Sounds like the full frame CMOS sensors will have some competition...


DALSA Presents New 48 Megapixel Image Sensor at IEDM Conference


11:00 EST Wednesday, December 17, 2008

WATERLOO, ONTARIO--(Marketwire - Dec. 17, 2008) - DALSA Corporation (TSX:DSA), an international leader in high performance digital imaging and semiconductors, was pleased to present its new 48 megapixel standard product image sensor for professional photography in a presentation at the International Electron Devices Meeting (IEDM) in San Francisco this week. IEDM, organized by the IEEE (Institute of Electrical and Electronics Engineers), is the world's premier forum for the presentation of advances in microelectronic and nanoelectronic devices.

Eric-Jan Manoury, a scientist at DALSA's Eindhoven R&D lab, gave the presentation titled "A 36x48mm2 48M-pixel CCD imager for professional DSC applications," highlighting not only the device's high resolution, but the advances it delivers in increased quantum efficiency, lower noise, and higher dynamic range compared to previous generation or competitor devices.

"This scientific presentation at the IEEE IEDM international conference demonstrates once again the world leadership which DALSA continues to provide in the field of large format image sensors for professional digital still camera applications," stated Dr. Savvas Chamberlain, CTO and Chairman of DALSA. "This is the first large image format imager with such high pixel rate (100 MHz) and with 74 dB dynamic range and 55000 electron signal capacity."

Built on DALSA's' new 6 micron pixel platform, the new device is available as a standard product. With the new performance benchmarks it establishes, including superior speed, dynamic range, and angular response, it is already generating demand from customers in one of the world's most demanding imaging applications: medium format professional digital photography.

DALSA's image sensors led the way in professional photography's shift to digital ten years ago, and this new device delivers even more of the performance professional photographers demand: higher throughput, exceptionally low noise, unmatched dynamic range, and better color.

To view the technical paper from the presentation, visit our website: http://www.dalsa.com/public/corp/PDFs/DALS...dings_final.pdf

If Dalsa sold half of that chip, Nikon could build a decent camera around it, and the title would become somewhat more accurate:

24MP DSLR's on the way

Meh, impress me with ISO, not with pixels.

Dalsa now has a sensor to compete with Kodak's 50MP, both basically 36x48.

Not sure what this has to do with DSLRs. These are medium format sized sensors.

Good for Dalsa. Now the goal should be clean 800 ISO, usable 1600 ISO and a digital back around $12-15K.

Six micron pixels compare with Sony's 5.9. There's likely to be the same high ISO noise problem. That gives up the main MFDB advantage of larger pixels with all that comes with that.

Puzzling news, all the MF companies likely to be willing to use this chip have recently announced their 2009 line of backs... some of them with other Dalsa chips.

Or would this be another sign of Nikon's upcoming MX being for real?...

Cheers,
Bernard

I'd settle for a 20 megapixel digital back for my FM

Frankly, I want ISO 25 more than I want ISO bloody big number.

This is not only the question of sensel size; the electronics plays a greater role.

1. CCDs are less noisy than CMOS,

2. the read out electronics of these expensive chips is very different from the others. Even the "others" are not equal; why do you think Canon's 1 serie and the 5d2 have lower noise, than the xxxD serie?

3. The quantum efficiency is claimed to be grater than the former versions.

It's a pity that such sensors are not available in DSLRs. It would be a great day if we could forget the dumb ISO setting and shoot away with a large dynamic range.

If you mean having just "dynamic range" and no ISO setting, I think that day is coming. Although it will always be nice to have the option to get a grainy image rather than no image. but point well taken. ISO is probably on it's way out. It still has relevance because of "grain" quality, but sooner or later, noise isn't going to be a problem, and ISO will fade away. Don't video cameras call ISO "LUX" for how much dynamic range they have?

48mp DSLR's? What are camera manufacturers going to do about their marginal 24mp 'compatible' lenses? Design a whole new bunch of them that are even bigger and better? I can't wait, and if I carry my bag on the other shoulder maybe they'll help me straighten up.


Steve

Nothing puzzling to me: Dalsa, like Kodak, is probably shifting to the new 6 micron cell size across all MF sensor sizes, and this is just making that changeover for the 36x48mm sensor size, having already introduced larger, more expensive sensors in "6 micron". And since Dalsa sells these sensors for other uses too, there is no need to hold the announcement until a DMF back is about to use the new sensor.

And has anyone said that models announced so far are the only models coming in 2009?

Surely no-one really expects the old, lower resolution cell sizes of 7 or 9 micron to stay around for much longer. Offering significantly higher resolution than 35mm digital offers is essential to the DMF market, through both lenses and sensors and 33MP vs 24MP is not much of a margin.

I expect back models that currently use the 33MP 36x48 sensor from Dalsa to be upgraded to this sensor at some point. And I expect prices significantly lower than for the larger "near 645 full frame" models, keeping the market alive for "48mm format".

Except in every single real world example since the advent of digital outside of special applications or labs...

