Home Made HD Cinema Cameras - Technical Discussion

[Wayne Morellini]

Rob S, I have a question. How tied to particular sensors, or frame grabbers, is your software going to be, and can it be adapted to work on any std interface (USB, Gige, Firewire etc).

By the way haven't heard back from you for a while.

[Wayne Morellini]

I have been reading a few things that I now realise might be significant to us.

Sony has announced developement of a 8 layer 200GB disk, and the release of 100GB disks in 2007 (so when on earth are they going to release the 200GB). Finale copies of 720p productions we complete in the meantime could be transfered to it (119GB+ for a 90 min 8bit) but with some compression it could be used for higher resolutions (268GB+ for 1080, and 1036GB+ SHD before compression).

The important thing is that they also announced that the PS3 will be shipped with a 54GB Blue Ray disk. They also aim the BR disk to be a alternative HDDVD format, I don't know which version they plan this to be.

http://www.gamesindustry.biz/content_page.php?aid=4534
http://www.tomshardware.com/hardnews...24_134741.html


This opens a new way for distribution. Movies could eventually be distributed for the PS3 type devices, using a on disc codec player that plays the movie (this all needs Sony's licensing though). Even though We could use a visually lossless codec, like cinefrom is, to do it (but with additional compression rates of upto 20:1 (equivalent to 5-10:1 Mpeg2?) for higher resolutions). That would enable even a short SHD 90min to fit on one PS3 disk. Big opportunity.

I think Sony has this pegged, they know that their projectors will deliver better than HDTV resolutions, and their disks can deliver the detail over HD-DVD, and the PS3 becomes a game subsidised cheap SHD player, giving them a clear advantage over the competition. If they make the format cell dependent (upgradable player codec) then anybody that wants to make a player has to buy cell chips. The number of PS3 like devices may not reach 100 million for sometime, but it is probably possible that they will satuate the S/HD hometheatre market before that, if they can deliver superior picture quality to HD-DVD, while everybody else can buy the DVD/HDDVD version. Or am I too far ahead.

[Brett Erskine]

Is anyone making box cameras with CCD's or CMOS censors that are larger than 2/3" (closer to 35mm size), do at least 1080i rez, and have virable frame rates higher than 24fps? Last time I checked you could only find something like that in a Dalsa camera but the sensor required too much light. Anyone know of any new CCD or CMOS box cameras out there that could be applied to this project?

[Aaron Shaw]

Why 1080i? Isn't one of the HD standards 1080 24p? I'd certainly buy one if it was available!

[Rob Lohman]

I think that is basically out of most people's reach money wise.
35mm CCD/CMOS chips exist, for a price. 1080p chips are becoming
available now, but they are twice as expensive ($2k+) as the
720p versions.

[Brett Erskine]

Rob-
Could you point a link or at least a name of the CCD/CMOS your talking about? If I understood you correctly on the price then $2k is a great deal for what you get. Please share.

Anyone else have info?

[Rob Lohman]

http://www.siliconimaging.com/SI%20Main3.html

It's either the SI-3300 or the SI-3170. I think AltaSens was
also coming out with a new chip soon. It might have been more
like $2.5K, not sure. If you check the other thread it is somewhere
in the last 10 pages I think:

http://www.dvinfo.net/conf/showthrea...threadid=25808

[Brett Erskine]

Rob-
I checked out the two models you listed and it looks like they are only 1/2" in size - not close to 35mm size. Perhaps you saw a different model. I'll keep looking.

[Rob Lohman]

No, it's actually what I meant. I DID NOT SAY there was a 35mm
chip out for $2K+. I said there is a 1080p chip out for that money.

A 35mm chip will probably be over $10K or something.

[Wayne Morellini]

There was a talk about:

http://videosystems.com/mag/video_shoot_tools_26/
JVC for under $2K, but everybod Video Systems must be too lazy to email me back about it to confirm this.

35mm cinema style is expensive as Rob says (if you do a search in the threads listed in the first post for Dalsa, Arri etc you might find some older pricing info). 1080 might also get much cheaper within the next year (though probably not from the cinema companies).

