HVX-200 best possible pixel arrangement and count

[ImBiggie]

Would everyone agree that our best hope is for 1280x720p?

The math shows below 1280x720p converted to interlance is barely less than 1920x1080i

720p 60 x 100 mbit = 1080i 60 x 100 mbit
720p 60 x 100 mbit = 1080p 30 x 100 mbit
720p 60 x 100 mbit = 1080p 24 x 100 mbit

Or

1280x720x60p = 55296000
1920x1080x60i = 124416000/2 = 62208000

Or

55296000 is 12.5% less than 62208000

--------------------------------------------------------------------------

Anything other than 1280x720p or 1920x1080i pixel count will result in panasonic making up pixels to get true HD.

For example 960x1080i will result in Panasonic interpolating 960 pixels to 1280 or 1920 depending on format.


Happy to hear your opinion!
:engel017:

 

[Barry_Green]

Would everyone agree that our best hope is for 1280x720p?

No, probably not.

The math shows below 1280x720p converted to interlance is barely less than 1920x1080i

In a raw square-pixel format, your math is correct. Both formats push approximately 60 million pixels per second through the television.

But the recording format doesn't record all those pixels. HDCAM and HDV both record 1440x1080, and DVCPRO-HD records 1280x1080. In 720p mode, DVCPRO-HD records 960x720.

For example 960x1080i will result in Panasonic interpolating 960 pixels to 1280 or 1920 depending on format.

No, not true. Interpolating is entirely different from pixel shift/spatial offset. Spatial Offset is a way that the CCDs are aligned, and a way that they are sampled, that results in a significant resolution boost while not "interpolating" any new pixels at all; it rearranges the pixels so that they give maximum coverage.

 

[Barry_Green]

I've been giving this a lot of thought, and in the end all I can say is: none of us are camera engineers (at least, none that I know of... maybe there are some CCD engineers here...) so this is just speculation...

Pixel shape on the CCD is only one aspect of quality video. There are two mutually exclusive elements that you have to factor in: bigger pixels = better sensitivity, better low light performance, better dynamic range, better signal-to-noise ratio. But lower resolution.

Smaller pixels = better resolution. But: smaller pixels = worse sensitivity, worse low light performance, worse dynamic range, worse signal-to-noise ratio.

You can't have both. You choose one or the other, or make your compromises in-between. If you go for higher pixel count, you will necessarily, irrevocably be giving up latitude, low-light performance, and adding noise.

So, the question becomes, how can we get the most pixels on the chip for resolution, while keeping the pixels as big as possible? That's really the trick. Big pixels for light gathering, but lots of them for resolution. It's easy on a bigger CCD, as you can pack in a lot of pixels, and the sheer size of the real estate on the CCD means that each pixel is physically bigger, therefore providing better performance.

But in a small camera, it's a real challenge. Already we're seeing the FX1, even after employing pixel shift (which is *not* interpolation), delivering sensitivity that's two full stops slower than the PD170.

So: what are the challenges, and what are the factors we have to work with?
The challenge is to deliver a system with the best resolution, best dynamic range, best sensitivity, and lowest noise, all at the same time. And it has to deliver that performance in two different resolutions: 1080i/p and 720p.

The factors we have to work with: the codec samples 4:2:2 at either 960x720, or 1280x1080. Any higher resolution is basically irrelevant, because the codec is going to compress the image down to these sizes.

Based on what we know, it would seem that the absolute highest resolution necessary to deliver the full potential of the DVCPRO-HD codec would be 1280x1080. Any more resolution would be wasted, and would contribute to making the pixels unnecessarily small, thus harming overall video performance.

But wait -- there's more! 1280 x 1080 CCDs would actually still employ pixels that were smaller than necessary. At 1280x1080, that would be sufficient to deliver a 1280x1080 frame at 4:4:4 color. We know already that there's no need for 4:4:4, because the codec is going to compress it down to 4:2:2 anyway. So native resolution at 1280x1080 would in fact be overkill, and would again be making the CCD pixels unnecessarily small, hurting video performance in a fruitless quest for resolution, because that resolution will be discarded by the codec anyway.

