The Seitz 160 Megapixel Camera - a scam?

Seitz 160 Megapixel Camera

I was going through my log files the other day and I noticed a bunch of new sites referring to my 222 Megapixel Machu Picchu Image. Usually that's nothing special, the picture got linked on many sites. However I like to check most incoming links and what I found reported there was pretty amazing. My image was used as an example to demonstrate the resolution of a 160 Megapixel Digital Camera, that is expected to hit the market in 2007. While I had to manually stitch 35 images of my Canon 20D camera, this bad boy will be able to achieve a similar resolution in one shot. Looking at the picture above I do have some doubts as to if the optics can handle this kind of resolution though. Considering that for my 20D, I will not get a significant improvement beyond about f/16 (Limits of Diffraction), the image sensor must be gigantic. From the picture, it seems the lens is not nearly big enough for such a large sensor.

The sensor of my 20D is 24mmx18mm. The Machu Picchu Image did use a 7x5 matrix (7 columns, 5 rows). Lets subtract 20% for overlap, the effective sensor size was 0.8*24*7 x 0.8*18*5 = 134.4mm x 72mm. Lets assume the pixel size of the 20D and this bad boy were the same, I would need a sensor of 134x72mm for the Machu Picchu Image and a lens large enough to cover this sensor edge to edge (no vignetting). This lens doesn't seem big enough. A smaller pixel size would certainly accomplish this, but then the Limits of Diffraction dictate that you will only be able to obtain a very small depth of field.

I have seen many articles popping up about this camera and despite the size of this camera, noone has ever dared to question the limits of the optical resolution. In the heads of most people a digital camera is always limited by the sensor alone, but that is simply not true (and a reason why I believe that even most consumer point and shoot cameras above 7MP are limited by their lenses already).

From my experience with the Machu Picchu Image, there is another limitation to consider here. Lets assume that the sensor really is about 134x72mm (yes I know that would be for the 222 MP model, but who says they are not already cooking that one up). Since 134 is approximately 3.7 times larger than 36mm (standard film size), such a sensor would have a crop factor of 1/3.7. This means in order to obtain the same Field of View as a 50mm lens on a normal film camera, we need a 50mm x 3.7 = 185mm lens. Unfortunately longer lenses compress depth of field accordingly, which means a high f-stop number is required (an effect well known to the digital SLR crowd). A consumer point and shoot camera usually has a crop factor around 2.5, which means those cameras have very small sensors and are capable of achieving great depth of field. As a consequence, this giant here would need an even higher f-stop, again pushing the Limits of Diffraction.

From their website we can find out that the camera is a 6x17 widescreen format. For an equivalent Medium Format Camera this would mean 60mm x 170mm. It would seem pretty hard to fit this kind of sensor behind this lens, but maybe not. The dimensions are hard to judge. A 22MP digital back sensor is about 48x36mm, which leads me to believe that 60x170mm should be about right. How they can fabricate such a large sensor at a reasonable yield is beyond me (yield is inversely proportional to area due to probability of impurity). It is certainly possible, but not at this price (A professional digital back doesn't cost much less either).

Here is some more information about the camera:

In case you were wondering about the street price, you should expect around $32k. Not that bad considering a 1DSM2 (Canon) runs at around 8k. The ISO range of the camera is 500-10,000. Now this gets me a little curious. How can this be true? Professional SLR cameras usually offer a range of 100 to 3200.

The Image Sensor has been developed by DALSA Corporation exclusively for Seitz. Whatever that means, I have never heard of Dalsa before. However I do know that Kodak once tried to develop their sensor but didn't do so well. Currently there are Canon, Nikon and Sony. Many other cameras like Konica Minolta are using the Sony sensor and outside these three there are not many manufacturers. The reason is simple, these three companies really have the edge when it comes to sensor development. It remains to be seen how good the Dalsa Sensor can be.

EOS 30D Digital SLR Camera, Body Only - 2.5'' LCD

The camera offers 48-bit images (this means 16-bit per channel). Not bad, considering the Canons are doing 12-bit and some Minolta are doing 14-bit. However i doubt that the camera will really deliver this kind of dynamic range. There is a reason Canon limits themselves to 12-bit. That reason is noise (or signal to noise ratio). Why should they record more bits if they are buried in noise anyways.

Well i am not going to conclude that the camera is just a marketing stunt. Noone would be crazy enough to announce something they cannot produce. The technical data is really impressive and Seitz may just have found a way to push the envelope a little further than most thought would be possible.

 

This is a scanning (strip-shaped) sensor

This is a scanning (strip-shaped) sensor. 170mm X 60 mm probably means something like a 5mm X 30mm that moves (as does the lens) during the exposure. That's why it takes one-second for the complete exposure. Check out http://luminous-landscape.com/reviews/cameras/better-light.shtml and http://www.betterlight.com/ Better Light has a scanning back that has just 8,000 x 3 “effective pixels” on the sensor but 15,990 sampling positions and thus gives 384 megapixel images. Basically automates what you did with your 20D on a huge scale. Seitz has built this kind of a system into a portable hand-held machine unlike others who are doing it just for digital backs on view cameras.

Thanks for the correction

Thanks for the correction Vivek,

that relieves the problem of manufacturability, but brings with it a whole bunch of other problems.

If the lens really moves, even some of the optical constraints (at least in one direction) might be removed. However one still needs to be able to resolve the full line resolution.

