The 1080P mode vs the 720P mode!

[David Heath]

I'd prefer to keep out of the 'which camera is better' debate, except to say that I'm sure all four of the comparable models will be found good enough to earn their respective owners money. But a few words of general theory regarding the whole concept of 'pixel shift' and resolution.

At first sight, it sounds like magic - how can resolution be magiced out of nowhere, how can a system deliver more resolution than the sensor posseses? Perhaps the key is to realise that what it can deliver is increased LUMINANCE resolution. What it effectively does (in a one dimensional sense) is convert five three-colour pixels (say) into ten from the perspective of luminance only. Hence instead of (R+G+B),(R+G+B),(R+G+B),(R+G+B),(R+G+B) it will give (R+B),(G),(R+B),(G),(R+B),(G),(R+B),(G),(R+B),(G) - ten (derived luminance) pixels instead of five.

What it won't do is give the same level of chrominance resolution - but that's not normally important as chrominance is recorded at a lower resolution anyway, and the eye is less sensitive to chrominance resolution than luminance. It also depends on the pixels being smaller than ideal, the theory showing best results would be obtained if they are only half as wide as the inter pixel spacing. If the pixels could be made as wide as they are spaced (obviously desirable for sensitivity) pixel shift wouldn't theoretically work.

It's easier to see how this all works in the horizontal dimension (and that's how the Z1 gets 1440 from a 960 chip). The suspicion is that Panasonic have employed it vertically as well, and that's where the geometry gets interesting........ (It also leads me to think that 4:2:2 recording there gives no advantage to 4:2:0 - the chroma resolution isn't there in the first place, for the reasons given earlier - but that's another story......)

[Barry Green]

Originally Posted by Steve Mullen

You've got to retest all the camcorders.

Sounds like a good idea.

Why, if the HVX200 can produce these numbers, were YOU not able to get these numbers when other people and other camcorders were around?

Don't know. Adam and I have been scratching our heads over that one ever since it happened.

Especially since I can point it at an EIA chart and get 730 lines out of it.

1) The only independent measures of the HVX200 show it to have 550x540 resolution while the HD100 measures 700x700 -- exactly as does a Varicam.

Yet that wasn't recorded resolution, that was E-to-E. So HDV didn't have a chance to take a whack at it (although it probably wouldn't matter; HDV would excel at a static b&w resolution chart) and the Varicam's subsampling upon recording didn't come into play either. That test measured live camera heads. Which is largely irrelevant, it's recorded definition that matters to 99.9% of us.

2) IF you want to continue talking about 1080p CCDs

Never talked about a 1080p CCD. I said the image was scanned off the CCD at 1080p, which is what Panasonic said. As far as I know nobody's ever said that the CCD has 1080 vertical pixels; the only thing they've said is that it employs vertical spatial offset (aka "pixel shift"). I would find it very surprising if the CCD actually had 1080 vertical pixels on it. In fact I'd be floored if it had more than 720. But that doesn't stop 'em from employing spatial offset to scan a 1080p image off of the 720 (or 576 or whatever else you want to guess at) pixel CCD.

-- then YOU need to explain why/how a 1080p CCD produces less measured resolution than do 720p CCDs.

I can explain it by three ways: pointing out that a) nobody said it was a 1080p CCD, b) pointing to the EIA that shows the HVX delivering 730 lines and the JVC doing about 530, and c) pointing to a 492-pixel DVX delivering about 750 lines vs. the 720-pixel JVC doing about 530.

It ain't about the count of the pixels, it's about what the system does with them. Final measured definition is the only thing that should matter here, and even then luma resolution (which is *all* we're talking about here so far) is only one component. I don't know how many of us are shooting black & white, but I'd venture to say there's not a whole lot of us doing that.

So when you put it all in context, and look at the actual images, what do you get? You get six of the eight of us at that test choosing to buy the HVX. You get one who decided to keep his JVC instead of buying the HVX he was contemplating. And you get one who decided to buy a second Canon instead of buying the HVX he'd been contemplating. So 75% of us there chose the HVX and chose that that's where we'd spend our money; even some of those of us who already had one of the competing products.

Actions speak a whole lot louder than theoretical pixel discussions, and those are the actions that those of us who were there took.

[Barry Green]

Originally Posted by David Heath

The suspicion is that Panasonic have employed it vertically as well

That's not a suspicion, that's confirmed. Panasonic has said that they employ spatial offset both horizontally and vertically.

[Chris Hurd]

Originally Posted by Steev Dinkins

I know this may not get addressed here, but is this thread done yet? ;)

Sadly, it's over-done.