XDCAM vs CINEFORM quality

Back to post

Comments

jabloomf1230 wrote on 5/29/2009, 6:08 PM
I didn't say that he blamed anybody. Cineform 2.8 is still licensed to SCS and the problem must have gone away in the mean time, because all subsequent versions of Vegas, up to V9, work with the newer versions of CFHD.

As you are aware, even Vegas Pro 9 is not the most up to date NLE, with regard to programming. It still uses VFW for its codec interface, instead of DirectShow. Vegas doen't take advantage of CUDA or any other video card accelerator software, like even very inexpensive NLEs, such as TMPGEnc XPress do.

Now to average user, if Vegas works for them, who really cares about the innards? But it does pose a burden on anyone trying to make codecs work with Vegas and it wouldn't surprise me that at some point, there will be no new versions of VFW codecs. Or maybe, as someone suggested in another thread, MS will stop supporting the VFW interface altogether.
blink3times wrote on 5/29/2009, 6:21 PM
"and the problem must have gone away in the mean time,"

Sure did ... didn't it.... the problem just.......... miraculously disappeared!!!!
Which is to say exactly what about David Newman's post that you speak of????

Look... my bottom line here is no big secret... CINEFORM dropped the ball... no body else did. New blu manged to update.... Excalibur manged to update..... Ult S managed to update.... mercalli managed to update.... they ALL did... except of course for Cineform, who indecently sits there with all these excuses (roughly) aimed at SCS. Well.... 8.1 has been out for a while now.... where's the cineform update and when do we stop with the excuses?

And yes I'm well aware of VFW. I have made many a post on the subject and as with others, wonder why SCS hasn't updated it yet.

But this is all neither here nor there... the point of this thread is MXF vs cineform... and from my point of view, the FREE mxf codec is looking like a steal of a deal over cineform.
apit34356 wrote on 5/29/2009, 6:22 PM
"Prospect4K performs better than HDCAM SR and it's the codec of choice for at least one high end post system and several recording systems" the HDCAM SR codec been around for a long time compared to Prospect4K.

"The errors are hard to see, I shifted the image by 1 pixel up/across and then back again between generations apart from the initial transcode from XDCAM EX to XDCAM 422. As this was green screen test footage shot as HQ MXF at 25p I applied a CK FX to highlight what is going on when it's switched to Show Mask Only."

Well, shifting pixels( the whole screen) around even for pixel is not a true test for generation coping, especially with compression codecs.

Bob, if your testing 4:2:2 compression, a one pixel shift can "fall out", behaving more like 4:2:0; assuming the complete frame is shifted, you are losing 1920 pixels at the top and 1080 at the side and adding new 1920 pixels at the bottom and 1080 at the left side. This is going to change macro boxes.........etc.... Plus there is the green screen issue ................................................
farss wrote on 5/29/2009, 6:38 PM
This is going to change macro boxes.........etc.... Plus there is the green screen issue ................................................ "

That was the point of the exercise. Note however that in the initial transcode I did not do this and there was still a signficant change in the image.
This is not a criticism, just an effort to see what is really going on. Eyeballing the results my eyes are not 'golden', I cannot see any difference. As I've learned though what we can see and what compositing software sees can be two quite different things.

The MXF 422 codec looks very promising and as far as I can see more than adequate for many uses.

Bob.
apit34356 wrote on 5/29/2009, 7:10 PM
Bob, I think you pointed out that the 32f pipe did not behave as expected in some earlier testing, maybe the 32f needs a closer look, especially the 10bit handling.
farss wrote on 5/29/2009, 7:45 PM
Glenn has tested the 32bpc float in V9 and updated his web site, I don't see much point in duplicating his work and I'm pretty pushed for time at the moment. I have a DVD with 120 pages of text to edit for a client, GROAN.
I really wish I had more time, more money and better test gear to devote to testing stuff. Heck, I wish SCS would tell us more too. Some of the issues I've had to address using Vegas would have been so much easier to work through if I'd had more information at hand. I'm also a total bust at note keeping. I think I've "solved" some problems several times over, over the years.

