Pair of Interesting Articles: Nordahl HDTV Reports -- formats and cams

[Chris Hurd]

Pretty interesting take on things... check this out:

Part 1: What is the future for HDCAM & DVCPRO-HD? by Tore Nordahl

"we compare the established formats of Sony HDCAM and Panasonic DVCPRO-HD."

Link is for Part 1 is http://www.coax.tv/DefaultJAN17.htm

Part 2: Is Sony's XDCAM HD just HDV; and Canon, JVC & Sony's HDV camcorders vs. Panasonic's HVX200, also by Tore Nordahl

Link for Part 2 is http://www.coax.tv/Default.htm

To my eye, this is where it gets really interesting (partial quote): "Because the DVCPRO-HD progressive format limits the bandwidth to 960x720 from the full 1280x720... it does not make sense for Panasonic to have full count 1280x720 CCDs as the recorded image quality on the P2 memory cards will be limited to the DVCPRO-HD progressive format’s reduced luminance bandwidth of 960x720 anyway!"

Some great reading at these two links. Enjoy,

[Robert M Wright]

When you stop and think about it, 960x720 (luminance) resolution isn't really a huge step up from PAL SD.

[Chris Hurd]

Yeah, that's been said here before... but in the long run, numbers simply don't matter. I threw out these links just to show a different perspective on things. The HVX200 gives us the single most affordable way to shoot DVCPRO HD, and that's what counts most, above everything else.

[Barry Werger]

Originally Posted by Chris Hurd

To my eye, this is where it gets really interesting (partial quote): "Because the DVCPRO-HD progressive format limits the bandwidth to 960x720 from the full 1280x720... it does not make sense for Panasonic to have full count 1280x720 CCDs as the recorded image quality on the P2 memory cards will be limited to the DVCPRO-HD progressive format’s reduced luminance bandwidth of 960x720 anyway!"

But that's just 720p mode... not to open that whole can o' wormies...

[David Andrews]

Thanks for the links - the clearest descriptions of the differences I have read so far.

As an aside, the WRAL-TV station referred to in the first article is, according to posts on the Canopus Edius forum, installing/using the Edius Broadcast nle for a new HD system run over a network.

[Steven White]

Hm. I'm not sure I believe his description of the FX1/Z1U chroma. He says it's 480 pixel shifted to 720... However, it's a 3CCD 960x1080 system, and should therefore have close to a full 4:4:4 960x1080 chroma grid to work with.

-Steve

[David Heath]

Originally Posted by Chris Hurd

Pretty interesting take on things...
.........
To my eye, this is where it gets really interesting (partial quote): "Because the DVCPRO-HD progressive format limits the bandwidth to 960x720 from the full 1280x720... it does not make sense for Panasonic to have full count 1280x720 CCDs as the recorded image quality on the P2 memory cards will be limited to the DVCPRO-HD progressive format’s reduced luminance bandwidth of 960x720 anyway!"

Some great reading at these two links. Enjoy,

It is a pretty interesting take on things, and one of the best (and most accurate) pull togethers of technical matters I've seen. The danger begins when some of the technicalities are seen as prime motivators for "X is the best buy because of these figures", rather than as background to the general subject.

That said, and referring to Chris's quote above, I slightly disagree with the authors conclusion, since it only refers to 720p - DVCPRO HD will resolve higher horizontal resolution than 920 in 1080 mode (1280). In practice, horizontal pixel shift will provide that extra resolution (for luminance), so I'd probably agree that 960 pixel chips with pixel shift were a sensible choice.

I'm not too sure about one aspect of the authors following paragraph:

But, remember, there is one possible post processing advantage with DVCPRO-HD over JVC: it is 4:2:2 while HDV is 4:2:0. Well not quite, as DVCPRO-HD in 720p is really 3:1.5:1.5 because of the L960-C480x720 ,......

We're led to believe that the HVX pixel shifts vertically as well as horizontally, which is a good way of gaining vertical luminance resolution from chips which may not seem to be capable of it from initially looking at the numbers. One aspect of this approach is that luminance resolution will be higher than chrominance. That's fine - it's how the eye works - but does bring into query why bother with a 4:2:2 recording system (equal luminance/chrominance res vertically), when the output of the front end has much lower vertical chrominance res (relative to lum) in the first place?

