HD For Indies

High Definition Video for Independent Filmmakers
A How To Guide for Digital Filmmakers
Welcome all! This is my blog to share my latest research,
thoughts, etc. on utilizing HD for independent filmmaking.

YES, I am available for consulting
Contact me at mike@hdforindies.com

All content copyright 2004-2007 Mike Curtis.

Wednesday, December 07, 2005

More Thoughts on RED camera

UPDATED-NOW with 10 & 12 bit math. Duh - problem with napkin math, you forget stuff...

Random notes that have occurred to me about the RED camera:

-If using film style lenses and traditional PL film style mounts, I can't see how this ISN'T a single sensor CMOS, and that implies a Bayer pattern filter. RAW off of that at 4520x2540 would be (let me check my Photoshop here):

4520x2540 8 bit grayscale (Bayer pattern, color comes after demosaicing it to figure out where the RGB pixels were, just like a still camera): 10.9 MB/frame.

Thus 10.9*24=261MB/sec for 24p. Oh, wait, it does 60p, so that's a max RAW video data rate of 654 MB/sec. Shazaam! That'll fill up your hard drives pretty quick! No wonder they spec dual fiber channel link! Converted to RGB (multiply by 3) is 1962 MB/sec - almost 2 GB/sec. OK, that's just a whooooooole new class of storage needs. That's 7TB/hr. Yikes! That's about $13K or XServe RAID per hour of footage. The quote is obvious - "We're gonna need a bigger boat." 654 MB/sec is a LOT. Even a 4 gigabit fiber channel connection has a THEORETICAL link speed of 500 MB/sec....so those dual link fiber channel connectors, if they are really going to kick out RAW 2540p, are probably already 4Gbit apiece.

But wait - who does 8 bit and is serious? We're talkin' at least 10 bit here. So 10.9*1.25=13.625 MB/frame. (10/8=1.25)
So 13.625*24fps=327 MB/sec at 24fps. 30fps=409MB/sec, and 60fps=818MB/sec. Ouch. Multiply those by 1.2 to get 12 bit values - so 392.4 MB/sec for 24fps, 491 MB/sec for 30fps, and 981 MB/sec for 60fps. Sheesh, nearly a GB/sec? Ouch!

So hmm...how to back up those kinds of storage? I recently read about holographic storage with a transfer rate of 160 MB/sec and 300GB per disc, but that's probably read not write speed, and 300GB is only a few minutes worth of footage - hardly an efficient storage medium.

The best solution I know of that's shipping is LTO3, a tape backup format, that backs up 400GB to a $100 tape at either 50 or 100 MB/sec, I can't recall which right now. Autoloaders that will hold 6 tapes are about $6K or $7K last I checked, but again, that's 2.4TB, about 20 minutes worth, and it'd take all night to back up I'd bet.

Their domain is registered in Eastsound, WA, I'm aware of no serious camera geeks up there.

Oddly, it says copyright 1999-2004 - have they been underground working on it THAT long? Or is this some leftover corporate entity that has just been floating around and is now being used for this?

-if they are talking about massively variable frame rates, data recording, etc, that sounds like they are freed from tape velocity constraints if they can record in one frame increments to hard drives. of course, the devil is always in the details - under what circumstances can you or can't you shoot variable frame rate? Can you shoot ramped frame rates, programming a "move" of frame rates blended with adjusted exposure settings? etc. etc. etc.

-the comments are still flying on that original post, so follow the mad speculating there.

-or check out the speculation over at DVInfo.net, too, there's some smart folks speculating away like mad.

As Chris Hurd, site founder, says:

"Ready, get set... speculate!"

