4:4:4 10bit single CMOS HD project

[Jason Rodriguez]

Not necessarily. At NAB this year there's was a RAM recorder called the CineRAM that I saw everywhere with the big cameras because it was very portable and could record dual-link HD at 1920x1080 (around 215MB/s). You could stick up to 32GB in it via normal PC2100 DIMM's. I believe they're also working with ARRI for a design that attaches to their camera for portability.

So with a good RAM recoder, we won't have such problems wondering what hard-drives we're going to use.

For instance, right now on pricewatch you can buy 8GB of RAM (8 sticks) for $1900. That's two minutes from the Altasens chip at 24fps. The only problem is that there are no motherboards/computers that are small and support this amount of RAM (Supermicro boards are not small, nor are G5's).

I did see a frame-grabber card that had an on-board RAM disk, but it was serious $$$$'s.

[Wayne Morellini]

Sorry to hear this Obin. Listening to everybody here, I got to hoping that this would not turn out to be the case. It was plain that cameras shouldn't have been ready for a major production before the software was. Steve has offered upgrades, maybe he can send you some screen shots. If you put "Sponsored and Filmed on Silicon Imaging Cameras" in the credits, they should think of giving it to you, as sponsorship.

Sumix is, supposedly, putting more effort into supporting us, but we are yet to see anything.

I can hear those Jerks over at other forums mocking us now, but in a few months we will be in a much better position. Everybody take heart, if you look at pictures of old cinema systems the cameras were so big and heavy that they would require real cranes and trucks to carry them, yet blockbusters were shoot with them. A big system is just a hassel, not a death sentence. When we get the cases and finale software, we will be beaut.

Wish you the best.

[Jason Rodriguez]

Check out these Solid-State Disks:

http://www.bitmicro.com/products_edisk_transit.php

They can go up to 230MB/s sustained. Wow.

They can also store up to 152GB or something like that, although I'm sure that's a really expensive disk at that size. We'd only need around 8GB.

Okay, here's something that's not vaporware.

http://www.cenatek.com/store/category.cfm?Category=9

This can hold up to 4GB for $3,000. It fits on a PCI slot, but it will sustain up to 115MB/s, I don't think two SATA drives (while being cheaper) will do that right now.

[Obin Olson]

Laugh all they want! it's a Joke as it stands but I will fix that as soon as I can ;) and I am sure all on the list will do the same ;)

[Jason Rodriguez]

Ben,

How many f-stops of dynamic range do you think you're getting?

Or another way, do you think you're getting a useable 1000:1 dynamic range with the IBIS?

[Wayne Morellini]

From previouse posts in the threads (usually Steve's).

Micron is 10 bit, I Steve said there was 12-bit support. Data is packed in multiple of 8bits (interface is not serial) 10 bits goes to 16-bits. Maybe for some strange reason 10 bits was also packed into a 12 bit mode.

The 48fps, is because to have a 48th/s film shutter, we have te read the chip at 48fps (doubling the peak badwidth on the interface, but it is read out even faster than 48th/s (I think, it is in the thread messages) then the 8bit 16 bit packing makes things even worse, which is one reason why USB is not really suitable for 720p. Gigabit Ethernet cameras, and compression should solve this.

Obin overcame problems with his camera being dark in the shadows by pulling up the shadows out of the 10 bit values, maybe the SI can be induced to move it's range, even in the 8bit mode.

We have asked Steve about pixel packing and cheap memory buffer to smooth out peak bandwidth requirements to actual recorded fps before.

Maybe we should have approached Cinelerra to do the software, they have a HD capture system allready, it would only need to be converted to cameralink etc. If you want a quick solution it might still be the best option, they might even do it for free (Opensource). They either are going to say "yes, soon", "yes, eventually", or "no, get lost".

www.heroinewarrior.com

Rai, that is good suggestions on the Matrox, I wasn't aware they did cameralink cards. I have talked to Robs' and SteveI about getting help before, even to pay them for their work so they can work on it more, but they haven't expressed any interest yet. If Matrox would support it that would help.

You have a deadline, and obviously your company has some money. If you need something now, I posted a link in this thread (or the Viper thread) on the RAID camera unit they used to film Starship Troopers II. Cinelerra might also be ussuable with something like a blackmagic HD-SDI capture board (but I don't know if they have Linux support). It will cost a lot more money for these, maybe you could rent some of it and use us our system for your next production.

I will probably be emailing you over the next few days about manufactured parts (cases etc), if you can produce them, or help in design, it would benefit us.

Thanks

Wayne.

[Juan M. M. Fiebelkorn]

Ben, Obin and some others here using cameras.

I've seen (thank you Ben ) that the IBIS chips have a great quality.I've never seen a DV camera with so good quality at the shadows as the images Ben posted!!
About sensitivity: You cannot know how sensitive the cheap could be using those crappy lenses.They're too small to gather enough light.The only way to test this correctly would be using a SLR lens (50 mm would be okay) with an optical adapter (a lens) to reduce the projected image to the most suitable size for this chip.
This little trick would give you more light to work with.
Remember how PL Cinema lenses look like.When you look at their front lens it looks like a 6x4 one...
Also the faster they are, the bigger their front lens.

[Wayne Morellini]

<<<-- Originally posted by David Newman :BOXX is too pricy for this market, so we have been thinking of ways to introduce a lower cost product that doesn't hurt the higher-end market -- we are still working on this marketing issue -->>>

Good to hear, David. I don't think your high end system is too pricey, as there will allways be some people with the extra money that want to shoot the best quality is SHD+ etc.

