XDCAM vs CINEFORM quality

Back to post

Comments

David Newman wrote on 6/3/2009, 8:47 PM
No, while there is a new file API with 9.0, we aren't going to start supporting until after 9.0a -- which fixes the half res playback bug. We want a stable Vegas platform for CineForm before we start build more on it. First 9.0a, then ASAP the x64 version (working in house) and then 32-bit float API support and more speed.
LarsHD wrote on 6/4/2009, 1:19 AM
1.

David: "4.8.6 was old compared to 5.0.7. "

Lars: Not *that* old... ;) And why do you mention this? Was it OK at the time of 4.8.6 (couple of weeks ago) to have chroma smear / shear of that magnitude?



2.

David: "Chroma fixes for Vegas are still coming."

Lars: What "chroma fixes" specifically are you referring to here? Chroma problems I already know or something *you know* and I don't yet know but will find out in the middle of a production and be dissapointed that I didn't know before... ;) ? Or the color shift (towards magenta) that I found in 5.0.7 that was worse than in 4.8.6? Pls explain. Why not fix the color shift *now*?


(Forgive me for saying so but with such vague answers you could just respond with "everything will be better in the future".... :)


3.

David: "More speed with happen with 9.0a and the new API (to follow after that.)"

Lars: Sounds great! But: "More speed" = as fast as MXF 35 Mbps/ MXF 50 Mbps. Or faster / slower? What is "more speed" ? And in July or December?


---------------------------------------------------------------------


Some thoughts: It would be great if Cineform when posting on this forum found a way to separate "general sales talk" from specifc useful real technical information. Both are important for CF of course, but it gets a bit complicated when CF's sales dept and the scientific tech department gets mixed together too much.. ;).

Things like "will be faster, will be more fun, etc. don't really enter into the real - and most of the time - very *specific* performance and technical questions dealt with on this forum.... It's real detailed answers on real specific questions most forum members are looking for I think.

I'm running tests here on my system, presenting very detailed as good as I can. And I'm getting sales talk back as replies. That's not good I think.


Since a codec plays a central role in production I think I would like to see a situation with much better information, more reliable updates etc. It isn't nice to have to do extensive testing to find out how some things got better, some things got worse, ect....


It seems that the parameters and vital criteria involved with codecs and running multigenerations could / should be defined in a much better way.

If CF had informed clearly about the performance and problems, I for instance would never even have bothered to download the 4.8.6 codec.

I want to *read* on the version updates how accurate/inaccurate chroma performance is. For instance:

Luminance level: should stay within +/- X % tafter 20 generations.
Chroma level: should stay within +/- X % tafter 20 generations.
Chroma drift / tint: Should stay within XY etc...

Etc.

As it is now, each new update seems to be able to have it's entirely new personality, lacking any real useul informtion about significant deviations from the expected. One time color shift this way, another time that way. (Imagine how that affects post work if color corrections are based on one specific version and then replaced by another).

It makes me feel a unsafe updating to a new version of a CF codec. I have to make extensive testing before I use it as I don't know what to expect. Testing that CF should have done and presented. Not me.

I do appreciate David N's comments here on this forum, however I find out more about the product while testing it myself. Davids talk is more of sales talk and is most of the time non-specific in terms of technical real detail.

When answers to real tech questions are so unclear, the discussion tends to take more time, compared to a situation where real detaild tech questions are answered with real tech answers.

A tech question shold not be answered by deviating into sales talk ("will be more fun, better later! :) etc.

What I spent hours finding out about the chroma shift problems should actually have been Cinforms task of informing about / correcting / informing.

And of course, not releasing a codec at all that was so severely crippled as for instance 4.8.6... (that *WAS* a codec realeased by CF and tested by CF wasn't it.....? )

And what info from CF informed me that 5.0.7 should have this new chroma problem cured?

-------------------------------------------------------

MXF and uncompressed AVI for me now.

==========================

My Cineform experience is that the coding and development skills are impressive and very real. It seems basically like a real good codec!

But the information, updates and quality control just isn't good enough for me to rely on in my work.

(my experiences are only from NeoScene and from within Vegas)

==========================

I have now again around 180 GB of 5D2 footage that is supposed to be transcoded to MXF/AVI etc. Will I use Cineform for it?