:-)

-mike

Labels: Red

Posted by: Mike Curtis / 12/07/2005 09:04:00 PM

Comments:

i don't really understand why people get exited about this... personally i'd be much more interested in systems with higher dymanic range and bit depth than higher resolution.
i mean, what lens can really resolve more than 4K? and would we even want them to (michael ballhaus once said that when the zeiss people asked him about the new lenses (i think it was the ultra primes), he told them: "they're way to sharp, i have to put all kind of diffusion on top of it or you'll see all the wrinkles and/or makeup).

for me the upper practical limit at the moment would be 2560x1920, at 10 or 12bit, (yes, full aperture, for scope lenses, with the option to store only a 16:9 cropped area out of it) .. which allows for some stabilizing/cropping etc and still have a full 2K image.
that's still nearly 6MB/frame though, or 140MB/sec, or 500GB/hour for a raw stream, but the field mags and studio storage would be a lot cheaper.

but even that would be out of reach for most indies (why not shoot super16 instead?)..
i'm happy when the HVX comes out ;)
just a few random thoughts
++ chris ++

# posted by

chris : Wed Dec 07, 02:12:00 AM CST

I think people who are interested to have 35mm DOF would be interested in this camera. And we don't know the specs on dynamic range yet. Maybe we will be pleasantly surprised? I don't think anyone thinks the perfect camera has been offered to us yet, so why not keep hope alive that this camera will be a step closer to that ideal? Maybe it will sell for $10K. Who knows. But I, for one, like the idea that someone is out there pushing the envelope to deliver a product that is better than the Japanese Big Three are offering us. Each of us secretly hopes for better. Red is keeping my hopes alive. I will buy a Panasonic HVX 200. But I still wish for more. I hope these guys are the ones to make it.

# posted by

spieden : Wed Dec 07, 03:35:00 AM CST

4k rez is easier on the lens to resolve at 35mm than HD on a 1/3" chip! So that's not a problem here at all. Sure you need a decent lens, but that's a given, right?

And who says it won't have great dynamic range? It sounds like the sensor is like those on top end DSLRs, and they have >12bit dynamic range....

Graeme

# posted by

Graeme Nattress : Wed Dec 07, 07:36:00 AM CST

I am really not so excited either, but probably because of seeing so many things fail, pointlessly. This camera is probably not for most of us, but for the ones with the extra money to spend on it. The dvinfo thread has somebody claiming to be on the inside know of the company, and somebody else posted stuff about microHD pocketHD, but I doubt it is one of the those near $100 SHD digital camera overclocked ;) .

The logo on their front page, the shape looks suspiciously like some sort of camera body.

I think I have an solution for the 60 fps, you notice that it quotes low end data rates, the camera in obviously got compressed abilities. With compression, the entry price for the camera (in highly compressed, or low res mode) could be relatively low. Another solution is to use an 1/3rd pixel shift scheme to reduce data rates and costs, that would give you 206.65 MB/s at 60 FPS, 91.84 MB/s at 24fps (in this scheme you transfer a special pixel shifted version of the bayer pattern at reduced resolution, or three chip). Even 1/2 pixel shift (fairly normal type) would achieve 60 fps uncompressed. From my own research, an uncompressed 720p camera could be done for around $200, so the basic Red camera (no storage or lens) could have latitude to come under $10K or $5K.

New storage is going down in price, and 1 terabyte hard drives and Holodisks might be available when this camera comes out. If it requires full HDSDi*2 bandwidth then the array to run upto 10 drives in parallel is of a bigger concern.

From the other post. I think there is little likely hood that this camera might be related to our projects at Dvinfo, as one of the major project players doesn't seem aware of it in the thread mentioned.

# posted by

Wayne Morellini : Wed Dec 07, 07:51:00 AM CST

Hmm... and Chris doesn't normally let us speculate at dvinfo, only "facts", from repectable news sources ;)

# posted by

Wayne Morellini : Wed Dec 07, 07:58:00 AM CST

Remember your point about how the other 4K camera that I posted about would require you to be tied to an array? Need I say more?

:)

# posted by

B-Scene Films : Wed Dec 07, 08:50:00 AM CST

Stuopid rambling: Let's say you set out to make an HD camera from a DSLR chip/sensor, one problem you will have is that SDLR sensors don't natively do fields, so you will have to get the chip to export full 60fps if you wanted 1080i. Now if you could get a DSLR chip to export well timed 60fps, well hot damn, you might as well go for the whole enchalada, and make the greatest digital camera ever made, hence the RED project. To me the problem is reliably and legally getting a DSLR chip to ouput 60 fps. From articles I have read about DSLR chips (dpreview.com), there are technical reasons why, say with my my canon DSLR, it can only do 3 rapid shots in a row. Wouldn't they have legal problems with hacking a canon chip and reselling? Also, why wouldn't canon do this themselves?

