Big Sensors

[J. Bernard Vallon]

I just came back from WPPI, (www.wppionline.com), and there is still no new news from Nikon about a full frame camera. I'm getting frustrated. All my gear is nikon, and i dont wanna switch.

Could someone explain to me why exactly it is so difficult to build big sensors? I'm really excited about RED because its full frame (for s35). I'd really like the new hasselblad h3 because its full frame for medium format, not for resolution or anything, but because of the shallow depth and low low noise. But they are so damned expensive!

Why doesnt nikon or canon (or red, for that matter) make a good medium format sized sensor camera? I dont understand the problem. Shouldnt nikon be able to build a 50mm x 50mm sensor and blow canon out of the water? I understand that more silicon means more money, but is it really by $2000 a chip...or more!?

[chuck colburn]

Nikon does not make medium format cameras or lenses for that format.
Nor does Canon.

[Petr Dvorak]

film 35 mm full frame is smaller than 35 mm photo full frame

[J. Bernard Vallon]

Nikon does not make medium format cameras or lenses for that format.
Nor does Canon.


film 35 mm full frame is smaller than 35 mm photo full frame

I'm fully aware. "Full Frame" is a term used to describe a sensor when it is the same physical size as its film based equivalent.

Therefor the Canon 1ds, at 36x24mm, is a Full Frame camera (equivalent to 35mm still photography)

The RED is full frame because 22.2x13.7 (correct me if im off) is equivilent to S35

and the H3 is a full frame (almost) medium format camera because 48x48mm is pretty darn close to medium formata full frame.

My question is, why is it so difficult and expensive to build these chips big? Big chips isnt just better, its AWESOME. It seems to me, canon/nikon could sweep the market with a great big 40mmx40mm beast.

[Bruce Allen]

It's expensive to build big chips because if you get one nasty error, you lose the whole sensor. But if were making sensors 1/4 of the size, you only lose 1 out of 4 sensors. I'm sure we'll get to full 35mm size soon, though... I too expected sooner!

Note that the Red "full frame" sensor is actually more similar in size to the Nikon reduced frame sensors.

Bruce Allen
www.boacinema.com

[Tom Lowe]

How big is the Phantom 65's sensor? They claim 65mm DOF.

[Michael Schrengohst]

And RED claims to be able to swap out sensors? That is one of the reasons we are going this direction. I would rather pay $10K for an overhaul every few years than to have to "upgrade" and get a whole new camera and lense(s).

[Mark B.]

I think the important thing is to get usable footage. If you can do that with an s35 sensor, then maybe it doesn't make sense to go to a larger sensor. Any noise-reduction benefits of the larger sensor will be lost because there will be a larger surface area to light up. Also, the larger chips have falloff issues to deal with, since the chip's aren't as good at capturing light from off-axis angles.

If you REALLY want to complain about chip technology, ask "Why haven't the chip manufacturers adopted newer, more light-sensitive technologies, such as utilizing Indium Gallium Nitride?"

[Dominic Jones]

Nikon does not make medium format cameras or lenses for that format.
Nor does Canon.
Actually, there are Nikkor lenses for 6x6 (and probably 6x.4.5 as well, if you look around). They made a lot of lenses for earlier Bronica cameras, like the S, S2, S2A etc, before Tamron started making their own Zenzanon glass...

See here, for instance:
http://cgi.ebay.co.uk/BRONICA-S2-fit-Nikkor-H-50mm-3-5-EC-S-S2-S2A_W0QQitemZ120085614572QQcategoryZ3350QQrdZ1QQss PageNameZWD1VQQcmdZViewItem

I think the main reason that bigger sensors are generally not created is that they don't give much of an advantage for the disadvantage of extremely shallow DoF (which is very hard to work with). For a film camera, resolution is directly proportional to image area. With digital, it is not so - for a given resolution you will get a higher latitude and greater sensitivity (although this will be offset by the lens coverage needed) with a bigger sensor, but these will be far less drastic improvements in most cases than the resolution gain seen with film...

[J. Bernard Vallon]

"....Why haven't the chip manufacturers adopted newer, more light-sensitive technologies, such as utilizing Indium Gallium Nitride?"

http://en.wikipedia.org/wiki/Indium_gallium_nitride

neat

[Erik Widding]

Could someone explain to me why exactly it is so difficult to build big sensors? I'm really excited about RED because its full frame (for s35). I'd really like the new hasselblad h3 because its full frame for medium format, not for resolution or anything, but because of the shallow depth and low low noise. But they are so damned expensive!

Why doesnt nikon or canon (or red, for that matter) make a good medium format sized sensor camera? I dont understand the problem. Shouldnt nikon be able to build a 50mm x 50mm sensor and blow canon out of the water? I understand that more silicon means more money, but is it really by $2000 a chip...or more!?

There is a limit to the reticle size of modern semiconductor processes. For cost effective production of sensors the process needs to have as few tweaks as possible from a standard process. Cropped sensors fit within a single reticle. The larger sensors do not.

The larger sensors will require multiple exposures for each mask layer with different reticles. Alignment becomes an even bigger problem than it already is. Feature sizes need to be much larger than process minimums to ensure that the seam between these tiles is not a major contributor to fallout. In addition to these costs you have the amortization of a much larger mask set (2 or more times as many masks as a smaller sensor), over many fewer sensors.

Then we have defects. Defects are not fixed per unit area, but more of a probability of occurrance. The probability function is not linear with size of the die that it will not have a defect that renders the die useless. The probability is more on the order of the area raised to a power higher than one.

Wafers are round. Sensors are rectangular. You can use a much higher percentage of a round object if you use smaller rectangles.

Defect probabilities are higher at the edge, rather than the center of a wafer. This also counts against big sensors.

An S35 (fullframe movie camera) sensor versus 35mm (fullframe SLR) has one third the area. Ignoring packaging costs which is probably rather close for both, the cost difference for the same pixel format from one to the next, and the same defect criteria, is probably atleast a factor of ten difference (something like the square of the area). This cost issue is often addressed by tolerating greater defects.

Going from an 35mm sensor to a medium format sensor probably requires going from a slightly tweaked CMOS process to a CCD process. CCD is much more expensive. Wafer sizes for CCD are much smaller than CMOS, which only exacerbates the round/rectangle problem from above.

The Dalsa sensor that is in the Hasselblad camera is probably a $10K sensor.

Please don't take this as any more than a brief intro to the issues, with numbers that are for the most part wild guesses based on a possibly limited recollection of my days in the test group at IBM Burlington, where I was involved with the testing of DRAM/SRAM, ASICs and CCDs. That was about ten years ago.

[J. Bernard Vallon]

Thanks. Good information. Its nice to know that the sensor really is more expensive to produce, and that the camera companies are not creating artificial demand.