And to go to anything better than 720p means going to 1080p, not 1080i, and not 1080psf where progressive is broadcast as interlace, and such progressive for broadcast is often filtered vertically as 1080i is, so you don't get the full vertical resolution 1080p is capable of.

Of course, take into account the compression that's applied in broadcast, you'll get, given the low bandwidth allotted to HD broadcasts, a better picture with 720p anyway. There's no point in HD if all it does is increase the number of nasty artifacts you see.

Graeme

Again, that's a display quotient, not an acquisition quotient. Until we have 1080p60, for fast motion with clean detail, 1080i is the best we have.

I'll bow out of the discussion as it's getting rather stale, as we're discussing things that are not changeable by any of us and playing ring around the rosie with opinions. My apologies for having kept my end of it up this long. Might as well be discussing how soon we'll be at war with Iran.
so as my final comments....
We shoot both formats here. We have everything from the bottom end HD camcorders to full-time access to an LDK 5000. (Grass Valley's international service center is here, and their director of service for the US is one of my closest fishing buddies) For what we shoot, and for what our clients money for the most part, 1080i is the best we can offer them affordably.
Motocross looks terrible acquired at 30p in any flavor. Same goes for lacrosse, and especially same for the high motion military stuff (like when helicopters are brought in to catch Mars rovers parachuting to earth)
On the other hand, we've used the JVC HD100U on two corporate projects, and have been very, very happy with the results. We're also very impressed with JVC's BR 50 upscale of DVX originated footage.
I also believe that there is a lot of ignorance about conversion quality, and the quality of the tools used to make those conversions. As an example, FCP and Adobe PP2.0 can't scale to save their a$$. Premiere 1.5 does a pretty good job. I'd wager that roughly 8-10 hours a day of my daily life involves downconverting, with maybe 10 hours per week in upconverting depending on the client, so I'm pretty comfortable with most of the tools, having experimented with virtually all of them in hardware and software. Sony Vegas, Algolith, and Red Giant are all wonderful software tools. I use Vegas most of the time, but occasionally use Algolith if I end up working AE.
I might not be able to get as deeply into the math as some folks can. But I'm not by any stretch blind, we own several different HD displays and two projectors here, and have a fairly critical client list, two of whom offer HD authoring tools. If we were putting out the "lesser grade" that some folks call 1080i, or 1080p30 converted, or in a couple cases, 1080i converted to 1080p60, then believe me, both of those software companies would be crawling up our rectums. Especially our client in Germany, who is likely the best known h.264 company on the planet next to Quicktime's new, but slightly inferior QT wrapping of .264. When we ship them HDCAM, we've yet to have a complaint. We haven't shot anything for them in HDCAM, but that's how we deliver. Prior to that, it was all the loaner 5000, or our older 700. Cost went down with the HDV cams, client is pleased with that, because they also feel quality didn't coincidentally go down.
We're happy, they're happy, they've noticed issues (made up in their minds or not) with 720p, and they've specifically asked for 1080i converted to 1080p30 or 1080p60 (in one case only) based on what we've sent them, all footage from all three HDV cameras available as of December 05 (not including the A1u)
We're likely moving to the XDCAM HD platform for most of our work, I've had quite a bit of time with the 350 thus far, and am disgustingly impressed. We'll likely go with the 350 with the Fuji 38 lens for a main cam as soon as they start shipping.

This is where the whole discussion seems to be falling apart - while 720p60 is undoubtedly better than 1080i, which cameras shoot it? Isn't the HVX the first to give this option (I'm talking about in the sub-$10k dv market)? So how have we been able to shoot this for ages?

Which means the comparison between 720p and 1080i in this market is primarily one of temporal vs. spatial resolution. 720p30 gives twice the spatial resolution while 1080i gives twice the temporal resolution. While 720p may convert better spatially to 1080p than 1080i does, 1080i should likewise convert better to 720p60 than 720p30 will, at least if motion is a concern. If you can accept the P2 workflow then the HVX at 720p60 may be the best of both worlds (assuming the CCDs are capable of maxing the spatial resolution of the format) until 1080p60 becomes practical. Otherwise the better format is the one which best fits your subject - 720p for dramatic, scripted, static or controlled motion shots vs. 1080i for sports, action, reality/documentary, etc.