[Wayne Morellini]

OK fans of flash (with very deep pockets) 66MB/s compact flash 3:
http://www.tomshardware.com/hardnews...27_125452.html

Nanotech batteries, far superior, end of next year:
http://www.tomshardware.com/hardnews...28_182506.html

The ridiculouse: 1GB and 100GB email accounts, anybody want to send raw footage:
http://www.tomshardware.com/hardnews...28_182506.html

Underclocking CPU for low power, interesting:
http://www20.tomshardware.com/cpu/20041001/index.html

Disclaimer:
I didn't say any of this was practical for the here and now.

[Rob Scott]

Wayne Morellini wrote:
There was a talk about:
http://videosystems.com/mag/video_shoot_tools_26/
JVC for under $2K

I'm pretty sure that was a misprint, and it's really around $20K.

[Wayne Morellini]

I hear you Rob. But still, it's fun to hope, while they refuse to confirm or deny it (funny, the xfiles was just on, the truth must be out there ;)

[Marto Lautz]

hey Waine I have overcloking an AMD and it is ease to do .
but a cirix 1.3 or a via embeded prossesor motherboard for low power.
also last year I re-cell some old anton bauer batteries I have with 9000 mA cells this things run for hours and hours my camera.

well I planing to build a 1ccd cmos raw to disk camera now please let me know any news and or advise.
regards
martin

[Wayne Morellini]

Martin, most of this stuff is just for interest until people have something current to discuss. Better batteries result is better recharge cycles and smaller batteries for lighter/smaller cameras. Recelling some of the older batteries will make them much more efficient at a cheap price (cheap Asian cell provider), but you might require more depending on how your system is setup. Some people want to use the most powerful CPU so the power savings (which can be quiet a lot) from turning the clock rate down on a normal cpu (not overclocking) to the minimum needed, especially dynamically, is of interest. We know of the low powered CPUs, but consensus is that only the Pentium M has enough Mhz (cost a fortune for the top models though), VIA will have better speeds by the end of the year or early next. Cirix 1.3, thanks I'll have to look it up.

So the thread looks at not only throwing together a camera system but how to do it better in the future. I only mentioned the nano tech batteries because they said it was going to be available next year (which is an incredibly short time span in battery research terms).

Now if you goto the first post I have set up links to the threads where these projects reside, and the website:

http://www.obscuracam.com/

Now as for the other questions I'm sure that the guys doing the cameras will be by shortly to post advise on what hardware performance is needed for the latest version of the software. One chip bayer 1080p24 should be possible on some drives (read through posts to find drive review sites that give you min sustained rates) two drives would give you better frame rates. No three chips yet or SHD single chip (that would require 4 of the fastest drives at least). Sumix plans 3chip and on camera compressed versions (which for 3chip would be most excellent). Silicon Imaging also hopes to do compression someday.

Generally two camera companies are in the running: Silicon Imaging, and Sumix. At the moment only the newer editions of the sensors are any good (lots of smear/flare, low saturation/niose etc problems depending on the sensor), and they are introducing cameras based on the professional Altasens sensors, that will give you most of the performance you need at 2/3".

Big tip, most of these cameras don't include on chip frame buffer and readout time is faster than shutter speed. So when you require high shutter speed the frame needs to be transfered across the interface at least that speed. Further most cameras only transfer in 8 or 16 bit chunks (so 10 or 12 bit pixel takes up 16 bits). So you need to choose your interface/capture card with high enough data rate to run at the speed you want. Some people aim for PCI-X (not PCI express version) for at least 266 Mb/s. Hard Disks are not much affected as they only recored the average frame rate. Another big tip, USB 2.0 tends to be a CPU hog and not very reliable (they can program around this but at CPU cost). Gigabit/FirewireB (I've been told) Ethernet delivers similar to normal sustained real world PCI 100 MB/s. But those two interfaces require the latest/best drivers. Silicon Imaging has some for Intel based GigaE.

Rolling shutter is the other big problem, where pixels are read out one line at a time while still gathering light. So any movement makes moving objects slant (as it moves between scan lines). Few cmos sensors have a solution to this, the fillfactory has Global shutter sensors (the whole frame is locked down and read out), and Altasens has a faster readfing mode. In the project people are trying to decrease read out time raise shutter speed to minimumise this.

Thanks

Wayne.

Re-edit:

I've seen your other post, and wasn't aware you had read up on it. But maybe Rob would like to insert the above as the second or third post of the thread for new people starting from "One chip bayer 1080p24 should be possible on some drives".