So how could we optimize video performance, while making the pixels as big as necessary?

That's the challenge, and I imagine the Panasonic engineers will be up to the task. Using a straight native resolution chip (such as JVC chose with the HD100) seems wasteful. They could go with some sort of hybrid system, where the green CCD is 1280x1080 but the red and blue CCDs are 640x1080, using pixels twice as wide... that would give better performance to the chroma, at the expense of the luma... don't know how well that might work out, but it's an idea.

In the end though, I expect they'll use spatial offset. Spatial offset is not some sort of lame "interpolation" technique. Interpolation is what you get when you resize a PhotoShop photo, or when you take a scanner with 600DPI "optical" resolution and try to scan a picture in at 9600 DPI -- those kind of techniques just kind of mathematically "makes up" pixels. That's not what spatial offset does. It is a technique for reading the CCDs that makes for more resolution; by offsetting the green CCD 1/2 pixel from the red & blue, it creates another luminance sample, so you get a genuine increase in luma resolution.

The key thing to consider is that spatial offset (what was once called "pixel shift") is a legitimate technique to increase luma resolution -- the BBC, in a white paper, showed that spatial offset can increase resolution to a maximum of 1.414 times (or, the square root of two), with a more practical limit being a 33% increase in horizontal resolution. What do you lose by employing spatial offset? Some chroma resolution -- to the tune of losing 1/2 pixel of chroma resolution accuracy. But, keep in mind, the codec is going to throw away half the chroma resolution anyway! So, in effect, you lose nothing. As far as actual, practical, usable footage, you lose nothing at all, you just gain (unless you're talking about 4:4:4 recording, which none of these cameras can do).

Employing horizontal spatial offset, they could make a 960-pixel CCD deliver 1280 pixels of resolution. The BBC's estimate of 33% increase in resolution is exactly the difference between 960 and 1280 (960 x 1.333 = 1280). So, ostensibly, you could get native horizontal resolution for both DVCPRO-HD modes by using 960 spatial-offset pixels across; you get either 960, or 960 spatial-offset to 1280. Either system will deliver full luma resolution, and full 4:2:2 chroma resolution.

And, by going with 960 instead of 1280, they can make the pixels 33% bigger than before. Bigger pixels means better low-light performance, better sensitivity, better dynamic range/latitude, and better signal-to-noise ratio. Because of all that, I really can't see why they'd make the CCD any more pixel-packed than 960 wide. There wouldn't be any real-world benefit at all to going with a more densely-packed chip, yet there would be definite detriment to the recorded video if they went with a more densely-packed chip.

So I predict 960. Maybe they'll go 1280, but I'm guessing 960. I would be stunned if it was more than 1280, since DVCPRO-HD records a maximum of 1280 wide anyway.

Okay, so now on to the vertical -- how many pixels vertically will it have? Well, that's a point of speculation. The Sony went with 960x1080 -- attempting to use spatial offset to boost the horizontal res from 960 all the way up to 1440, but using native pixel height for the vertical. That seems like a reasonable approach; so the HVX may very well be 960x1080 as well.

However... can spatial offset be used vertically, to increase resolution like it does horizontally? That's an interesting question. And Panasonic apparently believes it can, because they built the HDX400, a 1080i camera, with 720-pixel chips. Apparently they employ vertical spatial offset to make the 720-pixel-chips deliver 1080 pixels of resolution. Not quite sure how -- I don't really understand how vertical spatial offset can work without compromising some color resolution.

And that's what puzzles me about Sony's approach. Assume for a moment that they'd used vertical spatial offset as well; thus building their CCDs off of 960x720. They could have had pixels that are up to 50% larger than what they chose, and presumably the only drawback they would have faced would have been, again, a chroma sampling offset of 1/2 pixel. Yet HDV, vertically, only records half the chroma anyway! In other words, they could probably have gotten all the benefit, all the same resolution, better overall performance, and lost nothing (regarding vertical resolution). Sounds to me like they left a lot of performance "on the table".

