Why 4:2:0 ?

Status
Not open for further replies.

naveenmedia

Active member
So why does Sony use 4:2:0 for color sampling rather than 4:2:2 ? From my understanding, the latter is FAR better than the former.

How is XDCAM-HD footage for accuracy, and especially keying?

Or is there something about 4:2:0 in HD vs. HDV that I'm not aware of?

Naveen
 
So why does Sony use 4:2:0 for color sampling rather than 4:2:2 ? From my understanding, the latter is FAR better than the former.
A lot depends on whether you're talking about an interlace or progressive system. There's a lot more logic to using 4:2:0 with progressive images, as the subsampling is then truly symetric, in a way it isn't with interlace.

Also bear in mind these are ratios, and the EX is a true 1920x1080 recording. If one job offered you 25% of the takings, and another 33%, which would you go for? But what if it was 25% of $100, versus 33% of $50........? Overall figures can be more relevant than ratios.
 
So why does Sony use 4:2:0 for color sampling rather than 4:2:2 ? From my understanding, the latter is FAR better than the former.
Of course it is. It's twice as much color info.

Sony uses 4:2:0 because they are trying to cram their HD signals into the tiniest bandwidth possible. Going to 4:2:0 is just another a form of compression; they throw away half the color information.

Or is there something about 4:2:0 in HD vs. HDV that I'm not aware of?
You're not missing anything. 4:2:0 is inferior to 4:2:2.

With that said, 4:2:0 in progressive is not nearly as nasty as 4:2:0 in interlaced; interlaced 4:2:0 is hideous. Progressive 4:2:0 is better than interlaced 4:2:0. But neither holds a candle to genuine 4:2:2.
 
I hear that 4:2:2 is possible via HD-SDI out. But doesn't that mean that just the codec is 4:2:2, for improved color grading?

If the cam was hooked up to a computer going directly out of the HD-SDI (which is quite improbable), would one get a true 4:2:2 colorspace?
 
The codec is 4:2:0. The internal processing is 4:2:2, and apparently the HD-SDI output is 4:2:2. But as it records, it subsamples down to 4:2:0, as a form of compression -- discarding half the color information from the internal 4:2:2.

Yes if you hooked up to a computer (or a DVCPRO-HD tape deck or HDCAM-SR tape deck) you would get true 4:2:2.
 
I hear that 4:2:2 is possible via HD-SDI out. But doesn't that mean that just the codec is 4:2:2, for improved color grading?

If the cam was hooked up to a computer going directly out of the HD-SDI (which is quite improbable), would one get a true 4:2:2 colorspace?

Yes. However, this llooks a lot more probable these days. You can have a laptop record the HD-SDI signal with a Kona IO HD card, or the upcoming COnvergent Designs capture box, wich captures the HD-SDI signal to CF cards and converts it to whatever open source codec you would like...
 
Yes. However, this llooks a lot more probable these days. You can have a laptop record the HD-SDI signal with a Kona IO HD card, or the upcoming COnvergent Designs capture box, wich captures the HD-SDI signal to CF cards and converts it to whatever open source codec you would like...

And how much would that cost? (The cards/any necessary software)
 
About 5000 for the CF recorder. So a canon g1 and xdcam ex, with these recorder are theoretically capable of fantastic 160 mbit mpeg 2 intraframe 4:2:2 pictures, making these 2 cameras the fixed lenses cameras to beat in therms of picture quality. A G1 goes for 7000 USD right now, if I'm not mistaken. A XDCAM EX goes for 8 000. So a package should be 12000 USD for the Canon g1 option, and 13000 for the EX.

A fantastic deal for native 1920x1080 4:2:2 160mbits mpeg2 intraframe portable recording solution, in my humble opinion.

However, the CF recorder is not available yet, so this is all theoretical, of course.
 
Of course it is. It's twice as much color info.

Sony uses 4:2:0 because they are trying to cram their HD signals into the tiniest bandwidth possible. Going to 4:2:0 is just another a form of compression; they throw away half the color information.

You're not missing anything. 4:2:0 is inferior to 4:2:2.
True if you compare like with like.

