The general idea is to have the camera as a seperate system
from your processing computer so you kind of download the
things you shot from your camera (much like you capture now)
and then clear the harddisks for your next recording.

Let's consider this. If you where to record in 1280x720 @ 24 fps
in 10 bit the camera is sending us 16 bits which gives us a frame
rate of 42.19 MB/s. However, we will at least pack this before it
goes to the harddisk and this will yield 26.37 MB/s.

Now you put a 200 GB harddisk in your system (or two 100 GB
drives for exampl) and you can then record for 7766 seconds or
129 minutes which is just over 2 hours.

I am working on a fast and simple compression algorithm to
hopefully lower the bandwidth a bit to perhaps allow recording
without raid and a bit longer recording times.

Just to give you an idea of the massive amount of data: if we
where to decode this to FULL RGB without ANY COMPRESSION
then you where looking at a datarate of 126.56 MB/s or for
200 GB recorded will expand to 960 GB [2 hours].

So that basically rules out uncompressed for final edit use as
well and we will need to look at some form of compression there
as well. Some people have opted to use real-time bayer in a
codec, but I'm not sure about the quality vs. speed on that one.

Spencer: personally I would not pitch such a system at this point
in time. It isn't ready and it will not be in a while to come. It is all
in the development stages and nobody has even shot a movie
with any of this. It is not ready for primetime at this point in time.
As you correctly identified the workflow is still somewhat of a
problem.

Steve: regarding the SI-3330, can you read this chip out at a
resolution of 2048 x 1152?

Rob:
Native is 2048x1536 so anything less can be read out. Frame rate is the problem. You would have to run the chip at over 70MHz to get to 24fps. I'll try it later today if I can.

Spencer,

For what it's worth, I was editing on a PC platform for 7 years and just now bought a G5 with FCP HD . . . and let me tell you . . . .

It's like I died and went to heaven. Stability is rock solid. User-friendliness is awesome. Capabilities are awesome. Speed is awesome.

Macs are awesome, only 2 months with it and I am a complete and total convert. If you can afford it, the Mac is WELL WORTH IT.

Just IMHO.

Re: Feasibility of the whole Workflow
 

I haven't used a Mac extensively in several years, but I've heard very good things about FCP and it supports 10+ bits per channel directly, so it's a good match for a camera like this.

In fact, this weekend, a cousin of mine (actually a cousin-in-law, who oddly enough is also named Rob!) is going to be in town and we're going to try out some clips with FCP HD. I'll let you know how it goes. Unless i can get native QuickTime support in there in the next few days, we'll probably be transcoding from 16-bit TIFF sequence into QuickTime.

--- EDIT ---
Minor update to the development blog

Thanks for all the replies,

Rob L:
200GB disk for 2 hours sounds pretty good. We would rarely shoot more than 4 hours per day, so a couple 200GB disks wouldn't break the bank, which is nice to hear.

I completely understand that it's in the test phases now, but people seem to be moving rather quickly. My father is an electrical engineer who has helped me homebuild anything from jibs to body mounted steadicams, so this would be a welcome challenge to him and myself. The movie is still very up in the air, and shooting would not happen untill around this time next year, do you think that's a little too wishful thinking still? It's cool if it is, we'll just be forced to make the most of that dreaded SD :( hehe

Laurence:
I hear ya, I'm on a g4 with a 23" cinema display right now at my internship...the thing is most definitely impressive. I can't bring myself to switch completely to using a mac though, and cannot justify owning two expensive rigs at once. I just want to make sure that PC's work well with quicktime editing, 'cause myself I've only worked in Microsoft DV avi's.

Thanks again for your responses
Spencer

ASA?
 

Does anyone have any ideas about how to determine equivalent ISO/ASA for a sensor at various gain settings? I guess I could purchase the ISO standard, but I'm not sure if it would help, since it's meant for film.

I set up a new thread for this if anyone has any ideas.

Thanks!

BTW, Rob,

If you want to keep things RGB instead of YUV, use the 10-bit RGB codec from Blackmagic. You'll have to download the version 4.x version for the HD Pro card, but just install the RGB quicktime component instead of the whole driver package. If you need to, I can email you the codec.That's lossless compression, right?

BTW,

a quick question based on Spencer's questions:

Do you think by November/December we'll be able to start shooting with this system (as a beta-tester)?

I've found all the parts I need to create a very compact, portable, shoulder-mountable camera system, the only thing that is in question is software. I'll probably be using 4 2.5" SATA drives in a RAID 0 config using the dynamic disk (software RAID) options in Windows XP. Also it will have PCI-X, so It can handle the Altasens. A Pentium M at 2.0Ghz will round out the processor power. An Anton Bauer HyTRON 120 will give plenty of power for this system, since as of right now we're hovering around 60W of total power. And instead of a viewfinder I'll use a touchscreen from Xenarc. So if I was to build this, and combind it with the Altasens, do you think that shooting would be possible by December/January (provided the Altasens is out in October)? I'd like to start off by shooting a 10-minute short with this system.

