design and produce that 45x30mm sensor exclusively for one very low volume product. Maybe this is also why Olympus originally chose Kodak for its 4/3 sensors, before Panasonic saw the potential for a far higher volume through the transition to Micro Four Thirds, which I suspect was in Panasonic's plans from the beginning.

Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: bjanes on January 28, 2011, 01:09:43 pm
I'm glad you asked the question! If I were the emperor I would have:
...
Had Adobe implement their products with a full 16 bits

FWIW, the Adobe part you mention already exists for Camera Raw and Lightroom (>= 16 bits).

ACR may have a full 16 bit path, but as far as I know, Photoshop is still 15+1 bits. That is sufficient for all my photographic work, but sometimes I want to work in Photoshop with full 16 bit files from DCRaw or Iris, and the 15+1 is annoying.

Regards,

Bill
Title: Re: FF CCD better for highly off-perpendicular light with microlenses?
Post by: ErikKaffehr on January 29, 2011, 03:50:49 am
Hi,

+1

There is a lengthy comparison between D3X and M9 by Erwin Puts here:
http://www.luminous-landscape.com/essays/dxomark_sensor_for_benchmarking_cameras.shtml

My take is that experienced photographers like Michael Reichmann and yourself are very satisfied with the image quality of the M9.

My guess is that using CMOS instead of CCD would improve high ISO performance and possibly shadow detail. On the other hand, integrating a CMOS sensor is probably quite different from doing the same job with CCD. AFAIK Jeonptik was developing camera electronics for Leica M8/M9 while Leica developed the firmware for the S2 with help from the Uiversity of Technology in Cologne (?).

I'd suggest that Leica would have a great benefit from the lenses. They used to be truly excellent, at least what I can see from MTF curves so they would transfer very good fine detail contrast, enhancing texture over noise. This characteristics would also cause a lot of aliasing. Aliasing is essentially "fake detail".

This is very well described in a document from Schneider Optics: http://www.schneideroptics.com/pdfs/whitepapers/optics_for_digital_photography.pdf of a screen dump of an illustration from that document is enclosed. Please note that all information in the sensor response is fake. Targets is black and white and contains 20 black lines but image contains only 12 black lines and lines with "invented" gray tones. The image looks perfectly good, but is no reproduction of the test target.

Theoretically, a high fill factor would be beneficial in reducing aliasing effects. This would also benefit CCD vs. CMOS, but once both CMOS and CCDs have microlenses this may matter little.

Best regards
Erik

thanks for taking the time for this lengthly answer! eleanor

Title: Re: Why not a CMOS sensor in the Phase One IQ series?
Post by: Bart_van_der_Wolf on January 29, 2011, 06:11:28 am
Regardless of sensor type or manufacture, what about the effect that the signal A/D converters have on the result? We never seem to see this discussed or measured anywhere, but in audio (which is another of my passions) the quality of A/D converters when recording music is quite crucial to the end result.

Hi John,

Besides the analogies, there are significant differences between audio and image signals. One of the differences is Photon shot noise. At low signal levels, the Signal to Noise ratio is dominated by Photon shot noise. In audio it's the amplifier/circuit noise that's the killer.

Another issue is the "well depth" of sensors. With a maximum well depth of 50,000 electrons, and a (mostly) read noise of 10 electrons, we have a dynamic range of 5000:1, or a little more than 12 bits of real DR. A 14-bit DAC would be adequate, and cheaper than a 16-bit DAC. Of course I assume that the DAC and circuitry doesn't introduce huge (>1.5 bits) amounts of noise itself.

As sensors get denser with a smaller sensel pitch, and no real progress in storage capacity (which mostly scales with surface area, not depth, for now), the storage capacity per sensel will rather decrease than pose a challenge for DACs. There will still be benefits from low noise circuitry, but it's not the only source of noise when photons are involved.

Added to that, the optical system can limit the scene DR that is projected on the sensor to, say, 9 stops quite easily. There may also be other factors, e.g. pixel uniformity that are a priority to address. There are lots of improvement possibilities, but there may be diminishing returns involved with some of them as well.

Cheers,
Bart