But problems come if you compare the Ex codec with DVCProHD on the basis of colourspace, as that's NOT comparing like with like. They are ratios, and meaningless without the reference.

At 24fps progressive, the EX codec records 1920x1080 samples/frame for luminance, so 960x540 (518,400) each U and V pixels recorded per frame. Corresponding figures for DVCProHD are 480x720=345,600 (for 720p) or 640x1080=691,200 (for 1080p) - so in these particular cases we see a 4:2:2 system can actually have considerably LESS colour info per frame, or only 33% more in the best case.

And in the case of the 960x540 chips used in the HVX200 and 500, much of those extra recorded pixels are wasted anyway. Pixel shift only enhances luminance resolution.

Colour space comparisons only have sensible meaning if like is compared with like.
 
Thanks, but I still don't have a full sense of exactly what this box does.

I gather it will convert an SDI 10 bit 4:2:2 signal thru firewire directly into a Mac and it will be already converted into the Apple Hi Res Codec. Does that sound right. Will it then already be a quick time file and ready to play in FCP or does FCP have to do an import of some kind first?

I am guessing then that this $3500 box would allow you to shoot green screen or anything else you were concerned about with an EX-1 that was 10 bit 4:2:2.

Its either that or lug a Mac Pro tower to the set with a Kona card. ( Actually a spare tower with a card might not be much different expense, though a bit heavier.)

Its actually not that much money if you do a lot of green screen or for a rental house to get if the EX-1 becomes very popular. Actually that's assuming that you need the 4:2:2 to get decent green screen.

What else does this thing do?

Lenny Levy
 
Last edited:
Lenilenapi, to really get the specifics of what this box does, just call AJA or a reseller, and ask a sales person. Unless someone here knows. Having never used the box (we considered buying one, but are going with a tower and card instead), I don't know.
 
At 24fps progressive, the EX codec records 1920x1080 samples/frame for luminance, so 960x540 (518,400) each U and V pixels recorded per frame. Corresponding figures for DVCProHD are 480x720=345,600 (for 720p) or 640x1080=691,200 (for 1080p) - so in these particular cases we see a 4:2:2 system can actually have considerably LESS colour info per frame, or only 33% more in the best case.
Huh? Why are you comparing DVCPRO-HD 720 to XDCAM 1080, and drawing an arbitrary and incorrect conclusion from that? (and when did DVCPRO-HD enter into the discussion at all, anyway?)

XDCAM EX 720: 640x360 = 230,400 chroma samples
DV100 720: 480x720 = 345,600 chroma samples.

XDCAM EX 1080: 960x540 = 518,400 chroma samples
DV100 1080: 640x1080 = 691,200 chroma samples.

Like for like, bro.

In any case, this is all still impossible to compare because XDCAM's pixels may or may not even be the accurate pixels for that particular position depending on how heavily the codec is being taxed at the time.

Colour space comparisons only have sensible meaning if like is compared with like.
Agreed. So let's compare 1080 to 1080, and 720 to 720. And someday let's have Sony make a decent intraframe codec so we can compare intraframe to intraframe.

But back to the original question -- there was no comparison of DVCPRO vs MPEG. The question was: why would Sony choose to use 4:2:0 if 4:2:2 is better? And the answer is? Anyone? Bueller? Well, the answer is: they do, when they raise the price tag. The new XDCAM 50mbps version is going to be 4:2:2. Why? BECAUSE 4:2:2 IS BETTER. They just cheaped out on the EX with 4:2:0 even though they were already redesigning the codec specifically for XDCAM HQ, and they had plenty of bandwidth to store the full 50mbps signal.

So the answer to the original question:
So why does Sony use 4:2:0 for color sampling rather than 4:2:2 ? From my understanding, the latter is FAR better than the former.
The answer is: they lamed out. They could have, they just didn't. They had the opportunity since they were changing the codec anyway, and they have the technology, and they have the bandwidth. They just didn't do it.
 
