Panasonic site updates!

[Lawrence Bansbach]

Originally Posted by Barry Green

And you can do hot-swapping and perpetual recording with two 4gb's, whereas with a single 8gb you couldn't do that.

Putting aside the danger of jostling the camera while hot-swapping, I believe, unless I am mistaken, that you need at least three cards to do "perpetual" recording. Because it takes about as long to offload a card as to fill it, one must be offloaded and erased while one filling up: one actively being recorded to, a freshly erased one awaiting switchover, and one offloading. You should probably have a spare, too. So, yeah, the bundle with four 4-GB cards seems the best for indefinite continuous shooting with P2. But, boy, I'm sure the frequent swapping and the extra care not to jostle the camera could be nerve-racking. I think some HDD solution would be better for longer shooting.

[Chris Hurd]

Originally Posted by Lawrence Bansbach

I think some HDD solution would be better for longer shooting.

There are (or rather there will be) at least two different HDD options available to HVX200 shooters needing long-form recording. One is the FireStore FS-100 which connects to the camera by FireWire; the other is the CinePorter CP-2 which connects via the P2 card slot. Both offer at least 100GB of uninterrupted recording for less than the cost of two 8GB P2 cards, so take your pick.

[Lawrence Bansbach]

Originally Posted by Jeff Kilgroe

I've also read some various speculations that having both a horizontal and vertical shift is somewhat counter productive -- so I don't know if it would be shifted in both directions.

Looks like a half-pixel shift in each direction. In "Another Reader Report from DV Expo; more info on Panasonic HVX200" (HD for Indies), poster Kevin Shumacher reported from DV Expo West: "To clarify a couple of things regarding the HVX-200; I also attended Jan Crittenden Livingston's demo, where Jan did not indicate the resolution of the CCD's but did state they used both a vertical and horizontal 1/2 pixel offset 'to enhance luma resolution.'"

I'm a little confused. Wouldn't that mean a CCD with an active pixel count of 960 x 720? Yet in this dvxuser post, Jan seems to debunk the 960 x 720 figure. Also, wouldn't a half-pixel shift both horizontally and vertically reduce color sampling below 4:2:2 in the 1080p/i modes?

[Jeff Kilgroe]

Originally Posted by Lawrence Bansbach

I'm a little confused. Wouldn't that mean a CCD with an active pixel count of 960 x 720? Yet in this dvxuser post, Jan seems to debunk the 960 x 720 figure. Also, wouldn't a half-pixel shift both horizontally and vertically reduce color sampling below 4:2:2 in the 1080p/i modes?

I suppose it could be both a horizontal and vertical pixel shift. I'm not sure I understand the reasoning behind Panasonic's secrecy over the CCD block. I would guess that it's because their censor is lower res than what the competition is using, but the first few clips seem to show excellent detail. IMO, the "karate" clip and the good 1080 "guitar" clip, even though edited and recompressed to make them web-manageable, still show many aspects that are superior to what I've seen from HDV cameras.

Panasonic has claimed that their CCD "scans natively at 1080p". So if it doesn't have 1080 lines of vertical resolution, then they could get into a little trouble for claiming that. It may be a grey area, but that would be pretty slimey to claim 1080p scanning if they need a vertical pixel shift to achieve the equivalent of such.

[Lawrence Bansbach]

Originally Posted by Chris Hurd

There are (or rather there will be) at least two different HDD options available to HVX200 shooters. . . . Both offer at least 100GB of uninterrupted recording for less than the cost of two 8GB P2 cards, so take your pick.

The FireStore FS-100, at least, will cost significantly less than than the current price of one 8-GB card (which, I believe, is $2,250). From a Focus Enhancements press release: "The FireStore FS-100 for the HVX200 is expected to be available in March 2006, and price will be less than $2,000."

[Chris Hurd]

Right, point being, if you need this option for long-form recording, it's there for you.

[John Jay]

Originally Posted by Jeff Kilgroe

I suppose it could be both a horizontal and vertical pixel shift. I'm not sure I understand the reasoning behind Panasonic's secrecy over the CCD block. I would guess that it's because their censor is lower res than what the competition is using, but the first few clips seem to show excellent detail. IMO, the "karate" clip and the good 1080 "guitar" clip, even though edited and recompressed to make them web-manageable, still show many aspects that are superior to what I've seen from HDV cameras.

