Michael, if you read this thread, did the Dalsa guy give you any more info on the specs of this sensor that would explain the exclusive nature of it?

It was clear after just the first day of shooting that the P25's dynamic range is noticably superior to that of the MKII, and every other photographic device that I've ever seen.

I'll have lot's more to say on this in a while.

Michael

Do we have any information on the performance, dynamic range, and image quality of the Mamiya ZD backs at all?  Supposedly, it's less than half the price of the P25, so, I have no idea what to expect.  Michael?  Anyone?

T-1000

No one outside of a couple of beta testers has seen images from the ZD yet.

But, it's not unreasonable to make certain deductions. It uses a Dalsa sensor and support chips identical to those used in other highly regarded backs from other makers. There's therefore no reason to assume any significant divergance in image quality from what we're seeing from others.

Michael

Probably not.  Michael's article has a photo of a wafer of 24x36mm ("35mm sized") sensors, which are 11MP and so about 2700x4000, but no photo of Dalsa's 22MP sensor or of the more mysterious 5Kx5K.


P. S. Sorry, ignore this! As explained below, there is a second photo, Fig. 4, of the 5Kx5K

(I got myself confused with the fact that there is no photo of the 22MP sensor for the Mamiya and other backs. I can guess what it would look like though; merge each pair of those 24x36mm 11MP sensors in Fig. 3 and you get four of the 48x36mm 22MP ones.)

Michael:

I really liked your article on Dalsa, the company making both still & video digital sensors. The 8.2MP cinema camera sensor is very impressive and intriguing to say the least. Would love to see some sample clips from that!  

But i have a question about the 5Kx5K sensor your guide only smiled about when asked about potential customers. You speculated they may be military since only a government could afford such expensive equipment.

Doesn't 5Kx5K = 25MP? Aren't some large format digital backs 25K already? A few at least? So why would this one be so prohibitively expensive that only government/military could afford it?

There are large format digital backs go to 6kx8k (48MP) and beyond, but they use scanning of a linear sensor: a 6kx8k back has a sensor with a single line of 6,000 pixels, which is moved across the frame and used to read each of 8,000 lines in turn. Not for action photography.


P.S. CCD's for special purposes like astronomy and other "off-planet" photography often have very large photosites, for the sake of great dynamic range and low-light ability once suitably cooled. Kodak makes CCD's with pixel spacing of 12 microns and even up to 20 microns. So that 5kx5k could be 6x6cm with 12 micron pixels, or even larger. Well beyond the size what any of us are able to buy.

don't mean to hijack the topic, but the thing i found most interesting about the dalsa manufacturing process is that the large sensors are built up from smaller sensors. there's been some discussion of that (oh, ok! mostly by me (and jack flesher)) on the other digital back forum here. anyway, i guess i don't understand what's involved in stitching, because it sounds like the yield of large sensors shouldn't be much lower than that of small sensors, if large ones are indeed stitched from small ones.

ted

thanks for the clarification, michael. i read the 'stitching' to mean that individual, complete small sensors were put together to make a large sensor, which says that the yield rate would be essentially that of the small sensor (ie, just replace the bad small sensor with a good small sensor in the array of small sensors). but if it's just that the large sensor is made in patches, then it's an all-or-nothing deal.

much like reporters who have to ask questions they know won't be answered (yet), i gotta ask... how does the dynamic range compare between 1DsII and P25?

ted

A stepper is a photo device. It uses precision optics to focus a light source through a mask to expose a photo-sensitive coating that was applied to the wafer in an earlier step. The exposed areas of the coating harden in response to the light, the unexposed areas stay soft and are then removed in a later step. The substrate is then etched, only in the unexposed areas, to form the circuitry of, in this case, the light receptors. The name stepper comes from the fact that the wafer is stepped under the light source on a servo driven stage.

I think the only limitation on sensor size is the size of the substrate and the number of sensors you are willing to toss to get one that works. There will always be some number of receptors that don't work. The bigger the sensor the harder it is to get one that has enough working receptors. It's all about acceptable yield.

I work with this stuff everyday and I'm amazed it works at all.

I hope that this will include some information or at least informed speculation about whether this "impossible" dynamic range technology will migrate to 35mm DSLR's in the forseeable future and for a little less than astronomical prices.  1ds/1dsMKII pricing has been astronomical enough.