But does Panasonic's implementation of vertical spatial offset work in a progressive system? Perhaps it works for interlaced, because a 1080 interlace signal only really has about 800 discernible lines of resolution anyway, so maybe it somehow scales the 720 pixels up... I don't know.

But, let's give them the benefit of the doubt and say that vertical spatial offset works with no harm to video quality. In that case, they could go with 720-pixel CCDs, and be able to deliver nigh unto native resolution in both 720p and 1080p modes. A 1/2-pixel vertical offset on 720 lines would yield 1080 lines (1.5x720 = 1080).

So, I can see an argument being made for a variety of chip sizes: 1280x1080, 1280x720, 960x1080, or even 960x720. I think that the clever camera engineers will be able to extract the maximum resolution from any of those configurations. Let's not forget that the lens will be a limiting factor as well -- at 1280x1080 you're talking about pixels that would be about as dense as 200 line-pairs per millimeter, and it's extremely doubtful that any lens can resolve that tightly; most max out at about 133lp/mm. So 960x720 with spatial offset may indeed deliver a full 1280x1080 pixel grid, while being able to handle all the resolution the lens is giving it, and also allowing for the maximum big-pixel size so it offers better latitude, sensitivity, and noise performance.

Ultimately it's all just technical jousting though -- the only thing that really matters is what the footage looks like. If the engineers decide they need 1280x1080 to deliver the best resolution in the final footage, then that's what they'll go with. If they find that they can deliver the sharpest possible image while using a 960x720 pixel CCD, I would assume that's what they would go with.

I expect them to use the configuration that delivers the best of all worlds -- highest resolution (but no higher than the codec requires) while still giving the biggest pixels. At least, that's what I hope they do. I want the best, real-world footage I can get, not "bragging rights" over who has the "biggest CCD".

 

[ImBiggie]

Very informative Barry.

The pixel count below are from panasonic AJ-HDX400 and AJ-HDC27F brochures.

AJ-HDX400
2/3" 3-CCD 16:9 1080i DVCPRO HD® Camcorder

CCD Elements: 2/3” HD Progressive CCD
Picture Elements: Total: 1370 (H) x 744 (V)
Effective: 1280 (H) x 720(V)
-----------------------------------------------------------------


AJ-HDC27F
VariCam® Variable Frame Rate 16:9 HD CinemaTM Camera with CineGammaTM Software

CCD Elements 2/3" IT 3-CCD (RGB)
Picture Elements 1,019,280 pixels (total, 1,370 x 744)
921,600 pixels (effective, 1,280 x 720)
----------------------------------------------------------------

If your theory is correct, why would Panasonic implement their top line cameras as 1280x720p? Why would they go through the hassle of creating a 1.1 million pixels CCD to later discard some pixels to 960x720p. From what I understand from their brochures, they picked 720p because it can easily upconvert to 1080i without making up pixels (pixel shifting). 1280x720p converted to 1920x1080i and vice versa is the same in Panasonic's book.


Your statement about bigger pixels being better might be right but doesn’t hold true with today’s digital cameras.

For example:

Canon d30 - 3.1 million pixels
Canon 10d - 6.3 million pixels
Canon 20d - 8.3 million pixels

Resolution is higher and noise is lower with each upgrade.

I agree bigger pixels are better than small pixels but there are other improvements that can make up the difference.

:engel017:

 

[evinsky]

Canon is using a substantial amount of image processing through thier Digic chips to negate the Noise in thier CMOS chip line. But CMOS can't yet effectively move data at an HD video rate when arrayed in a three chip format. I believe it has problems clearing that CCDs don't. This is not noticeable in a stilll camera going 3-8 FPS. Also Canon made the XL2 with 1/3" CCDs despite thier expertiese in CMOS. And the XL2 has only slightly better noise performance than the DVX but less sensitivity. I do believe Canon uses some post process to reduce noise because the XL2 has a smooth look instead of the DVXs grain look. Honestly I'll give up a stop or two of sensitivity if It means a great looking image. I ust don't see it in the FX though.

