Jump to content


Photo

Inherent limits


  • Please log in to reply
23 replies to this topic

#1 Paul Bruening

Paul Bruening

    (deceased)

  • Sustaining Members
  • 2858 posts
  • Producer
  • Oxford, Mississippi

Posted 06 November 2006 - 09:54 PM

Hello all,

I'm a little boggled about some aspects of a camera like Red.

How can they put a CMOS of any quality in a camera at that price? CMOSs are graded on their proximity to perfection. The ones with very few duff pixels go for a whole heck of a lot more than their ... red-headed step siblings (please forgive the pun).

Another thing: if the CMOS is Bayer patterened, is it really 4K? Don't you have to divide that CMOS real estate by the number of colors and or number of RAW files?

What can they possibly be storing that high of a data flow on during taking? Could two mirrored 10,000 RPM SATA drives handle that data flow? (Not saying that that's what they're using. Just speculating).

What about seeing what you're shooting. Last I heard, No one was making a 4K monitor since IBM dumped their 4K LCD works? (that's who Veiwsonic was buying their screens from when they knocked out a 4K monitor).

Does anyone make a CMOS with that much latitude? These guys don't make their own. They have to buy it, right?

Maybe I'm just so out of the loop I don't realize these problems have already been solved by clever tech folks somewhere. I don't know.

Paul
  • 0

#2 Michael Collier

Michael Collier
  • Basic Members
  • PipPipPipPip
  • 1262 posts
  • Gaffer
  • Los Angeles, CA

Posted 06 November 2006 - 10:23 PM

Hello all,

I'm a little boggled about some aspects of a camera like Red.

How can they put a CMOS of any quality in a camera at that price? CMOSs are graded on their proximity to perfection. The ones with very few duff pixels go for a whole heck of a lot more than their ... red-headed step siblings (please forgive the pun).

Another thing: if the CMOS is Bayer patterened, is it really 4K? Don't you have to divide that CMOS real estate by the number of colors and or number of RAW files?

What can they possibly be storing that high of a data flow on during taking? Could two mirrored 10,000 RPM SATA drives handle that data flow? (Not saying that that's what they're using. Just speculating).

What about seeing what you're shooting. Last I heard, No one was making a 4K monitor since IBM dumped their 4K LCD works? (that's who Veiwsonic was buying their screens from when they knocked out a 4K monitor).

Does anyone make a CMOS with that much latitude? These guys don't make their own. They have to buy it, right?

Maybe I'm just so out of the loop I don't realize these problems have already been solved by clever tech folks somewhere. I don't know.

Paul


I thought CCDs were graded by closeness to perfection, and CMOS chips are almost always perfect. they use the same production technology that virtually every chip is based on, which is why they are cheaper than CCDS that must be made at special foundrys. If there was a percent of error in most chips, then computers and every electronic device would be MUCH more expensive (and more prone to DOA)

Yes, if its bayer its not truely 4K, but its not a divisor thing, its much more complex than that. There are groups of 4 pixels that make up one color, 2 green, 1 red and 1 blue, but a computer can up-res each one, and using data from ajoining pixels of different colors, create an image close to the resolution of the native bayer chip. Its difficult to explain, but estimation and interpolation plays a huge role. True, each pixel does not have accurate data for all 3 colors, but using advanced algorythms it can come close to estimating what the other two colors are. This gives you 4K in areas that don't have a hard edge. But have high-frequency detail, or a hard edge with contrasting colors and those areas would look a little soft. Its not all bad. Imagine the crazy effects of a tweed jacket on a 3chip CCD. the errors of the bayer algorythm actually smooths that out and keeps the moore patterns from being present. Its not true 4K, but its also not 1K or 1.33K either. True resolution at that point depends on what your shooting. It would look better than a 2K chip in most situations I imagine.

They can get chips with that kind of latitude and resolution because they have contracted with a chip manufacturer to create them. Their chips are proprietary and they are keeping hush on how they get such latitude. Their camera (and most of their R&D costs) have gone into the chip. Its a camera that houses a sensor and a path to storage, thats it (all proccessing is done after-the-fact) which is partly why it can be so cheap. No complex FPGA chips that have to compute bayer, white ballance, matrix and setup in real-time.

