So, here's the thing ...

Michael_Bott

Michael_Bott

Well-known member
SO, here's the thing ...

I go to my local pro dealer to buy a copy of Procoder 2.0 and in there showroom there's this plasma screen running some footage shot in Monaco. It's pretty stunning and I recognise from the colour/range and clips/grabs I've seen on the web that it must be HDV. I also assume from the amazing resolution that it must be running straight from an HDV file somewhere. But my eye follows the cable from the back of the screen to ... A DVD PLAYER!

OK - perhaps it's an HD .wmv file running from DVD - or DivX - or something weird. I ask the sales guy. No - that is running from an SD MPEG-2 DVD.

Omigod I say - what was the workflow? FX-1 - Premiere Pro - Procoder - DVD he says.

Now, the evidence of my eyes flies in the face of what I have heard said here and on other forums - that if you are delivering on DVD, there's no point going to HDV because your footage is going to turn to SD mush as soon as you encode it. Based on all this I had recently decided that I was hanging on the my DVX until the technology was there to deliver a significant benefit to the customer from filming at the higher resolutions.

AS I am going to be filming a lot of landscapes in the near future and knowing how DV turns trees and texture to porridge, what I saw in that showroom convinced ME that going down the Z1/FX1 route was a no-brainer - FOR ME.

But do other folk's experiences tally with what I have observed? I just don't get the discrepancy between what I've seen and what has been said about downrezzing from HDV to SD and MPEG-2. Anyone share my feelings or can shed light on the anomaly?
 
My sentiments about downrezing is mixed. Although I continue to think in some way, that if properly done, it will look better then DV. But that's still debatable, at least for me, until I actually experience it.

But I suggest before you go the FX1/Z1 route, that you at least wait until NAB2005. Keep your options open!
 
Even wailting 'til NAB, there's still the problem that I can get a Z1 now for £3500 or a HDX100 or the JVC ProHD later for around £7200. That's not a level playing field.

But back to the question: yes, my thoughts were mixed until I saw HDV fed via SD DVD to that plasma screen. How come folk who must have seen their own results been so equivocal about the quality of HDV downrezzed to SD? It's a mystery to me. :shocked:
 
Partly - possibly.

What was clear was that the detail had been preserved - ie it was AVAILABLE to whatever the final display might be. A world away from my experience of encoding MPEG-2 from SD DV.

I'm not saying that everyone should be buying the FZ/Z1 , not at all. What I saw was SO VIDEO. But, for some projects I have coming up, where I had previously resolved to use the DVX based on what people had said here and elsewhere about the res not surviving from HDV to DVD, I am now convinced that the Sony, at this price, is the machine for the job.
 
Make sure you test before buying, and that you can duplicate what you saw and that the quality holds up on a variety of different tvs. I've also heard that downrezzing HD doesn't translate into higher quality. Perhaps Premiere does a pretty good job of it, or procoder. Who knows..
 
Two things. First, there are some DVD players, like the Avel LinkPlayer, that can play back high-definition content from a standard DVD. If you encode the video using H.264 or something like that, or WMV-10 (don't really know), they can actually read those high-definition streams and play them to an HDTV. So perhaps what you were watching was actually high-def footage.

Second, there are other DVD players that can do an HD up-rez on playback, perhaps that's what you were seeing -- on-the-fly, up-rezzed-to-HD footage. That's my guess as to what you were seeing.

Other than that, no explanation. HD downrezzed to SD becomes SD -- it gets encoded in a 720x480 (or, in your case, 720x576) data stream. You cannot fit two HD pixels into one SD pixel and have it look like it's still two pixels. Square pegs and round holes -- there are only 720x576 SD pixels, and if they encoded to a standard-def DVD, that's what they had to work with.
 
Yeah - that's all possible, but I think the guy was being straight with me about what I was seeing.

Forgive my thick-headedness, but given what you say (and I agree that a pixel is a pixel is a pixel) this is in the same line as that old chestnut: "Why do commercial films on DVD look so good?". I know there's an explanation, but I've never understood it and further, I don't get why the same argument doesn't (or shouldn't) apply to an HD source.
 
I'll take a stab and say that usually a Film is mastered in HD first then the MPEG2 created from that.

So the encoder is working with much much more information than an SD file.. and way more than an HDV file.

Also don't forget that studios and such use Hardware encoding which is a heck of a lot better than software encoding, and they also encode (usually) at very high bite-rates and have access to Dual Layer DVD's.

I really think that HDV would be a much better format if it used a smalelr GOP, and had twice the biterate...

And big films usually have had a lot of post work done to them, as well as they have been professionally shot.. which makes a huge difference.But in the end it is all still the same SD resolution.
 