 

[lacuna]

So, I can see an argument being made for a variety of chip sizes: 1280x1080, 1280x720, 960x1080, or even 960x720. I think that the clever camera engineers will be able to extract the maximum resolution from any of those configurations. Let's not forget that the lens will be a limiting factor as well -- at 1280x1080 you're talking about pixels that would be about as dense as 200 line-pairs per millimeter, and it's extremely doubtful that any lens can resolve that tightly; most max out at about 133lp/mm. So 960x720 with spatial offset may indeed deliver a full 1280x1080 pixel grid, while being able to handle all the resolution the lens is giving it, and also allowing for the maximum big-pixel size so it offers better latitude, sensitivity, and noise performance.

Thanks for the in-depth post Barry. To me, a novice in these areas, one implication of what you are saying here is that given the limitations of a 1/3" CCD camera with a low-cost lens package, the perceived resolution we actually end up with may not be a heck of a lot different between 720p and 1080p.

I know there are compression issues with HDV, but in terms of sheer resolution, the HD100 will record 1280x720 to tape from 1/3" CCDs of the same pixel count. The HVX should have a similar pixel count, albeit in a different configuration, and will record 33% more pixels per frame to P2. So the maximum resolution gain we could expect from the HVX would be 33%, but given lens limitations this much difference may be difficult to percieve.

So, while an HD 720p frame is 1280x720, less than half the area of a 1080p frame at 1080x1920, the discerable resolution that we actually end up with if comparing a film blowup from an HVX at 1080p and an HD100 at 720p may be in the same ballpark.

By this reasoning, resolution is probably not a key reason to buy the HVX over the HD100. Codecs and media are more important issues.

Okay, I'm not an expert, just an enquiring mind, so I want to know, am I way off the mark here?

 

[Barry_Green]

If your theory is correct, why would Panasonic implement their top line cameras as 1280x720p?

Why they chose 1280 in the VariCam and HDX I 'm not sure (like I said, I'm not a camera engineer)... I believe it's probably for the HD-SDI output. They may have chosen 4:4:4-caliber resolution rendering to preserve the quality of the HD-SDI output, wherein it would output a native 4:4:4 signal. 960x720 only comes into play on the VariCam when you record it to tape; internally I'm pretty sure it processes and manipulate the signal as full 1280x720, and outputs it on the HD-SDI port as 1280x720 at 4:4:4.

And, at 4:4:4 color sampling, that's the only time you really *need* native pixel resolution on the CCD. 4:2:2, or especially 4:2:0, doesn't need it.

As for the HDX, well, that's a camera that uses a 1280x720 CCD to render what would theoretically be a 1920x1080 image. Except that DVCPRO-HD doesn't work with 1920, so 1280 is appropriate (since the codec records the 1920x1080 image as a 1280x1080 grid). Like I said, I'm not 100% clear on the vertical issue, as to how it accomplishes adequate resolution from 720 pixels vertically (but I don't doubt that it *does* accomplish adequate resolution; the live display at NAB looked absolutely fantastic).

Plus, keep in mind those are 2/3" cameras! At equivalent pixel density, the pixels are still over 4x as large as the pixels on a 1/3" camera. Even with its densely-packed CCDs, the VariCam is still around 500 EI. That's due to the bigger pixels. There's less to gain on the 2/3" camera, vs. the potential tradeoff of chroma resolution that may be applicable in a potential 4:4:4 (or uncompressed analog?) output system. But I don't know enough about the VariCam or HDX to know for sure.

None of that matters for the HVX though. HD-SDI is not an option on the HVX; the only digital way to get data out of the camera is through DVCPRO-HD. And that means 960x720 4:2:2, or 1280x1080 4:2:2. Any more than that would be wasted. And, I do believe, even that resolution, at native CCD res, would be wasted, because of the chroma sampling reduction that gets done.

I'm with evinsky about the digital camera comparison. Digicams can take forever (relatively speaking) in processing their images. Video cameras can't. A Digicam might shoot at maybe 3fps. A video camera has to output images at 30 or 50 or 60 frames per second. Digicams can do all sorts of processing, noise reduction, aliasing, whatever they want... video cameras can't. They have to output to tape or P2 immediately and clear the pipeline, because there's another 59 frames coming in just this one second.

