HDV=quality loss

I shoot my movies with a Canon xha1s Camera (records on dv-tapes with the 1080 25p HDV standard). When I connect the camera with a component cable to the monitor I see a perfect quality hd-movie.

To capture the movie I am using the common FireWire connection. I choose the HDV capture settings in the project to capture the movie to FCP. When I compare the quality afterwards,in FCP there is a big difference. I am really thinking that I am losing a lot of quality during the capture process.Is it the square vs rectangular thing? Does it mean that any footage shot in HDV always will look a bit noisy in the end? I am using this camera for 2 years now but never got it the way I want it to look, ie the same quality as on the tv using component out.

No, you're not losing any quality at all, it's a lossless digital transfer if you're bringing in HDV

My guess is that you're looking at the footage in the FCP Viewer / Canvas instead of on a proper external monitor via a capture card. Those are proxy displays and not full quality.

Also, understand that HDV is 25:1 compression with 4:2:0 color space so it's already a very noisy and lousy signal to begin with compared to other HD formats. So you're starting at a handicap compared to DVCPro HD even.

Walter Biscardi, Jr.
Biscardi Creative Media
biscardicreative.com

Thanks Walter,
I am working on a matrox box and a HDMI monitor but guess I need to get another camera so. Where will it ever stop?

Quote
walterbiscardi
So you're starting at a handicap compared to DVCPro HD even.

HDV is puny 25 Mb/s; DVCPro HD is generous 100 Mb/s. But this compares apples vs. oranges since the HDV is interframe compressed while the DVCPro HD compression is purely intraframe. It's not unusual to gain a factor of 4 from interframe compression, so the HDV I-frames probably have as many bits as the DVCPro HD frames. Then since HDV is 4:2:0 vs. DVCPro HD's 4:2:2, the HDV I-frames actually have greater visual sharpness than the DVCPro HD frames. (When capturing with limited bitrate, there's the high cost for choosing 4:2:2 over 4:2:0, 33% more bits, to be considered.)

I'm not recommending interframe compression for video capture, but for many images the quality is surprisingly high quality for the bitrate.

Dennis Couzin
Berlin, Germany

Very important what Walter said on the monitoring. If you cannot monitor off a proper signal, you cannot gauge quality. The capture itself is lossless, because the signal does not lose a generation when captured natively through FireWire.

I dunno about the interframe vs intraframe debate. You can float a more coherent signal on lower bit rates if the signal is temporally compressed, but I always felt the image captured on XDCAM EX (full raster HD Mpeg 2 long GOP @ 8 bit 35 Mbps) or HDV (thin raster HD long GOP Mpeg 2@ 25Mbps) degrades more quickly than Dvcpro HD (thin raster I-frame DCT based compression@ 100 Mbps). But that said, HDV is a bit of an engineering miracle because it is able to carry an HD signal with Mpeg2 compression at such a low bit rate. However, HDV is usually artificially sharpened in camera so it looks nice and sharp, but not necessarily better for post production, but these are other factors...

The EBU once mentioned that mpeg 2 was about twice as good than I frame only compression, but yea, it's an apples vs oranges debate, because aggressive compression of both produces different types of artifacts. With intra frame compression (DCT variants) you get macro blocking and mosquito noise, with long GOP, the still images look presentable, but the image may disintegrate on moving objects especially when there are fast moving objects with a lot of motion blur. But mpeg formats usually employ both inter and intra frame compression. So meh, do your tests...



www.strypesinpost.com

How would XDCAM compare against these two for datarate?


dcouzin Wrote:
-------------------------------------------------------
> So you're starting at a handicap compared to
> DVCPro HD even.
>
> HDV is puny 25 Mb/s; DVCPro HD is generous 100
> Mb/s. But this compares apples vs. oranges since
> the HDV is interframe compressed while the DVCPro
> HD compression is purely intraframe. It's not
> unusual to gain a factor of 4 from interframe
> compression, so the HDV I-frames probably have as
> many bits as the DVCPro HD frames. Then since HDV
> is 4:2:0 vs. DVCPro HD's 4:2:2, the HDV I-frames
> actually have greater visual sharpness than the
> DVCPro HD frames. (When capturing with limited
> bitrate, there's the high cost for choosing 4:2:2
> over 4:2:0, 33% more bits, to be considered.)
>
> I'm not recommending interframe compression for
> video capture, but for many images the quality is
> surprisingly high quality for the bitrate.

Quote
mark@avolution
How would XDCAM compare against these two for datarate?

XDCAM, like HDV, uses MPEG-2 interframe compression. (Interframe compression generally implies interframe and intraframe.) XDCAM comes in many flavors, the strongest using 50 Mb/s. So XDCAM must yield better image quality than HDV which uses 25 Mb/s. Of XDCAM's 50 Mb/s, figure 12.5 Mb/s is providing the 1920 horizontal pixels instead of the skimpo 1440 of HDV (and of DVCPro HD too), and figure 12.5 Mb/s is providing 4:2:2 chroma subsampling instead of 4:2:0.

Sony threw down the gauntlet with XDCAM: video capture shall be interframe compressed! Video delivery is invariably interframe compressed: HDTV, DVD/BluRay. There are cases where it is nevertheless advantageous to capture without interframe compression -- when there will be certain effects -- but Sony's move makes sense for most video.

There's a lot of subtlety in the visual world that doesn't make it through interframe compression. Strypes gives a good example with blurs. Video is at risk of looking cartoonish versus "digital cinema". Note that the DCI standard uses purely intraframe compression.

Dennis Couzin
Berlin, Germany

Hi there, all you HDV haters.

I'd like to mention that, despite the common loathing for HDV, the Canon XLH1 series produces an excellent image, largely due to the Canon glass - or so I imagine.