Storage is simple. Last I heard it was a flash device, which to my knowledge can support throughput of 3MB/sec or somewhere around there. Gang enough of those together in a Raid assembly and you get the throughput your looking for.

As for seeing, who knows. They say they have a OLED viewfinder in the works, but I haven't heard if its 4K or not. Seems like it may work like the HDV cameras that have SD viewfinders, where you have 'focus assist' that just zooms into the center of shot (the center 480p of the image that can be shown in native resolution) so you can see well enough to focus, then click that off and frame with a low-res viewfinder.

All of these are really acedemic questions, because the proof is not yet in the pudding. Its tantilizing, promising and a bit cool, but nobody has used this on a feature film and come back to talk about it. Once someone with ,ASC tacked on their name tells me its fit for primetime, I will believe it. But until then given their claims it will find some use in something, if not major motion picture.

And no, your not out of the loop, they just have vested interest in making sure the loop never evolves past their front doors (or sometimes windows)
  • 0

#3 jan von krogh

jan von krogh
  • Basic Members
  • PipPipPipPip
  • 454 posts
  • Producer

Posted 06 November 2006 - 10:35 PM

How can they put a CMOS of any quality in a camera at that price?

wrong question - correct question would be - what happens if a monopoly for chemical (arri) and digital (sony) top grade cameras has over one decade to dictate the market.

CMOSs are graded on their proximity to perfection. The ones with very few duff pixels go for a whole heck of a lot more than their ... red-headed step siblings (please forgive the pun).

as you might have noticed, >10 MP CMOS sensor cameras can be bought for <1000-3000$ from nikon, canon, sony...

Another thing: if the CMOS is Bayer patterened, is it really 4K? Don't you have to divide that CMOS real estate by the number of colors and or number of RAW files?

CMOS design and bayer-filtering (which has evolved quite a bit over mr. bayers first algorithms of the 70ties) have come a long way in the recent 30 years. simply put: a 12MP / 5k physical res chip (like reds sensor) easily handles 4k.

What can they possibly be storing that high of a data flow on during taking? Could two mirrored 10,000 RPM SATA drives handle that data flow? (Not saying that that's what they're using. Just speculating).

depending what you buy, taht would be:
1st) redflash. pure RAM memory. rather costy, but rather fast as well.
2nd) disk-ARRAYS
3hrd) redcode - datareduced, but waveletbasing quality onto a single drive.
4th) HD_SDI, dual as it seems. pick your flavour of 1080 recording solutions. good enough for james cameron, george lucas, mel gibson, david fincher, michael mann, certainly not good enough to some of the regular trolls in this forum.

What about seeing what you're shooting. Last I heard, No one was making a 4K monitor since IBM dumped their 4K LCD works? (that's who Veiwsonic was buying their screens from when they knocked out a 4K monitor).

correct approach to monitor in post would be:
http://www.dcinemato...aspx?newsID=648
we here will probably stay with 2K projection.
on set the ibm panels you´ve mentioned could be good choices, but indeed there ae several 4k panel manufacturers - not only for radiology, millitary etc.
however, if you come from a film background: then you should know, that nobody needs on-set monitoring as the dp & focuspuller probably have it always right.
without sarcasm: you can even use a 1080p display (like the one i am writing this post on right now) and zoom in, if you want to benefit from the digital era instead of waiting for dailies.

Does anyone make a CMOS with that much latitude? These guys don't make their own. They have to buy it, right?

Yes, no, no.
The sensor is RED design, if they outsource the production (i would do that) or do it inhouse is a decision jim jannard has to take. however, its red intellectual property.

Maybe I'm just so out of the loop I don't realize these problems have already been solved by clever tech folks somewhere. I don't know.

there are lots of people who are in the loop and still refuse to understand, so relax.
  • 0

#4 jan von krogh

jan von krogh
  • Basic Members
  • PipPipPipPip
  • 454 posts
  • Producer

Posted 06 November 2006 - 10:51 PM

Yes, if its bayer its not truely 4K, but its not a divisor thing, its much more complex than that. There are groups of 4 pixels that make up one color, 2 green, 1 red and 1 blue, but a computer can up-res each one, and using data from ajoining pixels of different colors, create an image close to the resolution of the native bayer chip. Its difficult to explain, but estimation and interpolation plays a huge role.

please once and for all - can we restrain from using wikipedia and photopro today level generalisations?
a) how you weight the GRB-balance is choice of the design team, and red had their own sensor designed.
B) you might have hear of pattern diffusion and other rather high-tech appoaches of the early 90ties. as everybody doesn´t know, we might also suppose that a totally stupid design team -might- have opted to not use static patterns?

c) as you describe:

True, each pixel does not have accurate data for all 3 colors, but using advanced algorythms it can come close to estimating what the other two colors are.