Ben, I just turned my monitor way up above it's normal range, and like Obin's footage, your still's have a lot of details in the shadows. It's not too bad. I also have another tip, somebody mentioned it on a XL1s 35mm adapter thread, and I also asked about it in a thread last year. By making an 35mm adpator with suitable codensor, instead of projection surface, you can pipe all the light down to the chip (no loss from the projection surface). You don't get SLR DOF, but you get the angle of veiw and extra bright image.

Jason, I agree those flash disks are great.

[Juan M. M. Fiebelkorn]

Ben, the only problem I see with the images is their color balance (they are a little bit green )and that pedestal is way too low...

[Ben Syverson]

@Juan: You cannot know how sensitive the cheap could be using that carppy lens.They're too small to gather enough light.The only way to test this correctly would be using a SLR lens

That makes no sense -- SLR lenses are designed to cover a larger sensor size. They're neither better nor worse at gathering light. At 50mm, an SLR lens will be faster, but at 8 or 25mm, a C-mount lens will be faster.

And the 25mm Angenieux is an incredible piece of glass. The image wouldn't improve that much if I were using a $3000 lens.

@Jason: yes, you can capture raw bayer images. They also have a nearest-neighbor color mode, and a bilinear and laplacian mode as well. My plug-in still produces the sharpest and smoothest images out of all of them, although as we've discussed, there are artifacts that can crop up. If they get to be really bad, just go with bilinear or apply a very slight blur to the image after my filter.

@Wayne re: real-time processing: My Pentium M 1.6ghz can only capture at full-speed when set to raw bayer mode. In other words, it's not quite able to de-mosaic the image in realtime. Your hopes for a 1ghz machine that can de-mosaic, compress and write the files at 24fps are now officially crushed. :)

@All: yeah, the footage can still be manipulated a good deal. It may be 8bit, but it's a damn clean 8bit. Remember that until very recently ILM did all of their work in a custom 8bit mode. You can adjust the gamma of the channels a great deal before you see problems.

- ben

[Juan M. M. Fiebelkorn]

Ben, could you read again my post and give me a complete answer, so I can explain you what I'm trying to say? :)

For example: You have a 100 ASA stock, a sensitive area of 36mm x 24mm and you have a lens which gives you a light intensity A.
So if you CONCENTRATE (like a magnifier, sunlight and an ant)
that A intensity over an area of 18mm x 12 you are getting a higher intensity for surface unit, isn't it?

[Wayne Morellini]

<<<-- Originally posted by Ben Syverson : @Juan: @Wayne re: real-time processing: My Pentium M 1.6ghz can only capture at full-speed when set to raw bayer mode. In other words, it's not quite able to de-mosaic the image in realtime. Your hopes for a 1ghz machine that can de-mosaic, compress and write the files at 24fps are now officially crushed. :)
- ben -->>>

Re-read what I posted, I am still aiming for raw storage at 1Ghz, with optimised software, maybe compression. Still those real figures off of a celeron 416 are much more optimistic. The performance speedup for good programming, and optimal system setup (I sent Rob some links) canbe many times. It is not that simple, most things on PC's are crap and canbe greatly spead up by opotimisations and tweaking, end of story.

[Ben Syverson]

@Juan:So if you CONCENTRATE that A intensity over an area of 18mm x 12 you are getting a higher intensity for surface unit, isn't it?

You don't understand lenses at all. That's what a proper sized lens does. An SLR lens adapted to a smaller sensor does not "concentrate" intensity over a smaller sensor -- it's designed to deliver an image for a large sensor, so when you use it with a C-mount camera, you're only using the center of the lens -- the rest spills out around the sensor and is unused. You're wasting light, not concentrating it. That's why all the digital SLRs have a multiplication that you apply to the lens to figure out how wide you can get. For example, a 28mm is wide in 35mm still format, but with most SLRs, it's not very wide at all. The extra FOV is being thrown away. You need a good primer on optics.

Go look at a C-mount adapter. There's no optics on it. It adjusts your SLR mount to the C-mount thread. The video sensor is just acting like a tiny crop of 35mm film. That's totally worthless to me, since I like to shoot wide. They don't make many 8mm lenses for the 35mm market, and the ones that are out there are expensive, slow, and rather bad. (The Sigma jumps to mind).

A c-mount camera is designed to concentrate all the light it gets onto a smaller sensor. But there is a whole world of C-mount lenses, from 16mm film lenses to 1/3" video to 1/2" video to 2/3" video. Every lens family is specifically designed to cover a certain sensor size.

Now quit keeping me in the house -- I want to go out and shoot! :)

- ben

[Juan M. M. Fiebelkorn]

You don't understand me.I mean CONCENTRATE (put a lens between the SLR and the sensor), not WASTE.

Anyway you don't like what I'm saying :(

Please if someone here could explain him what I'm trying to say, it would be nice :)

[Ben Syverson]

Oh, you're talking about a "real" SLR->C-mount adapter.

I don't think that would work, or else it would have been done before. I don't think there's enough room to mess with that without changing the back focal distance or focal length. Why wouldn't the C-mount adapters on the market incorporate that? It would be a great selling point. I think it's not possible (or at least, not easy).

But if anyone has a way to do that, I'd be interested to see it.

As it is, SLR lenses are basically useless to me unless I want to shoot ultra-telephoto. The 25mm Angenieux is a magnificent little guy, and is telephoto enough for me.

- ben