# posted by

Neil : Wed Dec 07, 11:00:00 AM CST

FILM FILM FILM FILM, FILM FILM FILM, FILM FILM FILM FILM, FILM FILM FILM> as per Ren and Stinky...

Look, at these data rates there is now way that you would either not employ a codec (a bunch of cameras could employ a h.264 real time encode/decode on a single chip this year, ala Xilinx) or shoot film. The bottom line is that HD has it purpose and for this RED camera to pull of its claims would be either astonishing or using a grouping of some incredible (unobtainium)hardware and on board software systems (yep it would need to have a duplex aware switch for dual FC which means a software interface or OS and an insanely good codec that slaps the Blackmagic or the Avid codec's to bit dust)...I don't know all the details but if it smells like fish, and it looks, well like a web site I could put together in about an hour, and it tastes like (the bad taste I have had in my mouth waiting for a realtime 4:2:2 system with a apeture range, and latitude any where close to Film for the last two years)...it must be SH*%, or Fish, sorry...

# posted by

Anonymous : Wed Dec 07, 11:45:00 AM CST

graeme,
to clarify (as i tend to make a mess):

i was talking about the dynamic range in the scene, ie. contrast between shadow and highlight detail.
bit depth is something different, it's the ability to separate differences between very similair colour/brightness values. btw, i'm not saying that the camera would use 8bit, i just used that number because mike assumed that for his storage calculations... but 8bit on this resolution would be a joke.

as for chip designs:
yes, one can build chips that have a great dynamic range, chips that have high sensibility, chips with high bit depth, bits that are immune to blooming and smearing, chips that can run high frame rates and even chips that are cheap... but not all of the above together! so all i really wanted to point out (rather cumbersome i admit) is that for the same money you can get a 4520px chip, you could get a 2K chip with of much better quality (and save tons of storage space).

about lenses:
the 1/3" CCDs for HDV cams usually have about 1000px resolution (or about 200px/mm), the 2/3" CCDs of the F900 has 1920px (or 200px/mm)... the red (if it ever appears) has 4520px on 24.5mm (or 185px/mm).. so yes, the lenses have to resolve the same detail, but it's much harder to build a lens that covers a bigger area.. it's possible, but again expensive (and HD lenses already are expensive).

all in all, i just don't think it makes sense to build a camera with that resolution at this moment... it's like building a car which runs 400mph.
i'd take the D20 over this any day (if only becasue the D20 is available now and the red probably never will)

++ chris ++

# posted by

chris : Wed Dec 07, 12:04:00 PM CST

Lattitude of sensor and bit depth are coupled, to a point, and dsp bit depth too.

It's a bayer chip, rather than 3 chips, I'd guess, so you'd be wanting about half your 186px/mm, and anyway, there'd be an ant-alias filter in there to stop nasties. Sounds to me it's a 60p DSLR.

As for never avail.... Lets wait and see eh? I've got a feeling there's people out there who know a whole lot more about Red than they're telling, and they'll be under NDA and hence won't be able to say a thing....

Graeme

# posted by

Graeme Nattress : Wed Dec 07, 01:04:00 PM CST

Neil, why worry about fields? this thing is obviously going straight to disk. Obviously a RAW img sequence.

I think 4k is ridiculous, but 2k a good idea. The problem is this, what and how are you going to cut??? DPX sequence can be red just fine, but Smoke is the only editing (and mostly online) system that I've heard of that can handle img sequences.

Since QT can read img sequences, I wonder how hard it would really be for FCP to be able to read them. Now that would be sexy.

The problem is, without a realistic (TAPELESS) post workflow all of these new cameras are useless. But imagine offline editing, sending and edl (or preferably XML by that point) to the conform house, and screening (in 2k) immediately. You can sort of do it with certain apps, but img sequences and EDL's don't have the meta data to be able to do these things well...

They do really miss the point though. Lattitude is more important than resolution for any involved in production. Hell, I'd rather shoot on the DVX-100a and upress than try to make HDV look good.