I agree bigger pixels are better than small pixels but there are other improvements that can make up the difference.

In theory, yes. In actual practice, during 60fps recording? Probably not. The day will come, of course, when internal processors in cameras are fast enough to handle anything that the camera throws at them, and when that happens then yes you're right. But I don't think we're anywhere near that point today.

 

[Barry_Green]

one implication of what you are saying here is that given the limitations of a 1/3" CCD camera with a low-cost lens package, the perceived resolution we actually end up with may not be a heck of a lot different between 720p and 1080p.

That does, indeed, remain to be seen. One thing I've been concerned about is lens resolution and whether a lens can actually resolve enough data to show legitimate resolution at this density. The Z1 seems to deliver a solid 700 lines of horizontal resolution, maybe up to 800, and 800 lines of vertical, so it may not be that much of a concern. Maybe 1080p will be noticeably higher res than 720. Maybe not. But it remains to be seen.

So, while an HD 720p frame is 1280x720, less than half the area of a 1080p frame at 1080x1920, the discerable resolution that we actually end up with if comparing a film blowup from an HVX at 1080p and an HD100 at 720p may be in the same ballpark.

Only if you totally discard color resolution. People seem fixated on luma resolution at the expense of color resolution. They're both vitally important. Color makes a difference -- a big difference, most especially if you're planning to blow it up.

Let's ignore luma and just focus on color. The HD100 will record 640x360. The HVX records 640x1080. That's three times more resolution! Do you think that'll show up on the big screen? I sure do. I think it'll be a very big, massive noticeable difference.

By this reasoning, resolution is probably not a key reason to buy the HVX over the HD100. Codecs and media are more important issues.

Codec and resolution are intrinsically tied together. If the HD100 used 4:2:2, I'd think your comparison would be much closer. But 1280x1080 luma vs. 1280x720 luma, and 640x1080 chroma vs. 640x360 chroma, (plus no motion artifacts), puts the ball squarely and firmly in the HVX's court.

Assuming the camera lives up to the promises that the specs on paper are making.

 

[lacuna]

Your reply much appreciated Barry, highly informative (and relevant) as always.

I often wonder how you find the time to produce the volumes of detailed posts you put out, but however you do it, thanks!

 

[Sumfun]

Ultimately it's all just technical jousting though -- the only thing that really matters is what the footage looks like. If the engineers decide they need 1280x1080 to deliver the best resolution in the final footage, then that's what they'll go with. If they find that they can deliver the sharpest possible image while using a 960x720 pixel CCD, I would assume that's what they would go with.

Barry, thanks for the detailed explanation. I found it very informative, not just on the CCD issue, but also for the codec, lenses, etc.

I only wish that Panasonic would provide the best resolution in the final footage as you say, but I would probably add the caviat that it's the best resolution - for the money. I would think that 960x720 can do the job, but they might surprise us with a higher resolution CCD if only to play the specsmanship game with Sony and JVC.

Your statement about bigger pixels being better might be right but doesn’t hold true with today’s digital cameras.

For example:

Canon d30 - 3.1 million pixels
Canon 10d - 6.3 million pixels
Canon 20d - 8.3 million pixels

Resolution is higher and noise is lower with each upgrade.

Evinsky pointed out earlier that the still cameras only operate at ~3 fps, but I wanted to add that the CMOS sensors in the still cameras are also much bigger than 1/3 inch (8.5mm). The sensor in the Canon 20d is 22.5mm x 15 mm. I don't know the exact dimensions of the video CCD's, but I would say that their pixel size is probably smaller than the still cameras'.

 

[ImBiggie]

The HD100 will record 640x360. The HVX records 640x1080. That's three times more resolution!.

Is the 640x1080 an interlance signal. if it is then it's 1.5 the resolution of 640x360 progressive.


Sony has a codec that can take full avantage of 1920x1080i. I think Panasonic needs to upgrade their codec to take full advantage 1280x720p, not 960x720p.

Is there a link to the DVCpro HD specification.

