YC 4:2:2 or 4:1:1?

[Evan Fisher]

What about using a device like a Datavideo DAC5 DV to SDI converter or a Canopus ADVC 1000 DV to SDI converter. The claim is that they take a DV signal and send it out 4:2:2 via SDI?

[Graeme Nattress]

Because any device that takes DV and converts to 4:2:2 is not putting any extra information in - they're just interpolating the new colours inbetween with no "real" information there!

Graeme

[Barry Green]

Yeah, you can actually do that without hardware, depending on the DV codec you use. Avid's codec does chroma smoothing, and you can configure the Matrox codec to do so also. And Vegas includes a "chroma blur" option which will smooth out the chroma.

Very, very handy thing to do for up-rezzing or chroma keying. It doesn't create any actual new information, but it definitely smooths things out and delivers better results.

[Thomas Smet]

this experiment would be much better if there was a dv camera that actually had a component output. The I could get a full U and V channel. Although that still doesn't mean it is 4:1:1 or 4:2:2. I once had this ancient Panasonic 300 CLE camera here. I wish I still had it. It wasn't digital since it was 20 years old I think but it was a 3 chip 2 piece pro camera. The great thing this 2 piece camera is the SVHS deck could come off. You then had a 26 pin connector which could be converted to a component cable. I know it wouldn't help with my dv/analog capture experiment but it would have been a good way to compare a raw component uncompressed video with a dv video. I use a lot of high end cameras now but they are all 1 piece and the best video out they have is composite which is worthless. Anybody have an old camera head with a 26 pin connector. If you do you can buy a cable to get component from the 26 pin for about $50.00.

[Barry Green]

The firewire is a component output, just compressed to DV first.

If you really want high-res 4:4:4, consider the new Sony HDR-FX1. It's 1440x1080, but if you downsample it to 1/4 size (720 x 540) you could get 4:4:4 color resolution out of it.

[Thomas Smet]

yeah I have thought about the SONY HDV camera. I would be interested in testing out some 1080i HDV footage from this camera when it comes out. I figured there would be other benefits from 1/4 scaled HDV footage as well.

1. a psuedo 4:4:4 color YUV image. I say psuedo because it wouldn't be a raw pixel per pixel image. The 4:4:4 color image would be interpolated to get to that point.

2. 30p or 60p footage instead of 30i. You can either throw out every other line for for 30p or use each field line scaled down to get 60p.

3. Mpeg2 HDV compression artifacts would get smoothed out and be partly eliminated from the scale down.

4. Noisy footage from low light and high gain will get smoothed out.

If this works out the way it does in theory then the SONY HDV camera will have three shooting modes instead of 2.

1. Normal 4:1:1 DV.

2. Normal 4:2:0 HDV 1080i

3. Psuedo 4:4:4 SD at 720 x 540 anamorphic. To get real SD however the image would have to be scaled to 853 x 480 or 720 x 480 anamorphic.

[Evan Fisher]

Barry,

Do you know if FCP HD has a codec or filter that smooths the chroma?

[Graeme Nattress]

For FCP you have two choices I know of:

Apple's 4:1:1 keying filter, does a linear chroma interpolation to improve the colour sampling of DV,

and my G Nicer Chroma reconstruction filter, which intelligently works out what the chroma should have looked like and reconstructs it for you.

You can see my comparisons here:

http://www.nattress.com/filmEffectsGNicerTests.htm

BTW, G Nicer V2.0 will look even better still.....

Graeme

[Andre De Clercq]

In addition to Barry's posting I want to add that talking about 4:2:2 or 4:1:1 (4:2:0 for Pal and DVD) is irrelevant in terms of S-video (Y/C) or composite video. In best case one could estimate how comparable a DV compressed 4:... signal is to S-video in terms of resolution. The 4:...whatever is an information originating from a spacially sampled form of video images, S-vid and composite video are encoded (NTSC,PAL, SECAM) forms of non sampled (or reconstructed) video signals.

[Valeriu Campan]

Thomas,
I thought about a component output option for the Sony DSR 570P. I was wondering if that camera outputs from the CCU port a component signal and that signal comes before going through the DVCAM circuitry. Sorry for posting it here, just followed the discussion.
Thanks for any replies anyway.

[Thomas Smet]

One other great thing about analog or SDI capture compared to "DV" firewire is with DV your video will always be 8 bit per channel. With certain analog and SDI capture boards you can get 10 or 12 bit video per channel. This is a huge bonus for color correcting. Yes you may get slightly lower resolution or sharpness but you make up for it with the bit depth.

Analog can also fix aliasing issues you may get with dv actually being too sharp. Super sharp video is almost the best thing to have. Expecially with digital compression. The sharper the image the harder it is to compress and the more artifacts you will get.

Finally somebody mentioned that YC might only give you 4:2 color because only the luma is seperated and the U and V are combined into one C channel. This is true but the C is just stuffing both the U and V into that channel. Both U and V can be extracted again. The problem with YC is that some small amount of color information can be lost or it can drift or bleed during the conversion. Most of it is still there however. Well at least enough to get 4:2:2. If you have a bandwidth of 4 and you need to fit two channels in it you cut them in half to 2 each. If the 4:2 theory were to work then you would have to say that since composite puts all three YUV channels into one channel that you would only have 1.33:1.33:1.33 color space or a 4:0:0 or just 4. Therefore I think if you were to give a number system to YC it might be 4:4 where the second 4 is made up of 2 2's. Boy is this stuff fun.