# posted by

Joseph : Wed Dec 07, 01:16:00 PM CST

Some of the limitations in SLR cameras are practical (it's a bit cheaper to build it slow) but a lot political (building slow ones allow you to build fast ones and charge many times more, even if the extra cost isn't that great). Maybe they are using a higher speed version of a digital camera chip, maybe something else.

About Lens, one thing, Imax, so obviously the capability is out there to go far above the super35 format, and the cheaper SLR lens also are much bigger.

Latitude is a cracker (and better a good 1080p chip than poor 2540p chip) like criss said. But with cross licensing, and recent company buy outs, many features can go onto the same chip. Sensor technology is now in the 16 Mpixel to 32 Mpixel, 8 Mpixel issues are becoming a thing of the past. We have to remember that commercial SLR are into the 16 Mpixel range.

OK, now don't get excited about these, it doesn't say they are video sensors, but they are representative of the progress in the sensor industry and are supposed to have leading image quality at 33 MP:

http://www.dpreview.com/news/0511/05112202dalsa_chips.asp

We shall see what this thing is, but yes, the website does not look promising. I'm almost waiting for some idiot to turn up proclaiming that it was a joke they pulled.

Chris, is that you?

# posted by

Wayne Morellini : Wed Dec 07, 01:16:00 PM CST

According to the spec sheet on the Red site, they never mention it recording uncompressed, in fact they only mention their RED CODEC which apparently tops out at 100Mbps. Capturing 100Mbps wouldn't be that difficult even at the quoted 2540p image size.

Obviously, if the thing has dual-fiber outs, then we could feed it into some extrodinary array to capture uncompressed, but it appears that the RED system itself is designed around a 100, 80, 60, 50, 25, or 19Mbps stream.

harlan

# posted by

harlan : Wed Dec 07, 01:38:00 PM CST

Cmos sensors use a similar process as microprocessors. If you want to get 60fps out of an 11mp sensor that normally does 4 then you have to run it at 15 times the rated speed. Hence they will generate much more heat and break.

If they are using a Cmos it will likely not have come from a Dslr. If Canon or Nikon had been using sensors capable of high speed readout then. The current circuitry in a Dslr is max of about 100Mhz. So for 11mp @ 60 fps you need 660Mhz cmos at the minimum since most Dslr use a single output to reduce number of wires. So if they were running that fast the chips from a Dslr would have the power consumption of an old PII chip.

Only leaves that they really do have a "custom" sensor or a high speed CMOS (which normally have poor colour filters and poor light response compared to non high speed). If the sensor really is a custom job (which means panavision type funding since theirs is a custom fill factory chip) then they really will build a camera.

# posted by

KEith Wakeham : Wed Dec 07, 02:22:00 PM CST

wayne,
i'm hardly ever on dvinfo, and i dont have any respectable sources, just a bit of comon sense .. so i doubt i'm the person you think of ;)

the point with lenses is really that if you have a larger recording medium, the lens has to resolve less detail per mm... try to find a hasselblad medium format lens which performs reasonably on a 2/3" HD camcorder (and those are NOT cheap lenses).

joseph,
obviously you wouldn't edit with 4K file sequences but create offline proxies... building a custom work won't be the problem if you can afford to shoot with this kind of setup.

graeme,
how would latitude and bit depth be coupled? maybe it's a manufacturing thing?
my understanding is that i can have a chip with 15 stops of latitude and store it in a 8bit file format (admittedly it will look like crap if i have to do any grading, but that was excactly my point).
on the other hand i can shoot with a chip that clips after 5 stops but has 12bit, i could still benefit from the subtle variations if i want to crank up the contrast etc (that is, if i control my lighting ratios).
but obviously a chip with high dynamic range benefits the most from higher bit depth.

high latitude is really one of the most wanted features in digital filmshooting, and trying to achieve this with a 4K+ sensor (that would have to output 12bit) isnt going to make it easier.