No, because the CF varies all the time, I get unclear answers and color problems *vary* between updates and it's so... unclear... I don't feel calm with it.

Have I spent a lot of time really trying to get to know the Cineform product and communicate the various issues. Yes, definitely! Was that fast and easy? No.


Best,
Lars




--------------------------------------------------------

apit34356 wrote on 6/4/2009, 1:58 AM
Lars, I think Dan's comments on gains/fixes are general because the performance between Intel&AMD cpus can vary a lot based on models. Then the OS and general computer hardware all play in determining thru-put on encoding/decoding.

I think you have found some interesting problems with Chroma colour shifting in multi generations. Since Sony's MXFs appears to correctly handle 422 colorspace, I would not be surprise to find other NLEs have 422 colorspace issues that are being overlooked.

But Affordable Universal 444 codec is my goal. So Cineform get busy! ;-)
LarsHD wrote on 6/4/2009, 7:16 AM
"I think Dan's comments on gains/fixes are general because the performance between Intel&AMD cpus can vary a lot based on models. Then the OS and general computer hardware all play in determining thru-put on encoding/decoding. "

The big differences I noticed between codec 480 and 507 were on the same VP8 in the same XP 32 bit. And with the same Intel Q6600 CPU.

best
Lars
farss wrote on 6/4/2009, 7:38 AM
I think you'll find over NLE's might do it correctly.
Several independant vendors make a point that decoding Y'CbCr to RGB is problematic and ideally should be done to 10bit precision.One of them was Convergent Design who were quite specific about there not being much to be gained by recording in 10bit but they strongly recommended decoding the Sony 4:2:2 XDCAM codec into a 10bit pipeline.
The full implications of what they were saying are only now starting to sink in.

Bob.
LarsHD wrote on 6/4/2009, 7:57 AM
Bob, transferring my 5D2 1920x1080 footage to uncompressed 1920x1080 AVI files would be the optimal way of preserving footage image quality right?

(While requiring a suffiecent capable streaming drive of course).

Lars
Laurence wrote on 6/4/2009, 8:06 AM
I wouldn't go uncompressed when there are several perfectly good free lossless compression codecs out there that cut the size of your files substantially: Huffy, Lagarith, Quicktime PNG, etc.
Laurence wrote on 6/4/2009, 8:11 AM
Might it be fair to say then that the XDCAM 444 flavor may not be any better than the 422 one for a single generation from an HDV camera (since the extra color space wasn't there to begin with), but it will handle the extra generations required by the time you finish a project with color correction, transitions and text overlays, and give you a noticeably better final master?

Might it also be true that the extra generation that is required to go from your master to your delivery format (be it Bluray, SD DVD, or online Flash) will be better coming from an XDCAM 444 master?
LarsHD wrote on 6/4/2009, 8:19 AM
"I wouldn't go uncompressed when there are several perfectly good free lossless compression codecs out there that cut the size of your files substantially: Huffy, Lagarith, Quicktime PNG, etc."


How can I try any of these "lossless" compression codecs? From within Vegas? Do I already have access to those in VP9 for instance?

Best
Lars
Laurence wrote on 6/4/2009, 8:26 AM
Install the codecs that you want outside of Vegas first. To access them within Vegas, choose the avi format, go into the custom parameters and they will be added to the compression types list.

Here is a good place to start looking at lossless video formats (the external links at the bottom of the page):

http://en.wikipedia.org/wiki/MSU_Lossless_Video_Codec
Laurence wrote on 6/4/2009, 8:39 AM
It seems to me that what Lars has been seeing is the difference, not so much between the Cineform and XDCAM codecs, but the difference between multiple generations of Y'CbCr to RGB conversions done in 444 vs 422 color space. Is that correct?
hazydave wrote on 6/4/2009, 8:46 AM
I know CineForm was claimed to be "the same idea" as DV for HD. But keep in mind, computationally, it's an entirely different animal. That was also, I believe, intentional -- they're using Wavelet rather than Discrete Cosine transforms, which means you're not reinforcing compression noise going from MPEG-2 or AVC to CineForm. But that doesn't really saying anything about generational noise.

