Thought you might like to see how much of a difference it makes to shoot in DVCPRO50 mode vs. standard DV mode, on the same HVX.

We were shooting some greenscreen tests, and we shot the exact same subject in the exact same lighting with the exact same camera on the exact same settings; the only thing that changed was I changed the recording format from DV to DVCPRO50 and then had the actress do the same actions.

Here's an uncompressed frame extraction from the familiar DV that we all know (4:1:1 color sampling):
http://www.fiftv.com/HVX200/DV25-Green.png


Here's 4:2:2 DVCPRO50:
http://www.fiftv.com/HVX200/DV50-Green.png


And if you want to see the images blown up, here's an extraction of her right hand at 400% size:
http://www.fiftv.com/HVX200/DV25-vs-DV50.png


Remember, absolutely nothing else changed. These shots were taken within seconds of each other. This is not trying to compare one camera against another; these shots were taken on the same HVX200. The only change whatsoever was to change the recording format from DV to DVCPRO50.

Thanks for your posting, but a still life would be more indicative -- I would guess that the right hand is 'flapping' in the 4:1:1 shot and so an awful lot of what we are seeing is motion artifacting/interlace converted to stills problems -- or at least so I would have to assume.

Maybe a nice bowl of fruit with a shiny can of Coca Cola and a spray of delicate flowers against a textured background would help ....

Cheers,
GB

" but a still life would be more indicative "

with my experience with DVCpro50, i beg to differ...motion artifacts in DV are pretty appalling, but those same shots with DVCPro50 can look stunning...
even this grab as an example.. look at the jaggies on DV... and colour... well it doesnt need commentary...

A still life, shot for 15seconds would be a better indicator of compression motion artefacts, not just a still shot taken from DV...

Ive always loved DVCpro50... i always thought it was the bees knees but price was abhorent...
Im glad to see the format making headway into "this" market... at least it will be appreciated by those that will be using it the most.. and those that have never used it, will come to appreciate the difference...

from the looks of it, not many ppl have used DV50 which prolly explains the lack of interest so far as everyone is focused on DVCProHD...

But objects in motion are exactly where keying problems occur. That, and those high-contrast diagonal lines that are the cause of obvious blocking in DV. What an amazing difference!

Having never worked in DVCPro50, I am wondering if the apparent increase in latittude is simply due to the increased color sampling. I noticed in the stills of the lizard sculpture, there was also more saturation and more contrast. It's as if the highlights remain about the same, but the camera digs deeper into the blacks. I would have expected more saturation, but not that much more contrast.

I've never worked with DVCPro50, only DigiBeta and certainly in post for things like keys 4:2:2 sure is much better. However I wouldn't be too hard on DV, the DV conversions I've made from the same DB tapes look pretty fine. In fact by the time it's encoded for DVD it's pretty damn hard to tell the DB sourced material from the DV conversion of the same material.
However (big however) I'm working in PAL where the 4:2:0 DV sampling is very close to the 4:2:0 sampling used on all DVDs, I'd suspect working in NTSC there maybe a more dramatic difference.

The "bees knees"? Suddenly I feel the need to pollenate something...

These results aren't surprising and solidify what I've always said about digital imaging: Color is more important than resolution.

In fact, I'd rather have an HVX shooting DVCPRO-50 than an H1, Z1 or HD100 shooting HDV; you just can't beat Panasonic out-of-the-camera color.

2 I would guess that the right hand is 'flapping' in the 4:1:1 shot and so an awful lot of what we are seeing is motion artifacting/interlace converted to stills problems
Nope. This is from a 24P sequence and there is no interlacing of any type, and I chose frames that were roughly comparable where the flapping had stopped.

I'll do a still life shot later with some color items.

Thanks for the clarification, Barry. I'd have said that the vertical resolution appeared halfed -- exactly what happens when a codec like DV25 does when faced with significant changes between fields. But you would know better than I how these frames are shot, how the stills were grabbed, and why the vertical resolution fails so miserably in the example posted.

Perhaps a better way for me to state my wish is that both frames _were_ the same -- although you indicate they were taken only a short time apart, the one frame 'captures' motion & the other doesn't ... and the problem revealed is in precisely the area where motion was present. It would be educational to see how the DV50 codec stood up to the same motion, or conversely to see how the frames compared if no motion was present in either.

I am happy to stipulate that DV50 is better than DV25 -- just a pedant about what the evidence presented reveals & doesn't.

