4k - DV ?

[PMedia]

Hi There

I'm really thinking of getting this camera. Sorry if I haven't got this correct but will the camera film write down to DV quality straight out of the camera or does it down convert? I know what you all will say why bother with DV but I'm asking because these no piont filming 4k for a coporate that will end up on DVD. Much easier for backup storage if the files are smaller. We will do all our Television commercials at higher res of couse but just want to know.

In the end a company like mine is mad to go and buy a high end sony or pana Cos I'll still need to go out and buy lens... If a red camera can go from DV to 4k at this price I'm blown away.

I don't think television has been this groundbreaking since the change from black white to colour.

Who would have thought 10years ago we'd be tapeless, 8 core Mac, FCP2, the red camera wow!!!!!!!!!!

Cheers

Campbell

[Brook Willard]

No SD in camera. Shoot 4K, 2K or HD... downconvert to SD in REDCINE in post.

or...

Take HDSDI output to a converter box on set

[Gunleik Groven]

Is this bait? -;)

Don't ever telle Graeme that you intended to do this.
he'll cry!

Bad boy!

-;)

Gunleik

[Cail Young]

Alternately, the "RedRad" importer in FCP will let you log and capture while transcoding to SD - you can use ProRes 422 at standard definitions or if you're desperately low on space, DV.

[Jeremy Hughes]

Who would have thought 10years ago we'd be tapeless, 8 core Mac, FCP2, the red camera wow!!!!!!!!!!

You're going to edit RED footage in Final Cut Pro 2? Well I guess you can't afford Final Cut Pro 6. It's a shame. Pretty soon they'll have REDCODE RAW support. Is that why you wanted to convert to DV? :whistling:

I know you meant FCS2 and FCP6. :wink:

SD isn't obsolete yet. And will look incredibly high quality on RED. If shot at 4K and downsampled to 480p, it'll have all the natrual sharpening and still have that incredibly wide dynamic range and shallow depth of field.

[ChristopherKenworthy]

Film looks quite good on DVD, so I suppose Red will too. But don't ever bother with the DV codec. You can create SD DVDs from all sorts of codecs that aren't nearly as hideous as DV. Given that you can export Redcode to any codec you like, pick the biggest and prettiest codec your DVD-software can handle.

[Wayne Osborne]

I would have to agree with Christopher. Your production will only ever be as good as your weakest link.

[David Dennis]

It might be worth elaborating a little to the original poster, who I don't think understands what he's saying.

DV is something called a CODEC, or Coder-Dedocer. That is, DV takes a huge stream of images and compresses it to something managable.

If you take a DV camera, it could have the best image sensor in the world and the image will still be probematical. It can look pretty good for closeups because they don't require as much detail, but for a dramatic panorama, bushes will come out as fuzzy clumps.

Why? Because the DV codec has to throw out most details in order to pack the huge image data into the small amount of space it has. This is why using the DV Codec is going to look like an insult to the RED team, because you're taking their image sensor and hard earned effort to make it perfect and running it through a codec that will lose almost all that quality!

The good news is that codecs have gotten better, because computers are faster, and can do more sophisticated compression than they could back then, and disk space has gotten more affordable, so less compression overall is needed.

So in practical terms I believe REDCODE takes maybe 10x the space of DV and gives enormous quality improvements. When I first started editing in DV, a 11gb hard drive was all I could afford. Now you can get a 500gb hard drive system for the same price! So you can see that, despite what you might have thought, taking 10x the disk space of DV is no big deal with today's gadgets.

If you look at DV and you look at DVD, both are highly compressed digitally. But DV has to compress on the fly and so it can't do as good a job as the encoders for DVD. What this means in the end is that if you shoot using the RED codec and you encode the result to DVD, it will look every bit as good as 35mm film, if not better. In contrast, if you took the RED image and transcoded it to DV and then to DVD, it would look awful.

There is of course no problem at all editing the REDCODE on the FCP timeline. (Well, there is some kind of last minute delay in FCP6's implementation, but I'm sure that will be fixed through software update long before you actually get the camera). REDCODE will work just like DV does for you now, except that it will look 100x better.

I hope this helps people's understanding, and if I was wrong in any way, corrections are welcome.

D

[Bruce Allen]

Who would have thought 10years ago we'd be tapeless, 8 core Mac, FCP2, the red camera wow!!!!!!!!!!

Cheers

Campbell

I did! That's why I only the only DV camera I bought was a second-hand 1-CCD Sony PC5! And I never bought the fastest, most expensive computer... I went for cheap, value-for-money PCs, all the time, and tried not to upgrade too often. Because anyone calling a computer or digital camera an "investment" is nuts. It's a great, essential tool that just happens to go out of date and lose value incredibly quickly.

Here's my (very easy to do) prediction: in 10 years, we will be wireless, 64+ core, amazing amounts of graphics power and storage, yet sadly, we'll still be waiting for our computers because someone will have invented some cool new software feature that takes a while to render. Take a look at the research papers coming out of the top universities - they have figured out a lot of ways to use up all of those CPU / GPU cycles already.