I've been using YUV nearly-losseless for layered edits on short videos, but actually kind of out of frustration: I couldn't get CineForm (my old fallback from the early HDV days) to agree to 1920x1080/60i as a supported resolution (this is with Vegas Pro 9). I have also had some problems with SonyYUV occasionally crashing on render, and the obvious alternate, HuffYUV, doesn't work with Sony's smart-rendering. So I've been a little frustrated...and neither of these are terribly attractive for long videos (given the 2 minute video I'm working on comes out at 8-10GB with the lossless CODECs).

I think I'll give XDCAM a try!
Laurence wrote on 6/4/2009, 9:15 AM
The new Neo Scene codec is only $129 and can do full 1920x1080x10bit (albiet at 422 color space). You can also access it outside of Vegas with programs like VirtualDub. It is much better than the free version which comes with Vegas.

I must say that XDCAM looks really good and is a joy to work with though on my minimal tests so far.
LarsHD wrote on 6/4/2009, 9:45 AM
Thanks Laurence for the codec links. Playing with it right now :)

Lars
LarsHD wrote on 6/4/2009, 10:05 AM
1. Huffyyuv - with nothing in "configure" changed, plays back aruond half the speed of the uncompressed AVI files (1920x1080).

2. The Huffy configure panel isn't working in VP9. But works in VP8.


The Huffy file is smaller than uncompressed AVI for sure
About three times as big as MXF

But first run played real lousy over here...

Didn't ceck the actual quality.


My goal is:
2 streams of 1920x1080 during dissolve without frame rate loss :)



best
Lars
jabloomf1230 wrote on 6/4/2009, 11:57 AM
Lars,

Go back and look at Eugenia's original intermediate codec shootout again. Morgan MJPEG2000 is probably the only other reasonable alternative to CFHD that has a VFW codec. It's payware, but the price is right. Unfortunately the Morgan website is apparently fouled up at the moment, so I won't post a URL. I have a copy of the codec and it works with both Vegas 8 and 9, but it's a bit slow on the rendering end. It's definitely a better alternative than HuffyYUV or Lagarith.

J

LarsHD wrote on 6/4/2009, 12:45 PM
Hi J,

Does any of these codecs play back faster than the uncompressed AVI? And with no image quality loss at all?

I play 1920x1080 at full 29.97 fps fine with uncompressed AVI.


Lars
LarsHD wrote on 6/4/2009, 12:54 PM
Trying to visit that Morgan website and AVG and Google stopped before it allowed the site to load. Warned again viruses and mal.ware etc. Not a good start... Don't want that kind of companies...

Seems like uncompressed AVI is the best if disk streaming and disk space is no issue.

If disk streaming is "perfect" isn't then uncompressed AVI really the best there is? It doesn't require any decoding and therefore must be optimal of the hard disks deliver enough data?

I really feel Sony's MXF and uncompressed to be the only solutions that have impressed me so far... MXF is great if I need to work on the laptop.

I'm not going to spend more time on thírd party stuff for the time being... I think Vegas 9 has got what is required to work comfortably here.


Lars

farss wrote on 6/4/2009, 2:52 PM
I suspect so.
On top of that the "How do I see this in Vegas" question is kind of moot. Using the system you are testing as the measurement system is flawed. Also visually apparent generational loss is a kind of meaningless test if your ultimate goal is get accurate repoduction of what was in front of the camera.
The problem could well be compounded by the camera that Lars is using not being a 'video' camera. I have no idea what the implications of starting with out of gamut colors and illegal levels are in this discussion.

Bob.
Laurence wrote on 6/4/2009, 9:22 PM
I was just looking over http://eugenia.gnomefiles.org/2008/09/15/intermediate-codecs-the-face-off/Eugenia video codec shootout[/link] and noticed that the Cineform tests in this comparison were done at 4:4:4 which is only available on Neo 4K and above. Since the errors over generations are mostly due to repetitive Y'CbCr to RGB conversions, these tests would paint a much rosier picture of how much damage would occur over generations than is actually the case. I would love to see Neo at 4:2:2 in such a test since that is what most of us who have bought either Neo Scene or Neo HD actually have.
LarsHD wrote on 6/5/2009, 2:00 AM
Well.... then there's that strange tint towards magenta that Cineform surprised me with yesterday....

