So, here's the thing ...

John_Hudson said:
Then put up.

All this talk and no footage; show us something worth looking at. Even if it is in that nasty 60i.
Click to expand...
Wow! A challenge. Two things:
1) I'm only a few months into "filmmaking" as opposed to camera operation, so I'm sure my work is far from the level of competence of the bext DVX footage.
2) As long as that caveat is understood...fine, okay! I'll do it.

But I don't have the bandwidth to support the hundreds and hundreds of downloads I'd get (I already go through 500 megs a day). So either you can PM me with an address for an actual DVD for your viewing pleasure, or I'll send you an mpeg file and you can host it here for folks to download and burn to their own DVDs.

I'll be using "glue," the (very) short film from my site (you have seen it, actually, so it's not like i haven't posted *anything*).

EDIT: I should point out some things, just to make sure we all understand what I have here. Glue uses almost all natural light, which may or may not be interesting. Most importantly, though, several of the shots were zoomed in during post (thanks to the extra HD resolution), so while those particular shots won't blow you away compared to DVX footage, you *will* get to see what an effective extra tool the post-zoom can be...
 
Okay, I think I have been misunderstood again, and I really don't appreciate any insinuation that I'm misleading anyone.

I didn't ask for a world-class exhibition of FX1/Z1 footage. In fact, that would be counterproductive -- an image standing on its own does nothing to help us compare. All I asked for was any sort of evidence that a downrezzed HD image looks better, on DVD, than the same shot from a comparable SD camera. It can be good lighting, bad lighting, interesting subject, lame subject, who cares -- just so long as it's the same shot. Then you can compare. And it doesn't have to be a full DVD, it can be a single frame, it can be a few seconds.

Some people keep saying that HDV on DVD is, to borrow some words, "mind-blowingly superior", "night and day", etc. etc. etc. And I tested it, and the fact of the matter is that it's not. It's about the same.

So I'm asking for those that profess that it is, to demonstrate. Heck, Jarred and I went and shot a dozen locations with all three cameras, side-by-side, and DVD'd the footage (because DVD is the dominant delivery format). The results are that downrezzed HDV is about comparable. So where is this notion that it's so much better coming from?

Hyperbolic, I do so wish that you could give me the benefit of the doubt, and assume that I'm not trying to mislead anyone, or "gloss over" anything. Re-read the posts and you'll see that that's exactly what I've been saying. I do not believe that you can cram six pixels into three pixels and have it look "mind-blowingly better" (or maybe even as good as) a comparable system imaging those three pixels natively, at full res. That is the premise, that is the thesis that Jarred and I explored. And we came away with evidence, shown to everyone, that that was true. I'm not just spouting theory here, I went and tried it first.

Now, for those who keep saying HDV looks better, and that they've "seen it" -- first of all, on an HDTV, at full HD resolution, yes of course the HDV looks phenomenally better. Arguing otherwise would be lunacy. It looks like VIDEO, yes, but it's definitely in a league all its own.

But that's not what I'm talking about, and never has been. I'm talking about using the camera to deliver product, and DVD is, and will be, the delivery medium for the next several years. Has anyone else even tried to shoot the same scenes and compare them? And if they tried, did they try to "do it right", or just shoot on auto right out of the box? Jarred and I couldn't take the time to "do it right", there were too many variables, so we set the cameras on neutral settings, we had our hands full just getting the focus, exposure, and framing right. Out of the box the FX1 and Z1 both have the detail level cranked up and have stronger contrast (especially when compared against CineGamma on a DVX). Those two items can lead someone to say "wow, this looks better" -- but they have *nothing* to do with resolution! And one could always crank up the detail and the contrast on the DVX to match -- which is probably how Shannon and I would proceed.

So it comes back to resolution, which is the crux, the focus, the point of this experiment. Downrezzing a high-def image, to see if it results in a comparable SD image. Heck, you can do a version of that yourself with your FX1, all by yourself (even though it won't be truly valid*, it's at least a start). Shoot a scene in HDV, then shoot the same scene in DV. Then burn a DVD with both of 'em. Do you get a dramatically superior image from the HDV version? And if so, would you care to show us a still frame that demonstrates all this magical, preserved resolution which seems to defy the math? Not being a smart*** here, I am really curious, as I have run the tests and can't find any way to verify what people keep saying. And, with the HDX or JVC, I'll be in the same HD boat soon, and it would be awfully nice (or at least comforting to the wallet) to be able to demonstrate that it really, truly is better for producing the final product.

