If the chips is only 2/3'' then how different is it from CineAlta? Yeah sure CineALta footage is more compressed but for a love of God i can't see that compression even if i look really hard. So... Unless i got a deal of a century i would still do with CineAlta. Then again if this camera can use 35mm lenses without any adapters -- then yeah we got something here!

On using lenses designed for 35mm cine: 2/3 inch is only a liitle larger than 16mm (0.5inch diag). The 35mm frame diagonal is 1.22 inches (roughly 2x). Probably lenses for Super16 would be more suited (0.57 inch diag). Of course 35mm cine lenses could be used, but with half the FOV of using them on a film camera.

Edit: as Bob says, not an ENG camera. But it sure looks sexy to me! In particular I liked what the cineform guy (David) was describing.

I hope you are all reading and watch the behind the scene from the shoot of the feature "Spoon" which is using two Silicon Imaging camera. Read more over on http://indiefilmlive.blogspot.com

David Newman
CTO, CineForm

I'm just happy we're finally getting away from tape as a recording medium. Although, I don't think hard drives are a good idea here. Aren't those things sensative to even the slightest bumps ?

Really wouldn't want to alter my shots so I don't have to worry about jolting that removable disk drive.

I could be wrong here, what they were removing could have been a more rugged drive. But even if it is, why go from tapes with all it's moving parts to a hard drive with even more moving parts ??

Just get moving on those 10, 20, 50 gig memory cards. That's what's needed for the next generation of digital cinema.....

Just my thoughts...

Stan

Well, you'll see on the Spoon blog that frequently they're shooting head only. Makes a much lighter rig and the recording drive is away on the end of a cable. I doubt that you'd want to do more dynamic camera work than they're using.

And for anyone who hasn't worked it out that cable is only off the shelf Ethernet cable, standard ethernet rules apply, so your recording device can be 100M from the head.

But the 'recording' device is only a fairly run of the mill PC, what you record to doesn't have to be a HDD, it could be CF cards, P2 cards, any device that WinXP can write to fast enough will do, BD disks etc. And if you really feel nervous about any of those, well there's always RAID 1 or 5 disk arrays.

But the most important part is WHAT you're recording. Basically it's RAW data along with metadata that defines how you'd like the image to look, those decisions are not applied during recording and can be changed at will in post, pretty much the closet you'll get to shooting S16 film without shooting film.

Sorry to nag on about this camera but it's come onto the scene with appallingly bad marketing, silly guys probably spent all the money on engineering so there's no hype, no tent, no sign, no pay now and hope someday you might see a working camera in your hands.

Bob.

Farss, I think shooting two movies, with active blogging, is great marketing on a dime!
No vaporware issues, like Pan's secret pre-release camera issues, ( ie, hiding technical details of CCDs design). The issue with the SI is "raw data" approach in pre/post workflow ( a major $$$ saving), the small film production people should like this also TV production studios.

Stonefield -- "I'm just happy we're finally getting away from tape as a recording medium. Although, I don't think hard drives are a good idea here. Aren't those things sensative to even the slightest bumps ?"

Not at all. We see 2.5" drives in iPods and in laptops, with these drives taking 100G shocks. Far more reliable than tape. When flash is big enough that can be used too. Fact the camera doesn't require a HDD, it will record to anything on USB2 -- you can capture directly to a thumb drive today.

David Newman
CTO, CineForm
www.cineform.com

I guess your right, when you have lot's of choices of storage media, ya shouldn't be too concerned about flimsy drives.

I love the idea about capturing ( filming ? Heh ) in the RAW format. If I'm clear it's the theory of getting a solid, well exposed sharp image, then being able to manipulate it all you need in post. ( Can being able to manipulate something even DURING capture even be called "post ?" )

Coming from a 35 mm stills background, it's interesting how much I enjoy the post process of digital grading and FX. I still admire the idea of the cinematographer out there with his assorted filters, lighiting, lenses and capturing everything "in camera" the same way it's been done for over 100 years. But just as I'm done with film for stills and anxiously waiting to shoot digital stills, I find the idea of shooting RAW video very interesting.

Quick question....how would shooting RAW benefit people that like to use blue/green screen compositing ?

