Next generation MF sensors

[Daniel Browning]

First, note that Jim Jannard and RED have not yet delivered anything beyond cinema 35mm format, with the next step up probably to 24x36mm format, so still catching up with Canon, Nikon and Sony on CMOS size

Excellent point. It doesn't make sense to compare theoretical future products with products that have actually shipped. One of the things I am guessing that we'll hear from RED in 4 days is that the Medium Format backs are significantly delayed (18-24 more months), but I hope I'm wrong.

(and for all I know, someone like Sony or Dalsa is actually making these sensors for RED.)

RED has been pretty specific that they have their own (obviously fabless) sensor design team. Perhaps they are lying, but some industry insiders (like Eric Fossum) have commented that they're doing genuine and remarkable image sensor design.

Let us see what he and RED actually deliver in medium format sizes, and at what price.

Agreed.

(The dude owns a 1200/5.6)

FWIW, he sold it last year at a tidy profit. (It was not long after that when B&H put a 1200mm f/5.6 on their web site.)

[stevesanacore]

He hasn't. Words are cheap.

Time will tell. I just don't really understand his motives for moving beyond even the S35 (APS) sized sensor for movie work. That is his niche and he's doing well. He must believe the larger sensors are achievable to make those claims. But he will be overwhelmed with orders for Scarlet and Epic very soon, and don't see that he'll have any time or motivation to pursue a 645 camera.

[georgl]

To make my cynical "GREEN"-comparison a little bit more clear and to get rid of some of the myths regarding RED:

RED is a prosumer camcorder with a unique market position - it uses a Super35-sized (24mm wide) single sensor and a common bayonet usually known from professional systems but it's technological quality is closer to prosumer camcorders from Pansonic or Sony - although they use an entirely different technical approach to achieve this quality.
It's like a MFDB at a much lower price but with ergonomic/technical aspects of consumer-cameras - it doesn't fit in the scheme

Professional "cine-style"-cameras cost much more (50-150k$) due to extreme build-quality, service, optical viewfinders (ARRI), highly selected sensors (Panavision takes only one of 600 sensors from Sony!!!) but the main difference is their ability to record real, uncompressed RAW (compression works well for final images but not as an automatized process within the camera) created out of oversampled/"non-bayered" signals (that's the clue about "4k", it's interpolated - there is no current digital system capable of more than real 2k - photosites on sensor = pixels in the final images isn't appropiate in professional video). The RED uses strong internal compression (about 1:10) to handle the high data-rates.

RED uses an unknown CMOS-sensor, but none of it's technology or capabilities (resolution, color rendition, DR...) is surprising or unique - it's a good prosumer tool worth it's price to some people - but the only amazing thing is it's marketing. RED doesn't manufacture one single piece of equipment, their HQ is a distribution/marketing center.

Don't mix up technical with artistic quality, some of the greatest photographers used shitty equipment. In fact, just looking at Mr. Soderberghs two last movies makes clear why RED shouldn't be used for the big screen... Peter Jackson although decided against RED for his new film and choose 35mm instead.

The professional high-end-cinematography is quite stable and well-served by two (+ some tiny companies) legendary, engineer-driven companies giving cinematographers high-end-equipment which has little to do with the mass-market but is a nearly perfect tool for their demanding work. No marketing-talk, long lifetime, open prototype testing over years to get things right, choosing the right, not the cheap way...
Let's hope demanding photographers still get the same chance and aren't forced to use press-cameras in the future...

[eronald]

He hasn't. Words are cheap.

Actually this type of circuit can be designed with low manpower, like RAM, because basically you just need to design a cell and some amps.
I think I could probably still do such a design myself if I had to. It might take me a couple of years though

Edmund

[Khun_K]

may be Canon or Sony can be the next medoum format back mannufacurer by stiching 2 5D-II sensor togeher and make the exisintg medium format camera flexibility we all talked about?

[Daniel Browning]

RED is a prosumer camcorder...

Interesting. That's the first time I've ever seen someone take that position.