You might want to qualify that: it seems to be true at base ISO, but CMOS sensor noise levels tend to increase more slowly with ISO speed (probably due to applying ISO gain amplification on-chip, earlier in the signal chain), so for those who only care about high ISO noise levels, the opposite is true. I would phrase it as
"CCDs have greater maximum dynamic range than CMOS sensors".

As far as I have read, CCD is inherently less noisier, but cmos allows easier for on chip NR, and this can be cleaned up better, which, on the other hand, makes for the overly smoothed out cmos look and the crispier CCD look.

Bernie

I'm with Panopeeper. I'd love to see a DSLR with a slow-readout CCD optimized for absolute maximum image quality at base ISO. Heck I might even be willing to pay 8K for a D3x that had one of those in it, instead of yet another CMOS sensor.

Well, let's see what Canon have to say to this subject:

CMOS sensors generally have the disadvantage of generating more electrical noise than CCDs, which can result in poor image quality. There are unavoidable fluctuations in the performance of the millions of photodiodes and amplifiers incorporated into a CMOS sensor, and the tiny differences in performance result in noise in the output image....

Of course, Canon worked hard and quite successfully on reducing this disadvantage, but that does not change the fact, that Canon (and now Nikon) chose CMOS for other, even more important advantages: faster operation, much-much lower power consupltion than CCDs and couplet to this, much less heat generation.

Is it then surprizing, that the MFDBs are using CCD? They are working slower, consume power and generate heat like an oven, but deliver clean images. On the other hand, they are weak with higher ISOs.

This Dalsa paper summarizes the differences between CCD and CMOS and does not take sides, since Dalsa makes both types of chip. It would seem as if CMOS is better suited to 35mm style DSLRs while CCD is better suited for the characteristics of a medium format format.

The Canon paper I quoted from says

The bigger CCDs are, the more power they consume. Making them faster also requires more power. A Canon in-house comparison of CCD and CMOS power consumption found that with the very small sensor in point-and-shoot digital cameras, the CCD consumes 50% more power than CMOS. In the case of an APS-C size sensor, used in DSLR cameras such as the EOS Digital Rebel XT, EOS 20D and 30D, the CCD consumes more than twice as much power. With full-frame 35mm sensors, CCDs consume about three times more power as a baseline.

This factor, and the speed issue make CMOS the better candidate for medium format. The fact, that the MF sensors are CCDs despite these considerations is the proof, that another factor, the noise still speaks for CCD. Even the Dalsa paper mentioned above writes:

System noise: low (CCD), moderate (CMOS)

Btw, the Canon document is a "white paper", called "Canon's full frame CMOS sensors" from 2006. I don't have the link, I saved the document.

The Canon document can still be found on Rob Galbraith's website: http://www.robgalbraith.com/public_files/C...White_Paper.pdf

Wasn't the Nikon MX rumour fueled by Nikon advertising a new product of interest to medium format users - which turned out to be the D3x, doh! - and dpreview 'experts' seeing the advert and explaining that Nikon MX was a cert because the new tilt shift lenses covered a large image circle and hence the optics were suitable for a new range of MF lenses?

Or was there some actual solid evidence for Nikon MX?

The Nikon D200 uses a CCD chip. The D300 uses a CMOS chip. Comparing the two, we see that at ISO 1600 the D200 produces horrible results, with softness, and luminance noise. At ISO 1600 the D300 produces much better results. Even at ISO 800 D200 sensor noise and loss of sharpness are obvious. Only when Nikon went to CMOS were they able to match the low noise results from Canon cameras. In the real world CMOS rules.

I guess that's the reason of Canon putting CCD in all Powershots and Nikon in the Coolpixes, as well as Hasselblad and Phase One using CCDs in their newest backs from Kodak respectively from Dalsa.

Some comments from "Dalsa": http://www.dalsa.com/corp/markets/CCD_vs_CMOS.aspx

For a general purpose DSLR where things like high ISO performance, frame rate, and battery life are all important, I would tend to agree with you. But I think there would be a market for a DSLR that optimized low-ISO image quality even at the expense of some of those other factors. I won't be moving to MF anytime soon for a variety of reasons (not just cost), but if Nikon would offer a 24x36 "mini-MF" camera with similar imaging characteristics I'd for sure buy one.

So I guess ISO 800 and up is the only thing that matters in your definition of the "real world"? This may come as a shock, but you don't speak for everybody.

BTW using past-generation Nikon DSLR's is not a very good way to prove your point. There's a generation or two of R&D in-between the D200 and the D300. A better comparison would be the D2x and the D200. The D2x was CMOS, yet it was routinely slammed for high ISO noise and was in fact a bit worse than the D200.

It seems to me that if you don't mind having a battery that gives only 150 to 200 shots, or having a battery three to four times larger, then CCD's would be fine for 35ff cameras. But I doubt that most people would like that.