I can live with 1ds resolution for another year or two, but I'd re-arrange my priorities considerably for something like a 1ds with better than negative film dynamic range.  That would be worth more than 23% more resolution.

Summary of Mamiya ZD

Press Release

Sensor Specification

So, the answer is still NO then.  ???

Nothing there to give us any idea of the actual performance of the back/camera itself yet. We can quote specs up the ying-yang but that doesn't necessarily translate into picture quality. (and I want it to be great btw as I'd be on eof the first to upgrade from my 645M).

Ray,

I agree that with current technologies, technical considerations strongly suggest that the Kodak FFT CCD sensor of the P25 should be able to give greater dynamic range than the Canon CMOS sensor of slightly less than half its size (area), and the evidence from Michael and other sources is that this is the case.

But since we are comparing a  particular pair of sensors that use different technologies, it does not seem fair to dismiss such differences in the comparison.

One end of the dynamic range is set by highlight headroom, usually determined by electron well capacity. This seems to vary by a factor of two or more for pixels of the same size in different technologies, for a potential one stop or more difference in dynamic range. Surprisingly, the highest well capacity I know of is for the very unusual style of CMOS used in the FillFactory sensors of the recent Kodak DSLRs, followed by FFT CCDs (going by Kodak and Dalsa data sheets), and then interline CCD (Kodak and Sony data sheets) and more normal CMOS (though I have less reliable data sheets to back up this last claim, so I am going mostly on my understanding of CMOS design). Fuji's new version of SuperCCD is the dark horse of highlight headroom; will the S3's sensor really be several stops better?

The other end of dynamic range limit involves dark noise levels, which can also vary significantly between sensor types: the FillFactory CMOS sensor loses much of its advantage by having about the highest dark noise levels around; about twice as high as for FFT CCD's at a given pixel pitch, so another one stop difference there. I will avoid comment on the near-religious issue of CMOS vs CCD dark noise levels!

Bah, I want it all.  Extreme number of pixels, clean at high ISO, extreme dynamic range, small light camera, modest price.  How much longer do I gotta wait, huh?

Read Michael's article that this thread is discussing...

Read the article. The photo is there unless Michael just changed it.

I would guess that the 5kx5k reference is to the die size - eg compared to the 1kx1k.  That chip is looks bigger then 25mpx

The Dalsa representative wouldn't say anything about the 5K chip, only when I asked if it was for the military, he just smiled.

Smaller segements don't make any difference when it comes to yield. Yield has to do with the number of bad pixels on a chip. With small chips the chances of bad pixels is reduced. With large chips it becomes almost a certainty. They can be mapped out, but this becomes expensive if there are too many.

The smaller segments mean though that large chips are possible because the stepper motors that etch the pixel pattern typically can only move in 1K increments.

Michael

Optimistically, I take this as including an advantage over even color negative film, fullfilling the paper promise of the very impressive spec's of the high end medium format Full Frame Transfer CCD sensors.

I hope that the latest medium format backs will establish a clear advantage in image quality attributable to superior dynamic range, and thus put improved dynamic range at the top of many customer and camera designer wish lists for improvements in DSLRs, ahead of yet more pixels, or clean ISO 3200 images, or catching up with the 10fps rate of the fastest 35mm film SLR.

I'm not sure if I'm going to get into trouble again with Michael, but I'll risk it (and desist at the first warning  :) ).

Putting aside issues such as CMOS versus CCD, pixel size versus pixel pitch and differences in aspect ratio between the P25 sensor and the 1Ds sensor, essentially, for a given field of view the P25 employs two 9 micron pixels to do the job of one 1Ds pixel.

No smaller format can compete with that advantage. Whatever tricks can be devised to reduce noise in the smaller format and increase dynamic range, the same or similar tricks (or different but effectively the same) can be used to reduce noise and increase dynamic range in the larger format.

Yeah, thanks for the links, but this is pretty much the same crap we've been reading for the past 5 or 6 months.  

I have a strong belief that Mamiya's website just plain sucks.  At least Phase One posts samples.  I guess since Mamiya doesn't show us anything, there's a reason why we have Michael and Dpreview to do the previews/reviews for us.  I'll just have to wait and see.

T-1000

Is there a link to a photo of this 5kx5k Dalsa chip?

Thanks.

D.

Thanks.

The article is here>

http://www.luminous-landscape.com/essays/dalsa.shtml

...if you're interested.

I wonder if the forthcoming Mamiya, also featured, will take RB lenses?

D.