[Jeff Donald]

Thomas, is there any basis for your ramblings? You seem to be ignoring the replies of very knowledgeable members with expertise in a wide variety of areas. What is your degree in? Are you working in the industry?

[Thomas Smet]

if it is ramblings then why are you here reading and talking about it. You must have had some level of interest at some point to come here. I happened to just be asking if anybody knew about which cameras had a raw YC output and all of a sudden people came on here talking about how bad analog is. I asked a simple question which somehow turned into an argument by those defending DV. I never once bashed DV or praised analog but was trying to compare the two. Some people listed the faults of using analog and I was just pointing out that there are also benefits as well. I thought some of us were talking about ways of getting the best image quality but it seems some people do not like other people to talk about that stuff.

I was not being rude about any of this but trying to talk about higher quality color. If you do not like what I or others are talking about then you do not have to read any of it. I am not ignoring anything anybody has said. I know there is no way of trying to compare a digital 4:?:? signal with an analog but I was trying to come up with a way to compare everything. I am not a video engineer which is why I am on this forum trying to figure this stuff out. I do however have a BFA in visual effects and am trying to work a better image out of cheaper cameras for compositing . I am also currently starting up a software company and making 4 different video and vfx related programs.

I know digital is for the most part better. I also know starting with a high end digital format at 4:2:2 or 4:4:4 is better. Of course SDI is better. What I am getting at is how much better? It is easy to say something is better but how can you put a value on it? SDI will be better than component but by how much?

Many people have been quick to say analog wouldn't work for increased color resolution but so far I haven't had enough reason from them to feel the same way. All the points they made I agree with. I was however trying to bring up other points that nobody talked about yet trying to help the topic. I am sorry if anybody feels as though I have been "rambling"

[Jeff Donald]

Thomas, I'm glad you've come to learn. SDI (serial digital interface) is not a format. It is a pipeline. It is the means to move a signal from one device to another. Just like FireWire. FireWire and SDI are not formats. Formats are a standard for recording video. Today we have formats like DV, DVC PRO 25, DVC PRO 50, Digital Betacam, D1, D3, D5. These are all digital formats and their digital signals can be moved between compatible devices via SDI and in many cases FireWire.

The digital signals can be uncompressed, such as 4:4:4 video, or compressed into 4:2:2, 4:1:1, 4:2:0 etc. The first number represents the luminance channel and the next two numbers represent the chrominance channels. In 4:2:2 video, for every four luminance samples the chrominance channels are each sampled twice. Betacam SP is a form of CAV (component analog video) in which the signal is divided into 3 components, 1 luminance channel and two color channels, each carried on a separate wire. Each wire carries an analog voltage that varies with picture content.

There are various types of analog CAV signals as specified by NTSC or PAL, such as (Y, U, V), (Y, R-Y, B-Y), (Y, I, Q), and (Y, Pb, Pr). Again the first letter Y, represents the luminance signal and the next two sets of letters represent the color difference signal.

S-Video or more commonly Y/C is not true component analog video. The signal is merely separated in to one luminance signal and one chrominance signal.

[Graeme Nattress]

<<<-- Originally posted by Thomas Smet : One other great thing about analog or SDI capture compared to "DV" firewire is with DV your video will always be 8 bit per channel. With certain analog and SDI capture boards you can get 10 or 12 bit video per channel. This is a huge bonus for color correcting. Yes you may get slightly lower resolution or sharpness but you make up for it with the bit depth.-->>>

But DV is 8 bit per channel!! How can it be anything else?? FCP, for instance, will, if you ask it nicely, calculate a lot of it's filters in 32bit, It hink, thus you edit with compressed DV, but as soon as you apply a filter it gets fully uncopmpressed, all effects calculated in high precision mode, then back to 8 bit DV for layback to tape. 10bit etc. is only needed for the calculations. 8 bit, if properly dithered, has enough levels to accurately represent the video signal for most purposes. Yes, 10bit is better, but 8 bit properly done is good enough. Also, any SDI or ananlogue output from DV will be 8bit or from an 8bit DtoA. You won't get any more than that!

<<<--Analog can also fix aliasing issues you may get with dv actually being too sharp. Super sharp video is almost the best thing to have. Expecially with digital compression. The sharper the image the harder it is to compress and the more artifacts you will get.-->>>

DV is already pre-filtered before going to tape to remove the higher frequencies. If you do get a specific aliassing issue, surely it's better to write a filter to address just that particular issue rather than reduce the rez of the whole capture???

<<<--Finally somebody mentioned that YC might only give you 4:2 color because only the luma is seperated and the U and V are combined into one C channel. This is true but the C is just stuffing both the U and V into that channel. Both U and V can be extracted again. The problem with YC is that some small amount of color information can be lost or it can drift or bleed during the conversion. Most of it is still there however. Well at least enough to get 4:2:2. If you have a bandwidth of 4 and you need to fit two channels in it you cut them in half to 2 each. If the 4:2 theory were to work then you would have to say that since composite puts all three YUV channels into one channel that you would only have 1.33:1.33:1.33 color space or a 4:0:0 or just 4. Therefore I think if you were to give a number system to YC it might be 4:4 where the second 4 is made up of 2 2's. Boy is this stuff fun. -->>>

Y/C is inherently bandwidth limited on both chroma and luma.

Graeme