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

[Thomas Smet]

Does anybody know either from Canon or from their own tests if the composite or YC analog output on the XL2 outputs the video before it is compressed with 4:1:1 or after?

What I mean is uncompressed video is 4:2:2 and is what all CCDs produce. The camera then converts this signal to 4:1:1 to fit on the DV tape. Some cameras leave the video raw when sending the signal to the analog outputs so the video is 4:2:2. Other cameras however convert the signal right away before sending to dv tape or analog outputs.

Does anybody know how the XL2 handles this?

[Rob Lohman]

Uncompressed is 4:4:4, not 4:2:2. That is another form of
compression actually. At this point in time it is unknown how
the signal arrives at the analog outputs on any of the XL
series camera's. Testing might be in order with some high
contrast scenes etc. (and an EXCELLENT analog capture board,
because it would otherwise not matter at all).

[Thomas Smet]

yeah I know uncompressed video is 4:4:4 but not much suports that. In the video world 4:2:2 is kind of the high end standard and is considered uncompressed. Take a look at any uncompressed editing system such as the video toaster or dps velocity and their uncompressed is 4:2:2 going in and out of the board. Even though 4:2:2 has color compression it has no other form of image loss compression so it is considered uncompressed in terms of video.

[Rob Lohman]

I know it is being used a lot and it is considered near lossless
(or uncompressed), but it just ain't. I'm trying to avoid some
confusion here. I'm pretty sure the truly high-end packages like
combustion, flame, inferno, da vinci, digital fusion etc. all support
true 4:4:4 etc. So I'm assuming you are mainly talking in regards
to the broadcast sector?

Do you have a HQ capture card to do these kind of comparisons
with? Would be an interesting test for sure.

[Thomas Smet]

I do not have one here in this studio yet but will hopefully shortly. I do work with somebody who uses a Video Toaster but that is just broadcast standard at 4:2:2.

I totally agree with you about uncompressed being 4:4:4. Those are some very high end cards you listed. Most cards I have seen are only 4:2:2. At this point I think getting 4:2:2 even would be a lot better then 4:1:1.

[Rob Lohman]

Definitely. When I say HQ capture card I'm talking about one that
does 4:2:2 at least preferred uncompressed or at least a very
high bitrate MJPEG so we can actually check for the difference
between that and DV. Not some el cheapo TV capture card or
a dazzle for example <g> Just a professional capture card, 4:2:2
is fine for comparison!

[Damon Botsford]

I'm also very curious to find out what kind of video comes from the composite out. I have very little knowledge in this area as I've always used firewire for capture. Could it possibly be uncompressed raw video? (same question for Sony FX1 composite out) If so, what would be the best way to pull this into your NLE? I'm pretty dedicated to my Canopus setup and I'd like to stick with a Canopus solution... perhaps Edius Pro. I imagine if it was uncompressed, it would bog my 3.02ghz processor. Any thoughts?

[Graeme Nattress]

Uncompressed composite capture is daft, composite is going to look worse than firewire in the resolution due to a to D and D to A conversions. Even comming direct off the camera without going to tape, I doubt composite is going to look good.

Even if the S-video is 4:2:2 live, it's doubtful wether even a perfect capture of the analogue will be any better than a firewire capture after compression. It will also be hard to tell if the S-Video is 4:2:2 or 4:1:1 upsampled to 4:2:2 because even after recording to tape, the output from the camera will get upsampled and the chroma smoothed after the codec.

Finally, compression masks camera issues, like resolution, so even if you can pull a perfect analogue capture before DV compression, the picture may not end up looking any better because you get to see different issues with the camera that the DV codec masks.

Giving up the convenience of firect DV recording and capture for, perhaps, a marginal improvement at best, and the need to either tether the camera to a computer a portable digiBeta deck to record 4:2:2 is a bizarre way to make a movie...

Graeme

[Damon Botsford]

Graeme,

Thanks for your input. I had a feeling that was the case. I guess it's just wishful thinking. I've invested quite a few bucks into my NLE + computer and I think I've pretty much came to the conclusion that the XL2 via firewire is my best option for the next year or so.
I've been checking out the clips and frame grabs posted recently and it appears that if I learn how to utilize the XL2's functions, I can come up with some great images. Even with the new HDV cameras coming out, I'm really skeptical about maintaining a nice picture after major mpeg compression, editing (a whole 'nother story) and then recompression. Perhaps the format will be fine tuned next year. Until then, I can't wait to see some shorts made with the XL2.