Bob.
apit34356 wrote on 5/29/2009, 10:03 PM
"I'm also a total bust at note keeping. I think I've "solved" some problems several times over, over the years." ;-) know the feeling ;-) I need to check Glenn posting, but I wish SCS would publish a white paper on their implication of the video pipe.
LarsHD wrote on 5/29/2009, 11:15 PM
How can I upload screenshots?

farss wrote on 5/29/2009, 11:50 PM
"How can I upload screenshots? "

You cannot upload them to this forum. Upload them to somewhere like flikr and post links.

I wouldn't mind some advice on a good site for doing this either:)

Bob.
farss wrote on 5/30/2009, 1:04 AM
Images from my test are in an album here on photobucket.

Bob.
CClub wrote on 6/1/2009, 4:29 PM
I'm confused about the initial steps to take in getting either the uncompressed AVI or MXF onto your computer: didn't you have to capture as an m2t and then render into the AVI or MXF? I'm not familiar with another capture utility besides Vegas itself or Cineform, so I don't know if there's a way to capture right to either AVI or MXF and then use that as your editing "intermediate."

I think it'd be a great option to capture right to MXF; I would have no problem bypassing Cineform. But if I have to capture to m2t, then render into another format, I've already added a generation that I'd prefer not to. Whereas with Cineform, I capture right to AVI and then begin editing.
Laurence wrote on 6/1/2009, 4:51 PM
Cineform captures involve capturing m2t and converting it to Cineform codec too. It's just automatic so you don't notice it.

Keep in mind that there is no reason to rerender m2t into XDCAM format in order to start editing. I will continue to use m2t on the timeline for my original footage. Where I use Cineform is for any part of my video that needs to be second generation because of color correction, video filters, text overlays, transitions etc. For these parts you don't want to rerender into m2t because of the damage caused by multiple renders. That and for final masters. That way, if you prerender sections, they will smart-render into a larger finished product. For this way of working, XDCAM or Cineform should both work quite well.
CClub wrote on 6/1/2009, 5:02 PM
So the only time you are using the original m2t files is if you aren't adding any color correction, etc. and you're just smart-rendering back into m2t? If you add any color correction, transitions, etc., you're using Cineform or XDCAM files?
Laurence wrote on 6/1/2009, 5:16 PM
Something like that. Keep in mind that m2t does a data decompress and recompress of the audio every time you smart-render the video.

The way I'll use it is to pop the original m2t raw footage on the timeline, and from that point on, every time I render it will be to XDCAM. I might do one m2t smartrender at the beginning as I am getting rid of junk footage, but for the mostpart, it will be XDCAM every time I render. I do documentary stuff so there is a tremendous amount of initial footage which needs to be folded down into smaller and smaller bits until the final project is less than an hour. The smart rendering of both the video and audio in the XDCAM or Cineform formats is necessary for this sort of thing.

The smaller size of XDCAM is particularly enticing because of the quantity of footage on a documentary project. It looks like the quality is close enough (maybe even better) so that I don't have to worry about the tradeoff.

I'm loving being able to preview rendered sections at full resolution with the XDCAM viewer software.
CClub wrote on 6/1/2009, 5:53 PM
That is VERY helpful info... thanks. I do a lot of documentary work also, and I never even thought about doing a quick smart-render to get rid of junk footage. I always do two camera shots (primary: Sony V1U, backup: Canon HV20), which makes it easy to splice cuts together, but the wasted footage gets annoying.