Anyone with an FX1 or Z1 want to run that test, and just show us some 720x480 stills extracted from an MPEG-2-compressed DVD-compatible file? No fancy trickery and "unsharp" masks and whatever during the downrez, unless you also apply similar tricks to the standard-def footage... after all, what should be tested is the camera's performance and the conversion, not post-production techniques.

And Hyperbolic, please try to extend the benefit of the doubt -- if you think I'm trying to mislead someone or am pulling some shenanigans, I assure you that your understanding is wrong, and I'd urge you to re-read whatever it was I wrote and give it the benefit of the doubt.

* reason I say it wouldn't be quite as valid is because the FX1/Z1 aren't in quite the same league as the class-leading XL2 and DVX as far as when shooting regular DV. The DV image from them is fairly comparable, so the test will at least get us in the ballpark, but it'd still be interesting to see all three cameras pushed to their best and then compared.
 
...what i see here is a bunch of FX1 advocaters saying that the FX1 is worth saying "wow" about when its downconverted to SD.. and that its also worth saying "WOW" about when its downconverted to SD and played back on an HD monitor..

..OK. well .. nobodies saying your wrong about that. im sure it IS worth saying "wow" about. thats not the point. the point in that the DVX straight SD to SD on playback is ALSO worth saying "wow" about. and when played back on the same HD monitor would ALSO be equally if not more to say "wow" about. and i havent heard a single shred of real factual information/evidence/proof to support the theory/hope that the downconverted FX1 footage actually looks any better than the DVX. all i continually hear is people saying that the downconverted Fx1 footage looks amazing. ..ok again thats not whats in question.. whats in question is whether or not it looks MORE amazing than the DVX given the same display output.

ive watched enough SD DVDs on high def displays to know that most ANY SD footage looks better on a high def screen and is worth saying "wow" about. the problem here is that the people who've seen the downconverted FX1 footage and said "wow" didnt have some "dvx footage right there to play as well and compare and see that its just as wow if not more. barry has offered to make it happen. and im already hearing complaints about his offer to show people the reality of the situation. absolutely rediculous.
 
I'm used to seing my DVX footage at as good a resolution as it can get and my brain still went 'WOW' when I saw the FX1 footage. As I've been very careful to point out, I was looking from a very particular viewpoint - ie. I have projects coming up that will rely on landscape footage looking good. Now, withn the best will in the world, no matter what it's other qualities are (and they are astounding) the DVX, along with other SD cameras turns landscapes to a jumble. It's got to have something in the foreground to get it's mojo working. If my next shoot had been a drama shoot I wouldn't have given that screen another glance - I would have had the smug knowledge that my DVX is the complete bees knees and FX/Z1 owners can go and ....

BUT. I saw what I saw (didn't 'hear' about it) and not only is that crucial landscape detail getting recorded - lord save me - it's getting preserved through to DVD. And that means, for me, having to think again in order to fulfil what I have in mind for my upcoming projects.

From a DVX advocator .....
 
He raises an excellent point. You can watch HD shows -- particularly Jay Leno -- on SD TVs today -- tonight -- perhaps right now. I, too, watch "The Tonight Show" in both HD and SD, and I, too, agree that on SD, it doesn't look any better than any other show shot on SD cams.

"American Idol" is shot on the Varicam, according to Jan. Looks splendiferous in HD. Looks normal in SD.

These shows are shot by pros with tons of money and experience, so it's a fair test of the HD cams, I'd have to think.
 
Whoa, guys i'm really confused, if i agree with Barry (in theory i do) then films shot on 35mm and downconverteed to DVD should not hold any advantage, because as Barry was saying earlier a pixel is a pixel etc..

However the counter argument to this is 35mm >> SD DVD does not make any sense ethier; namely " it starts with extremly higher resolution" huh? thats a hypocritcal counter argument since the counter argument is saying : high rez footage downconverted to lower rez yields superior picture ?!?

Simply put the counter arugument is the SAME arugument as the HDV >> SD DVD arugument , BUT why is it that the logic of the HDV is denied yet the same logic is accepted when it comes to 35mm >> Sd DVD?

the way i see it is DVX is 1/3" and FX1/Z1 is 1/3" they are gonna be that same, in the long run.

me thinks its all about the SIZE of the CCD / 35mm negative plane.

Ok let me try to bash out some cowboy theory as to why downcoversion should work:

(a) Imagine a circle (infinte points) now pick a point in that circle call it P_actual

(b) If we were to resolve this image with a higher lines then these lines will be "closer" to P_actual thus when average the two nearest lines we get P_aprrox_HD

(c) If we resolve using lower rez, then the lines are going to be "futher" apart yielding P_aprrox_SD

(d) now niether P_approx_HD and P_approx_SD will ever get the exact same value of P_actual , HOWEVER P_approx_HD should be closer in value than P_approx_SD


arrrrr.... any way who cares? quickest way to see for yourself:

get a DVX footage on to your computer , download .mt2 or wmv or whatever from net/friend/your own FX1 (in HD of course) and plug your computer to your SD TV or burn DVD and play on your TV, THEN you decide.
 