[Thomas Smet]

The fact is that DV and firewire were intended as a consumer format. It just so happened to look good enough for people to start using it for pro use. Most high end video actually uses analog connections and not firewire. Of course they do not use S-video or composite but they do use component which does look very good. The funny thing about firewire is that everybody thinks that is whats doing the trick for them seeing good video. Do you hook your firewire cable to your TV? When we are actually watching our video we shot we are hooking the cameras or decks up to our TV's with either composite or S-video/YC. At that point do we sit there and say man the stuff I just shot looks like crap? No we think it looks very good. Hooking up your YC cable to your TV and seeing high quality video is no different than hooking that same cable up to a high end capture card. The one major downside to editing with analog however is the generation loss of quality. This will only hurt you if you plan on going in and out numerous times. This was a big issue in the days of tape to tape editing but with non linear editing it isn't really an issue anymore.

If you capture with analog compared to firewire your footage will be very slightly softer. You can however gain more color detail in return. You see a normal dv frame of 720 x 480 only has a color resolution of 180 x 480 pixels. The human eye has a hard time seeing this color information loss but it is there. If we can have a camera with true 4:2:2 analog output that will at least give us 360 x 480 pixels of color information which is a lot more detail. 4:4:4 is even better because then the color channels have the same 720 x 480 resolution as the luminence channel.

Finally when you use dv and firewire the compression ratio is 5 to 1. That means 80% of the video information is thrown away. Yeah it might look ok to the eye but 80% is a lot. If we can use our analog outputs we will get slightly softer video but all 100% of the information will still be there. I will try to do some tests with my old XL1 with my friend who has a Video Toaster and let everybody know what I find.

[Graeme Nattress]

Thomas, pro video does use Firewire, look at DVCPro50, and DVCProHD for that - AJ100A deck from Panasonic. Pro video is also all digital - using SDI (4:2:2) or HD SDI (4:2:2) or dual link HD SDI (4:4:4) for signal transfer. Pro video no longer uses component for transfer of signals - that went out with the ark...

Capturing DV over S-Video does not uncompress the picture, does not make it 4:2:2 and does not improve the picture. DV is, as you say, 5:1 compressed, and 4:1:1 but nether of these are a problem, as you still end up with a better picture than the old BetaSP analogue standard - yes - they made DV too good!!

Yes, when you send DV over S-Video you get a softer picture - that's loss of resolution to me! And you don't get 4:2:2! You get 4:1:1 interpolated to 4:2:2 which does not give you any extra resolution, just makes it look nicer. It's really easy in your NLE to do a similar interpolation and change the codec of your DV file to a 4:2:2 or 4:4:4 compressed or uncompressed codec and get the same effect, without making the picture softer. Indeed, I did experiments to compare 4:1:1 with it interpolated to 4:4:4 with it reconstructed to 4:4:4 using an algorithm I invented:

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

and being improved now.

Your comments about the TV and S-video are interesting..... Yes, people generally do hook their TV to their deck via S-Video and a good picture emerges, but if, instead, you had a DV deck with SDI output or a monitor with firewire input (yes, they do exist) and you connected to your monitor digitally, you'd get an even better picture still. Once video is digital, you don't want to, for quality reasons, convert to analogue at any point, and indeed, because we see analogue, and it must be finqally converted to analogue for us to see the video, you should leave that to the last possible moment for the best possible quality.

There is no SD camera for 4:4:4 video, but there are some HD cameras that do 4:4:4, like HDCAM SR and Viper.

[Thomas Smet]

I was not talking about recording to tape and then trying to get uncompressed from that. Of course once it's on tape it is already compressed. I am talking about a raw video signal from the camera live without recording to tape. "Some" cameras actually do give us a 4:2:2 uncompressed video output.

I do not think the fact that one high end deck with firewire means it is a standard. The deck you mentioned is very new and the only one of it's kind. The fact is up until a few months ago firewire was not in the high end market. It is slowly starting to work it's way in but in a different format. Firewire is just a cable connection. It is capable of 400 Mbits/s. DV is only 25 Mbits/s so clearly firewire could be used for more than it is. If you are talking about getting dvcpro50 or Hd from the panasonic deck well then we aren't even talking about the same thing anymore. You can't really compare a 4:1:1 dv signal and a 4:2:2 dvcpro50 signal by saying they are both coming from firewire. That fact is however for 98% of us DV users firewire will only give us 4:1:1 because that is how it is hardwired into our cameras/decks.