... and one should mention as well, that FILM has usually 3 layers which usually BELOW each other which usually leads to, ok... you have got the idea i suppose.

This gives you 4K in areas that don't have a hard edge. But have high-frequency detail, or a hard edge with contrasting colors and those areas would look a little soft.

wrong. this only depends on HOW you tune the dealiasing. you can go for sharp, you can go for velvet, you can go for... thats design decision.

They can get chips with that kind of latitude and resolution because they have contracted with a chip manufacturer to create them. Their chips are proprietary and they are keeping hush on how they get such latitude. Their camera (and most of their R&D costs) have gone into the chip. Its a camera that houses a sensor and a path to storage, thats it (all proccessing is done after-the-fact) which is partly why it can be so cheap. No complex FPGA chips that have to compute bayer, white ballance, matrix and setup in real-time.

the red has FPGA which do bayer etc in RT, btw how do you suppose the RT HD_SDI should work without?
what you misunderstand here is that there is redcine, an application which allows to customize parameters when decoding the raw later on - if you wish so. we, that is for sure, will often use the hd-sdi and go directly to hdcam or hdcam sr.

All of these are really acedemic questions, because the proof is not yet in the pudding. Its tantilizing, promising and a bit cool, but nobody has used this on a feature film and come back to talk about it. Once someone with ,ASC tacked on their name tells me its fit for primetime, I will believe it. But until then given their claims it will find some use in something, if not major motion picture.

David Stump, Warner Bros, Peter Jackson, Otto Nemenz, David Fincher......
http://www.cinematog...mp;#entry136287

And no, your not out of the loop, they just have vested interest in making sure the loop never evolves past their front doors (or sometimes windows)

they plan prototypes ~30 days, delivery 90-120 days for the cameras for the first 1000s orders. if they can do that, my 2 reds are booked out together with some of our zeiss & angenieuxs...
  • 0

#5 Chris Kenny

Chris Kenny
  • Basic Members
  • PipPipPipPip
  • 264 posts
  • Other

Posted 06 November 2006 - 11:15 PM

Hello all,

I'm a little boggled about some aspects of a camera like Red.

How can they put a CMOS of any quality in a camera at that price? CMOSs are graded on their proximity to perfection. The ones with very few duff pixels go for a whole heck of a lot more than their ... red-headed step siblings (please forgive the pun).


The chip is basically the same as what you can find in a digital SLR camera in the $2-5K range, but it can handle more frames per second. The reject rate on CMOS chips isn't particularly high, as far as I'm aware. It may be higher for a chip that has to operate as such fast speeds; I'm not familiar with the technical issues in any detail.

Another thing: if the CMOS is Bayer patterened, is it really 4K? Don't you have to divide that CMOS real estate by the number of colors and or number of RAW files?


It's not that bad. Modern de-mosaicing algorithms can recover something like 70% of 'native' resolution. Red might do a little better. Real horizontal resolution will probably be around 3K.

What can they possibly be storing that high of a data flow on during taking? Could two mirrored 10,000 RPM SATA drives handle that data flow? (Not saying that that's what they're using. Just speculating).


They're using a compression algorithm called REDCODE at about 12:1, to squeeze 4K bayer down to 27.5 MB/s, which is then recorded to a single 2.5" drive (they're also selling a flash magazine). They've reported the codec is essentially visually lossless in their testing, and the sample frame they've posted which went through the codec looks very good. REDCODE is a wavelet compression algorithm. Wavelet compression algorithms don't exhibit most of the objectionable artifacts that the more common DCT codecs (like JPEG, MiniDI, DVCPRO HD, MPEG-2, etc.) do. If you want to play around with wavelet compression, Photoshop supports JPEG 2000, which is a wavelet algorithm.