Cheers,
Geoff Baker

I'm a skeptic, but the photos don't lie.

How much do you attribute to 4:1:1 versus 4:2:2?

The photos aren't lying, but neither are they photos of the same event ...

GB

thanks for the post barry!

The photos aren't lying, but neither are they photos of the same event ...

GB
They couldn't be photos of the same event because they are the same camera.

There's no interlacing going on, these are progressive captures.

I'll do some still-life stuff later, non-green-screen stuff.

Some years ago DV magazine did a stress test of popular digital codecs compared to Betacam -- they used a repeated motion event to create a 'same event' comparison -- a merry go round or some such. A table top version of the same would make a perfect 'still life' sized equivalent, if you are so inclined.

I remain convinced I'm seeing a halfed vertical resolution in the motion area of the DV25 image ... how exactly is the 'progressive' nature of that recording created?

GB

Scanned off the CCD progressively, recorded as 24P carried within a 60i stream. Imported to Vegas where the pulldown is removed, resulting in restoring the original 24P frames. Frame was copied to the clipboard and then pasted into PhotoShop.

Scanned off the CCD progressively, recorded as 24P carried within a 60i stream. Imported to Vegas where the pulldown is removed, resulting in restoring the original 24P frames. Frame was copied to the clipboard and then pasted into PhotoShop.

Barry,
Can you give me a quick rundown on how you use DV50 in vegas and what codecs or add-ons you used? Also.....can you preview out via firewire to an NTSC monitor? If so is it downsampled to DV25?

Thanks

Imported to Vegas.
Vegas??? Barry... you still hangin in there with the PC Federation? *smile* I thought you went over to the Dark Side??
How did you capture and prepare the DVCPRO-50 stuff for Vegas?

Also....I'm gonna call you about that Vegas gig.

- ShannonRawls.com

Barry,
Can you give me a quick rundown on how you use DV50 in vegas and what codecs or add-ons you used?
Files were converted using DVFilm RayLight software, and the free Matrox DVCPRO50 codec. Marcus van Bavel of DVFilm is a ninja, and he supplied a beta-test upgrade of Raylight that includes DVCPRO50 and DV MXF file conversion.

So the process is: buy & install raylight. Download the Matrox codec. Copy the files to the hard disk through the USB2 port on the camera. Convert the files using RayMaker; then edit as normal.

RayMaker works pretty well for DV50, I can almost (ALMOST) get a stream of realtime playback on my aging P4 2.66GHz system. It's about 28fps playback rate for a single stream.

Also.....can you preview out via firewire to an NTSC monitor? If so is it downsampled to DV25?
Vegas only does DV25 firewire preview. You can put high-def on the timeline and it'll downconvert to DV25 for a firewire preview.

Vegas??? Barry... you still hangin in there with the PC Federation? *smile* I thought you went over to the Dark Side??
Well, no, didn't "go over" per se; I am currently running both platforms. I tried Raylight/Vegas on my existing system, and... man, there's just no comparison.

I also finally set up the Mac G5, and it plays back multiple streams of high-def at full frame rate with realtime transitions inbetween! It's just like editing DV, but the frame size and data rate are 4x as high. I don't have a RAID either, I'm just running off the internal drive. Don't know what those Apple guys have done, but man, this is the bottom-of-the-line G5 and it's way ahead of the PC as far as DVCPRO-HD playback/editing goes. We ran some HD on an FX guy's dual xeon 3.06ghz system, and it was pokey slow by comparison.

I'm looking at Edius Broadcast too for the PC, but they're recommended minimum editing stations are a bit pricey for my blood (about 3x as expensive as the Mac was, recommending a dual Xeon 3.6 or better with RAID storage), so I'm going to work with the Mac for a while and wait and see what happens with Vegas. I keep using Vegas because it's so unbelievably convenient and it's just so awesome -- the learning curve on FCP has not proved to be quite as simple as I'd hoped. It looks more like Premiere than I would like; once I fled Premiere for Vegas I have never wanted to see Premiere again. But it's hard to argue the performance; the Apple just has it all over the PC apps when it comes to DVCPRO-HD editing.

