LarsHD wrote on 5/26/2009, 4:26 AM
I added some footage and some test bars and ramps etc on the time line. I rendered it all to:

a) uncompressed AVI (1920x1080)
b) xdcam 50 Mbps 1920x1080
c) cineform 4.8.6 codec 1920x1080

I kept the original events on the time line and then after that dropped a) b) and c) on the timeline to compare and analyze.

The uncompressed AVI was of course identical to the original PC generated bars etc. So this is best for quality. Playing back multiple streanms of uncompressed AVI 1920x1080 is not fun.

I wasn't able to actually *see* a difference in quality in the camera footage I had between the Cineform and XDCAM codec. Perhaps a slight, slight difference on a black / white striped shirt on a model, textile, fibres, fine details. But hmmm... very little difference..

both codecs have - as expected as they are compressed versions of the original - a small dgree of artefacts, and I'm not sure which of them has the most acceptable level of artefacts. They're both very good I think. The blacks are turned down just a little bit in Cineform. you see the black going below the line wher as in the original, uncompr and XD it says where it should.

For me the XD cam codec wins when it comes to producing smaller files and playing back at a slightly better FPS.

In perceived image quality the difference is minimal... If I render to 1280x720 WMV and upload, will you tell me the difference?

And the fact the XDCAM exists today doesn't make it less attractive... ;) For VP9 I mean... *and* it runs in *both* 32 and 64 bit so this make XDCAM somthing I can use now.

XDCAM codec after all makes it possible to work. Today. Cineform was nice with Pro 8. And I don't know about the future.

Overall performance with *XDCAM codec and Pro 9* is clearly *much* better than with *Pro 8 and Cineform* (frame rate and smoothness)

So staying in Pro 8 for the sake of Cineform is definitely no option for me thanks to XD.



farss wrote on 5/26/2009, 5:47 AM
Unless you've got seriously expensive monitors and test equipment trying to eyeball these kinds of things is impossible. Monitors also introduce banding, noise and artifacts. How they interact with other issues in the image can totally skew what you see.

You can use Vegas to subtract two images. What remains is the difference. This is a quite reliable way to compare images however I don't know if this is a terribly valid way to judge the visual significance of the differences.

LarsHD wrote on 5/26/2009, 5:59 AM
OK, found out how to set it up to subtract. Well, both the Cine and XD are of course displaying a little "differential-data" but it's - as you indicate - difficult or vague as what conclusions to draw from this... It's very little indeed.

XD looks like something I can use for sure. Having done some more streaming tests, Vegas seem to handle XDCAM and the hard drive streaming very well.

Laurence wrote on 5/26/2009, 7:42 AM
First generation XDCAM and Cineform should be very close. If you end up rerendering a few generations, the Cineform should look the same and the XDCAM will start falling apart. Cineform is much like the Sony DV codec in that it was designed for multiple generations of rendering. The tenth generation should look the same as the first. That is the beauty of this sort of codec.
LarsHD wrote on 5/26/2009, 8:21 AM
First generation XDCAM and Cineform should be very close.

Lars: True. They are.

Laurence: If you end up rerendering a few generations, the Cineform should look the same and the XDCAM will start falling apart.

Lars: Wrong. I just ran 6th generations of each (adding a little "1, 2,3, 4, 5, 6 generation" etc text so that it would actually rerender. XDCAM is *A LOT* better. Cineform is the codec falling apart here. Not XDCAM. I don't think you have actually performed test...

Laurence: Cineform is much like the Sony DV codec in that it was designed for multiple generations of rendering. The tenth generation should look the same as the first. That is the beauty of this sort of codec.

Lars: If you add text it will force the rendering to render from scratch again and you will see how it makes a new version of its artefacts every time. Artefacts will be intrepreted as new data and new artefacts on artefacts etc. Its is *very* obvious in Cineform...

Lars: The following things are clearly *very* visible on the 6th generation of Cineform:

1. Blue, bluish, magenta colors are slid to the left overlapping other colors (check on the color bars). There is a horisontal chroma skew in Cineform that surprised me. the cyan color bar is sliding over into the yellow color bar causing a blurry transition. Not so in XDCAM.

2. When you see the white text against the color bar, especially white against red, there are black halos around the text and various artefacts. Much, much less on tenh XDCAM.