[Keith Wakeham]

Nice read and looking forward to part 3. As Steven mentioned, the sampling of the sony z1u / fx1 may be incorrectly identified but unless a sony engineer tells us the truth it is hard to know but the end result is still the same - the chorma has to be interpolated.

[Ken Hodson]

I found this interesting as I never had heard that befor. It might explain why Canon scored so much higher or resolution tests then the other HDV cams when live output tests were performed. It appears then that it is probably a lot closer to the other HDV cams when capturing to tape.

"Canon pixel shift increases perceived resolution to nearly 1920x1080 full count!. However, the near full count is only available on the HD-SDI output"

[Ken Hodson]

Originally Posted by David Heath

I'm not too sure about one aspect of the authors following paragraph:


We're led to believe that the HVX pixel shifts vertically as well as horizontally, which is a good way of gaining vertical luminance resolution from chips which may not seem to be capable of it from initially looking at the numbers. One aspect of this approach is that luminance resolution will be higher than chrominance. That's fine - it's how the eye works - but does bring into query why bother with a 4:2:2 recording system (equal luminance/chrominance res vertically), when the output of the front end has much lower vertical chrominance res (relative to lum) in the first place?

I have seen this exact same issue being currently discussed on the DV.com forums. Some say that because DVCproHD is only capturing 3/4 of the 720p format that it should be listed as 3:1.5:1.5 others have stated that the first number is not a reference to full luma resolution but strictly a chroma reference. I am torn myself. There does need to be a way to reference that DVCproHD is not full 720p luma, so there for would not hold as much chroma info one would expect when 4:2:2 720p is listed. Maybe we need a whole new system?

[David Heath]

Originally Posted by Ken Hodson

I am torn myself. There does need to be a way to reference that DVCproHD is not full 720p luma, so there for would not hold as much chroma info one would expect when 4:2:2 720p is listed. Maybe we need a whole new system?

Agreed to the last sentence. I have no concern about such as DVCPRO HD being described as 4:2:2 in isolation - that's quite valid, and a simple ratio is being referred to. What I do disagree with is when direct comparisons are made with it and a 1280 recording system, and no attempt is made to explain that the '4' means a different thing in each case. If DIRECTLY compared to such as HDV1, '4:2:2 v 4:2:0' has no validity.

Whether we talk of '4:2:2' or '3:1.5:1.5' doesn't affect the real point I was intending. Pixel shift means colour resolution will be less than luminance - so what's the point of combining a camera front end using vertical pixel shift with a recording system that is capable of equal vertical lum and chrom resolutions?

[Tim Dashwood]

Part 3 is out:

http://www.coax.tv/Default.htm

[Ken Hodson]

How does pixel shifting effect chroma rez? And by how much?

[David Heath]

Originally Posted by Ken Hodson

How does pixel shifting effect chroma rez? And by how much?

Can I refer you back to my last paragraph, post 7 this thread?

"One aspect of this approach is that luminance resolution will be higher than chrominance. That's fine - it's how the eye works - but does bring into query why bother with a 4:2:2 recording system (equal luminance/chrominance res vertically), when the output of the front end has much lower vertical chrominance res (relative to lum) in the first place?"

That said, you should be cautious about quoting and referring one set of specs, whilst ignoring others, and the whole colour space issue is a prime example of this. Marketing men look for simple ways to get their message across, and "4:2:2 good, 4:2:0 bad" would obviously suit that philosophy. Unfortunately, the truth is far more complex. I frequently hear "DV50 is so much better quality than DV25, it's because it's 4:2:2". The first statement is undeniably true, the second a gross and misleading oversimplification. 4:2:2 may indeed help the quality gain, but the reduced level of overall compression (in DV50 v DV25) may well have as much if not more effect. But that would rather spoil the ad mans one liner, wouldn't it?

[K. Forman]

What I got from those articles, is the guy hates tape. What I would like to know, is how do you go about getting 4:2:2 out of these cams? Aside from the ProHD tapes, once it is compressed to tape, you have DV 4:1:1, right? I suppose if you are in a studio, running hd-sdi into a pc is possible, but what about being in the field- literally? Out in the middle of nowhere, what is practicle and affordable?