Commercial DVDs looks so much better because the image they're starting from is a much higher quality image. Note: not necessarily higher resolution, but higher quality. You're talking about professional cinematography, professional art direction, the best film transfer, the best color correction, etc etc etc.

When those films are transferred to DVD, they're scanned in at 720x480. DVD resolution. They're not scanned in higher-res and then scaled down.

Now, if you were to point an SDX900 at that same scene, and make a DVD from it, and point a $300 DV camera at the same scene, and make a DVD of it... which do you think will look better on DVD? The SDX footage, by a mile. Even though the SDX is the exact same resolution (720x480), the better lens, better chips, better processing, better everything in that $25,000 professional camera will make for much better footage than the little camera would deliver. And it has nothing to do with resolution. For that matter, an SDX will make a much better looking DVD than a DVX will, and both are capable of resolving about the same amount of detail. But the SDX costs 8x as much, and makes better-looking pictures.

To make a fair comparison, you have to compare equal elements in the production chain -- comparable quality lenses, comparable quality production values, etc.

The resolution issue is a false argument. HD won't look better on a DVD than a comparable 2/3" SD camera... a CineAlta image shot in 60i isn't going to look better, on DVD, than a Digital Betacam image shot in 60i and transferred to DVD. The CineAlta image will look better than a DVX image, because you're comparing a much superior camera to a lesser camera, regardless of the resolution issue.

But if you compare equivalent lenses/chips/processing, as Jarred and I did, you'll find that a high-def FX1, a standard-def XL2 and a standard-def DVX all produce about the same quality image on DVD, with the XL2 and DVX holding the edge in light sensitivity, latitude, and 24p imaging capability. The lenses, chips, and processing hardware are all roughly comparable between those three cameras. The FX1 image on DVD was comparable in sharpness/detail with the other cameras, but not better. Of course, on an HDTV the FX1 was miles more detailed than the others, but on DVD, no.

I know what I'm saying will sound like blasphemy to HD proponents, but it's a) the truth, and b) verifiable. The resolution issue is a red herring. DVD is a 720x480 pixel grid. Scaling 1440x1080 down to 720x480 is not going to give you more detail, it will inherently do (at the very best) a comparable job, and more likely it'll be a slightly softer image (due to aliasing and pixel-blending). You cannot fit two pixels of HD data into one pixel of DVD data. It cannot be. And, as we tested, it is not.
 
Ok, take a 1 megapixel still camera.

Take a picture of something.

Take a 16 megapixel still camera.

Takea picture of that same thing.

Take the 16 megapixel still and shrink it to 1 megapixel.

Which image looks better?


The second 1 megapixel picture not high resolution, it's high defenition. Same res as the first, but way more information compressed into the same area.

16 blocks per one block. It chooses the nicest of the blocks, and thus the image looks better.
 
Test would only be valid if all other elements in the imaging chain are the same. The way to test that would be to take the same camera and do both shots (so you'd need a camera with variable resolution to pull that off), and shoot the same shots like that, and then shrink the 16-megapixel version down.

Done properly, they should look about the same, with the 1-megapixel version being just a bit sharper because it hasn't had to go through any downrezzing/anti-aliasing/resizing process.

And, again, the comparison can never be truly right, because if all other things were equal, the 1-megapixel camera could use much larger CCD pixels, giving it better light response, sensitivity, and latitude. Using the same camera for both pictures would just put the 1-megapixel image at a slight disadvantage.

Think of it in a much smaller, and more crass version: a 1-pixel image, vs. a 4-pixel image. Downrez the 4-pixel image to a 1-pixel size. Which will be a more accurate representation? A blended/shrunk/anti-aliased version, or the original raw shot? And, even if the blended version ends up with a slightly more accurate color, would it be in any way "sharper"?
 
Michael_Bott .....

You just opened Pandora's Box for everybody to explain to you the scientific reason why what you saw wasn't what you really saw. *smile*

However Michael....I feel you and KNOW EXACLTY what you saw. I see it everyime I shoot with my Sony Z1. I have gave up trying to expalin it to people on the internet. Now, it's Show and Tell. So far.....everybody who witnesses the amazing HD picture that was DOWNCONVERTED to DV, still to this day....leaves with their jaw dropped. (like you did at the store!) *smile*

If it walks like a duck....talks like a ducks.......some people will swear it's a chicken.

- Shannon W. Rawls
 
Personally, I'd love to see some screen grabs side by side of the DVX and downconverted Z1. I don't have access to the Sony though :(. Would anyone be interested in trying this out for me (and others I am sure)?
 