...it's technological quality is closer to prosumer camcorders from Pansonic or Sony

May I ask you to expound on these technological qualities in more detail? When I think of ways in which they compare and contrast, here is what I identify:

• Camcorder sensor size: Mostly 1/4", 1/3", 1/2", and maybe 2/3" (depending on your idea of prosumer). That's 4mm to 10mm wide. It does not allow for very much control over DOF, even at f/1.6 on 1/3". (DigiPrimes on 2/3" is sufficent DOF control for most movie work, but they probably wouldn't consider that prosumer).
• RED ONE Sensor size:24.4x12.2mm. That allows more control over depth of field as well as the use of a wider variety of lenses (some that are decidedly not prosumer, like the Master Primes).
• Camcorder resolution: the best prosumer camcorders (and even the $250,000 F35) only have 1920x1080 files. If you have the bare minimum amount of anti-aliasing (2.5 pixels per line pair, though I much prefer at least 3), that's only 432 lp/PH and 768 lp/PW. Of course, many of the prosumer camcorders allow their images to be contaminated by aliasing artifacts; if you count those artifacts toward resolution, you can get up to the maximum of 960 lp/PW. (Granted, most viewers actually like and prefer aliasing, but I still don't think it should be forced on everyone.)
• RED ONE Resolution:  144 l/mm *measured* resolution. That's 1750 lp/PW, over twice as much as an anti-aliased 1080p camera (768 lp/PW). Look at the proof: http://www.cinematography.net/natress-red-res.html
• Camcorder compression: There's a bit of variety here, from 1440x1080 8-bit 4:2:0 MPEG-2 25 Mbps (3 MB/s) to 1920x1080 h.264 8-bit 4:2:0 17 Mbps (2 MB/s). The most expensive cinema/video cameras use HDCAM-SR, which is 1920x1080 10-bit 4:4:4 440 Mbps (55 MB/s).
• RED ONE compression: 4480x1920 raw wavelet-compressed 12-bit 42 MB/s. The number of pixels is over four times higher, bit depth is four times higher, and the bitrate is an order of magnitude higher. Furthermore, the quality is even superior to HDCAM-SR, without even considering the resolution advantage (i.e. downsampled to 1920x1080).
• Color accuracy: Less metamerism and better Delta-E measurements than any 3-chip camera.
• Frame rate: No camcorder that I know of can hit 120 FPS at 2K.
• Luma and chroma aliasing. Look at the zone plate above for proof that the RED ONE has low aliasing. The Genesis/F35 have far worse aliasing, it's bad enough that high end cinematographers notice and dislike the artifacts.

There are certainly some ways in which a prosumer camcorder is superior to the RED ONE, but I think the above difference place the RED ONE firmly in the realm of digital cinema.

Professional "cine-style"-cameras cost much more (50-150k$) [...], the main difference is their ability to record real, uncompressed RAW

Can you name even one full length feature film that was shot entirely on real, uncompressed raw? There might have been one or two (I've never heard of them), out of tens of thousands. Even 30-second commercials have a hard time justifying the astronomical costs associated with uncompressed data capture (over 2,600 Mbps for DALSA), which is one of the main reasons why DALSA's camera division died. A lot of feature films use highly processed non-RAW formats like HDCAM-SR, which already has the color matrix, black clip, noise reduction, tone curve, saturation, contrast, and other image modifications applied to the data and converted to 10-bit log. RED ONE, on the other hand, only has wavelet compression.

(that's the clue about "4k", it's interpolated - there is no current digital system capable of more than real 2k - photosites on sensor = pixels in the final images isn't appropiate in professional video).

I kindly disagree. RED ONE measures 3.2K just fine, and that is far higher than 2K. See the proof for yourself:

http://www.cinematography.net/natress-red-res.html

Have you seen the zone plates of other cameras like the F35? Not only is their luma and chroma resolution far lower, but they have aliasing at lower spatial frequencies as well.

The RED uses strong internal compression (about 1:10) to handle the high data-rates.

HDCAM-SR uses 55 MB/s for a paltry 2 MP 10-bit, wheras RED ONE uses far superior technology to get 9 MP 12-bit into 44 MB/s. By the time you downsample down to 2 MP 12-bit for comparison with HDCAM-SR, you will not see *any* compression artifacts (if indeed it was possible to see any at 9 MP; most people do not).

RED uses an unknown CMOS-sensor, but none of it's technology or capabilities (resolution, color rendition, DR...) is surprising or unique

Funny, because Eric Fossum said RED ONE's capabilities were quite remarkable and indicated the future direction of sensor technology. Please don't be offended that I take his word over yours.

Peter Jackson although decided against RED for his new film and choose 35mm instead.

Well, some other directors/DPs found that it was sufficient:

http://www.red.com/shot_on_red/

Plus, I think their reels are just excellent:

http://red.cachefly.net/redreel/RedReel_h264_720.mov

http://red.cachefly.net/RedReel500.mov

Let's hope demanding photographers [...] aren't forced to use press-cameras in the future...

