JVC Micro-4K Prototype

[Steve Mullen]

THIS DESERVES IT'S OWN THREAD.

The JVC FalconBrid DSP used in its coming 1080p60 36Mbps H.264/AVC camera can handle 4K2K video. Link to JVC 4K2K prototype camcorder:

JVC GC-PX1 High Speed Concept Camera

Watch the video at the bottom.

YouTube - JVC 4K2K Camcorder and TV

YouTube - JVC videocamera 4K una novità del CES 2011

It looks like JVC can PROCESS 4K2K but not yet compress more than 2MP. That would imply a different codec may be within the chip. What's 4X faster to compress than H.264? They could license the CineForm codec. Or, something from Sony.

The CMOS chip must be able to run at 60fps reading out a 16:9 8MP window on the chip. In turn this means the OLPF can be optimized for 10MP for both stills and 4K2K so all the artifacts we get from the other cameras will be gone.

[Glen Vandermolen]

The linked article states that the sensor is possibly 1/4 the size of a 35mm sensor, or into the "point-and-shoot" category. Does anyone know what that relates to in comparison to video camera chips? Is it bigger than a 2/3" chip?
If it's comparable to a 2/3" chip, that would make this a mini-Scarlet. Well, sort of. No RAW video, but 180/60P at 36mbps is pretty cool.

edit - OK, it has a 1/2.3" sensor. Typical consumer camcorder size.

[Steve Mullen]

In case it's not clear, and you didn't watch the videos, there are two different camcorders that will get the DSP chip. The PX1 is one of them. It is not a 4K2K camera. It has a 1/2.3 chip.

The 4K2K prototype is built in an HM100 body. In this case, unless JVC plans to reuse the body, the sensor size isn't relevant. Since JVC buys it's CMOS chips from Sony, the JVC pro division can replace the tiny chip with an APS size depending on the size of the camera package. (Assuming, of course, Sony will make a big chip with only 10MP.)

-------

The point of my posting is neither the PX1, which is detailed in the VG10 forum, nor the prototype 4K2K camera!

The important item is the DSP chip. That is REDs secret sauce and why RED works differently than all the other big chip cameras. It is the key to the Japanese being able to make a RED.

It also is the camera's codec, so it determines HOW 8MP at 60fps can be recorded. Will it be RAW video? Lightly compressed? Because this is 4X more data than anyone but RED stores, it determines the type of storage that will be used.

Lastly, will the DSP work in 4:2:0 or 4:2:2 or 4:4:4 RGB? A consumer version would likely use only 4:2:0 but if the chip can do more it gives the pro division an option for a much better camera.

If you watch the video the JVC presenter seems clear the DSP is going to be used in real products. Since the PX1 is going to be put on sale next month, it seems likely the 4K2K camcorder will also come to market.

[Lawrence Bansbach]

Originally Posted by Steve Mullen

The important item is the DSP chip. That is REDs secret sauce and why RED works differently than all the other big chip cameras. It is the key to the Japanese being able to make a RED.

If Sony and Panasonic are any indication of the Japanese manufacturers' response to Red, none of them -- including JVC with its 4K2K technology -- seem to have any idea of what people want. Sony's F3 is somewhat more expensive than the projected price of the Epic-S, and more than double that of the 2/3-inch Scarlet. Further, the F3 is only 1,920 x 1,080, whereas Red's forthcoming cameras will start at 3K.

Pansonic's AF100 is more price-competitive with the Scarlet, except it's an 8-bit camera with a heavily compressed codec recording, again, only 1080p. Want 4:2:2? Pay anywhere from $2K to $9K for an external recorder, and even then it will still be only 8-bit. Oh, yeah, and the lens is extra. There goes the low price.

JVC's 4K2K doesn't seem to be 4K at all, but Quad HD. That's better than Scarlet's 3K, but it seems to rely on a heavily compressed codec as well. And it works only at 60p. Last I checked, the film standard is still 24 fps. Sure, 24p isn't everything, but not having it limits the technology's utility.

More interesting to me is the CMOSIS CMV12000: 4,096 x 3,072-pixel, APS-C class global-shutter chip. It's supposed to be able to run at up to 300 fps full frame. Shame it's only 12-bit maximally -- 14- or 16-bit would be better. And to my knowledge, none of the big players will be using it.

In short, while JVC's technology is interesting, on price, features/specs, and codec, Red seems to win.