Finally all I wanted to know is what cameras actually do give us the uncompressed signal before going to tape. Regardless if it is softer or not that is my problem to deal with. I just want to see if I can get a raw 4:2:2 with no interpolation. Why does it bother you for people to try different ways of getting a different image.

[Thomas Smet]

by the way I am currently working on a capture utility to capture the luminence from the firewire cable and the U and V channels from YC for cameras that do have 4:2:2 YC output. This will give you a live capture utility that will give your Y channel the same quality and sharpness of DV but the higher yet slightly softer resolution of the 4:2:2 analog color. I know it isn't as good as dvcpro50 but it would be better than 4:1:1 DV. Softer doesn't always mean less resoultion. If I use a soft dissusion filter on my camera I am not getting less resolution just a softer picture. In my opinion 360 x (400 for theoretical resolution loss) is better than 180 x 480.

[Graeme Nattress]

Thomas, I don't mean to get argumentative, but as you can realise from my R&D, this is a subject that interests me - getting the most out of affordable video formats!

I guess what I'm getting at is that without looking at a circuit diagram, there's no real way to measure from the outside what the S-Video is outputting - either 4:1:1 or 4:2:2, because even if it is 4:1:1, it gets upsampled to 4:2:2, and how easy is it, at that point, to tell wether it's real 4:2:2 or upsampled to 4:2:2?? Next, is the bandwidth of the S-Video output capable of sending a full 4:2:2 chroma?? And finally, are the losses in the conversions and analogue cable capable of extracting back 4:2:2 from the S-video if indeed it starts life as true 4:2:2?? I'm not saying that things are impossible, just that there are an awful lot of "ifs" with no guarantee of success. Do the digital to analogue converters that take the signal from the camera head have a full 4:2:2 bandwidth even, or are they bandwidth limited. I guess they must be, to some extent. Measuring that with a test pattern and osciliscope will give you an idea of wether the rest is worth pursuing.

If you can find an affordable DV camera with 4:2:2 out of the S-video, live, so to speak, you're right that capturing the luminance over firewire will be better than luminance over S-video. I think such a capture program will make for a very interesting experiment, comparing upsampled 4:1:1 chroma, with reconstructed 4:1:1 to 4:4:4 chroma to captured via S-video 4:2:2 chroma and luminance. I'd be very keen to see the results.

Just as I'm trying to push the bounds of DV using software after it's been captured, it's great to know that others are working on their own methods, using different means!

Graeme

[Barry Green]

Quite a spirited discussion.

First of all, it could be said that you could never get 4:2:2 out of an s-video port, because S-video irretrievably mixes the chroma signals together, so it could be stated that you'd get (at best) 4:2 out of it (if anybody follows what I'm saying).

DV is compressed and recorded as component video, and s-video is a form of composite video. DV records Y:Cr:Cb, but s-video transmits ony Y:C. So you're going to suffer a signal loss just by it being s-video in the first place.

Second, the basic premise Thomas has is that some cameras are outputting the s-video signal prior to compression. The question I have is, is that true?

First, let's point out that there are two types of compression going on: 4:1:1 chroma decimation, and 5:1 DV. The 4:1:1 is not counted in the DV compression (because, technically, if you count the bits up for a raw data stream vs. a DV stream, the overall compression ratio is more like 8:1).

If the signal is decimated to 4:1:1 when the image processing DSP gets ahold of it, then the entire argument is moot, because all video going out that s-video port is unquestionably post-DSP.

Furthermore, it was proven many years ago that the Sony VX1000's s-video port came AFTER the DV compression engine, which in some ways only makes sense: the s-video port has to play back info from tape, so the DV uncompressor has to be doing its job prior to the signal hitting the s-video port. Does it really make sense that there are two different paths to get to the s-video port, one avoiding compression and one after uncompression (for playback)?

For that reason alone, I'd be willing to wager that you're not going to get the higher chroma sampling of 4:2:2 out of any DV camera's s-video port.

But until someone either tests it (someone who has a 4:2:2 analog capture card) or traces some camera schematics, the whole debate cannot be resolved.