Mysterium sensor

It's CMOS sumfun... says on the Red website

"4520X2540 pixels...DATA TO BURN At the heart of RED lives the 12-megapixel Mysterium™ CMOS sensor, Super 35mm sized, with unparalleled fidelity and flexibility."
 
Graeme, I believe 66dB is a shade under 11 stops- We've already heard third-party accounts that the Red IBC footage 'appeared' to be demonstrating lattitude as high as 12-15 stops (IIRC), so can I ask-- Is 66dB already being achieved? Is this figure your target "engineer's spec", or is this number already achievable, and likely to be exceeded?

Sorry to push for more details, but obviously this 66dB figure was quoted long before you had a full working sensor and demo images :) It's already entering new territories (varicam, etc. can't come close with a flat curve), but if you can eventually push the snr past say 78dB/13stops then you're really doing something very very significant! That's more than you can get from a lot of film stocks, especially when you consider that red will specialise in pulling detail from the deep shadows.
 
The only spec we stand on is >66db. Any other numbers are from people in the industry who have seen our footage and made their own observations. If we have spoken these words, it should have been qualified by "we have heard these numbers from industry experts". We have NOT measured the dynamic range yet. But it sure looks incredible to us as well as others.

Jim
 
Yes, with a good algorithm you will get superb colour resolution from a bayer CFA. You will get very good colour accuracy too.

Measuring dynamic range is tricky, and no two people give the same answer. I've attempted to measure the snr, and it does look indeed to be around or greater than 66db, but again, it's tricky to measure.

However, an snr of 66db does equate to 11 stops mathematically, but the perception of dynamic range is influenced by a great number of other factors, hence the comments from the expert observations.

All we at RED will say though, is >66db.

Graeme
 
Beyer pattern isn't perfect, moire can rear it's head, but it's very widely used .. It's good, it's proven, it's the way everyone would expect Red's CMOS to function. digiSLRs, etc. use it of course. Remember that this is 4K-- if you're going to zoom in to an individual pixel and look for beyer related problems then I'm sure you could point out where the interpolation doesn't work perfectly, etc., but it's *4K* .. is the lens 'really' resolving the full 4K? what codec are you converting to afterwards? are you printing to film, or projecting ..? These things are *dramatically* more impact than the subtle details of how beyer works, it's very low-order ... The pixels at 4K are so small, I feel that it's a non issue. To be honest, I feel that even pixel-level noise at 4K becomes less of an issue- the pixels are so small that your projector/filmprint/etc will struggle to resolve them anyway.

The majority of us will be down-converting to 2K or 1080p, and all of this comes moot.
 
There's never any need to see moiré unless you remove the OLPF. Remember even 3 chips cameras or foveon like cameras need an OLPF too to stop aliassing.

Indeed, with the high resolution we have, and as proven by the top end Nikon and Canon DSLRs, bayer CFA ruining picture quality is not something we have to unduly worry about.

I remember compression was thought of as evil. But once you embrace the fact of compression, work with it, and use it properly, it's not evil any more, but a useful tool. Once you learn how to use compression at it's best, bayer CFA at it's best etc, they cease to be problems, but open the doors to new solutions.

Graeme
 
Yes, I see now. I see great image quality and lower bit rates. I've seen the EOS 1Ds Mark II pictures, they're great. I'm just suprised that there is no noise.
 
Cheers for the reply Graeme!

And you're right about compression-- there's no doubt that (for example), a 720p uncompressed signal is vastly larger but completely inferior to your 4K wavelet signal...

Could you please clarify something for me? If the sensor is performing with an snr if ~66db, then is this the absolute maximum range that you can pull from the sensor? What I mean is, if a current camera is resolving approx 6 stops with a video/linear curve, ~7stops with low knee and cine type gamma, then the sensor would have to be operating with a snr of >42db .. despite measuring 6 stops linear .. is that right, or wrong ..?
 
sorry if my terms were very slighly muddled- by 6 stops "linear" I mean "6 stops in real-life shooting with a standard linear/video response curve", not a measure of the dynamic range of the clip's actual linear response/output.

hopefully you'll know what I'm getting at ;)
 
Last edited:
66db would show you the max dr you can pull from the sensor, but not the max perceived dr, as that's made up of perceptual factors.

Because other cameras don't offer raw, you can't do the same comparisons I'm afraid.

Graeme
 
True graham, but remember that the DVX rates approx 7 stops max (if you're being generous), while the andromeda mod pulls way more- I've seen 9.5 quoted ... It's this loss that's confusing me, how can panasonic have managed to lose all this lattitude? The sensor seems to be capable of at least 57dB. Obviously the 8bit output format can't hack it, but surely the curves should have done better with that signal than 6-7 recorded stops. This is what I can't get my head around :) And that's why the 66dB Red measurement sounds really conservative when you're already pulling images that pro DOP's feel demonstrate at 'least' 12 stops (even if you don't have the same problem of cramming into a tiny bitdepth). Cheers for listening!
 
Jannard said:
Did you ever see the colors of the Canon 1Ds MkII? Single CMOS bayer sensor.
Click to expand...
Holy Moly. If the Red cam has anywhere near the same quality as the 1Ds MK11 Im going to jump off a bridge.

I just did a photoshoot for my fashion label a couple of months ago using the MK11, samples attached. Its a huge image, retaining the original dimensions so make sure to have it zoomed in. If this is the quality were talking then.. Ummm... Speechless.

other.jpg


And no its not retouched... So dont worry about the loose strands of hair flying everywhere.
 
Last edited:
Oh yikes.. I didnt know it was going to be that huge on the page. I hope I didnt stuff up the forum page.
 
"If the Red cam has anywhere near the same quality as the 1Ds MK11 Im going to jump off a bridge."

be very careful of what you ask for - you just might get it !!

so do you have a bridge in mind ?
 
You must log in or register to reply here.
Back
Top