The CIneform RAW format is compressed, so technically it throws away from information as it compresses the image. However, it uses wavelet compression instead of DCT compression (i.e. DV, HDV). Whereas DCT adds things like mosquito noise and other artifacts, wavelet sort of "blurs" the fine details image (AFAIK). At mild ratios, it's hard/impossible to see the artifacts. Wavelet is generally more efficient, and takes more CPU to decode.
*HDV can be horrible to decode since it's interframe compression and not intraframe like DV and cineform. Just because it's not wavelet doesn't mean it decodes faster than wavelet.

Quick question....how would shooting RAW benefit people that like to use blue/green screen compositing ?
Cineform RAW implies a Bayer-like single sensor, which won't provide the full color resolution of the frame size of the sensor. So the SI sensor has 1920X1080 pixels... it won't give you 1920X1080 color resolution. Actual performance likely comes down to its implementation.

A more ideal sensor would be a 3CCD design that's 1920X1080, or a Bayer pattern that's highly oversampled (sort of like Genesis, and what Red is proposing).

Compression-wise, wavelet compression should be a reasonable cost compromise since it doesn't introduce nosie like DCT does (*this is just my guess, I haven't worked with footage from the SI camera). Obviously no compression would be ideal, but by that line of thought everyone would be shooting on 70mm film (or better?).

At SI's price point, I think it would be a top contender for green screen performance.

My understanding is that a 3 CCD 1920 x 1080 block creates more problems than it solves. Maintaining optical alignment is one problem and secondly lenses should really be designed specifically for the task. I certainly don't understand the finer points of the lens issue but it seems you can't put a film lens on the front of a 3 CCD camera and get good results. I think that explains why every digital film (still or moving) camera has gone with a single chip design.

As you've rightly pointed out with a Bayer chip at 1920x1080 you can never get full color res, interpolation is used. Even so technically 3 CCD cameras such as the Z1, XL H1 and HVX 200 aren't using Bayer sensors they are using 2D or 3D pixel shift, unless I've missed something this is pretty much the same thing, the RGB sensor elements are not providing the values from the same point so again interpolation is required.

However the real trick with this system (as I understand it) is to move that interpolation from prior to encoding to post encoding and recording. So the efficiency of the Cinefrom compression is vastly different to what's happening when it's used with say HDV.

With CF and say HDV, bandwidth is being used up encoding interpolated data. By having access to the raw data from the imager and only encoding what the imager sees and leaving the demozaic function for post more actual true imager data can be recorded. One advantage of this is a very significant improvement in latitude.

Also there is some history involved. The imager in the SI camera was an existing design, already in use for industrial imaging systems. I think the imager in the Dalsa camera comes from the same background. Designing an imager from the ground up and getting it built with reasonable yield is no cheap process, so cost wise the camera benefits greatly from work already done. Ideally I'm certain a bigger sensor would be better but all the associated costs have to be borne by the user. And there's another problem, more pixels means faster A->D conversion, even more so with progessive scan. Eventually you simply run out of steam and have to start splitting the block into 2 or more parts like the HD100 does and that can introduce another set of technical problems.

Of course in the end all that matters is how it looks on the big screen and how well it composites and from what we've seen so far it holds up very well against systems costing 10 times the price.

Bob.

On lenses and prism blocks: putting glass behind a lens affects the design of lens corrections. You'll remember that the Bolex RX cameras needed RX lenses because of the rectangular optical block between lens and film (the block diverted light into the TTL viewfinder). Using non-RX lenses was OK at smaller apertures. It will be the same with 3CCD cameras. Don't know about arrangements in the Canon XL-H1, but I note there are specifically designated H1 lenses; however the adapter for 35mm still lenses contains no optics.

My understanding is that a 3 CCD 1920 x 1080 block creates more problems than it solves. Maintaining optical alignment is one problem and secondly lenses should really be designed specifically for the task. I certainly don't understand the finer points of the lens issue but it seems you can't put a film lens on the front of a 3 CCD camera and get good results. I think that explains why every digital film (still or moving) camera has gone with a single chip design.
I think that was only when you wanted to get higher resolution... i.e. 4K. Although I'm not sure now.
I think (but I'm not sure) the Sony HDCF950 has 1920 X 1080 CCDs.

I believe the reason you can't put a film lens on a 3CCD cameras is because the prism block is long. The distance between lens and sensor has to be kept at a certain length.