Wavelet compression shouldn't cause problems with color correction, but it'll be interesting to see how well it holds up for pulling keys.

Also, remember, the camera shoots RAW at 4K, and performs a 12-bit lin -> 10 bit log conversion. So they're basically compressing a single 10-bit monochrome bayer-pattern image for each frame; they've not compressing three channels of 12-bit color.

You can shoot uncompressed RAW as well, but this requires an external RAID (probably 8-10 3.5" drives, if you want to be able to use their full capacity rather than just recording on the "fast" parts of the disks).

What about seeing what you're shooting. Last I heard, No one was making a 4K monitor since IBM dumped their 4K LCD works? (that's who Veiwsonic was buying their screens from when they knocked out a 4K monitor).


There's probably no really compelling use case for actually monitoring at 4K on set. The camera can output anything up to 2K 4:4:4 RGB via HD-SDI (and possibly also via HDMI) with a custom LUT applied while simultaneously recording 4K RAW. Many people will probably just use 1080p monitors.

Does anyone make a CMOS with that much latitude? These guys don't make their own. They have to buy it, right?


It's not an off-the-shelf part. They don't physically make it, I'm sure; very few companies actually do that. They were involved with its design, however. How much was done in-house and how much was outsourced in unclear, but isn't really important; they paid for the R&D, they own exclusive rights.
  • 0

#6 Max Jacoby

Max Jacoby
  • Sustaining Members
  • 2955 posts
  • Other

Posted 07 November 2006 - 05:49 AM

David Stump, Warner Bros, Peter Jackson, Otto Nemenz, David Fincher......

I hope you are not suggesting that these people are endorsing the Red. David Stump will be testing it and these other people might have reservations, but that doesn't mean in any way that they are already planning to shoot their films on it. These people are all rich enough to buy a Red as you and I would buy a ? 40 digital still camera, so I wouldn't read anything into their reservations.
  • 0

#7 jan von krogh

jan von krogh
  • Basic Members
  • PipPipPipPip
  • 454 posts
  • Producer

Posted 07 November 2006 - 05:58 AM

I hope you are not suggesting that these people are endorsing the Red.

certainly not. they are still waiting for your approval, mr. jacoby,

David Stump will be testing it and these other people might have reservations,

yeah, they are stupid....

but that doesn't mean in any way that they are already planning to shoot their films on it.

also, different from young mr. jacoby here, these people have so much time to waste...

These people are all rich enough to buy a Red as you and I would buy a ? 40 digital still camera, so I wouldn't read anything into their reservations.

hey mr. jacoby did you .ever. had 17.500 spare so far? for two cams that would be 35.000?
  • 0

#8 jan von krogh

jan von krogh
  • Basic Members
  • PipPipPipPip
  • 454 posts
  • Producer

Posted 07 November 2006 - 06:11 AM

These people are all rich enough to buy a Red as you and I would buy a ? 40 digital still camera, so I wouldn't read anything into their reservations.


ah, i forgot to mention - if red confirms the delivery date, we will have 1 or 2 arri bl overstock here.

i would sell the BL IIcs for 1500/2500 and the BL IIbs for 1250/2000.

interested buddy?
  • 0

#9 Max Jacoby

Max Jacoby
  • Sustaining Members
  • 2955 posts
  • Other

Posted 07 November 2006 - 06:27 AM

Wow Jan

For someone who likes to avdocate professional language your post just exposed you as the hypocrite that you really are. You are obviously over-enthousiastic about the Red and I'm afraid your post is devoid of logical arguments.

As a potential customer your biased and insulting posts do not entice me to ever consider renting any equipment from you and your company.
  • 0

#10 Gavin Greenwalt

Gavin Greenwalt
  • Basic Members
  • PipPipPip
  • 225 posts
  • Other

Posted 07 November 2006 - 06:31 AM

1) certainly not. they are still waiting for your approval, mr. jacoby,
2) yeah, they are stupid....
3) also, different from young mr. jacoby here, these people have so much time to waste...
4) hey mr. jacoby did you .ever. had 17.500 spare so far? for two cams that would be 35.000?