I'm not sure I can move away from Cineform, as I never liked all the steps getting 24p out of my HV20 without using Cineform, but I'm definitely going to keep my eye on what Eugenia finds out from her input above re MXF vs Cineform. Thanks.
David Newman wrote on 6/1/2009, 7:36 PM
In recent years we have focused more on the high end markets for compression, as the compression itself is excellent for multiple generations. However Vegas does something we don't recommend for good multi-gen, RGB to YUV transforms and back again over each generation. In 444 RGB or native YUV modes in tools like Premiere and FCP, we avoid colorspace switches as much as possible (or use 32-bit float precision when one is need.) Yet Vegas requests RGB 4:4:4 data from a 4:2:2 YUV source, in 8-bit, and there is issue with multi-generation. If you use the CineForm RGB 4:4:4 codec in Vegas all is fine (I've done 50+ generations.) So errors seen above after x-generation are accumulated colorspace transcoding errors -- any smear is due to leaving 4:2:2 to 4:4:4 upsample on (that is not multigen safe -- turn it off in the control panel.) It seems XDCAM is doing a better job at colorspace conversion, so we intend to fix it as the CineForm wavelet core is far better than MPEG2, and we expect Vegas 9.0a will solve the remaining issue we face. Eventually if the Vegas market desires, we will implement with Sony's help the native CineForm importer for 32-bit I/O which leaves colorspace issues in the past.

David Newman
CTO, CineForm

P.S. New builds (v4.04 and Neo Scene 1.3.1) with Codec 5.0.7 out tomorrow. More quality, more speed, more fun (for those using First Light.)
farss wrote on 6/1/2009, 8:40 PM
Thanks Dan for keeping us informed.

Bob.
LarsHD wrote on 6/2/2009, 12:51 AM
Thanks David for replying and comfirming my findings.


David: In recent years we have focused more on the high end markets for compression, as the compression itself is excellent for multiple generations.


Lars: It is not excellent for multiple generations at all as far as I can see wen using it in Vegas here... Running multiple generations is where I see it fail here...

Lars: ****QUESTION 1*** I see CIneform is applying a bit of sharpening to image. I see this already at the first generation of Cineform. Is this "sharpening" a deliberate sharpening or is it artefacts that are produced as a result of the smear and color problems you say Cineform has presently?

David: However Vegas does something we don't recommend for good multi-gen, RGB to YUV transforms and back again over each generation. In 444 RGB or native YUV modes in tools like Premiere and FCP, we avoid colorspace switches as much as possible (or use 32-bit float precision when one is need.) Yet Vegas requests RGB 4:4:4 data from a 4:2:2 YUV source, in 8-bit, and there is issue with multi-generation. If you use the CineForm RGB 4:4:4 codec in Vegas all is fine (I've done 50+ generations.)

Lars: ******QUESTION 2:* Can I achieve multiple generation results *today* that are better than MXF?

David: So errors seen above after x-generation are accumulated colorspace transcoding errors -- any smear is due to leaving 4:2:2 to 4:4:4 upsample on (that is not multigen safe -- turn it off in the control panel.)

Lars: *****QUESTION 3*: David, exactly what settings in Cineforms dialogue (from within Vegas) should be used in order to prove that Cineform is actually working for multiple generations? Please list what boxes should be checked / unchecked.

David: It seems XDCAM is doing a better job at colorspace conversion, so we intend to fix it as the CineForm wavelet core is far better than MPEG2, and we expect Vegas 9.0a will solve the remaining issue we face.

Lars: ****QUESTION 4:* Does this mean that *today* it is *not* possibe to do multiple generations with the Cineform codec that can compete with MXF?

David: Eventually if the Vegas market desires, we will implement with Sony's help the native CineForm importer for 32-bit I/O which leaves colorspace issues in the past.

Lars: ****QUESTION 5* . The MXF *plays* back better than your codec (fps wise), for example I can run 1920x1080 footage and see a dissolve between *two* video streamswithout a drop in frame rate. With Cineform that is not so. Will anything be done to actually improve the playback performance of Cineform when used in Vegas? The MXF handles very well and even if Cineforms multigen color space problems will be fixed, the MXF XDCAM will still be very attractive to work with. And... if I really want the theoretically best possible quality when rendering, I can always throw in my uncompressed AVI files. Where does then Cineform fit in in this scenario? A slower codec than MXF and not as good as uncompr. AVI. What is the advantage with the Cineform codec here?


