I wonder if RED will be at Cine Gear Expo this year. If there's one place to market a new camera to the feature/high-end production community this would be it.

Agree completely. IBC too of course, but Cine Gear Expo is right up RED's alley.

http://www.cinegearexpo.com/

Chris,

Probably you know the most, can you answer the technical question for me.
I just read about RED digital camera and I could not attend NAB2006 this year.

From Steve Gibby to Jim Jannard interview:
>60fps in 2,540p, which is 4,520 x 2,540 pixels – bigger even than the 4K spec for Digital Cinema. We’re flexible in the signal format: from 2,540p, 4K and 2K and down to 720p. We’re also flexible on the data streams that you can output: RAW, 4:4:4, 4:2:2, or you can record to REDCODE, which is our Wavelet codec…Its even flexible on high-speed data output, so for example, we can offer a choice between Infiniband and any other high-speed interface that comes along in the future. The idea is that this flexibility enables you with as much capability as is possible, in the smallest package available – and, priced right, I might add.<


1)
CMOS censor always have heating problem and how does it deal with it that small RED camera body? Will be liquid cooling system with low noise fan in it?

2)
Transferring image from image censor (CMOS) to RGB 10 bit: 1,024 will huge technological challenge. They are only 4 companies, (not Sony, Panasonic and JVC.) they are Arri, Panavision, Thomson and Dalsa and their companies set for the Digital Cinema Initiatives, LLC compliance of 4K or 2K: Arriflex D-20 (2,880 x 1,620 for 1.78), Dalsa Origin (4,096x2,048 for full res.) Panavision Genesis (4,644 x 2,160 for full res.) and Thomson Viper stream system (sorry, I don’t know). Their censor chips came from various sources and all of them had to solve the CCD or CMOS chips’ heating problem; receiving analogue light exposure from lenses to image sensor chips convert to digital, an internal computer electronic circuit algorithm (uses higher than 10 bit: 1,024, respectably14 bit: 16,384 to 24 bit: 16,777,216) re-sizes to 10 bit: 1,024 out put. That will be the biggest challenge. Do you know any information about censor chip of RED and who would supply to RED? How much bit rate do they use analogue image to convert (Data Signal Processing) to digital 10 bit: 1,024 by sampling rate?

3)
Also, non-compress signal 4:4:4 and 4k need infiniband, which Jim Jannard mentions, That’s understandable and to do the right transfer to the storage server, which need transfer rate 200Meg to 0.5Gig per second and have to have 100 terabit to 200 terabit space to shoot full length feature.
It’s not 4:1:1 or 4:2:2.
Final product needs 200Mbit/sec and 190,000Gig for 12bit:4,096 @ 24P Digital output.
Reel-Stream Andromeda does 4:4:4, but 1/3 chips that different story.


Each those corporations spent 100’s millions to develope the Digital Intermediate Film Camera for at least 10 years.
It’ll be a big challenge for any corporation to deliver full size 35mm censor 4k-output camera right now and RED's suggesting price.

Even mass-produced Still camera like: Canon EOS-1DS Mark II, 16.7 Mega pixel (is capable of shooting up to 4 frames-per-second for up to 32 JPEG or 11 RAW images.) costs 7k and could not capture more than 5 raw images per second yet.

If Jim Jannard delivers first batch of the finish products in the end of 2006, I’ll be shocked and which will be a Nobel Prize story for me.

But if this is real, I’ll wait for the RED, instead to get Sony XDCAM350 (both same price range).

Thanks

Hi Chosei,

I'm not Chris, but I think I can give you a couple of answers:

There's probably a reason for the mystery surrounding the Mysterium sensor. RED doesn't want the competition to know too much about it - and thus far what's on the RED website has been the only details I've seen about it. IF anyone knows more about it, I'm sure they've signed NDA's that prevent them from shedding any light about the subject. So I don't think you'll get the info you're asking for.