1) Out of line
2) He wasn't implying anything of the sort
3) What the hell?
4) He didn't say he did. He said they do... and they do.

I would wager that Mr Stump is being payed to come into the Red laboratories. He was also payed by FXguidePHD to shoot green screen material for them... that's how a professional DP makes a living... by shooting footage for clients.

Just because people have "talked" to red doesn't mean they've signed anything, endorsed anything or even seen any footage.

Within the limited amount of knowledge we have I would be just as accurate to assert:

<Wild Speculation and not actual quotes>
David Stump said it was "An asshat toy."
Peter Jackson said "I've seen better pictures with my lens cap on."
Warner Brothers immediately signed a contract forbidding its use on any of their productions
and
Fincher just fainted from the stench.

</speculation>

I'm not suggesting these are their opinions, all I'm saying is. Don't put words in people's mouths. Especially when they can't easily defend themselves.
  • 0

#11 Stephen Williams

Stephen Williams
  • Sustaining Members
  • 4708 posts
  • Cinematographer
  • Europe

Posted 07 November 2006 - 08:04 AM

There's probably no really compelling use case for actually monitoring at 4K on set.


Chris,

Depends if you want to check for focus or not.

Stephen
  • 0

#12 Stephen Williams

Stephen Williams
  • Sustaining Members
  • 4708 posts
  • Cinematographer
  • Europe

Posted 07 November 2006 - 08:19 AM

ah, i forgot to mention - if red confirms the delivery date, we will have 1 or 2 arri bl overstock here.

i would sell the BL IIcs for 1500/2500 and the BL IIbs for 1250/2000.

interested buddy?


Jan,

Are you selling Arri BL's or Arri II's or don't you know the difference?

If they are BL's, I will take them off your hands at that price, when can I check them out!
If they are II b/c you will be lucky do get that much.

FWIW I saw Max's award winning film last Friday, 35mm Anamorphic, it kicked ass. I had a round trip of over 500km / 300miles well worth the journey.

Stephen
  • 0

#13 Nate Downes

Nate Downes
  • Basic Members
  • PipPipPipPip
  • 1638 posts
  • Florida, USA

Posted 07 November 2006 - 08:21 AM

Jan,

Are you selling Arri BL's or Arri II's or don't you know the difference?

If they are BL's, I will take them off your hands at that price, when can I check them out!
If they are II b/c you will be lucky do get that much.

Stephen


Um... actually there was a BLII with cs motor on the market. I'm debating on buying one, torn between it, a Mitchell BNCR and a Kinor 35H.
  • 0

#14 Phil Rhodes

Phil Rhodes
  • Sustaining Members
  • 11937 posts
  • Other

Posted 07 November 2006 - 09:58 AM

Hi,

I think what's important to realise here is that they haven't broken any "inherent limits." They haven't really done anything terribly clever; it's a camera head. No recorder, no lens, no support gear, it's just a camera head and there are dozens of those available already.

So the answer to the entirely reasonable question "what have they done to overcome these factors" is nothing at all.

Phil
  • 0

#15 Chris Kenny

Chris Kenny
  • Basic Members
  • PipPipPipPip
  • 264 posts
  • Other

Posted 07 November 2006 - 10:27 AM

Chris,

Depends if you want to check for focus or not.

Stephen


Well, we don't actually know how easy or hard it might be to check 4K focus on a 1080p screen. To the best of my knowledge, nobody has ever done it; I don't believe the Origin, the only other 4K camera in the world, has live 1080p output.

Red is promising a pretty impressive focus assist; hopefully that will solve the problem either way.

Hi,

I think what's important to realise here is that they haven't broken any "inherent limits." They haven't really done anything terribly clever; it's a camera head. No recorder, no lens, no support gear, it's just a camera head and there are dozens of those available already.

So the answer to the entirely reasonable question "what have they done to overcome these factors" is nothing at all.


Red can't be properly called a "camera head". There is an internal slot to record to a flash magazine, which makes it a fully-fledged camera body. And all the necessary processing is done on-board to allow the camera to record to 2.5" hard drive magazines, which cost under $1000 and weigh about a pound. Red almost certainly could have made room to insert these directly into the camera, but you gain more flexibility by having them external.