++ chris ++

# posted by

chris : Wed Dec 07, 02:36:00 PM CST

Keith, I agree that a custom built sensor of quality is really hard to imagine. I still think the RED is based on a still digital camera sensor. The resolution points to that, and so does the 60fps. The point I was trying to make before about fields is that 60fps would be required output from a dslr sensor if you want 1080i b/c dslr doen't do fields. I can't help suspecting that this project stemmed from the idea of taking the still digital camera sensor (a huge industry to pull from) and using it to make an HD camera (something I myself always wondered if it was possible). They sucessfully got a dslr to output 60fps, and said, wow , we can take this well beyond an hd camera.

# posted by

neil : Wed Dec 07, 03:38:00 PM CST

Chris makes a good point. Our chip latitude doesn't need to be much more than our file format can support.

I really don't know what to make of this particular camera. The specs seem to be somewhat self-contradictory, until you get rid of the RAW component. If it's not actually storing the RAW data, then the Bayer pattern likelyhood makes more sense, and their RED codec makes more sense. Also, some form of compression makes this a lot closer to the realm of possibilty.

# posted by

Isaac Botkin : Wed Dec 07, 04:23:00 PM CST

I admit I wasn't making much sense, I wasn't talking about offline proxies (that part would be easy). The major problem is that instead of having ~1000-2000 shots to manage (for a feature) you've got *millions* of dpx files to manage. I was reading something about an dpx straight from camera (a viper?) workflow (or lack thereof) where there was simply no rational way to conform.

All of this technology (vaporware as it is) is meaningless unless there can be a *useful* way of dealing with img sequences.

I would love there to be a nice wrapper, (*.isw) that can be opened and and messed with. Imagine how windows handles .zip files. It's a single file, but that can be opened (with a txt based header and img sequence inside). I guess just using actual folders are ok... anyone know how smoke handles that stuff?

# posted by

Joseph : Wed Dec 07, 04:30:00 PM CST

On the smoke, it "sees" an image sequence as one file in the interface, albeit if you ever went to the command line, it would be quite a mess in any given directory.

A lot of software now can read the timecode in a DPX header file and make a conformed sequence, i.e., Nucoda, Scratch, etc.

Also the blackmagic framelink utiliy "wraps" DPX sequences in a QT wrapper with a 10-bit 4:4:4 RGB codec. The only problem is that for a large sequence, it can take some time to wrap and unwrap the seq.

# posted by

Jason Rodriguez : Wed Dec 07, 11:03:00 PM CST

Keith

I'm not suggesting they are using the SLR or custom versions, but that the chip might also be available in different commercial versions with more outputs for higher data rates. We can see that Fill-factory did one for one of the movie cameras (I forget which Arri or Pana) and another for Kodak, similar technology different format and different output capability. They could even be waiting for the official release of the chip (like other people we have known).

> So for 11mp @ 60 fps you need 660Mhz cmos at the minimum since most Dslr use a single output to reduce number of wires. So if they were running that fast the chips from a Dslr would have the power consumption of an old PII chip.

660Mhz is done now days at less than 1 Watt (I know a person that has a processor that does 2.4Gz on 20mw, I'm not joking, it is the way they have to design the PC/server processors that produces the power consumption). For an interface 660Mhz is hardly a dent on the state of the art, but good marketing to charge extra for, pin count from too many outputs is the worry. As more outputs means more potential for parallelism, you should view it in power consumption terms, like a 41.5Mhz DDR2 chip (although in this case you can shift data by very wide slow internal busses into a high speed interface). Using 1/3rd pixel shift, even some existing SLR sensors are fast enough.

Chris

> the point with lenses is really that if you have a larger recording medium, the lens has to resolve less detail per mm... try to find a hasselblad medium format lens which performs reasonably on a 2/3" HD camcorder (and those are NOT cheap lenses).

To avoid a long debate about the resolution limits of lens, the point was where ever there was a lens that could handle UHD resolution, and there is.

Yes, latitude is different from bit depth, latitude is the difference between the top and bottom, bit depth is merely how finely it is chopped up into levels. So you can have two 8 bit cameras, one with 100 times the latitude of the other one. From what I understand, the native latitude makes an difference, as with the same lens, sensor QE, SN and sensitivity, you will have to step down the iris nearly 7 stops on the lower latitude sensor to get equivalent clipping to the larger latitude camera. 12 bits will merely allow you to divide the top and bottom value of the 8 bits, another 16 times, allowing you to adjust the iris and ND filters to manipulate the latitude to give another 4 stops (as long as Signal to noise ratio high enough not to wipe out the extra bits).