Second, we all realize that building RED and making it work as advertised, for the price advertised is an extremely challenging undertaking. I doubt Jim would have embarked on this journey if it was easy. I'm guessing that the sheer impossibility of this project is partly what makes him want to do it.

The fact that it hasn't been done before, doesn't mean it can't be done.

Yes I too would be kind of shocked if RED ships by the end of the year. But I'm making a reservation, because I want to believe in this thing. I'm willing to buy into this dream :-)

Barend

Hi Chosei,

I am also not Chris, but maybe I can add a few things.

1)
My understanding is that CMOS dissupates much less power than CCD (Still after a great deal of improvement). Though we have not been given many details about the chip itself.

2)
My understanding is heating due to external light is minimised by IR filtering. This done even sunlight is barely warm. Conversion from 12bit, if the chip outputs this (It has been hinted) would probably be done to 10bit with a look up table. I don't see the difficulty here.

3)
Infiniband has been mentioned as an example, I don't think this has been finalised yet. 4k at 4:4:4 will not be available. 4k and 2540p will be RAW, colour mask (probably Bayer at 4:2:0).

I'm not Chris either, but..

I know a little about sensor architecture. CCD's can't hit much higher than 40MHz clock rate or else they will easily be damaged. Cmos can hit much higher but I think the limit is around 200Mhz with the new processes. Dslr normally have one output and to get the highest quality clock lower to reduce noise from electronics. This is without any cooling, so using multiple outputs from a sensor or having a high speed onchip multiplexer means that heat shouldnt' be an issue.

Normally CCD's gain and AD conversion happens off the chip and normally CMOS happens on the chip, but Panavision has a CMOS with 8 outputs (not the chip in genesis, the QuadHD from Panavision SVI) that require external AD conversion. They call this ACS. RED owns the Sensor IP so who knows what they got in that thing.

RED's sensor is a closely gaurded secret that isn't on the market, they had it designed and built for RED only. They will probably only release a handful of specs and we may never know who makes it, but its pointless anyway because nobody outside of the RED team will ever handle the sensors or data sheets.

Actually their is a neat way to pack bayer data onto a HD-SDI line. I experimented with it a while ago and Jim could actually reprogram the HD-SDI dual link lines to actually feed raw bayer and not require infiniband. He might not because its out of spec for the SMPTE docs but it can be done. If he did an 12bit log to 10bit linear conversion it would be insane. Total datarate for the RAW 4K output over this method would be about 300MB/s but it would be limited to 30fps. You would need dual dual-link HD-SDI to do this for 60FPS and that would be around 600MB/s. Talk about intense data.

Now back to more on topic.
As for IBC and CineGear Expo, I hope they go and show off a little more, keeping us up to date on progress. Maybe I'll be able to make it their since I couldn't get to NAB.

EDIT: Forgot, when I mean raw over dual link i mean two full 4:2:2 links not a 4:2:2 + 0:2:2 link. And I'm not 100% sure if this is sanctioned in SMPTE 372, but I think it is, and Sony SRW-1 is dual 4:2:2 not 4:2:2+0:2:2 so I think it might be.

And as I have pointed out elsewhere the Canon EOS-1DS Mark II sensor is much larger, 2 to 3 times, than the Mysterium. 35mm still and 35mm movie do things differently.

You guys gave me an answer already; I knew new CMOS improved dramatically in past 3 years. Red’s CMOS could do the A/D converter in itself, will be a significant evolution and it might be real sensation to the film world.
I’ll wait to see the RED till Cine Gear Expo and hope to see further developments.

Exciting!

And as I have pointed out elsewhere the Canon EOS-1DS Mark II sensor is much larger, 2 to 3 times, than the Mysterium. 35mm still and 35mm movie do things differently.