[Steve Mullen]

Good, but somewhat needless, defense of RED.

1) 60fps is the maximum frame rate and needed by those of us who want 1080p60 -- not some last Century film look. No reason to assume the chip can't run more slowly. And, you want 8 MP 60P for SLO-MO.

2) You'll note the DSP has two video specs. One is for video processing. One is for the codec. I assume they are connected together for the tiny P1 camera.

3) You raise an interesting point about quad HD. This chip can compress dual HD for 3D. (I assume that is 2 streams of 30p (60i).) But, let's assume they have someway of compressing four 2MP frames and recording them on, perhaps, a pair of SD cards. Now further assume they provide an FCP plug-in that combines the 4 streams into one 4K2K ProRes 422 HQ stream. (Or, quarter samples each stream for one FullHD AIC stream that we edit as a proxy.)

Do I really care HOW the 8 million pixels got to FCP? By the time I scale it down to FullHD, I'm going to have something that looks very nice. (And, if I can write 4 streams back to the camera, I can feed JVC's 4K2K HDTV via 4 HDMI cables -- which is pretty much what one had to do with HDV until there was BD.


This is why the chip -- and not any specific camera is important. Without a DSP that can process 4K2K from a 4K2K sensor at full motion speeds, RED would be simply yet another HD-DSLR that skips lines and interpolates pixels to get HD video. They key is getting 4K2K from the sensor to the recording media.

Doing it at $2000? Priceless!

[Lawrence Bansbach]

Originally Posted by Steve Mullen

Good, but somewhat needless, defense of RED.

Not defending Red at all. My point is, at all but the consumer price points, Red's vertical integration of the tech will produce, at least on paper, the best image and at the lowest price. It's not that that Sony, JVC, Panasonic, and Canon (whom we haven't heard from in this area, save for the "hairdryer" concept camera) can't compete with Red on both the spec and price fronts, but they just don't seem to want to.

60fps is the maximum frame rate and needed by those of us who want 1080p60 -- not some last Century film look. No reason to assume the chip can't run more slowly. And, you want 8 MP 60P for SLO-MO.

True, there's no technical reason the chip couldn't run at 24p/25p or 30p. But the manufacturers tend to hobble products to provide certain features to certain markets. Sony, for example, for the longest time needlessly omitted 24p from its consumer and low-end prosumer cameras. Let's hope JVC is a little more forward-thinking.

In its defense, Red listens to its customers, albeit sometimes grudgingly. For example, after Scarlet 3K was announced at NAB in 2008, there was an almost immediate request for an interchangeable-lens version. Jim Jannard was notably peeved. Yet it happened. The problem is, in trying to give customers what it thinks they want, Red keeps tinkering with its designs. This may improve them, but has definitely resulted in many delays and significant price increases -- so much so that Jannard had to throw in the towel and concede he couldn't produce a consumer-priced camera.

This is why the chip -- and not any specific camera is important. Without a DSP that can process 4K2K from a 4K2K sensor at full motion speeds, RED would be simply yet another HD-DSLR that skips lines and interpolates pixels to get HD video. They key is getting 4K2K from the sensor to the recording media.

Doing it at $2000? Priceless!

At a consumer price point JVC might have a chance. I just hope that they don't strip out features, such as 24p or the ability to use a better codec, preferably one that works at 10 or 12 bits and at least 4:2:2.

[Steve Mullen]

I can see how the chip can be used in several ways:

1) A single 8MP chip's data are fed to the DSP where it is interpolated to 2MP (1920x1080). The reduction ratio is 4:1. These data are compressed at 36Mbps. All parts running at 60p so the stream is 1920x1080p60.

2) A 3D pair of 3.3MP chips running at 60p. First 1/60th second, each chip sends its odd lines to the DSP. The DSP inputs both streams and interpolates to 1080-lines (540-lines above and 540-lines below) and also interpolates 3840 pixels to 1920 pixels. The DSP sends 1920x1080 pixels to codec as odd TALL field (2MP). The reduction ratio is again 4:1. Compressed and stored as above/below 3D odd field.

Next 1/60 second, each chip sends its even lines to the DSP. The DSP inputs both streams and processes and sends 1920x1080 pixels to codec as even TALL field (2MP). The reduction ratio is again 4:1. Compressed and stored as above/below 3D even field.

So in 1/30 second, two 1080-line interlaced frames have been recorded. Each chip stream is 60i compressed at 18Mbps. Or, is it 36Mbps?