3. Ringing effects / overshoot / undershoot. Check the waveform monitor and set it in "composit mode" and watch how distorted the color bars are, How the lead in edge and shapes are warped. Compare to the original clean bars generated in Vegas and then look at XDCAM that is much much cleaner than Cineform. The difference isn't small it is very obvious. this is a little like comparing square wave signals, test signals on a wav-file and then looking at the encoded mp3 file. Similar type of degradation of the signal.

4. Levels. What was 100% white has gone down to 96-97% in Cineform while XDCAM remains intact. Cineforms black that was 0% in first generation is 3-4% below black after 6 generations.

Cineforms 1st generation exhibits a light sharpening in 1st gen, the start of the bars show a little "peak" that perhaps some perceive as "nice". (check the wavform in comp mode) However, multiple generations show that such side effects don't turn out to be nice after a while.

XDCAM s HD 422 MXF just simply turns out to be a much better companion for me as I can really use this not only for transcoding my camera files, but also when making sub-files for effects etc and bring into the project again. Cineform - as I've experienced it today - is only good for one initial go. Not for several generations.


Between XDCAM and Cineform, it's definitely XDCAM for me. Thanks for the suggestion to try XDCAM today!




When I ran the test I used Cineform 4.8.6 codec in best and high settings. I used SONY MXF HD 422 1920x1080 50 Mbps.

Sony files are slightly *smaller* than the Cineform files.

Laurence wrote on 5/26/2009, 10:55 AM
I like you Lars. You and I have something in common: that is we listen to people telling us what the common knowledge is, but then we do our own tests to see if the common knowledge is something we should accept or whether a different course is actually the better one to pursue. I have not tested multiple generations of Cineform against the same number of generations of XDCam to see if I could see quality differences between them. I was just reciting the common knowledge of what should be the case. You've got me curious enough to do my own tests though.

I do know that there are some Cineform settings tabs that should be checked a certain way for multiple generations. I am in the middle of a render right now so I can't say for certain what they are, but I seem to remember that it has something to do with expanding color space on playback (which looks good on one generation but leads to multiple conversions on successive renders). Make sure that you have this setting right on your comparative rerender tests.
apit34356 wrote on 5/26/2009, 11:31 AM
Thanks LarsHD for the rendering test series. As I posted in another thread, DXCAM codec is a Sony's BT codec, meaning it must be commercially usable or the XDCAM camera line would be long dead. It does not surprise me that it did well. I think a lot of people equate HDV with XDCAM which is not the case at all.
Laurence wrote on 5/26/2009, 11:42 AM
One really good thing about the XDCAM format as compared to HDV is that the audio is uncompressed. That means that you can smart-render the video without worrying about successive data compressed audio degradation.
LarsHD wrote on 5/26/2009, 1:49 PM
Thanks Laurence ;)

And thanks for remininding me of the different settings in the Cineform dialgue that comes up when rendering to CF in Vegas.

Unfortunately these settings do not make the generation by generation degradation any better. Problem still exist as I described it in my post above.

Actually, if you make just one single generation and blow up the wave form monitor you'll see what direction the Cineform transcoding is taking you... (choose the composite view for a richer view of whats happening).

For fun I made 15 generations... By then the difference between Cineform and MXF is *HUGE*. This probably means something?

I have tried all possible combinations of the 5 checkboxes but I simply cannot get any quality improvements. Sonys MXF file / codec is clearly superior after 6 generations.

Thanks for the encouraging words - making multiple tests of 6 generations of renderings is rather boring. On the other hand, when working and editing it always feels good to know that the methods you are using are the right one and that they are chosen as a result of real life tests.

I had completly missed this MXF thing. I thought it was only something that had to do with a specific camera and it wasn't in my mind as a format I could use for transcoding 5D2 MOV footage into files that played nicely in Vegas. Now I know! :)



PS. Laurence, if there is any particular combinations of CF settings that you feel should affect my results, pls let me know and I'll run the test again.

Or if Cineform/David N. pehaps have a suggested setting that will change / improve my test results?
apit34356 wrote on 5/26/2009, 2:01 PM
Cineform made HDV workable and offered a lot smaller files than AVI. So, I hope Cineform resolves the V9 for all its customers.

When possible, I prefer 4:4:4 uncompress, a little of the mid 2000's workflow. But many of the new workflows between 3D and video permit more compression, so more options are nice, I think. ;-)
farss wrote on 5/26/2009, 4:27 PM
My best efforts at trying to break the XDCAM 422 codec were interesting. So far it gets 9/10 and Vegas 9.0 gets 1/10, oh dear.