Last edited:
...
But back to the original question -- there was no comparison of DVCPRO vs MPEG. The question was: why would Sony choose to use 4:2:0 if 4:2:2 is better? And the answer is? Anyone? Bueller? Well, the answer is: they do, when they raise the price tag. The new XDCAM 50mbps version is going to be 4:2:2. Why? BECAUSE 4:2:2 IS BETTER. They just cheaped out on the EX with 4:2:0 even though they were already redesigning the codec specifically for XDCAM HQ, and they had plenty of bandwidth to store the full 50mbps signal.

So the answer to the original question:
Quote:
So why does Sony use 4:2:0 for color sampling rather than 4:2:2 ? From my understanding, the latter is FAR better than the former.
The answer is: they lamed out. They could have, they just didn't. They had the opportunity since they were changing the codec anyway, and they have the technology, and they have the bandwidth. They just didn't do it.


From a technical standpoint, 1920hx1080v at 4:2:2 wouldn't work well at 35Mb/s, but it would work at 50Mb/s. (35Mb/s would have too much data to compress cleanly.) Now at 1280hx720v 30p or 24p, 35Mb/s might be possible...

To me, it appears to be more of a marketing issue than a technical issue. Sony hinted of higher end models showing up at NAB. It seems likely that the higher end models will have the 50Mb/s 4:2:2 mode.

Firmware upgrades of the camera are still possible and Sony did say that an upgrade will be needed to support the 32GB cards when they come out and additional features could be added. Other than the 32GB SxS support, Sony is NOT saying what additional features might be added. So, a 50Mb/s 4:2:2 mode, might be added, BUT we can't say for sure either way until Sony says what the firmware upgrade will add.

IF Sony is waiting for the 32GB SxS card to support 50Mb/s 4:2:2 mode, it seems logical, because at that rate, 1 GB gives us only 2.25 minutes of video. Thus an 8GB SxS = 17 minutes, 16GB SxS = 35 minutes, and 32GB SxS = 70 minutes.

On the other hand, maybe Sony doesn't want to make the EX-1 camera so good, that sales of the higher end models are lost to the lower end model.

Maybe this is their ace in their back pocket in the event that other camera companies come up with improved models in the same price range....


Bob Diaz
 
Huh? Why are you comparing DVCPRO-HD 720 to XDCAM 1080, and drawing an arbitrary and incorrect conclusion from that?
Because some people have said to me, comparing the two examples above, that "oh, it's 4:2:2 v 4:2:0, no contest" and been genuinely surprised to have the subtleties pointed out. They haven't appreciated that colour spaces mean ratios, not actual numbers. Which is why I started by saying that what you said was true - but with the like for like qualifier. I'm sure you are well aware of that - but not everybody else is. And if it isn't stressed, if like is not compared with like, *then* comments like "It's twice as much color info" in isolation are what lead to arbitrary and incorrect conclusions.

And as for DVCProHD 720 to XDCAM 1080, then that reflects the camera front ends. The Ex has 1920x1080 chips, the HVX200/500 have 960x540. The EX front end is matched to 1080 codec recording, the others can't do it full justice. Use the EX at 720 and there is considerable detail loss relative to 1080 - with the others the loss is much less, certainly in the vertical direction. What's the point of throwing bandwidth away trying to record detail that isn't there? The comparison was XDCAM 1080 to DVCProHD 720 *AND* 1080, anyway.

For interlace I fully agree that 4:2:2 is highly desirable. For progressive it may be desirable, but not at any cost, and not if it means an overall lower resolution.
 
On the other hand, maybe Sony doesn't want to make the EX-1 camera so good, that sales of the higher end models are lost to the lower end model.
All well put, Bob. I'm actually surprised that the EX does full 1920x1080 recording versus the 1440x1080 of the other XDCAM-HD's regardless of colour space. The XDCAM dsic is a double edged sword for Sony - desirable because many users like the "consumable media" concept (whilst still being tapeless), and a millstone because of lower max data rates than solid state. (Even SD cards, let alone SxS, CF or P2!!)

I foresee interesting developments at NAB.
 
Status
Not open for further replies.
Back
Top