I think this time next year I will have a complete camera system that is easy to use and costs a bunch less then any other HD system out now..not to say that EVERYONE will not have a mpeg2 HDV cam by then but ours will be a professional RIG not some HDV garbage..so yes I think you will have a system by that time ;) I sure hope we do!

BTW i have seen a test of the 3300 RGB from Silicon Imaging..it looks VERY good...NO smear at all guys! I will have one in my hands maybe as early as next week...I will keep everyone posted...we are still working through the software, no big news yet...

Jason I need priceing of realworld micro boards with P M CPU..do you have any that are forsale NOW ? not some future design?

BTW make sure it has DUAL HEAD graphics so we can have a viewfinder AND a touchscreen LCD....

Jason I have a 1024x768 touch screen ready for our camera..it's good but not as bright as a standard non-touch screen...resolution is great for such a tiny screen...what is the res on the one you found?

here is an awesome item:

http://www.grandtec.com/wirelessVGA.htm

for transmission of preview screen from camera to a monitor for follow focus and or a Director monitor...awesome...I will build a custom follow focus system for our camera that will use standard 35mm SLR manual lenses and a wireless focus system like I build for the dvx100...sweet!

Obin, the SBC83810 from Axiomtek is a 5.25" board (5.75"x8" or the size of a 5.25" enclosure/CD-ROM, etc.) with Pentium M, PCI-X, 2 SATA ports, MiniPCI, PC104, USB, VGA, LVDS (dual head graphics), 1GB of RAM, 40-pin and 44-pin IDE connectors, and costs $489. They can ship one to you in a week if you want it.

If you need more SATA ports, you can add the SATA-2R miniPCI card with two more SATA ports from Kontron. That's what I'm planning on doing with a total of 4x2.5" SATA drives from Fujitsu, and then stripping them using Window's built-in software RAID. I'll then also add a 1.8" 20GB drive from Toshiba for the system drive.

Contact David Ramil at ramil@axiomtek.com and tell him I sent you. He's our sales rep since we're both in the Virginia/NC area. They're in California, so be concious of the time differences.

BTW Obin, what are you going to use for a viewfinder? One of those Kopin EVF's? Also I haven't found any small-sized LVDS touchscreens, most of those interfaces are around 15" or more for LVDS. The small 7" touchscreen I've found is the Xenarc, and it's nice because it includes this keyboard emulation utility and right-mouse button emulation, so you can do without a keyboard.

Something tells me that Obin r-e-a-l-l-y wishes that the HDV thing didn't exist ;)
I like using it as a reality checker. Did you guys like the dollar bill shot? ;) Not bad for 1.1 Mpixels!

Personally, I wish that they didn't use decades old mpeg2 compression, but what the heck. It sure beats the web cam DV stuff.
I am looking forward to a 2 Mpixel plus single chip high bit depth camera. I too will set up a camera system to use such an industrial style camera. Just wait for a Taiwanese company to produce something. Like what they did to the flat bed scanner market. Sumix ? Bring it on.
-Les



<<<-- Originally posted by Obin Olson : I think this time next year I will have a complete camera system that is easy to use and costs a bunch less then any other HD system out now..not to say that EVERYONE will not have a mpeg2 HDV cam by then but ours will be a professional RIG not some HDV garbage..so yes I think you will have a system by that time ;) I sure hope we do!

BTW i have seen a test of the 3300 RGB from Silicon Imaging..it looks VERY good...NO smear at all guys! I will have one in my hands maybe as early as next week...I will keep everyone posted...we are still working through the software, no big news yet... -->>>

<<<-- Originally posted by Jason Rodriguez : That's lossless compression, right? -->>>

Definitely Jason! I have it running for 8-bit at the moment and
will be testing it for 10/12 bit and how fast it can compress this
weekend. I'm not ease to throw much numbers your way but
it looks like it can do 40% reduction on average. Again this is
for 8 bit files and we'll have to see how well it does in 10/12
bit and how fast it will be in C or in Assembly (hopefully fast
enough for a realtime implementation => it was "designed"
with that in mind, so hopefully that will show as well)

Steve: the reason I'm asking about 2048x1152 is that this will
be a 16:9 frame on the 4:3 sensor. If we can drop some bayer
blocks and lines this shouldn't be too hard to transform to
1920x1080 or 1280x720 in real-time and maintaining the full
field of view. The only question is how good it will look and I
hope to test that out with your original 2048x1536 file this
weekend.

BTW Steve, did you get my e-mail?

It's really hard to estimate a time frame with so many variables, but I think it's very likely. A lot of it depends on how much rework will be required to support the AltaSens camera.
Production quality? Hard to say. It might be suitable for a ten-minute project, but I'd have a "Plan B" if I were you :-)

We'll see how it goes over the next few months.

Well, the 10-minute film could be "on-going" if you know what I mean. With my own camera it doesn't have to be shot in one weekend :-)

What kind of compression Rob?
Arithmetic, Huffman (fixed table, dynamic),RLE, Dictionary variant.....
Could you tell us more?