I took a series of random still images as an images sequence into a 1080p25 project. I rendered to XDCAM 422. Result held up remarkably well, you'd almost think this codec is I frame only as mpeg-2 long GOP should really be falling apart under this kind of stress as every frame is totally different. Then again the images were lacking much high frequency detail so the DCT compression would have had an easy time of it.

Much bigger issue though. Trying to compare the rendered output to the original frame I found Vegas renders frame 11 as solid red from track 1, frame 12 it decodes correctly but sees the source frame on track 2 as solid black. Turn the Invert FX off and the problems go away.

This is an extreme synthetic test however that's the way to test software to shake out the bugs. That's why the military just love open days, let the kids play with their new hardware and see what they can shake loose. V9 fails the shake test.

jabloomf1230 wrote on 5/26/2009, 7:19 PM

There are some interesting comments in this blog which describes the testing of various intermediate codecs:


You might want to contact the author and see if she would submit the XDCAM codec to similar testing. Eugenia posts here periodically, so she may have seen your thread and thought about testing the XDCAM codec already.
Eugenia wrote on 5/26/2009, 9:13 PM
Is XDCAM coming with Vegas Pro 9? I have the box, but I haven't installed it yet. Without the codec easily available, I can't test it.
Justin Young wrote on 5/26/2009, 9:21 PM
It's in V9, and probably V8. Render as a MXF file, the XDCAM templates are there.
Eugenia wrote on 5/26/2009, 11:33 PM
Thanks, I might do it this weekend. I am behind in writing 3 other reviews this week, so this will have to wait a bit.
jabloomf1230 wrote on 5/27/2009, 10:38 AM
I tried a little "eyeball" testing of 6 generations of the XDCAM codec, using Vegas 9, 64 bit and it does work very well to preserve intergenerational quality. XDCAM HD 422 is MPEG-2, long GOP, at least according to Sony's specs:


With a quad core computer or better, you should get close to RT preview frame rates in Vegas, before applying FX.
cliff_622 wrote on 5/28/2009, 7:21 AM
Anyone know if XDCAM HD 422 is a 10 bit codec?

This is an excellent thread!

kairosmatt wrote on 5/28/2009, 7:45 AM
If you want to convert AVCHD to XDCAM, is the best way to do this to render the clips from the Vegas timeline? Or is there an app that will do this?

Barefoot Joe wrote on 5/28/2009, 7:50 AM
"If you want to convert AVCHD to XDCAM, is the best way to do this to render the clips from the Vegas timeline? Or is there an app that will do this?"

I posted a similar question in a little more detail. C'mon guys...spoon feed us! (and thanks)
Laurence wrote on 5/28/2009, 8:21 AM
The easiest way to do this is with a script like VASST Ultimate-S which has a function for this among many other things. Vegas does the actual rendering, but it is all set up and automated from the script.
LarsHD wrote on 5/28/2009, 11:01 AM
Another comparison, just 1st generation...

MXF 35 Mbps codec - file size 245 665 KB
MXF 50 Mbps codec - file size 668 416 KB
Cineform 4.8.0 codec - file size 2 154 219 KB
Uncompressed AVI - file size 20 336 576 KB

White text and shapes against a red background and artefacts:

MXF 35 Mbps - slight "jpg-like" artefacts on the red surface near the white borders.

MXF 50 Mpps - very, very slight "jpg-like" artefacts near the white borders.

Cineform 4.8.6 - even less "jpg-like" artefacts near the white borders.

Uncompressed AVI - no artefacts at all

So here I made a test where I noticed a little thing that was *better* in Cineform than in MXF.


The multigeneration test I ran a few days showed quality degradation where I think MXF was a clear winner.


Comparing and testing compressed material 1st generation is complex and isn't as easy as drawing some curves ans saying this codec is better than that codec etc... It depends on each individual situation.

There may be situations where on the first generation during one extreme condition or one specific combination of shapes / Siemens Resolution Stars (test chart) etc. one codec can appear slightly better than the other.

After multiple generations things usually get a little easier to judge.

That said, first generation of Cineform has some sharpening artefacts that first gen of MXF doesn't have... Cineform produces a little black shadow around white characters etc...


MXF files are really small! It is amazing how these small, easy to handle files perform so well I think! And Sony has this nice XDCAM player that is quite good to have on the desktop!