I'm going to quote Barry here. Anhar, maybe you missed his earlier response.

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.
Click to expand...
 
Barry_Green said:
All I asked for was any sort of evidence that a downrezzed HD image looks better, on DVD, than the same shot from a comparable SD camera.
Click to expand...


Well, that I don't have yet. But I will endeavor to get it going ASAP, and hopefully Shannon et al. will do the same.
 
Hey Isaac, thanks for that link! Walter's is the first test I've seen that verifies what Jarred and I found. I mean, I know what I've seen, I know what I've tested and I know it's right, but it sure is nice to see someone else verify it.

Anhar, I've gone on at great length to describe why the film-to-DVD argument is a total red herring. It's not about resolution, because film isn't a digital medium. Film is an analog medium, when it gets captured to video that's when it gets digitized. It gets scanned at the native resolution of the target medium, vs. the HD argument where we're talking about taking an image scanned at higher resolution and then digitally resizing it down. It's like comparing apples to oranges.

The argument, boiled down into its most tiniest comprehensive nutshell, is this: can you scale a high-res digital image down to a lower resolution, and preserve more resolution, than a native version generated at that lower resolution? And I contend that it's pretty much impossible.

I mean, let's just run tests in Photoshop... let's totally eliminate lenses and chip size and contrast ratios and production values and everything like that. Let's just create an image in PhotoShop at low res, then a higher-res, more detailed version of the same image, scale that higher-res version down, and see if it retains detail over the low-res one. It won't. It can't.

That's all I'm talking about.

Oversampling and downrezzing will yield lesser results than just sampling at the desired resolution in the first place.

Here's an example from the XL2...
XL2-16x9-vs-4x3.JPG

What this image is demonstrating is the difference between sampling a 720x480 image into a 720x480 frame, vs. downsampling a 960x720 image into a 720x480 frame. In the 4:3 shot the chip uses 720x480 CCD pixels to generate a 720x480 image. In 16:9 mode the chip uses 960x480 pixels and downsamples them into the 720x480 image.

Obviously the shape of the frame is different because the 16x9 version has to fit 33% more pixels into the same space, so everything is horizontally compressed. But which has more detail? Which is more clearly resolved? Does the 960-pixel version lead to more detail in the 720x480 frame? Quite clearly not. Look at the zeroes in the 600 number -- the 720x480 looks a lot more detailed than the 960x480. Look at the vertical bars above 700 -- the 4x3 is resolving them much more cleanly than the 16x9 version. Obviously. Because you're trying to cram too many pixels into too few, and you're getting aliasing and blending.

It's unavoidable.

You cannot preserve more detail by downsampling; the most detail you can possibly have is as much as the target medium is capable of resolving. And imaging it at the native resolution will inherently deliver clearer, sharper results than imaging it at a higher resolution and downsampling it (all other things being equal). It's just logical.
 
oh ya guys?

well.. my dad can beat up your DAD!

so bleh!

as a side note.. I agree with Barry
 
But Barry, there is a problem here in your example, when capturing a real life object it has infinite points (think about a cirlce or ball).

Barry i think you didnt quite understand my cowboy theory on downrezing, in your example your using something that has a image pattern with discreete points. this ceratinly aint the same as an object with infinite points.

camera at target rez and camera at higher rez are not going to resolve that real world ball/cirlce in the same way, i'm thinking according to my cowboy theory that even after the down rez it will have values that are closer to the "real" values than the values obtained from the camera at target rez.. can you at least see some logic in this?

Gonna run some downrez tests of my own, (in a weeks time, if i havea ny time!, gotta finish my thesis first, what on earth am i doing in this forum anyway :)

Anyway Barry, i've got enuff respect for you, but i've gotta prove it for my self, with my OWN real world tests.... so guys please dont bite my head off :)
 
Flintstone said:
Barry, wouldn't that chart look somewhat better with a DigiBeta? Less compression, 10bit, etc, etc...
Click to expand...
Surely it would. Everything about a DigiBeta camera is superior -- the lenses, the compression, the chips, the DSP, the processor, everything. DigiBeta is pretty much recognized as being at the peak of SD recording technology.
 
