Back when I was involved in HDTV matters (in 1986-89), there was a raging argument about pixel counts. I dredged up a paper, done by NHK in their early pioneering work on basic HDTV issues. It showed that, ideally, the best perceived overall resolution came about when pixels were "square"--that is, the pixel spacing was the same along the x and y axes. But it was by no means critical.

In the late 80's, American and European HDTV developers were terrified that Japan was going to crush the remaining video manufacturers, and were determined to do anything to throw down roadblocks to slow the NHK train. One of the roadblocks was the insistence on having 1080 scan lines, so that resulting pixels would be square (originally, the NHK HDTV standard was for 1035 active lines x 1920 active samples per line). So that's how 1080 came about.

So, really, the horizontal sampling rate can be pretty much anything. Anything less than 1920 will give you fat pixels and reduced horizontal resolution, but it's not that big a deal.

Just something I read trolling around the forums last night. DVCPRO HD (100) is only 1220 x 1080, it uses even more 'strecth' on the pixels than HDV and perhaps startlingly the FX1 looks better than the output from the DVCPRO HD systems. I'd have to assume the optics on the Panny cameras would be way better and all else being equal would yield much better images than the FX1, perhaps though a lot is lost due to the higher level of pixel stretch.
I think the critical test was when DVCPRO HD and HDV were downresed to DV, the DVCPRO HD developed noticeable noise on the edges, this may have been more to do with FCPs resampling algorithms than either format, hard to say for sure. Still it does show that the FX1 can hold its head up high even against some very expensive kit.
Bob.

Spot I guess I don't really know how the .909 pixels work in DV then, for if the 1440 pixels are "streched" to fit 1920 as you say then that sounds dishonest to me as there are still 1440 streached pixels and not 1920 square ones. Am I mising something? I just ordered your HDV book so I hope I will up to par on this stuff soon.

Mike S

Mike, you've got it right; the pixels are stretched. And it's not dishonest; it's just the way it works.

DV is 720x480, but it fills a rectangle that is proportional to 654x480. This means that the pixels are only 0.909 as wide as they are tall. In this case they're squished instead of streched. For wide 16:9 DV, 720x480 is still used but they're stretched out to fit a rectangle that is 853 wide. It's still exactly the same number of pixels in the frame, but they're displayed wider instead of narrower. This allows different screen formats to be recorded with the same data format.

But when you strech the pixels to fill more space you are loosing spacial resolution while saying 1440 yeilds 1920 implies more resolution. This is what I see as dishonest. Why are they using pixel counts (prorated ones to) to represent size and ratios? Seems it would be better to have all pixels nominaly square and then say the width is 1440 pixels streched to an increase of 1.3 times. If I have an image called 1920x1080 and I blow it up I expect to be able to count 2,073,600 dots on the screen.

Ah, but now you're talking about resolution, which is an entirely different matter than just "pixels." For resolution measurement you use a resolution chart (the one that has the converging lines). And yes, it's true that when measured with a resolution chart, a system having fewer horizontal samples will have a correspondingly lower resolution. But it's still a heck of a lot better than the 480x720 DV resolution.