A scanning back will always have a problem with moving things (like leaves of trees), due to the long exposure time.
During a stitching process (what I did), these things can be corrected by the algorithm (using the data from one picture and not both in the overlap region).

Anyways, its a good point that you brought up, but I am not sure a camera like this would be of any practical use to a nature photographer then, unless you can guarantee your subjects to be static.

Also, the test image shown on one of the sizes was a sports shot (windsurfers) which pretty much lead me to discard the idea of a scanning sensor. I guess they just made the picture up.

Moving the lens will bring a whole bunch of other problems. I don't know how they can do this with someone shooting a handheld shot (like in the marketing picture), without blurring the whole picture. The mechanics will have to be extrordinary (very precice), but it can be done (some Slide scanners offer resolutions above 4000 dpi so it seems doable, just crazy).

EDIT: If you follow the link in my article (technical data), it seems they claim a minimum exposure time of 1/20,000s, which is not possible with a scanning sensor. I should have had a closer look on this page, since they are also saying the pixel size is 8umx8um (that means for the quoted resolution a theoretical (or equivalent) sensor size of 60mmx170mm (what I already speculated).

Andre

The Gigapxl Project

Hi!

While this isn't digital, it is a 1 gigapixel camera, you might be interested reading about it.

http://www.gigapxl.org/

they've got some incredible images from it...

Ben

Thx

Thanks Ben,

I have read about this many times :-)
I agree that their images are really excellent.
Still, I am wondering how they define 1 Gigapixel if their camera is not digital.

Bah

Andre, your position in your original post and your comment clearly come across as very negative and prejudiced. Being so biased really hurts your credibility since you are sitting here taking pot shots at whatever you can. For example, when you learned more about the camera you rebounded with whatever negative statement you could: "Anyways, its a good point that you brought up, but I am not sure a camera like this would be of any practical use to a nature photographer then, unless you can guarantee your subjects to be static."

It would make you appear far less arrogant if you simply posted your questions without having already made your mind up that this camera sucks.

They're claiming minimum exposure times of 1/20,000, yet you are prejudiced enough to call them liars. To wit: "Also, the test image shown on one of the sizes was a sports shot (windsurfers) which pretty much lead me to discard the idea of a scanning sensor. I guess they just made the picture up."

What's worse is that you disparage it without bothering to closely read their page on the product.

The worst is that you are, in essence, a liar. I quote: "It remains to be seen how good the Dalsa Sensor can be." Sorry, I didn't realize a public corporation with $200 million in revenue had to check with Andre Gunther before their sensor was determined as "good." But then, you are so uninformed that you claim you hadn't even heard of Dalsa.

Therefore, you definitely shouldn't be sitting there casting a dark cloud over something you admit you aren't well-informed or well-read about.

I will remember to include

I will remember to include irony tags in my posts in the future for people like the previous poster who obviously have no sense of humor.
For example my statement on a made up picture was simply a way to point out that this can hardly be a scanning back.
I wasn't aware that anyone would actually take this seriously.
At least I have learned that I should not assume that my readers get my irony at first glance and should make it stick out a little more.
One more word about the sensor: Check out the history of Kodak Cameras (who are worth close to 7 billion by the way). Money doesn't make good designs, people do. One of the best sensors is done by Foveon who aren't publicly traded.

"Therefore, you definitely shouldn't be sitting there casting a dark cloud over something you admit you aren't well-informed or well-read about."

Why not? I let you cast a dark cloud over me on my own website. Everyone is entitled to their own opinion and so am I. I don't care wether you like that or not.

Did you see that I asked questions and used qualifiers such as remains to be seen?

Seitz 6x17

Andre,
I have been in contact with Seitz re their 6x17 camera and they assured me the sensor was 6 cm high. I don't know how many rows of pixels it has though, but they have obviously achieved a breakthrough in scanning speed relative to the other scan systems (Betterlight, Phase One etc) which can take 35 seconds or more in sunlight. Dalsa is located in Waterloo, Ontario, about 100 km west of Toronto,and they are one of the major chip makers.

The sensor moves on the Seitz camera, and the lens stays fixed. The lens looks like one of the super wides, perhaps a 90mm or 115mm, either of which would easily cover the format. Some of the large format lenses are surprisingly small, I have a 240mm f9 fuji lens which covers 8x10, but takes a 52mm (I think) lens cap and weighs only a few ounces. Some of the early wide angles were extremely small, but also very slow ie. f18 as in the Zeiss Protars.

Hopefully this new technology will migrate to some of the 4x5 scan backs. This kind of speed would open up a whole new world of photographic possibilities for 4x5 digital capture.

Harold

That would be a really

That would be a really amazing accomplishment, considering the sports shot they advertise with and the shutter times they specify?
I am not sure how they can move the sensor in 1/20000 seconds. Also this requires multiple exposures (each row one exposure).
So if the sensor were 100 times smaller (in the long direction) than the maximum resolution, this would require 100 exposures (I am not sure I am being very clear with this statement).
So despite the obvious mechanical challenges they must really have a ultra high speed sensor.
I am not going to speculate anymore, since I would just give more fuel to those attacking my questioning nature.
As I said, I'd like to wait and see what this thing can do.
Did Seitz confirm that it is a scanning sensor? It wasn't clear from your post.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd>
  • Lines and paragraphs break automatically.

More information about formatting options