However the real trick with this system (as I understand it) is to move that interpolation from prior to encoding to post encoding and recording. So the efficiency of the Cinefrom compression is vastly different to what's happening when it's used with say HDV.
I believe Cineform is efficient because it uses wavelet instead of DCT compression. Interframe compression (like HDV) would be even more efficient, but really bad for editing.

Recording compressed RAW instead of compression 4:2:2/non-RAW/whatever is a trick specific to Bayer(-like) designs.

Even so technically 3 CCD cameras such as the Z1, XL H1 and HVX 200 aren't using Bayer sensors they are using 2D or 3D pixel shift, unless I've missed something this is pretty much the same thing, the RGB sensor elements are not providing the values from the same point so again interpolation is required.
With pixel shifting you can get similar problems to Bayer pattern sensors.
However, there are 3CCD cameras that don't pixel shift (i.e. Sony F950). They have significantly less color aliasing, and full "4:4:4" color resolution (same color resolution as luminance resolution).

One advantage of this is a very significant improvement in latitude.
That's a result of changing the camera's transfer function / not applying gamma compression. Any camera can have this feature... and it's a good feature.

I think the imager in the Dalsa camera comes from the same background.
As far as I know, Dalsa designs high-resolution imagers for the defence industry and other specialised uses.

SI gets their sensor from Altasens, which targets a broader market. Unfortunately, Altasens is not shipping a 35mm-sized sensor.

Of course in the end all that matters is how it looks on the big screen and how well it composites and from what we've seen so far it holds up very well against systems costing 10 times the price.
What have you seen btw?
From what I've seen of the SI camera (the WMV-compressed Spoon footage), it's not as impressive as what "Broken" did with the DVX100. This is for non-technical reasons.

"What have you seen btw?"

Only the Spoon footage and the camera in operation at NAB.
There's a lot more footage up now and some of it uncompressed tiffs.


It's pretty hard to judge without having access to the raw footage and the grading tools plus looking at it on a LCD I guess kind of negates the whole concept without playing around with the grading.

Bob.

Update: We just shoot a short movie for the 48 Hour Film Project using the Silicon Imaging camera. Check out the project details and download the HD film from http://www.cineform.com/48hour/

David Newman
CTO, CineForm

Aside from the way it captures video, the fact that the camera body can be upgraded/replaced at your local computer store is a great thing for low budget movie making. That and using an affordable off the shelf gigabit Ethernet switch doing double duty as a realtime video switcher.

Unfortunatley, like Greg, I've yet to see anything that surpassed the current crop of HDV cameras in visual quality, and that was looking at the stuff at NAB. I do think they are getting better with each improvement in software, and you really can't use highly compressed wmv from uncorrected dailys as a benchmark.

What gets me excited is how they are using off the shelf components and innovative software to do all this. Aside from the camera head itself, no high priced proprietary stuff. Even if the camera doesn't end up doing well, I hope it inspires others to create siimilar products. I'm looking forward to the day $100K tape decks are museum pieces.

btw, it uses S16, not 35mm. Silicon Imaging has been recommending Arri Primes.

That and using an affordable off the shelf gigabit Ethernet switch doing double duty as a realtime video switcher.
I don't believe a gigE switch will do that??

Switching to me has two different meanings... in a computer networking context, and in a live video production context. Different things.

Unfortunatley, like Greg, I've yet to see anything that surpassed the current crop of HDV cameras in visual quality, and that was looking at the stuff at NAB.
It would seem to me that the SI camera would offer a leap in quality over HDV.
- More dynamic range captured
- full 1920x1080 raster captured
- (speculation) less objectionable compression artifacts than DVCPRO HD or HDV... lets you push the colors around more in color grading. DVCPRO HD is noisy IMO.

"DVCPRO HD is noisy IMO"

I don't think the codec itself is noisy, just that the most popular camera currently using it is noisy. DVCPro HD should handle noise better than HDV however it's almost inevitable that at some point in time the video is going to get hit with mpeg-2 compresion and that is never kind to noise or film grain.

Many people describe DVCPRO-HD as being "less compressed than HDV," but it would be very reasonable to argue that the opposite is true.

Intraframe compression is far less efficient than MPEG-2, so the 100Mbps are no match for overwhelming the ugly DCT compression, leading to artifacts that can look very much like other noise.

DVCPRO-HD is not great for post work either, other than for convenience (low demands on the machine).