Panasonic has claimed that their CCD "scans natively at 1080p". So if it doesn't have 1080 lines of vertical resolution, then they could get into a little trouble for claiming that. It may be a grey area, but that would be pretty slimey to claim 1080p scanning if they need a vertical pixel shift to achieve the equivalent of such.

I too suspect the CCD is the smallest pixel count of the current crop of HD Camcorders, hence the secrecy.

They claim "Progressive to interlace conversion, cross conversion and down conversion all start with the 1080p/60 scan." I take that to mean the CCD is bi-directional pixel shifted 0.5 pixels and the analog signal is treated as 1080p and then a scanning frequency of 148.5 MHz is performed - ie it is treated as if it were a 1080p progressive CCD. The game is given away in the small print of the newspaper the guitar man is holding, which is quite soft in concert with the fairly large DOF present in the shot.

Also because everything is derived from the 1080p scan the 720p will not be as sharp as it could otherwise be, but will have a slight noise reduction benefit.

[Barry Green]

Originally Posted by Lawrence Bansbach

Because it takes about as long to offload a card as to fill it

Well, that depends on what you're using to offload. On the P2 Store, yes it's about the same. But if you were using a fast laptop with a very fast hard disk, you can offload much faster than that -- the theoretical maximum is that a 4gb card could be offloaded in 50 seconds. It'd take a RAID of hard disks that could sustain an 80mBps data rate, but it *could* be done...

For a practical matter, yes it'll take about the same amount of time. Even so, I still think the two-4gb is a far superior solution to the single-8gb. We shot for four days with just two 4gb's, and after one was full I'd pull it and start the offload and the cameraman would continue with the other card. No waiting, no delay. Rinse, cycle, repeat. Had we had just one 8gb we could have shot for twice as long, but then we would have had to shut down the production while offloading. That, I can imagine, would have been most irritating.

But, boy, I'm sure the frequent swapping and the extra care not to jostle the camera could be nerve-racking. I think some HDD solution would be better for longer shooting.

Almost assuredly.

[Barry Green]

Originally Posted by Lawrence Bansbach

Wouldn't that mean a CCD with an active pixel count of 960 x 720?

No, because CCD pixels don't bear a 1:1 relationship with frame pixels. This is a common misconception, but in simple terms the CCD pixels are not directly related to the frame pixels at all. There's CCD pixels, which get sampled into the internal format, and then that gets processed into the recorded format.

So an HDX400 has a 1280x720 CCD, which gets sampled off the chips and into the DSP at (I believe) 1920x1080, and then recorded on tape at 1280x1080. The chip's resolution isn't directly related to pretty much anything else in the chain.

Resolution is important, obviously, but it's just one component in the whole chain, and optimizing for resolution means compromising on other aspects of the image chain.

CCD pixels are an oft-quoted statistic, but it's tough to figure out what they really mean because the system has no way to access them. CMOS is different -- with CMOS the system can access the individual pixels. But with a CCD, the pixels get sampled into an analog signal, which then gets digitized as it comes off the chip. There is no way to get at the contents of a CCD pixel; it's all output as a continuous stream of voltage, which then gets sampled. Think of a microphone -- it senses air pressure and turns that into voltage; then an audio sampling circuit will go in and sample that voltage stream at a 16-bit, 48khz rate. Yet we don't ascribe "bits" to the microphone, do we? Do we question whether a microphone is 16-bit or not? A CCD is pretty much a "video microphone" -- a microphone senses air pressure and turns it into voltage, a CCD senses photons and turns them into voltage. The actual fineness of resolution with which it can respond would be nice to know, but obviously doesn't bear a direct 1:1 relationship with the digitized signal.

[Barry Green]

Originally Posted by Jeff Kilgroe

I'm not sure I understand the reasoning behind Panasonic's secrecy over the CCD block. I would guess that it's because their censor is lower res than what the competition is using, but the first few clips seem to show excellent detail. IMO, the "karate" clip and the good 1080 "guitar" clip, even though edited and recompressed to make them web-manageable, still show many aspects that are superior to what I've seen from HDV cameras.