Red is delivering a full solution here; they're not just capturing 4K and leaving it to someone else to figure out how to record it or deal with it in post.
  • 0

#16 Stephen Williams

Stephen Williams
  • Sustaining Members
  • 4708 posts
  • Cinematographer
  • Europe

Posted 07 November 2006 - 10:30 AM

Um... actually there was a BLII with cs motor on the market. I'm debating on buying one, torn between it, a Mitchell BNCR and a Kinor 35H.


Nate,

At the price Jan is indicating, if it's a BL II, it's an easy choice IMHO.

Stephen
  • 0

#17 Laurent Andrieux

Laurent Andrieux
  • Sustaining Members
  • 1527 posts
  • Cinematographer
  • France

Posted 10 November 2006 - 09:41 AM

Um... actually there was a BLII with cs motor on the market.


Of course there is a ARRI BL 35 II (that is crystal sync), not any BL II B or C. ARRI 35 II B's and C's are non BL (blimped) cameras.

One should know - as Stephen asked - if one is talking about an ARRI 35 BL II or an ARRI 35 II B or C...
  • 0

#18 Stephen Williams

Stephen Williams
  • Sustaining Members
  • 4708 posts
  • Cinematographer
  • Europe

Posted 10 November 2006 - 11:03 AM

Of course there is a ARRI BL 35 II (that is crystal sync), not any BL II B or C. ARRI 35 II B and C is a non BL (blimped) camera.

One should know - as Stephen asked - if one is talking about a ARRI 35 BL II or a ARRI 35 II B or C...


Hi Laurent,

Pleased to clear that up.

So Jan, How many BL II's will you sell me for between 1500 - 2500 ?

Stephen
  • 0

#19 Andy_Alderslade

Andy_Alderslade
  • Basic Members
  • PipPipPipPip
  • 1055 posts
  • Cinematographer
  • London, UK

Posted 10 November 2006 - 12:30 PM

FWIW I saw Max's award winning film last Friday, 35mm Anamorphic, it kicked ass. I had a round trip of over 500km / 300miles well worth the journey.


I once had a seminar teacher who used to use that expression to describe the films of Scorcese, Cronenburg and even Nicholos Ray - so it most be good!

Is it showing in a festival anytime soon Max?
  • 0

#20 jan von krogh

jan von krogh
  • Basic Members
  • PipPipPipPip
  • 454 posts
  • Producer

Posted 12 November 2006 - 10:31 AM

Hi Laurent,

Pleased to clear that up.

So Jan, How many BL II's will you sell me for between 1500 - 2500 ?

Stephen


my bad - they are both 35 II B.

i indeed thought the 35 II B & C were called BL II B & C, and the later then 35BL & BL4 -sorry, it are the non-syncsound non-blimp.

from the 2 units of the 35 II B, one will be sold >jan 07, the other one -might- be modified with
http://www.intervalo...i/rev-arri2.stm
and stay for special purposes shooting or will also be sold.

both haven´t been seen much use since ~2001/2.
However added new 12V powercells set for the motor in august 06 and both are in fully working order.
last maintenance was 04.

the camera tech (who *ahem* different from <me> *ahem* seems to know every single arri model) will prepare the offer with serialnr., photos, etc.

ps. i wanted to add the list of mags, motors, mount-adapters etc, but after this first bummer i suppose i will better have the tech make the list.
  • 0


Aerial Filmworks

FJS International, LLC

Ritter Battery

rebotnix Technologies

Broadcast Solutions Inc

Opal

The Slider

Media Blackout - Custom Cables and AKS

Tai Audio

Rig Wheels Passport

CineTape

Gamma Ray Digital Inc

CineLab

Visual Products

Wooden Camera

Metropolis Post

Technodolly

Willys Widgets

Paralinx LLC

Glidecam

Abel Cine

Metropolis Post

The Slider

CineLab

Technodolly

Wooden Camera

Gamma Ray Digital Inc

CineTape

Aerial Filmworks

Paralinx LLC

Glidecam

Rig Wheels Passport

Broadcast Solutions Inc

Visual Products

Ritter Battery

Willys Widgets

FJS International, LLC

Media Blackout - Custom Cables and AKS

Opal

Tai Audio

Abel Cine

rebotnix Technologies