Agreed.

[aaanorton]

Excellent point. It doesn't make sense to compare theoretical future products with products that have actually shipped. One of the things I am guessing that we'll hear from RED in 4 days is that the Medium Format backs are significantly delayed (18-24 more months), but I hope I'm wrong.

I'm gonna say 6 - 12 months. Just because I choose to. But I'll wager you, Daniel, 1 PayPal buck to this end.
By the way, I was just on my way to check out your sensor comparison test over at REDUSER. Thanks for posting it!

Don't mix up technical with artistic quality, some of the greatest photographers used shitty equipment.
Peter Jackson although decided against RED for his new film and choose 35mm instead.

No offense, but doesn't either of these sentences negate the integrity of the other?

Thanks,
-c

[Daniel Browning]

Thanks.

I'm gonna say 6 - 12 months. Just because I choose to. But I'll wager you, Daniel, 1 PayPal buck to this end.

You're on.

[stevesanacore]

To make my cynical "GREEN"-comparison a little bit more clear and to get rid of some of the myths regarding RED:

RED is a prosumer camcorder with a unique market position - it uses a Super35-sized (24mm wide) single sensor and a common bayonet usually known from professional systems but it's technological quality is closer to prosumer camcorders from Pansonic or Sony - although they use an entirely different technical approach to achieve this quality.
It's like a MFDB at a much lower price but with ergonomic/technical aspects of consumer-cameras - it doesn't fit in the scheme

Professional "cine-style"-cameras cost much more (50-150k$) due to extreme build-quality, service, optical viewfinders (ARRI), highly selected sensors (Panavision takes only one of 600 sensors from Sony!!!) but the main difference is their ability to record real, uncompressed RAW (compression works well for final images but not as an automatized process within the camera) created out of oversampled/"non-bayered" signals (that's the clue about "4k", it's interpolated - there is no current digital system capable of more than real 2k - photosites on sensor = pixels in the final images isn't appropiate in professional video). The RED uses strong internal compression (about 1:10) to handle the high data-rates.

RED uses an unknown CMOS-sensor, but none of it's technology or capabilities (resolution, color rendition, DR...) is surprising or unique - it's a good prosumer tool worth it's price to some people - but the only amazing thing is it's marketing. RED doesn't manufacture one single piece of equipment, their HQ is a distribution/marketing center.

Don't mix up technical with artistic quality, some of the greatest photographers used shitty equipment. In fact, just looking at Mr. Soderberghs two last movies makes clear why RED shouldn't be used for the big screen... Peter Jackson although decided against RED for his new film and choose 35mm instead.

The professional high-end-cinematography is quite stable and well-served by two (+ some tiny companies) legendary, engineer-driven companies giving cinematographers high-end-equipment which has little to do with the mass-market but is a nearly perfect tool for their demanding work. No marketing-talk, long lifetime, open prototype testing over years to get things right, choosing the right, not the cheap way...
Let's hope demanding photographers still get the same chance and aren't forced to use press-cameras in the future...

I am not sure how to politely say this but it doesn't sound like you have any clue about what you are saying. Next time RED is at a trade show such as NAB, stop by and watch their 4K projected sample reel all taken from stock footage and other real DPs shooting with their camera and you will be blown away from the quality. The RED certainly had issues, but it's being used quite often by many filmmakers with outstanding results. 35MM movie film, being very high quality color negative stock, will be hard to beat for latitude and detail and many DPs simply refuse to shoot digital out of insecurity.

RED's new models Epic and Scarlet  with delivery dates to be announced in a few days will probably settle any and all arguments about both their technical and image quality.

[BJL]

Actually this type of circuit can be designed with low manpower, like RAM, because basically you just need to design a cell and some amps.

... and then you have to be able to read them all out fast enough without adding too much noise, which too-high clock rates can do to analog signals. After all, read speed limitations have held back the frame rate-pixel count combinations of DSLR's for years, being the most likely reason for even many top-of-the line DSLRs offering lower frame rates than even many moderately priced film SLR's. That used to be true even for the models intended specifically for high frame rate work, in particular the 5FPS D1H, and is still true with the mere 5FPS of the 1DsMkIII and D3X.

[georgl]

Like I said, the RED doesn't fit the scheme, therefore any comparison with other systems is a bit tricky. I think it's great that a halfway affordable 35mm-sized digital camcorder is available, many cinematographers (especially students, low-budget-work...) don't have to struggle with strange image converters to simulate 35mm-DoF anymore and REDRAW gives them more flexibility.