The MXF 35 Mbps file 1920x1080 plays back real well even on a small little 5400 rpm laptop! I even ran a dissolve betwwen two 35 Mbps streams and the frame rate dropped from the full 29.97 to around 23 fps. 1920x1080 on a laptop and dissolves! Amazing. Cineform doesn't get close to that.

In the laptop again; preview/half/quarter even the 50 Mbps plays back fine at full 29.97 fps!

The MXF files plays back much, much better in Vegas on a small laptop than Cineform. And footage *looks* the same.

On the big studio PC with Raid-0 etc MXF files allow me to have more real time effects and dissolves etc than when using Cineform.

Now if I want to work with 1920x1080 footage and not making smaller lighter proxy files but still want truly ultimate output when rendering can't I just then use the MXF files, 35 or 50 Mbps, render out from the 50 Mbps, OR use the 35 M while working and 50 Mp for rendering.

Or, if I truly, truly want optimum quality - throw in the uncompressed AVIs when rendering.

When doing multiple generation exports/rendering for effects and complex montages etc. I think MXF 50 or uncompressed AVI is best.

For me I feel the MXF 35 and 50 and truly practical files to deal with. They don't cost anything to make. They play super well. They are small.


Now, are there *any* drawbacks with the MXF codec here that I have missed? Please let me know!


If the original footage is 1920x1080 and the end product is supposed to be 1920x1080 it is nice to have 1920x1080 size when working with the project. Text size stuff and other things get simpler. And also making various adjustments get easier.

Even with Cineform I have done 960x540 versions just to make the project run really smooth while working. With MXF I seem to get good performance while staying at 1920x1080 and good quality.


So while doing a lot of testing here and getting pretty impressed with MXF, I'm anxious to hear from other users. It looks so good to me that I feel this is something I really can use instead of Cineform.

But I want to make sure I don't miss anything.

Laurence wrote on 5/28/2009, 12:45 PM
I just did a .mfx test render. One thing I notice right off the bat is that it plays beautifully at full 1920x1080 resolution with the VLC media player and looks absolutely incredible.

I was really hoping I could convert it directly with VirtualDub or the DivX convertor. No luck unfortunately.

None-the-less, even at the highest quality setting, it is a fraction of the size of Cineform codec avi and previews and renders very quickly. Audio is uncompressed and can be 16 or 24 bits with up to 8 audio tracks. Rendering is lossless unless something changes in the frame, and with the quality of the rerenders, it seems like a very practical way to work.

Yeah, I want to echo Lars here: is there something we are all missing here as to why we shouldn't all be using this codec?

jabloomf1230 wrote on 5/28/2009, 12:58 PM
Now, are there *any* drawbacks with the MXF codec here that I have missed? Please let me know!"

The drawbacks don't appear major, but I am also waiting until Eugenia gets a chance to do her testing. Here's my thoughts. Other than VLC, no other media player will presently play the XDCAM HD 422 clips. Not even codec appliances, like graphedit or GSpot will do it.

Most NLEs (other than Vegas) will not handle the clips either. Both the Premiere and After Affects documentation clearly states that CS4 does not handle them natively. You need to buy an expensive converter from MainConcept to do so. Cineform presently will not convert the XDCAM HD 422 clips either.

So, in summary, if you are working solely in Vegas, this might be a good approach. But if you need to transfer the clips to someone not working in Vegas, it's not a good idea.
kairosmatt wrote on 5/28/2009, 1:26 PM
I have an HF-100 clip that is 11.32 MB, recorded at highest quality (Max bit rate of 24, but its variable).

When I rendered that to XDCAM EX 35Mbps it comes out to 11.83 MB. Something must be wrong, because if file size is based on bit rate, then the EX should be larger!

Resolution and framerate are the same. Field order and PAR are also the same. The only difference I can find is that clip attributes go from 1920x1080x16 to 1920x1080x32.

Any thoughts?
apit34356 wrote on 5/28/2009, 1:37 PM
"Yeah, I want to echo Lars here: is there something we are all missing here as to why we shouldn't all be using this codec?" ;-) Since V8 XDCAM codec has been available, but SCS has been slow to educated the benefit of this codec for non XDCAM owners; this may be a choice to not interfere with 3nd party codecs and avoid annoying Sony BT division with non XDCAM users complains about trying to using 4;2;2 correctly.