Another factor is where the R&D dollars are going. CCD sensors cost more than CMOS sensors, because CMOS sensors are made on the same lines as any other chip. But CCD's are made so differently, that they need special lines. Because of that, the cost to manufacture is much higher, without many other benefits for most people. As for CCDs being used for most compact cameras, that's still true, but won't be for the future. These chips are so small, and so many cameras use them, that the cost is reasonable. The cheaper electronics in those cameras outweighs the slightly higher cost of manufacture for those small chips.

Because every db of noise reduction is needed for those small chips, CCD is still the way to go. we notice just how bad the higher ISO performance those cameras have, due to the small sensor, the cheap electronics, and the CCD.

As far as MF goes, the story is opposite to that. Price is almost of no object, and the best electronics is used, often with some form of built-in refrigeration to keep those large chips cool. CCD's get much noisier when they heat than do CMOS.

Eventually though, everything will move to CMOS, as their performance is rising faster than is that of CCD.

It's like film. Years ago, when digital began to get useful in the way of quality, about the time the Canon D30 came out, my Kodak partners (we processed professional kodachrome film, the only lab to do so) and I discussed the future of film and digital.

Their opinion was that film had come about to the end of its possible development. There was room for improvement, but no major performance enhancements were possible without much more effort that was affordable. Digital was seen as advancing so rapidly, that there was no way the money could be spent on major film development efforts.

The same thing is coming true for CCD vs CMOS. I've read number of papers on this over the past few years. CCD has come a long way, and it's felt that it doesn't have that much more to go. A few years from now, and its advances will peter out. but CMOS is still in its infancy, and has a way to go.

There are actually some newer technologies in the lab that will give better results than either, but they are a ways off. They'll likely appear just when needed, as this is usually what happens.

what I would like to see is a DSLR with a 35mm sized sensor (CCD or CMOS) that did not have a bayer array and shot in Monochrome only where each pixel would only see and record brightness levels and not worry about colors/RGB. It would basically be just like shooting with B&W film, and I could use real filters on the lens. Canon, Nikon, Sony etc could just put this sensor into a DSLR body that they already make. Which would help keep costs down

Where did I say that ISO 800 and up is all that matters? Where did you get that crap from? I suggest you learn to read my posts and do not put words into my mouth.

I use a D200 with which I am very happy. I do not need high ISO performance. I was addressing a separate point made by someone else.

The point about different generations has of course some merit. But the D2x has 13MP compared to 10MP in the D200. So it is a generation earlier, has a significant increase in pixel count, and still performs almost as well. Okay, far from proof, but it does look as if in 35mm cameras, CMOS has in practice, for whatever reasons, lower noise.

Yeah, me too -- that would be awesome.

I'd like to see an intelligent alien race, personally.

In theory, there's no reason why each sensing site couldn't be used a single "pixel" in the image for B/W. It's done that way for color but it's the filters place above those sites that screws that idea up. So you really do still need the Bayer algorithms.

The question is whether there would be a big enough market for them to supply special sensing modules for this. They could easily do so if the market were big enough.

The advantages for B/W would be very noticeable. for one, no filters would increase sensitivity by at least one stop, more likely two,, lowering noise, and increasing dynamic range. Then you wouldn't need the antialiasing filters as there would be no need for de-mosaicing, which means no color aliasing causing by the Bayer transforms (though you would still get moire from the interaction of the sensor array with fine detail). Sharpness would increase dramatically.

Kodak made such a beast a few years ago. The idea worked great, but they didn't sell enough of them to make it an economically viable market niche...

Here you go:
Monochrome P45

I've spoken to a few people about that one. I'd like to know someone who's interested enough to actually buy one. but even though I know a lot of people who do medium format digital, I don't know anyone who's interested enough to plunk down that much for what would be, today, a specialized back, only good some of the time. Lloyd himself tells me it's too pricy for him, though if there's enough interest, he might try to borrow one from where he gets his samples for reviews, and rents his other stuff from.

I do work with Leaf, but they have told me that the market is SO small, they can't be bothered. Even if they would, the price would be comparable.

A 5D mkII for an extra $1,000 might sell enough, but it's hard to say. They would have to do a special front glass without the anti-aliasing filter, but with the fluorine coating for the new anti dust system. Making just a few of those glasses, and just a few of the entire modules would be expensive. Whenever you remove a component from the assembly line for a "special" the price is much more than what it's worth, because of the extra labor involved.

Everyone was expecting the crossover for MF to occur at this generation. It didn't. If they decide to move to back-thinned sensor that might win another couple generations for our unlikely survivor, and by then, who knows, some other technology might come up to nix CMOS's time in the MF sun and replace CCDs.

Edmund

Of course, by "eventually", I mean the next few years, as I mentioned, there are newer technologies in the labs. But the newest tech just being understood, won't be available for some time.

I'd like the same , a real B&W MF sensor

P45+ monochrome is too expensive

http://www.mega-vision.com/products/Mono/Mono.htm

The real deal at significantly lower prices (IMO) than the P45 Mono.

Note - because there's no Bayer pattern, you only need actual film resolution to get well, film like results. So that's 2460*3690 @ 3600DPI. A modest 9MP. The 11MP version would cover that quite nicely, with plenty of room to spare.