Oh yeah, and in 10 years' time those people sitting in the coffee shops writing their scripts will have much cooler looking Powerbooks but they still will spend too much time browsing the internet and writing comments like this one I am writing right now instead of listening to the way people really talk and writing great dialogue...

Or else forumspeak will become how we talk? Scary...

Bruce Allen
www.boacinema.com

[MikeCurtis]

FYI - DV isn't a good match for DVDs - while the pixel resolution matches at 720x480, the color sampling is a mismatch - DV is 4:1:1, DVD is 4:2:0...so you're ditching even more of the color sampling data. Use DVCPRO50 at least.

[Vincent Rice]

Personally I would keep a 720p workflow until the final Compressor stage.

[Nick Shaw]

DV is 4:1:1, DVD is 4:2:0

Not true for PAL (which I believe New Zealand is). PAL DV is 4:2:0 720x576 like PAL DVD.

Nonetheless, every re-compression cycle between source footage and end delivery degrades the image quality. Dropping to 4:2:0 only at the last stage can only be a good thing.

Even when I'm working with DV sourced material, I do my final renders all 4:2:2 uncompressed, and encode to DVD from that.

[GlennChan]

Even in PAL there's a slight mismatch in the 4:2:0 schemes.

In PAL DV, the 4:2:0 is co-sited both vertically and horizontally.
In DVD, 4:2:0 is co-sited horizontally and interstitially sited vertically.

To explain this clearer:

If the chroma is co-sited, its center is exactly over a luma pixel. On re-construction, the existing chroma values are sort of blurred together to form the missing chroma values.

In interstitial, the center of the chroma is between luma pixels. Scaling is done in a block-like fashion. This is the fastest re-sampling method / least computationally expensive.

If you mismatch the two, then the chroma will get shifted by 0.5 pixels... though you don't really notice this and ironically it helps to fix other problems with chroma subsampling. It looks better than it should. (Both effects are very subtle.) It doesn't look quite as good as doing things right in the first place.

2- For the best quality (from a technical standpoint), you want your project to be 4:4:4. Chroma subsampling degrades quality whenever you have multiple generations. Drop down to 4:2:0 at the end when you encode to DVD.

As well, a lot of decoders and encoders don't follow the rules (i.e. they go interstitial instead of co-sited)... so you get minor hits in quality there.

3- Um... it gets more complex from here. Because when you encode the chroma, you should be applying sinc-like FIR filters to avoid aliasing (though this creates ringing artifacts). ITU-R Rec. 601 and Rec. 709 specify filter parameters for the design of these FIR filters. Most encoders don't do this.

That probably doesn't make sense unless you understand the basics of resampling. Basically when you re-sample, you have to pick your poison:
-aliasing
-less than full resolution / blurriness
-ringing artifacts / edge halos

You can't avoid all of these things, you have to pick at least one of them or some combination of them.

On downsampling, aliasing are the spurious detail that shows up. See the pictures at
http://www.cambridgeincolour.com/tutorials/image-resize-for-web.htm
("moire artifacts")

If you apply sinc filtering, you get perfect resolution and zero aliasing, at the expense of ringing / edge halos. However, sinc filters have an infinite response and take up way too much computational power to figure out. You can however approximate a sinc filter; this takes less computations and you pick up minor aliasing (and reduce ringing somewhat).

Also see:
http://www.cambridgeincolour.com/tutorials/digital-photo-enlargement.htm


So um.... yeah. More than you ever possibly wanted to know about chroma subsampling. The ideal solution and a simple solution is #2: Work in 4:4:4.

[Nick Shaw]

More than you ever possibly wanted to know about chroma subsampling

Very informative Glenn. We all come to this forum to learn from the expertise of the other members. Too much information is certainly better than not enough. I'm a strong believer in understanding how things work 'under the hood' rather than just following rules of thumb when I don't understand why.

[GlennChan]

Oh wait I'm not done yet. :D

Typical filtering is done in the gamma-corrected domain, which causes shifts in luminance. To get luminance right, you have to do your calculations/processing in the linear light domain. So remove gamma correction (that gives you linear light), do your processing, then add gamma correction back in.

This concept is the same idea as luma versus luminance.
http://poynton.com/papers/YUV_and_luminance_harmful.html

This phenomenon can be seen in the attached images... the red loses it strength, so the overall image doesn't look as saturated/vivid/the reds aren't as pure.

2- The difference between luma and luminance can also be seen in the attached images as the green/purple areas getting darker.

3- The attached images also show aliasing occurring (it doesn't apply FIR filtering to reduce aliasing). Your monitor is probably introducing aliasing too if you have a CRT, so zoom in via photoshop or something.

It's the SonyYUV codec in Vegas; it's a codec that mismatches interstitial siting and co-siting.