Canon EOS-1DS Mark II

• 36 x 24 mm CMOS
• 16.7 million effective pixels
• 4992 x 3328

RED

• 24.4 x 13.7 mm
• 11.4 million effective pixels
• 4520 x 2540 pixels

Comparison of sensors:

Ansi Super 35 (35mm film)
Image area = 24.9mm x 18.7mm

Typical 2/3” CCD sensor
Image area = 9.4mm x 5.3mm
Pixels = 1920x1080
Pixel pitch = 4.9 x 4.9 microns

Arri D-20 CMOS sensor
Image area = 24.9mm x 18.15mm (full width of S35)
Pixels = 3018 x 2200
Pixel pitch = 8.25 x 8.25

Mysterium CMOS sensor *
Image area = 24.4mm x 13.7mm (S35 size)
Pixels = 4520 x 2540
Pixel pitch = 5.4 x 5.4 microns

*Note: This is the initially announced Mysterium sensor proposed specifications. RED has announced no final specs for the sensor. It's possible that the final specs for the sensor may differ.

Gibby
www.cut4.tv

Comparison of sensors:

Ansi Super 35 (35mm film)
Image area = 24.9mm x 18.7mm

Arri D-20 CMOS sensor
Image area = 24.9mm x 18.15mm (full width of S35)
Pixels = 3018 x 2200
Pixel pitch = 8.25 x 8.25

Mysterium CMOS sensor *
Image area = 24.4mm x 13.7mm (S35 size)
Pixels = 4520 x 2540
Pixel pitch = 5.4 x 5.4 microns

So the image area is smaller than the arri...yet the pixels are more???
not sure I follow. In short, are you getting the full meal deal or not... :-)

peace

dalen

I'm a techie, but not an engineer. I'm sure Graeme Nattress or one of the engineers on this board could provide some insight on that.

I'm a techie, but not an engineer. I'm sure Graeme Nattress or one of the engineers on this board could provide some insight on that.

thanks...as im not a techie or an engineer...just an artist with a dream. :-)

peace

dalen

If ever there was a Graeme question, this one is it!

Ansi Super 35 (35mm film)
Image area = 24.9mm x 18.7mm
Arri D-20 CMOS sensor
Image area = 24.9mm x 18.15mm (full width of S35)

a 35mm gate is NOT 16x9 it is 1:33 ( basically 4x3) plus you need room for audio track .... in super 35 forget the audio track ...
a frame in 35mm uses 4 perfs ( sprockets) height for 1:33 to get 1:85 they cut off the top and bottom of the frame in the projector gate so you only see a area equal to 3 perfs .. on some 35mm camera you can have them set to record 3 perf so it uses 25% less film but then you have to do a optical print master so it ends up back on 35mm using 4 perfs for theaters to be able to project it ... sooooooooo you might say 16x9 is more equal to 3 perf 35mm

thanks...as im not a techie or an engineer...just an artist with a dream. :-)

peace

dalen

I hear ya. I try to balance four things: artist, techie, manager, and journalist. I need them all for what I do...

Perhaps a little less on the technical side, but, when you consider that the RED sensor is providing a pristine original 4k image- compared to film's print and scan and print and print... RED might boost the cosmetic surgery industry in Tinsel-Town. ( :

Perhaps a little less on the technical side, but, when you consider that the RED sensor is providing a pristine original 4k image- compared to film's print and scan and print and print... RED might boost the cosmetic surgery industry in Tinsel-Town. ( :



Well at 4k, you're still limited to what you can show your feature on. Right now it's either output to film, where you have to go through the same steps for a release print.

Sony's 4k projector is another option, though I hear it's not ready for prime time and not sure what theater is using it yet.

Digital projection in general is going to eliminate the issues with the generation loss of a release print.

If you scan a film negative at 4k, and digitally project at the end, it should be a pristine image directly from that film negative, minus the compression used for the digital projection file.

Perhaps a little less on the technical side, but, when you consider that the RED sensor is providing a pristine original 4k image- compared to film's print and scan and print and print... RED might boost the cosmetic surgery industry in Tinsel-Town. ( :

haha---yes, yes---so true

lol

Like we need it! oy!