> high latitude is really one of the most wanted features in digital filmshooting, and trying to achieve this with a 4K+ sensor (that would have to output 12 bit) isnt going to make it easier.

There are some technologies that get a large latitude, they have been out for a while, so there is some sensors that support them. Fill factory Ibis uses a double sampling scheme to get a lot more latitude, and Smalcamera uses some sort of gain scheme to do likewise (actually the Sony HC1 might also).

# posted by

Wayen Morellini : Thu Dec 08, 03:20:00 AM CST

Has enyone ever thought about this concept?

The sensor doesn't have to output 60 fps... 5 or 10 fps is enough (imagine a full frame CMOS like the one in the Canon D5).

Just use more than one CMOS.

Use a mechanical moving mirror (modified version of the ones used in SLR Cameras) to reflect the picture to the different sensors in sequenz. It will be a mechanical shutter.

Each sensor uses its own processor (like in the canon D5) to render JPGS and write to a unified RAM-Cache...
transfer the data from RAM-Cache via FirWire 800 to a normal striped HDD-Recorder (like the LaCie BiggerDisk Extreme). That would be more than fast enough.

All standart CanonEOS Optics could be used. Yes, all standart 35mm lenses are UltraHighDef by nature. (See picture quality of the Canon D5, it will leave a Arri D-20 in the dust)

The JPG sequenz can be stichted together in Adobe AF or a custom software to any wanted format.

Yes, the mirror has to be extreeeeeeemly precise, but "jitter" problems could be eliminated by tricky software deliverd with the camera.

the number and quality of the different CMOS sensors will give you a varity of possible quality, framerate, resolution.

This should deliver outstandig picture quality at a resonable price.

A Canon D5 costs 3000 USD... so I presume that the sensor and processer will go for 1.500 USD multiplied by... let's say 8 sensors will go for 12.000 USD or less. The Disk will go for 2.000 USD. The Body and Intefaces for another 1.000 USD.... what the mirror will cost? I have no Idea... but let's think positive.... 2.000 USD

This totals 17.000 USD

You want even better Image quality.. use different CMOS sensors

Please, give me your thoughts on this... why should this concept don't work???

Nikolas
(sorry for my english... I'm german and not used to write in english)

# posted by

Nikolas : Thu Dec 08, 06:44:00 AM CST

Regarding the issue of latitude and look, don't forget that the recently released animation movie "The Corpse Bride" was entirely shot with DSLR Canon EOS-1Ds cameras.

The Canon EOS-1D is capable of shooting 8 megapixel images in excess of 8 fps on a single CF card. The resolution curently required for theatrical projection is ~2k. These cameras employ curtain shutters and a mirror that goes up and down, two factors that can limit a higher frame rate. Using some decimation, taking the sensor out from this environment, please consider the possibilities...

# posted by

Valeriu Campan : Thu Dec 08, 06:59:00 AM CST

nikolas,

sorry to break your bubble, but you're off by a factor of 10 or so... to make a commercial product the way you plan it you'd end up with a lot of custom work and R/D, like building the mirror, ironing out the jitter issues, getting the timing right, writing custom software... also components will be much more expensive: 2000 for a disk that can record 400mb/sec? body for 1000? i dont think so!

the whole setup would be 100'000-200'000 i guess
++ chris ++

# posted by

chris : Thu Dec 08, 08:51:00 AM CST

Nikolas

Yes, this was suggested sometime ago over at dvinfo, was that you?

There are a number of people trying different things out there, you might as well have a go at it, treat it as a hobby and see how far you get.

Somethings look way more complex than what they are, other things look far more simpler than what they are, but one thing that helps is looking at previous products that have their own solutions, like mirror mechanisms in laser projectors, Nintendo Virtual Boy, laser printers, mirror prisms (lookup prisms links over at the original dvinfo 35mm SLR lens adaptor threads you may find prisms with triple output somewhere). So you can see there are designs out there already that may help solve some problems.