I'd love to see a single frame of downconverted HD (from any source, not just the Z1) that shows any superiority (in any way) over a comparable SD camera shooting the same footage (with comparable settings, of course, or with fully qualified shooters getting the best from each camera). That means VariCam vs. SDX900, or Z1 against DVX, etc.

People keep saying it like it's the gospel or something, but I believe it is an impossibility, and I don't think anybody's ever demonstrated it. Because it cannot be.

I believe in it enough that I'm willing to put $500 on the line, as detailed in the other thread -- I'll put a DVX up against an FX1/Z1, shooting the same image, and making a DVD of the footage. Whatever tricks are used in shooting or in post for one camera would be equally available to the other, but obviously I'd like to avoid post manipulation as much as possible -- this is about testing cameras, not post techniques. Final footage on DVD, split-screened side-by-side. If a group of viewers in a blind test (not knowing which clip is which) picks the Z1/FX1 footage over the DVX's as being "sharper" or "higher resolution," I'll fork over $500 to the FX1/Z1 shooter right then and there. And I'll make it easier -- the FX1/Z1 can shoot in HDV mode, of course, and the DVX will be in Squeeze mode/24p.

I figure if I'm wrong, I'd rather learn about it sooner, so I can quit being wrong.

But I'm not wrong. I'm sure of it.
 
I agree Barry. It's a claim that never made any sense to me. Hence the reason I want to see some tests! Just to make sure I'm not missing something.
 
I don't want the $500 cause this is all too easy. I work with a lot of HiDef stuff, 3000x2000 4:4:4. To put that onto DVD I can downsample to 720x576 and encode from that or I can directly encode from the HiRes images. The results are a dramatic improvement if I go direct from the HiDef images yet following the logic it should be the same.
It isn't the same because I've bypassed the DV25 compression and I'm encoding to a different form of 4:2:0 than that used in PAL DV.
The same thing happens with Z1 footage. If you shoot HDV and let the camera / deck downconvert to DV25, capture that and encode for DVD it looks pretty good.
If you capture as HDV and encode directly to DVD you're way ahead. You've bypassed not only the DV25 compression but also the 4:2:0 sampling. Decent downsampling algorithms can resample the HDV footage to give a better color sampled image (at least 4:2:2) to pass to the encoder, result is a better looking image.
Barry is right, commercial DVDs are made from DigiBetacam transfers from the 35mm. What he didn't mention is they're sampled at 4:2:2 SD res, something a DVX100 isn't capable of but something a Z1 is capable of if you have the right software. Actually I suspect some are done direct from film scans to avoid the stretched pixels in SD. Again a small but significant improvement in the final output results.
 
Hi Shannon - I thought I'd hear from you :thumbsup: . I didn't mean to open Pandora's box, though I guess it was kind of inevetable. But I hope the issue may get resolved here none way or another (though I don't really wan't to see Barry light by $500:happy: ).


I was moved to submit this thread not by a a close judgement call - this wasn't 'well, that looks slightly better', this was WOW - HOLY ****, what's going on here ...?

There's no way I'm doing a 'this camera is better than that one' routine. I love my DVX and I love the image. Despite what I've seen nothing beats the DVX in it's own terms.

Needed to say that.

However, for SOME projects I have coming up, if I can get that quality to my customers on DVD - then it's a no-brainer.

I must say that I had resolved to stick with the DVX not just based on what others were saying, but on my own experiments. I have downloaded the raw FX1 footage from VASST, put it on a Vegas timeline and rendered to MPEG-2 via Manconcept and I was not impressed. How did you encode to get your results in the FX1/DVX/XL2 test Barry? Perhaps the current incarnation of Mainconcept isn't up to the job? I have put the same clips through Procoder and although I don't have a plasma screen, the results I can see on my monitors suggest that the results I saw at CVP were not a fluke.

Maybe this debate is not important to everybody, but if you've got real DVD clients/customers for whom a significant WOW factor could make a difference in your pocket, then we must be able to make informed, powerful choices.

Respect to you all ...

(all the above written before Robroysyd's post)
 
Yeah, this is Pandora's box, alright! There's no easy way to demonstrate the difference unless you see it with your own eyes. Screen grabs are fine, but they're not even close to the same experience -- it's all about watching the actual downconverted footage playing/moving on an actual DVD player on a regular TV.

The FX1 footage downconverted for DVD just looks better. Forget the formulas and just find a way to see some of it. It's not 100% totally kickbutt better, but it looks noticeably better than SD-originated footage to the human eye, IMHO. Eventually, I bet all the HD-DVX buyers will start saying the same thing!

Maybe I should offer to send out a DVD of some of my stuff so people can see what I'm talking about on their own TVs... the web versions at http://www.hyperbolicfilms.com are optimized for the web, naturally.
 
You must log in or register to reply here.
Back
Top