I've shot boatloads of stuff on an XLH1 at 24f 1080 and I have never - that's never - detected any of the artifacts (that I have seen illustrated by other HDV cameras) messing up rapid movement and so forth.

3 features shot on a Canon of similar vintage to the OP's camera have all got through QC and have played on cable all over the world. Sure - I see macro blocking in very low light situations, but I did want to mention the upside to that particular series of cameras.

But Canon service sucks.

Best,

Harry.

Hi Harry,

Not bashing Canon cameras (the firewire and tape interface protocols are a little crummy though). The lenses on the other hand, are great. They work like magic, but they are not magic. The Canon 5Ds also look great (because they can use prime photography lenses), and has to a large degree, created a niche market in the video world, but it shoots to 8 bit long GOP 4:2:0 in H.264. The Deadliest Catch was shot on Sony HDV cameras and they look pretty good on TV too, but no doubt the camera guys are really good.

There is some limitations inherent in the codec, and because of those limitations, it makes it very hard to do heavy graphics work with HDV or XDCAM formats. I'm not sure about macro blocking in low light situations, but you will see a boatload of noise. Artifacts on movement may be hard to spot in most cases (some are pretty evident), but for some situations (eg. rolling shutter), you will have caffeine loaded roto people distorting masks on fast moving objects to try to fix those problems.



www.strypesinpost.com

> The Canon 5Ds also look great (because they can use prime photography lenses), and
> has to a large degree, created a niche market in the video world, but it shoots to 8 bit long
> GOP 4:2:0 in H.264.

The Canon DSLRs were never good because of their compression. I just shot my first non-documentary-type project on my Canon 60D and I was also stunned at how poorly the shots responded to speed changes (after transcoding to ProRes, of course), even nice even numbers like 200 per cent and 50 per cent. I also had to compensate for moire issues, even though I was shooting on 24p. I would happily blame this on my handling on the camera, except I also saw all these issues on a large-budget commercial I'd edited which was shot on the 5DmkII. Motion (both camera and subject) just isn't the strength of these cameras.

The DSLRs are good because of their large sensors, rich colours, excellent still-photography lenses, ability to use digital ISO. And I'm quite simply addicted to the manual White Balance controls, allowing me to adjust the hues of the shot right there at the shooting stage.

I suppose prosumer video camcorders never went for the swappable lenses because they usually put in a mechanical zoom. I wonder how they would fare if they could accept still-photography lenses like the DSLRs, because that piece of glass really makes an enormous difference.


www.derekmok.com

I could be wrong about this as I am not a cinematographer or a lens manufacturer. My guess is that traditional video cameras can't use many prime lenses because they use 3 CCDs (and the light has to pass through a prism splitter to obtain pure RGB values). The Canon 5Ds use a single CMOs sensor, so the light focuses on the same plane. And this is why traditional video cameras need a 35mm adapter so they can use 35mm lenses.



www.strypesinpost.com

I agree absolutely.

Harry.

strypes, There is nothing to prevent a manufacturer of 3-sensor cameras from also making a set of interchangeable lenses which have the compensatory aberrations for the prism system. It's no harder to design or manufacture such lenses than lenses for imaging onto a single sensor. Lenses don't have to be made in huge quantities to be affordably priced.
Similarly in the late 1950s a Bolex 16mm cine camera used a behind-the-lens prism requiring a set of special lenses to be made for just that camera. At least three competing lens manufacturers made them, and except for ensuing confusion over when one could get by with standard lenses the problem was solved.

Dennis Couzin
Berlin, Germany

Well, the system prior to the RED, then the DSLR craze, was that everyone was shooting with DOF adapters. One of the biggest problems was that you lost a bit of light doing that. Not too cool, especially when most of the folks were shooting with the HVX200, which used 3x 1/3" CCDs, which meant that the shots went soft pretty quickly.



www.strypesinpost.com

Indeed, strypes,

I have seen many photos of gear-heads proudly brandishing 1/3" cameras with these massive railed "rigs" groaning under the stress of supporting those DOF adapters, but I have never heard of anyone ever shooting anything of substance with them.

And they aren't cheap.

Harry.

A 35 mm adapter is optically funky. It consists of a lens mount with a diffuser in the mount's image plane plus a "field lens" (to reduce the light loss from diffusion) plus a "closeup lens" to allow the native camera lens to re-image that first image plane onto the camera sensor(s). The adapter gives what it promises: the DOF expected for the lens mounted onto the adapter and the view angle of that lens. Generally it cuts image quality. Image degradation due to the first lens, the diffuser, and the combination of the field lens, closeup lens, and native camera lens adds. (However, the native camera lens is immune to out-of-field flare.)

It's sad that some videographers can't part with the shallow depth of field of 35 mm cinema. It's just a trick employed in certain kinds of "story telling", an affectation really. Video could be a new medium if it dared.

Dennis Couzin
Berlin, Germany

I don't know. I both agree and disagree that video should be a new medium, and not cling onto the aesthetics of a different medium. The experience of the video and even web video is very different from that of the cinema. So it really shouldn't be seen as a form of cheap film, but unfortunately, web video has become the new catch phrase for "I have no budget". Which was what many people saw video as- a cheaper medium than film. And I cringe when I see video cameramen shoot video in a way that they wouldn't dare if they were rolling at $100 bucks a minute, and the millions of amateurs who shoot a video drama like they were doing electronic news gathering.

On the other hand, the shallower depth of field helps the mind focus on the essential bits required to tell the story. I was watching a video where stu Maschwitz was talking about 24p. I don't necessarily agree with him on frame rate, but I agree that story wise, you tend to take out what is not required to tell the story. I often say that what does not help the story, takes away from the story.



www.strypesinpost.com