Thanks again Barry.

 

[Barry_Green]

640x1080 would be in 1080/24p or 1080/30p, so no, not interlaced. It is 3x the resolution of 640x360.

Sony's only codec that takes full advantage of 1920x1080 is in the super-ultra-mega-expensive HDCAM SR line. Not even the $100,000 CineAlta records 1920x1080 -- it subsamples to 1440x1080. Only HDCAM SR records the full 1920x1080.

There will be no updating to DVCPRO-HD. They've come out with a system that's top-to-bottom cross-compatible with their whole broadcast lineup, from the VariCam down to the HVX, including the HDX and the AJ-HD1200A. It's a complete product line, fully cross-compatible. If they changed the HVX it'd be an orphan product.

As for specs on DVCPRO-HD, I don't know if there's any quick reference. I know I found a paper on it at broadcastpapers.com... (at least that's where I think it was)...

 

[Don Tucci]

Hi whats the Divix code how and where does it fit in?Or is it only a post production code ?one of the dvx users had a grab it looked good.

 

[Barry_Green]

DivX (and its open-source competitor, Xvid) are post-production MPEG-4 codecs. They're not used in cameras, but they're used in streaming video delivery, etc.

 

[thisiswells]

Barry--

Could you elaborate on the 640x1080 and 640x360?
I haven't heard those numbers tossed around...

 

[Barry_Green]

Certainly -- sorry if it wasn't clearer before.

People toss around the numbers 1280x720 and 1920x1080, without necessarily clarifying that they're talking only about LUMA resolution. Black and white. Any objects in color may have significantly lower resolution.

For example, in a 4:4:4 system (such as D-5 or HDCAM SR), the luma (black & white) and the chroma (color) have identical resolution. But in HDCAM, or DVCPRO-HD, or DVCPRO50, or HDV, or Digital Betacam or DV or basically any other digital recording system, color is recorded with less resolution. In some cases *significantly* less.

For example: in NTSC DV, the black & white signal is recorded using a 720x480 grid. In fact, in all the codecs we're discussing, black and white gets recorded using the pixel resolution that we commonly talk about. So DV, HDV, DVCPRO-HD, they all record a black and white image like this:

That's an extraction from a frame created in PhotoShop, which consists entirely of a 1-pixel-thick black line on a white background. That picture was cropped into DV, DVCPRO-HD, and HDV resolutions, and run through Vegas and output through DV, HDV 720p, and DVCPRO-HD 1080 codecs, and it always came back pixel-perfect as compared to what it was fed.

But color's totally different. Due to DV's 4:1:1 color sampling, NTSC DV's color resolution is actually recorded using 160x480.

In HDV 720p, the black & white portion is recorded using a full 1280x720, yes. But the color is recorded at half that resolution on the horizontal, and half on the vertical, so the color resolution is actually 640x360.

In DVCPRO-HD 1080, the black & white portion is recorded using 1280x1080. But the color is recorded at half the horizontal resolution, and full vertical resolution. Meaning, the color gets recorded at 640x1080 resolution.

I used this picture as an example: a green background with a red line drawn through it, exactly as before but instead of black & white (luma), I used color (chroma):

Here's a sample of each after the codecs get through with it (blown up to 300% size):

This time the picture consisted of a one-pixel red line against a green background. In DV, you can see that the resolution is *nowhere near* what it was for black & white. Each DV chroma sample is four pixels wide, and one pixel tall (720/4 = 160, giving chroma resolution of 160x480). But DV can't even resolve that the line was red, because it has two colors that it has to try to reconcile (both green and red) in the same 160x480 pixel. So it just kind of pukes and gives us a slightly-red green. You can see that the luminance is still being tracked at full resolution (by the presence of the darker line) but the chroma is very low res.

Same with HDV -- just as low chroma resolution as DV, but in a different pattern. Instead of 1/4 width and full height, HDV's chroma res is 1/2 width, 1/2 height. So you get the same boxy pattern, but you still can't tell that the line was actually red, as HDV still has to balance that there's a red and green pixel in each square, and it can't resolve both. So it chooses a middle ground color. Each chroma pixel is twice as wide and twice as tall as its corresponding luma pixel. Which means that in its 1280x720 frame, you're actually getting chroma resolution of no better than 640x360.