But many people (especially those with strong affinity to the internet and knowing RED only from there) have high expectations, they expect a "revolution" making previous technology seem outdated and obsolete because "RED has a professional camcorder with 4 times higher resolution (4k = 4 times 2k) and a ten times lower price tag than other camcorders" so when the first "still-photography"-REDs arrive we don't have to bother with overpriced, low-performance MFDBs, Nikons, Leicas or Canons because a billiioniare invested billions in a product to save the photographic world from greedy global-players...

This is simply not the case with the R1! It's a different concept but shows similar qualities (some aspects are better, some worse) to existing solutions in it's price-range - none of it's technology is unique, it's approach is.
What's the exact definition of "prosumer" vs. "professional" in this market? I fear there is none, but a few technical facts:

- it's sensor is a bayer-filtered CMOS-design with 5,4µm-pixel-pitch - nothing special about it (size, speed, sensitivity), just like many other designs on the market - go to DALSA or Cypress and they make you one to your needs.

- it's not capable of more than 2k (which is nearly 1080p size-wise) without interpolation, just like other cameras within it's price-range (in fact, there are only 4k-prototypes available)

- it's not capable of more than 60fps with HD (2k is using a small portion of the sensor and it's interpolated "HD" just like the tiny 2/3"-single-sensor cameras from P&S or Ikonoskop)

- it's build-quality is comparable to other cameras in it's price-range

So while it offers advantages over many 2/3"-cameras within it's price-range for certain applications (DoF, 60fps progressive) it's not a professional "digital cinema"-camera:

- no uncompressed output

- no optical viewfinder

- no professional build-quality (have you ever seen the skilled craftsmen of Zeiss and ARRI doing their job?)

In fact, the "digital cinema"-cameras from ARRI/Panavision aren't even supposed to be used for cinema, but mainly TV. I simply wanted to get the name "Peter Jackson" out of this game for this discussion. I'm worried that some great filmmakers "fell" for 1080p-video (even Cameron and Fincher) and constantly making weird comments about their technical choices.

I'm sure it's a nice camera and in the right hands it's capable of nice results (although I don't like the color rendition and limited DR of ANY digital camcorder, even the D21 or Genesis), but that's also the case with other digital cameras for a similar price - "4k" and all this "revolution"-talk is just a marketing phrase. They sell their "own" lenses instead of "overpriced" Cooke/Zeiss/Panavision (although they're designed and manufactured at different standards) - that's not a revolution, that's a loss of professional quality standards (when Zeiss/Cooke/Panavision were forced to compete in price with reducing quality - happened in other industries, too).

RED = 150k$ worth professional "digital cinema"-camera like ARRI/Panavision/F35 is simply a wrong equation - different standards, qualities, technical aspects - you get what you pay for. It's like comparing the D3X to the A900 or the 5dMk2 to the 1dsIII just because they're technical data sounds comparable at first sight. It's great that those "cheaper" solutions exist but we have to be careful that we don't ruin our quality standards - that's my primary fear.

We're a bit off-topic already, but the point is: based on the technical background of RED and the performance of the R1 we cannot assume any major "revolution" in the future from this company.

[Graeme Nattress]

RED doesn't manufacture one single piece of equipment, their HQ is a distribution/marketing center.

Which just shows how little you know...

HQ does indeed have a warehouse for distribution. Marketing at RED is a single individual (Ted) and he doesn't work out of the HQ, but instead travels the world talking about RED, demoing stuff, handling the REDucation classes etc.

Also in the HQ is a small 4k cinema for showing demo reels and our own internal testing.

Most of the offices are electrical / electronic / mechanical design. There are also testing areas for cameras and sensors and lenses. There is a small R&D / design area.

There are manufacturing facilities there also - rapid prototyping for design visualization and CNC equipment for shot run or prototyping.

There is also a shop.

And there are the tech support people, right there alongside the engineers so that they can get the answers to any questions beyond their immediate expertise.

Graeme

[Daniel Browning]

- it's not capable of more than 2k (which is nearly 1080p size-wise) without interpolation, just like other cameras within it's price-range (in fact, there are only 4k-prototypes available)

All that matters is real world results. In the real world, R1 hits 3.2K resolution, as proven in the measurements linked above. That is far greater than 2K. Your emphasis on pre-interpolated "resolution" is mistaken.