Well, there it is in a nutshell. That's been exactly their point. The sensor probably does have a lower pixel count, but the fact is that the resolved detail is definitely higher than the HD100 or FX1 (haven't had a chance to do a side-by-side with the XLH1 yet). So had they announced the number prior to showing the footage, people could have just looked at numbers and decided that one was better than the other. Which would have been an incorrect conclusion to reach. So instead of engaging in what Jan calls "a numbers war", she said they'd release footage instead and let people judge based on the pictures, not the spec sheet.

Maybe once it's been out a while and established, maybe then they'll release the number. But for now she said they don't want to distract from the images.

Panasonic has claimed that their CCD "scans natively at 1080p". So if it doesn't have 1080 lines of vertical resolution, then they could get into a little trouble for claiming that. It may be a grey area, but that would be pretty slimey to claim 1080p scanning if they need a vertical pixel shift to achieve the equivalent of such.

Well, no, because executing a 1080p scan doesn't have a 1:1 relationship with the pixels on the chip. The HDX400 has a 1280x720 chip, yet it scans 1920x1080 off of it. "pixel shift" isn't some sort of photoshop-style uprezzing algorithm. It's a spatial offset technique that leads to actual, verifiable gains in resolution. The BBC says that it's capable of a maximum of around 1.414x gain in resolution, with a more practical maximum being about 33% more resolution.

[Les Dit]

Barry, CCD's do transfer photosite charge in an analog manner, but you should not confuse that with the 'old world' way of thinking, as in a megahertz type of rating for resolution. I'm not sure that is what you were hinting at. CCD's have a very distinct and effective charge transfer from each and every photosite out to the external ADC. The effective samples may then be stretched and scaled to any size they want, but the native resolution is still what it was. As long as the optics does a good job of , say illuminating only the odd photosites, the resolution is defined by that limit. Of course no optics can do that with 100% contrast between odd and even avalable pixels, but they try.

The edge enhancement 'they' apply to the resized image do make the image look sharper to the eye, but look horrid to film people that are used to looking at film.
This whole uprezing reminds me of the days of cheap flatbed scanners that claimed all kinds of DPI ratings, but they were just interpolating ( resizing/scaling) the data. A lot of BS.

It will all come out in the wash when someone points the cameras at a standard resolution chart with line pair patterns. Up sizing that and then sharpening won't pick up the finest line pairs that the original ccd source didn't capture. Aliasing aside, it's all about the ccd dimensions , the optics, and the compression that it undergoes.

Let's see how some line pair charts look like! It would make a great shootout.
Now only if they saved more that 8 bits/color, that would wipe out the 16mm film cameras for good!

[Don Donatello]

"Also because everything is derived from the 1080p scan the 720p will not be as sharp as it could otherwise be, but will have a slight noise reduction benefit."

from my observations thought 720p had more noise then 1080.
switching to 1080 didn't really increase resolution but maybe a little less noise - now the resolution might be because viewing on 720p monitor ?

[John Jay]

Originally Posted by Don Donatello

from my observations thought 720p had more noise then 1080.
switching to 1080 didn't really increase resolution but maybe a little less noise - now the resolution might be because viewing on 720p monitor ?

this is an example of where theory and practice dont necessary tally - another example is the notion that pixel shift can increase resolution just take a look at the following jpeg

http://rapidshare.de/files/9046674/pixel_shift.jpg.html

it shows the effect of resolution increase in the luma from the co-sited pixels at A, pixel shifted 50% at B and the contribution to the luma of the extra resolution at C. Ok there is a resolution increase but it can hardly be seen and is obtained from a huge penalty to overall image contrast; a bit like 'robbing Peter to pay Paul'

Sony use a horizontal pixel shift but the resolution charts show it has little effect and therefore no practical benefit.

[Robert Graf]

I left feedback on Panasonic's website about updating the message on the HVX200 page about the delivery date, which has been corrected. I also notated my observation that an online Operating Manual is missing. Today, I was grateful to receive a personalized email from Panasonic's Corporate Brand Marketing Department, which stated that it will be added to their site "in the coming weeks."

So hopefully that answers someone's question on this subject.

Cheers!
Bob

[Chris Hurd]

Thanks for that update, Robert!