Search for a good reasonably priced camera, attach an external hard drive to each camera (there is memory card to hard drive adaptors out there) to save that cameras images, synchronise the cameras to click one after another in rotation, and save while in parallel.

The big problem is the editing software and the prism to make it work, otherwise it is relatively simple. Keep the price down by aiming for 24 fps, and reduced resolution and using compressed modes. Don't expect not to spend thousands, and a lot of time (an extreme lot more to commercialise it) though I wonder what you could do with $50 dollar cameras and a \\\ series of 33% and 50% partial reflective first surface mirrors with a full one at the back (optical mirror eliminates unwanted reflections) with a camera at the bottom of each (big mirrors fro field of view, and big rig). I don't think it is worth it except as a DIY project.

# posted by

Wayne Morellini : Thu Dec 08, 09:55:00 AM CST

Thanks for your thoughts Chris...

2000 for a disk that can record 400mb/sec????

were Do you get 400mb(sec (is it bits or bytes... I know the Film guys usualy talk bits)

Lets Do some math:

One frame at 2K resolution will have (JPG high quality) about 1 MB (lets use bytes).

Thats easy, it will leave you with 24 MB/sek for cinema (this camera is a cinema solution idea) and at 60 MB/sek for a pretty good slomo or 1080p 60.

Lacie Bigger Disk Extreme via FireWire 800 performs with:
90 MB/sek (still bytes) while reading and
70 MB/sek while writing

yep, this is supposed to be sustained.

Even if this doesn't work... lets use a very big buffer cache. 4 Gigabytes of RAM gives you 3 Minute Shots... the Data could be transfered "as fast as possible" to free the cache. Need longer shots... get yourself another 4 GB in normal Dimms.

sound sufficient to me????

The 1TB (yeees bytes ;-) version goes for 799 USD
not fast enough??? use two

http://www.lacie.com/products/product.htm?pid=10451

Tru, the mirror is the main problem and hardest to develope. Actually... nowbody will develope this... lets look for a workaround. Is there any existing system that could be used... ore modified to be used. Where are our physics with a good Idea???? Something like... use LCDs to mirrer light in different directions... use a arri shutter and paste the CMOSs on a rotating wheel (well that'll hurt them) or whatsoefer. Ideas please.

Your right on software etc. and the producer hasn't earned anything yet... it's just a concept. And 17.000 ist still a lot cheaper that a arri 20-D.
This concept cam would still be interesting for 25.000. By the way the Guys building the "Drace" do exactly that INCLUDING softwear to read date from CMOS, what we don't have to do.. where using the original Canon signal processors.

I know, I know... there is a lot of unsolved problems. This is just for the love of the idea.

Nikolas

# posted by

Nikolas : Thu Dec 08, 10:18:00 AM CST

Hello Wayne,

nope... that wasn't me.

You're right in all points. Thats why I'm not going to do this... I'm not a hardware guy... just a user...I would expect some major company who builts cameras but no film cameras to do it... just to get into the market.

Hello Nikon... anybody listening.

Probably I'll end up with a HVX

BUT I WANT MY CINEMA ;-)

# posted by

Nikolas : Thu Dec 08, 10:23:00 AM CST

nikolas,

ideas are always nice.. ;)
but you're positioning your concept pretty unfortunate if you go for a 2K camera that records to jpegs... it will be too expensive for the home user and not good enough for the professional filmmakers.
what gets people exited is to record more than 2K uncompressed, to be more precise 4520px at 12bit raw with the red.. and that's where the 400MB/sec comes in (yes, that's nearly half a gigabyte A SECOND).

seriously, you would never buy such a camera anyway but always rent it, and if you factor in the costs for lenses (and you dont wanna use an old angenieux zoom with a 4K chip), storage, and postproduction you suddenly might find the D20 quite competitive with the red.

but i wouldnt mind to get proven wrong and somebody coming out with a 4K 10bit raw camera for 20K though ;)
++ chris ++

# posted by

chris : Thu Dec 08, 06:31:00 PM CST

HD For Indies

Advertisements

Great HD Links

Archives

Wednesday, December 07, 2005

More Thoughts on RED camera