...in the previous codec the tint was in another direction of the chroma spectrum... Still under the same component/RGB conditions...

...so I think there's something in quality control, precision missing here. I've struggled enough with that codec now. I've simply decided to not use Cineform anymore. Too many various issues with it and it just takes time to test, discuss, worry etc.

MXF is *way* superior in handling and uncompressed AVI ultimate in rendering for the best for full quality if for some reason MXF shouldn't do it. But MXF does very well. MXF is a an absolutely great codec.

And hard drives and hard drive space cheap these days (AVIs) and staying within the Sony/Vegas world is very convenient.

I looked at the Eugenia page and the result from the "all afternoon" she spent on testing lots of codecs. Wish my testing time could have been squeezed in one afternoon... ;) It doesn't at all cover real production aspects of this topic which I'm sure wasn't her intention eiether. Watching Excel bar graphs indicating "quality" can be fun but really doesn't answer the real life questions at all... So that page really didn't give me anything useful in this particular case.

And I don't care how good Cineform say it works in *other* environments or if things were "different", hearing things about 200 fps etc. etc. etc. I'm in Vegas, I really love Vegas and I just want it to work here and now. MXF and uncompressed AVI does just that. Works well in Vegas here and now!

It's what it looks like on the screen and how it *feels* when working with it on RAID-0 setups and how it feels winding backwards and forwards on the timeline and making dissolves etc that counts. And knowing colors are accurate and that updates are reasonable sensible and that I don't get "hidden" color shifts like with CF.

And...! also how easy it is to call support and get instant friendly help. Sony is outstanding here - I don't think I have ever experienced support that is so fast and friendly. My experiences from third party codec manufacturers simply can't compete here.

Sony MXF and - if needed occasionally - uncompressed BIG BIG AVI and a new real fast RAID-0 and I'm really happy now.

It is VERY easy that testing and technical struggle takes away a lot of creative time. With uncompr AVI and MXF and a good RAID-setups I can spend time editing and not worrying about mysterious codec updates...


Best,
Lars



LarsHD wrote on 6/5/2009, 2:05 AM
Bob: "The problem could well be compounded by the camera that Lars is using not being a 'video' camera. I have no idea what the implications of starting with out of gamut colors and illegal levels are in this discussion."


Lars: The codec problem with Cineform hasn't anything to do with the 5D2 footage. It's from other footage too, and it's from test bars, stills, patterns, shades, resolution charts in b/w in color etc. etc. Definitely not linked to the fact that 5D2 can produce colors and levels that may be out of the typical broadcast standard gamut (which many, many dedicated video-only cameras do too). The problems I've revealed with CF can be revealed using no camera footage at all or any camera footage or only test bars etc... I've been running lots of other cameras footage and the issues are completely separated...


Best,
Lars

BTW - I'm not discussing codecs anymore... I've made up my mind and I'm using Vegas for editing now... ;) Also, I see lots of !" I think" and "could it be" and "maybe it is" etc. etc. Run real tests like I have done. Spend several hours/nights/days suffering and report *real findings* and say *how* you did the tests. There's way to much philosophic "maybe" talk here...

Make *REAL* tests!

Speculations tend to confuse. If anyone thinks a certain camera will confuse a certain codec to work differently (which it isn't in this case) then bring in that camera and shoot the same scenes with that camera and another camera and then compare the results... ;)

Make *REAL* tests!
Serena wrote on 6/5/2009, 2:39 AM
Well spoken. In these discussions it is always important to be absolutely clear whether one is speaking from specific knowledge ("I know") or from general principles ("I think"). Tests are, of course, open to mis-interpretation about underlying phenomenon ("I deduce").
farss wrote on 6/5/2009, 3:44 AM
There's one thing I do know for a fact, this has been the most pointless discussion I've ever been stupid enough to become involved in. I've been accused of being a "measurabator" more than once but rendering static images through 20 generations is way outside the limits of my sanity.

I did a few tests, wasted 1 hour to prove to myself that something did exactly what I thought it would do and moved on. I also confirmed what I already knew, the codec that the camera uses does more harm to the image than anything else. I can live with that, much better to spend my time on what's in front of the camera and honing my skills using it.

Bob.