Now we're getting to the crux of the issue. There's more to a video frame than just the resolution, there's the color sampling. A higher res image provides more information to the subsequent downsampler. As I've said before, if I start with a HiRes still and encode direct to 4:2:0 mpeg-2 I get a noticably better result that if I downsample first to DV25 4:2:0 and encode from that. Resolution charts don't show you this at all and doing it in PS tells you nothing as it's working in 4:4:4.
What I'm saying is pretty easy for anyone to try, take any decent image form a digital still camera, you'll soon see what I'm on about, well it does make a difference using Vegas!
I should add that downsampling HiDef is NOT an easy task, even the best hardware boxes have issues such as jitter and moire, these things only happen when you start with a HD image, again this shows that something is going on.

Can I just add a small thing about the Z1 that we should all be able to agree on, forget the HDV hubris for a minute, the damn thing shoots 50i and 60i and if you don't want to live on the bleeding edge of HDV, no sweat, it'll shoot SD NTSC and PAL in 16:9 and 4:3. So here's something to get your teeth into, how good does SD NSTC shot on a DVX and converted to PAL stack up against SD PAL out of a Z1?

No one seems to have noticed this small but to me very significant feature of the Z1. All I can say is whatever Panasonic or JVC or Canon come up with in response to the Z1 it'd better do PAL and NTSC.
 
Anhar, please do test -- I'm not trying to stifle any testing, I'm trying to find the answer! I'm asking for testing, I'd love to see the results.

As far as infinite points vs. discrete points -- in the end we're still talking about a map of pixels, and digitally downsampling a larger map of pixels down to a smaller map. I don't know why having a complex-modeled shape would necessarily prove or disprove the bit-resampling portion of the argument -- perhaps it would, but I don't know how. And blending/averaging vs. rendering the best value at the target resolution... I don't see how using two pixels and averaging them together results in a better result than using the native sampling of that target area. I mean, I can see what you're saying, but in practice it doesn't seem like it would actually help. There can only be one pixel's worth of display on the target medium, surely imaging it at native res and capturing what the imager thinks, in the analog domain with the full frequency spectrum available, will produce a more accurate result than digitally combining two digitized samples...

I went to the simple grid because I was trying to eliminate a lot of the potential confusion -- we could get lost in discussions of lenses and chip size and DSP depth and things like that. I was trying to use a simple black and white example to boil the essence down to something we can easily grasp, and once we have that understanding, extending the size and bit depth of the pixel map shouldn't really have any effect, it'll still have to obey the same downsampling principles. Unless you're thinking of some specific reason?
 
robroysyd said:
Now we're getting to the crux of the issue. There's more to a video frame than just the resolution, there's the color sampling. A higher res image provides more information to the subsequent downsampler.
Click to expand...
Indeed there is. Does that information survive, is the question...

HDV is 4:2:0, and DVD is 4:2:0... but apparently not the same 4:2:0. According to Adam Wilt, there are two kinds of 4:2:0 -- the US version, which vertically interleaves the color samples, and the European version, which co-sites the color samples. So downrezzing HDV to DVD may or may not retain as much color as you'd think, in at least one of the territories. Then you have the issue that while the horizontal is an even divisor (1440 to 720 = 2:1), on the vertical it's not the same (540 to 288/240), so there will be some sort of averaging going on.

So here's something to get your teeth into, how good does SD NSTC shot on a DVX and converted to PAL stack up against SD PAL out of a Z1?
Click to expand...
Interesting question. Not having a Z1 on hand to compare, my guess is that the Z1 footage would be quite superior in terms of vertical resolution and color, and (if we're talking about PAL 50i) also quite a bit better in motion rendition (but shooting 24p and converting to 25p frame-for-frame should eliminate the motion rendition question). Shooting the native format should always be better than shooting in a different format and converting, all other things being equal.

No one seems to have noticed this small but to me very significant feature of the Z1. All I can say is whatever Panasonic or JVC or Canon come up with in response to the Z1 it'd better do PAL and NTSC.
Click to expand...
Agreed -- I think the time has come. Sony's proven it's possible, hopefully all other manufacturers will deliver it as well.
 
"Let's just create an image in PhotoShop at low res, then a higher-res, more detailed version of the same image, scale that higher-res version down, and see if it retains detail over the low-res one. It won't. It can't."

Very good point... My inexpensive Canon Powershot G5 captures (to my eyes, subjective, yes I agree) better images at 1600x1200 than it does at its' highest setting, something like 2544x????. I cannot exactly figure out why, but the lower-res images have less noise and the colors appear more vivid since there is less noise.

My reference shot was a picture of the oak floors in my apartment on a tripod and the only change was the resolution in the camera. Converting that higher resolution image to 1600x1200 is noisier than if i had just captured in the lower quality from the onset. Very good point, Barry. Thanks.

brian wells
 
You must log in or register to reply here.
Back
Top