Best,
Lars


PS 1. Pls if you want elaborate on the "sharpening" that I see in your codec presently. I suspect it may not be actual sharpening per se, rather a side effect of the smear and chroma horisontal shift that I see, is this correct? The SIemens res stars seem rather unaffected but color transitions are severely affected.

PS 2. So when I purchased NeoScene to use it for my 5D2 footage when editing in Vegas I was actually paying for something that Cineform knew was not all right here...? Or was it my finding that revealed the problems with multigen prob / chorma skew 7 shifts etc?

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

David: "....any smear is due to leaving 4:2:2 to 4:4:4 upsample on (that is not multigen safe -- turn it off in the control panel..."

Lars: OK, I turned it off and already after 1-5 generations is looks dreadful. Is turning this "off" supposed to do anything better when rendering Cineform in Vegas 8.0c? It seems not (?).

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




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

David Newman
CTO, CineForm

P.S. New builds (v4.04 and Neo Scene 1.3.1) with Codec 5.0.7 out tomorrow. More quality, more speed, more fun (for those using First Light.)
CClub wrote on 6/2/2009, 6:19 PM
David,
I'm confused... you stated "If you use the Cineform RGB 4:4:4 codec in Vegas, all is fine." But then you state, "any smear is due to leaving 4:2:2 to 4:4:4 upsample on... turn it off in the control panel." Can you tell us exactly what you would recommend for settings in the Configure Control Panel in Vegas? Thanks much, James
Laurence wrote on 6/2/2009, 7:31 PM
OK, this makes me think of a new question: How do different flavors of mfx stack up against each other when it comes to multiple generations of rerendering (non-smart-rendering)?
David Newman wrote on 6/2/2009, 8:06 PM
CClub,

Sorry to confuse you. Neo 4K and Prospect 4K have a native 4:4:4 codecs, Neo HD and Neo Scene has a 4:2:2 (like most codecs.) Vegas askes for 4:4:4 from a 4:2:2 source all the time, that is the causes of the issue -- already 95% addressed in-house. The control panel 442to444 setting is a filter for making the intepolation from 4:2:2 to 4:4:4 nicer (ProRES has the same option) yet this filter can't be safely recursed as it causes chroma shear.

Lars,

1) We add no sharpening.
2) Outside of Vegas or with a 4:4:4 codec, yes. Fixing 4:2:2 in Vegas now.
3) Neo 4K, set 4:4:4 encoding mode.
4) Use 4:4:4 or Premiere or FCP or AE, etc. We have a bug in Vegas.
5) Vegas is using a slow API, we hope to move beyond that. It was faster in earlier versions, and half resolution except for 9.0 (fixed 9.0a) is faster with CineForm. It is weird that Vegas can stuggle to play 1080p29.97 on modern PC*, whereas under DirectShow (same core code) will playback over 200fps on my i7.

We will support more Vegas enhancments if their is enough interest. We've seen a big up-tick is sales from lower end markets with Canon 5D MkII and very soon the Lumix GH1, we just don't know how many are Vegas users.

* Note: Neo HD 4.0.4 and Neo Scene 1.3.1 are faster overall, worth getting you free updates on these tools

David Newman
CTO, CineForm
LarsHD wrote on 6/3/2009, 10:24 AM
1 - CHROMA SHIFT / SHEAR / SMEAR - MUCH BETTER - SEEMS FIXED! :)

Todays updated Cineform codec seems to me *considerably* better than the previous Cineform codec. Much, much less chroma shear! The horisontal severe chroma shifts I was speaking about earlier practically don't exist anymore. And I also don't see what I perceived as a "sharpening effect". Apparently what I perceived was due to the chroma shift problem (?).


2 - COLOR BALANCE - SHIFTING TOWARDS MAGENTA :(
The latest codec from Cineform, on the 20th generation now show a color shift / drift towards magenta. A black and white image is truly black and white in the uncompressed AVI original and in the Sony MXF file. Cineform 5.0.7 codec (todays codec) is shifting the entire image visibly towards magenta. The older 4.8.0 Cineform codec shifted the entire image color spectra towards green / cyan. Comparing the orginal and 6th generation Cineform today and 6th generation MXF, the MXF looks more like the original than Cineform. So only 6 generations of Cineform tints the footage towards magenta / cold sort of tone.