hey guys - great points made by all. thanks. It reads to me that Graemes' view is theoritically correct , and many here are indebted to him for his theorical knowledge. Douglas' point is based on real world equipment and it's limitations. Obviously his clients are his judge but there's no sustitute for handson experience. Most here will never have the opportunity to use 1/4 of the equipment he uses.
What we see now with the available cheap hd equipment is they're all performing more or less equally. Maybe the next generation will get closer to perfection , i.e. 1080/60p. I guess what I liked about Joshs' article was that all of the cameras have achilles heels and no one should feel they need to go out and get the next best toy to shoot good video. I think alot of sony owners felt that canon, jvc , and panasonic were being touted as being sooo much better that the sony hdvs' were passe' and they needed to figure out how to raise 10k or just go back to shooting dv. The truth is last week I saw a mint arriflex bl with 10-100 zeiss and internal lightmeter sell for 1200 bucks. That's what these small, cheap hd cameras have done. I mean , who in their right mind would want to still pay $50 a footage/minute ? Now we can start filming instead of butting heads over formats and quality. Until the xdcams define the next gen we're all more or less in the same boat- 720p,108060i are just splittin' videohairs. thanks esp. to Douglas and Graeme for turning this thread around- Kurth

"last week I saw a mint arriflex bl with 10-100 zeiss....who in their right mind would want to still pay $50 a footage/minute ?"

well i'm in that $$ per foot area ... i looked/tested the 4 hand size HD camera's ... i wanted to BUY/spend but my wallet didn't jump out of my pocket .. i thought they were all OK ..but OK isn't good enough at this time ... so for short project, rented a spring wound 16mm bolex - scanned ( not telecined) at 2k res -now in post ...

>>UK home? Most of us live in a shoebox - and count ourselves lucky.... but I digress.
<<

David, come to the Washington DC/Northern Virginia area of the US, you'll feel perfectly at home. :)

And Douglas, if Sony offered financing, I would be looking at that 330 with a nice lense. I'm mainly interested in 24p, which should scale to 108060p just fine in the future. (For narrative work, not live stuff).

Graeme,

Since it appears you have a pretty good handle on digital video in general, I would like to share some thoughts. I am not a programmer, but completed a class in C, and I have been a student of compression techniques for quite awhile.

First, a question. If my understanding is correct, MP4 (like MP2) is based on the DCT. Is this also true of WMV9?. I ask because I can't help thinking that some sort of processing that could dynamically allocate data to spacial or temporal resolution as needed would be beneficial.

The process basically would be to capture frames into buffer (60p), then forgo resolution for frame rate as motion increases. It is obvious that we cannot see high resolution when there is motion.

Since you are into filters and such, how about a filter that could detect DCT blocks, then during high motion anti-alias based on the DCT blocks rather than pixels. It would make for a soft image of objects in motion, but it would be better than seeing the blocks. This would probably be easiest for the fixed 8x8 blocks of MP2, and could even be a standalone post processor.

Maybe this will get you thinking - either about what I have said - or that I am an idiot!

I thought MPEG4 was was like h.264 which is block matching based, neither wavelet nor DCT? I have no idea on WMV9.

Wavelet codecs go blurry when they compress too far, as opposed to DCT blocking artifacts, but if the video is blurry on motion, both tend to automatically be able to compress further as they're both frquency based and hence on a blurry image with less high freq info, they get better data rates, or you could allocate more to temporal.

h.264 has a smoothing option where blocks get smoothed, so I'm sure something similar could be put into a DCT based codec. Failing that, you can do some kind of artifact removal by looking for the characteristic blocks, but you're often just making the artifacts less visible, rather than removing them, as the data just isn't really there.

Graeme

Do you mean that 720p -- for instance at HD100, it can produce a better picture than 1080/24p at Sony's XDCAM-HD, for example? Or just to broadcast because compression reasons?

I'd say, all things being equal, 720p24 would compare very favourably to 1080p24 from the XDCAM HD, but I say that only because I couldn't get real answers on progressive resolution from Sony. They might have it giving full 1080p resolution, but then again, they might not.

Graeme

I suspect the truth is that whilst as a system 1080p/24 is capable of giving far better results than 720p/24, the front end of many cameras aren't capable of doing it justice.

Having an academic discussion about formats in isolation is one matter. Discussing actual cameras is another, especially in the case of 1/3" chips.

Graeme,

I remember reading that MP4 dynamically changed the dimensions of blocks, depending on similarity of pixels. They need not be square either. I guess I assumed it was still DCT processing. If somebody comes up with a good scheme to dynamically balance res and frame rate, it could end both interlaced and progressive scanning!

Leunami,

I think you have a very valid point. It's what gets delivered to the final viewer that counts. It does no good to create a super high resolution, then over-compress to meet limited bandwidth requirements. I have read that ALL broadcast 1080i is hor-res limited to 1440 for that very reason. Maybe someone can confirm or deny.

David, you're right that 1/3" HD is a great equalizer, as at that small chip size, they're all much of a muchness.

As for dynamically balancing rez and frame rate, that's not going to work, because frame rate is determined at the CCD / CMOS, and can't later be manipulated by the codec in any meaningful manner.

Graeme

Seems like you could always reduce it - easy to make 30p out of 60p. Throw away resolution where there is lots of motion. Throw away frame rate where there is little motion.

Sure, but I think the visual effect would be utterly bizarre!

Graeme