DVCPRO-HD does significantly better -- way better. You can actually tell that the line had some red in it, for example, which you can't even discern in the other codecs. With its single-pixel-height resolution, and two-pixel-wide resolution, it can much more clearly resolve the image. In a 1280x1080 luma grid, DVCPRO-HD can resolve color at 640x1080 resolution. Vastly superior.

Now, that's still not a truly fair comparison because it's showing each picture at full-pixel size. It's giving more credit to HDV and DV than they deserve, seeing as they'd have to be magnified more to fill the screen to show comparative resolution. But in relation to Lacuna's question, blowing it up to a theater screen, we'd have to take into account that each image would be magnified at a different ratio to fill the screen. And that's where the resolution differences will really come into play. Having three times the chroma resolution will give DVCPRO-HD an even bigger advantage than these pictures show.

At a given screen size, the HD100 will require 50% more magnification on the vertical to match DVCPRO-HD's resolution (both b&w and color). Less res, blown up more, will make the comparison even blockier.

I think it'll be no contest, DVCPRO-HD 1080/24p footage will be resoundingly more clean, clear, sharp, and resolution-rich than HD100 HDV 720/24p. No contest.

 

[lacuna]

That's a very forcefull argument Barry, thanks for taking the time to illustrate it so clearly.

Do you mind running some numbers by us on the HD100 implementation of 720p vs what we can expect from the HVX? I take it the HDV codec will record a 33% higher luma resolution (1280x720 vs 960x720) but a lower colour resolution, which means...

And given the fact that the CCDs in the HVX are likely to have more pixels horizontally than will actually be recorded by DVCPROHD at 720p, does the HD100 have any advantage in it's 1:1 pixel mapping of the CCDs at 720p? ie: do you lose anything by squeezing 1080 horizontal lines from the CCD into 720 lines in the codec?

Again I ask as I wallow in a sea of my own ignorance, but in this thread you've come up with some great stuff so I'm encouraged to ask again

 

[icicle22]

Also...since we are getting into nitty gritty details, Barry has said that the HVX is a DVX with 16x9 for those who want to stick with DV. I am not trying to argue but Barry himself stated in the 3-way shoot out that any SD camera is going to produce better images than a HD camera downsampled. So that brings up the question: How will the HVX or HD100 create a 720x480 image for dv from a much higher CCD? It must use some downsample algorithm or it could sample from a smaller part of the CCD. Either way is not ideal as the downsample will ultimately lose some clarity and to sample the smaller CCD would change the FOV or focal length to an unusable ammount.

Any comments on this? I am trying to be cautious as a camera intended for HD might have some drawbacks when trying to also cater tot he DV crowd. There is always a trade off and we should be cautious of the SD on these cameras until it is proven that they can indeed match or exceed the current prosumer DV champs like the DVX or XL2. It just seems that by design they cannot possibly do HD incredible and do SD as well as it should.

 

[thisiswells]

Thanks so much, Barry! Your explanations rock. It all makes sense now

 

[Gibby]

I think it'll be no contest, DVCPRO-HD 1080/24p footage will be resoundingly more clean, clear, sharp, and resolution-rich than HD100 HDV 720/24p. No contest.

Interesting stuff Barry. How about running those hypothetical numbers for HVX200 DVCProHD 720p24 versus HD100 HDV 720p24, keeping in mind, of course, that neither camera has shipped or been tested. Since the HD100 has the ability to output analog uncompressed 720p24 at 4:2:2 color resolution, how about also running hypothetical colorspace numbers for that 720p24 4:2:2 HDV footage against the 720p24 4:2:2 DVCProHD footage. Nobody should be surprised at all that 1080p24 4:2:2 footage has much more colorspace resolution than 720p24 4:2:0 footage, but how about a 720p24 4:2:2 (HDV) vs 720p24 4:2:2 (DVCPRoHD) comparison?

Gibby
www.cut4.tv