3 - LUMINANCE LEVEL IS BETTER BUT STILL GOING DOWN :(

Uncompressed AVI has its white at 100%. So does the MXF even after 20 generations. There are practically no luminance level alterations in the MXF file after 20 generations. In the "old" 4.8.0 Cineform codec the lúminance drop off was significant. In the "todays codec" from Cineform 5.0.7 there is still a luminance drop, but not as severe as before. But after 6 generations it isn't really an issue at all, the difference isn't really big.


4. ABSENCE O ARTEFACTS - MUCH BETTER :)

"Todays codec" is much better. Maybe the chroma shear have made things look so much better? No question that when things are right the Cineform codec is good. However, the shifting towards magenta is easily visible and is as bad as getting artefacts


5. CHROMA LEVEL ACCURACY

Here Cineforms latest codec is much better than the previous. But it is boosting some colors in level while lowering some other colors. Switch between a first gen color bar and 20th generations and you'll see. Still not perfect. Cineform after 20 gen very obvisously alters the relationship between the colors. Cineform *AND* MXF can be improved here I think... But the magenta shift isn't nice.



Summing up my experiences today:
============================

Cineforms codec 5.0.7 after 20 generations looks dramatically better than the previous 4.8.6 codec after 20 generations. The old codec was really bad. This looks much better.

Cineforms codec shift the entire image color tone towards magenta which isn't so good. Can this be fixed David?

Cineform is still slower to play than MXF. This is actually an important factor here...

Cineform has fewer visible compression artefacts, however after less than 20 generations this isn't necessarily an issue..

In my view, MXF is still very attractive. Why? Because if I only run 1-4 generations it still looks fairly good and plays SUPERWELL on my Raid-0 10000 rpm streaming drives.

I can play several streams of MXF and enjoy full frame rate during dissolves on 1920x1080 footage. This *IS* a great advantage. Very important to me. Gives me a smooth editing session.

Had Cineform made their files play a little better, as well as MXF, then I would probably have chosen Cineform (if magenta was fixed). What would happen if Cineform gave up just a little in quality and made the files play better? For proxy filoe work /oOffered alternative quality levels so I could swap back when rendering?

One concept I might try is having 4-8 10000 rpm drives in a raid setup and running uncompressed AVI only. And I wouldn't perhaps have to deal with these codecs altogether?

Or how much extra hard drive speed would CInefrom require to play the dissolves without a drop in frame rate? (David?)

Another route is using MXF, because it plays my 1920x1080 files really well, then render for delivery from that when that is OK. Then when I require 100% quality, I just throw in the uncompressed AVI files and things couldn't look better.

From the visual quality point of view, I think 2-4 generations of Cineform now is just fine. Except for the magenta shift. But not perfect and not (on my system playing dissolves 1920x1080) MXF is sitting ready waiting to be used...

What I will test further is: what are the differences 1-3 generations?

MXF really plays well. Cineform too, but not as well as MXF.

I think the biggest challenge for Cineform is how to make 1920x1080 files play as fast and smooth in Vegas as MXF...

I will test 1-3 gen and dissolves more.

And Cineform needs to get rid of this new magenta cast that indeed is visible rather quickly.

MXF plays better. MXF has no magenta cast after 1-3 generations.


best,
Lars

I ran all my tests today VP 8 in XP 32. Haven't checked them in VP9.
David Newman wrote on 6/3/2009, 8:28 PM
4.8.6 was old compared to 5.0.7. Chroma fixes for Vegas are still coming. More speed with happen with 9.0a and the new API (to follow after that.)

David Newman
CTO, CineForm
Jøran Toresen wrote on 6/3/2009, 8:35 PM
David, are you saying that there will be a new API for Vegas after the release of Vegas 9a? If so, what are the main modifications / enhancements?

Jøran Toresen