3) single 8MP (4K2K) chip running at 60p. 8.3MP is sent to the DSP. The 8.3MP are then sent, without interpolation, to the JPEG still-image compressor. It can JPEG compress 8.3MP images at 60fps. This would be a "motion picture" as RED likes to call it.

Depending on the amount of compression used, this system would be very competitive with Sony and Panasonic using 24Mbps AVCHD.

The fact 4K2K can be done, doesn't explain why JVC would do this as a consumer product. Moreover, the prototype was clearly a prosumer to pro level product.

My guess is that the consumer group is doing a 4K2K product it will share with the pro group. They won't sell many to consumers but will get tons of free PR for JVC just like they did with the first HDV camcorder.

[Steve Mullen]

I just realized that if they are going to reuse the hm-100 body the optical system is going to be geared toward a 1/3 inch chip. Which means the consumer 1/2.3 inch chip will be used. No shallow dof, but at least the chip isn't crammed with 16MP.

However you can bet there will be a 790 class unit that may have a bigger chip. It would make sense to use a 2/3 inch because going APS means no power zoom. And there are 2/3 adaptors for that want to use other glass.

What I didn't realize how far along 4K2K is in japan!

Nab was so focused on 3D but the other NEXT BIG THING is 4K2K.

[Glen Vandermolen]

A 790 (or now, the 750) with a 4K chip? Now, that will turn the industry on its ear. Think of the images it could create! Sign me up!

[Alister Chapman]

Sure and exactly how much will a lens with a high enough MTF to deal with 4k on a chip smaller than 2/3" cost?

While the sensors and chips are possible the optics are unrealistic. With a 2/3" sensor, to resolve 4k you'll have to have your lens more open than f4. Smaller sensors... for 4k... not really practical IMHO. Try making a zoom lens for a 1/3" or 1/4" sensor with an MTF high enough for 4k and best performance at f2.8. Not gonna be cheap.

4k needs a bigger sensor to make sense, then you run into the cost of fast glass again, but at least with a big sensor, diffraction is less of an issue, so lens speed (from a resolution POV) becomes less significant.

I personally think that Scarlet at 3k on 2/3" is probably close to the limits of what is practical. Well, 3K resolution as opposed to 3k pixels that is.

[Paulo Teixeira]

I wouldn't even be surprised if the chip is around 1/2 or slightly bigger.

At least we know JVC has the glass issue covered. The HM960 with a 1/2.3 chip has a lens that goes to f/1.2.

The codec for the 4k2k camera is not that bad at 144Mbps using H.264. You do need a decent computer.

[Steve Mullen]

While some know the lens may be the limiting factor I don't think this matters when it comes to marketing 4K2K. Who measured the mtf of the hd1 or the z1!

Thinking about resolution misses the key point of an 8MP flow from sensor to file. All the photocites in the 16:9 window are processed and compressed. Only RED can do this now. All the other hddslrs skip columns and/rows to drop the amount of pixels by about 6x. So even if the camera doesn't resolve the full resolution the image is going to look great.

[Evan Donn]

This also just highlights that we're currently in the middle of a transition period where the sensor capabilities have dramatically outstripped what can be practically/cheaply processed in a small camera. 2 years ago this was a real limitation and we got the combination of big, high-res chips limited by the need to line skip. Now with this chip we see the processing side catching up. Give it a couple more years and we'll hit a point where inexpensive hardware will be capable of pretty much any combination of resolution (within reason), frame rate, codec, etc - and the only limitations will be those placed in the firmware by the manufacturers.

[Steve Mullen]

Transitions a good word. But I wonder how many of those buying one of the hddslrs or even the vg10 or the coming nxcam realize most of those pixels they are paying for are never pulled from the chip. Interesting how this works. Even Panas af100 works the same way.

Looks like JVC gets to explain all this just as they did HDV. With JVC free from Pana and buying Sony sensors, we may see this working arrangement play itself out into the future. Good for both as Sonly needs super high volume products and JVC needs niche products.

[Arnie Schlissel]

But these aren't real cameras. These are "proof of concept", meaning "here's what we can do, but we haven't decided to actually do it." Just like the Detroit concept cars that they like to present at the auto shows.

It remains to be seen whether or not JVC will actually decide to create a production model based on either of these prototypes. They will gauge the response from reviewers and potential consumers and then make decisions based on what they project the market to be.