How did you capture and prepare the DVCPRO-50 stuff for Vegas?
You can't capture, at least not until Vegas (someday, hopefully) implements a capture app. You have to copy the MXF files manually from the card, and then run RayMaker on them. It's not any big deal, it's just another step. Direct MXF import would be magnificent though; this "import P2/unwrap/rewrap" process on the Mac is just a big waste of time to get a more limited system. Hopefully they'll announce a new version that just lets you edit native; it's the only flaw in their system (well, that plus the learning curve for us current Vegas exiles!) Right now Vegas feels like "home", FCP feels like a hotel room. I'll give it a shot though.

I'd be very interested to see the same tests done with a 50Hz camera, as DV25 is then 4:2:0. I don't deny the superiority of the 50Mb end product over 25Mb, but wonder how much it is due to the far lower general compression of that codec, as opposed to the differing colour space.

My reasoning goes that for DV25 there are two colour samples (one each U and V) for every four luminance, in DV50 there are four - so a 33% increase in no of samples, whilst the bitrate increases by 100%. Hence 2/3 of the bitrate increase is going towards lowering the general compression level rather than a colourspace ratio improvement.

I find another part of the images very interesting - namely around the girls left elbow. Enlarge the image until you can make pixels out and the compression artifacting (on DV25) is very obviously asymmetrical - far, far worse on vertical edges than horizontal. Which leads me to think that maybe it's the one sample in four going into the compressor that is causing the worst problems, and makes me wonder how 4:2:0 as in PAL DV25 would compare? My suspicion (based on the problem being so assymetrical in NTSC 4:1:1) is that the difference would be less, but I'd find it a very interesting experiment.

Barry...

Was the DVCPro 50 format written to tape or the P2 card? If to tape, that would be a pretty useful feature.

Thanks,

You can not record DVCpro 50 to tape with the HVX unfortunately. Only DV.

New photos, total still-life with no motion whatsoever, exact same subject in both. Again, this is the same camera for all shots, same settings etc., the only thing whatsoever that changed was the recording format.

DV:
http://www.fiftv.com/HVX200/Turkey-DV25.JPG

DV50:
http://www.fiftv.com/HVX200/Turkey-DV50.JPG

Blown up to 400% to compare:
http://www.fiftv.com/HVX200/Turkey-DV-DV50.JPG

Here's a shot where the DV50 was blown up to 200% size and put side-by-side with an HD shot:
http://www.fiftv.com/HVX200/Lemonade-DV50-DVHD.JPG

Barry, they are very good illustrations, and if anyone is in any doubt of the extent of the superior quality of HD just look at the last one.

It's left me thinking that an interesting comparison along the same lines would be to use a Z1 to shoot the same scene as Barry has done, but both in NTSC and PAL SD DV25, all else kept the same. Obviously the vertical resolution will differ, but the big difference between the two examples will be 4:1:1 for NTSC, 4:2:0 for PAL. The compression applied should be the same in each case as both have the same number of samples/second (720*480*30 = 720*576*25 = 10368000 for luminance), so any significant differences in artifacting should be due to colour space, and how the codec handles them.

Any takers?

I'm relatively new to the world of video and just bought an HVX as my first "Real" camera. After shooting with DVCpro50 and P2 for the first time I cannot imagine ever going back to dv25 and tape. The difference is night and day. Bigger even than the difference between Vegas and FCP in my opinion.

......Of course I have no real world experience to base that last comment on, but it sure did sound good. Here's to Panasonic!!!

well, the DV25 looks really horrible in the chroma channel, but people tend to forget what can be done with a bit of preprocessing of the footage (it's not really much exta work).

here's a comparison between raw DV25, smoothed DV25 and the DV50 after keying:

http://voon.dyndns.org/hvx/DV25vsDV50.jpeg

(or lossless compressed for those who want to take a closer look)
http://voon.dyndns.org/hvx/DV25vsDV50.png

++ chris

Chris, what kind of preprocessing did you use on that DV25 smooth? That looks as good as raw dv50 to me.. Will Pal DV25 achieve a better result then NTSC DV25? I'm considering a pal DVX100b, or HVX for DV50 since I'm sticking with SD for now..

it's really just some chroma reconstruction/smoothing before the key and a lightwrap after the key...

i updated the sample pic to do the same thing with the DV50 footage.
++ chris

Just another chroma key test.

It was done with primate 2 (after effects); I tried to keep the same settings for both images.

http://www.ahproductions.com/public/video/DV50-Green comp.tif
http://www.ahproductions.com/public/video/DV25-Green comp.tif

On this one I have applied deartifacting filter.
http://www.ahproductions.com/public/video/DV25-Green comp deartifacting.tif