Agreed.
The titles etc I would add at the very last stage. Quality loss on graphic elements are the most noticable. For vision if you use the Sony 4:2:2 MXF. The quality loss is small(ish) especially compared to the what the consummer AVCHD cameras do anyway.
Note also that MXF uses uncompressed audio and supports multiple channels of it.

Bob.

According to the Filmmaker's Handbook (the bible), in the old days three generations was considered the limit of usability for film copies.

Then it goes on to say something to the effect that high enough bitrate copies do not suffer from appreciable loss.

After I get jumped on over this, I'll go make a scan and double check it.

Codecs are not all built the same.

Some are mathematically lossless. You can re-encode 100 times and see zero difference (HuffYUV, Lagarith, uncompressed, Animation, MJ2k Lossless).

Some are visually lossless. Most of these hold up well for 5-10 generations. (ProRes, DNxHD, CanopusHQ, Cineform, MJ2K, etc.)

Some are somewhat lossy. I would not go past 2-3 generations on these (HDCam 422 is one I'd put here)

Some are very lossy (AVCHD, Mpeg2, etc.) so it's hard to generalize.

Isn't MXF Mpeg2 customized by Sony?

No.
MXF is a container. mpeg-2 was developed by the Motion Picture Engineering Group. It defines an encoding scheme and at several different profiles plus it has different stream types. With a high enough bitrate and 4:2:2 chroma sampling it's exceptionally good but of course file sizes become pretty large. GOP length also has a major impact. Some tape formats e.g. Digital Betacam use it but with a very short GOPand it's 10 bit which makes it non compliant with the standard. As such a capture from that tape format is not lossless.

You're quite right, the loss from printing film is very high. The best camera neg can manage 4K res, by the time it gets to a release print it's down to around 700 lines. This is just my opinion however I suspect the way film losses resolution through the optical process is totally different to the way things happen in the digital world. For one thing film does not have discreet fixed pixels.

Bob.

I've got my FMH by my side to defend against any "counterfeit pope's" that may come out of the woodwork. ;>)

The FMH kind of hedges on this stuff too, but what I'm getting is that the format you edit in determines the loss rate.

It says highly compressed codecs deteriorate much faster when edited and re-rendered back even to the same codec, which makes sense.

It seems to say that AVCHD deteriorates at about the same rate as film if edited natively.

So I don't understand the cry for better AVCHD support, getting a faster computer to build a lossless or near-lossless proxy is the way to go.

That certainly motivates ME to convert to MXF now.

It doesn't matter what you EDIT in. It matters what you use for rendering and final. But, the idea of "edit" encompasses much more than editorial these days, so it's convoluted.

So at least you've come to understand why pros don't worry at all about AVCHD support in the NLE. It's an atrocious codec to work with. Slow, highly compressed, poor color structure, etc. Proxies are one way to go. Transcodes into a final format are another depending on machine speed.

I'd find the Sony MXF filetypes much more useful as proxies if I could dictate my frame size, bitrate, or frame speed. But they are totally inflexible, and that won't work for all jobs.

Yeah, yeah, yeah, by "edit in" I meant compressing/decompressing using the online workflow method.

Yeah, that's what turned me off too. I think there's only two choices of bitrate and ZERO choices for anything else.

It would be nice to be able to tweak it towards the final format. It's as if Sony, in its infinite wisdom, has pre-thunk the issue for us. To be fair, perhaps they feel that this is the best way to "mix and match" clips from different sources more efficiently or consistently.

I'm sure the Vegas engine is optimized for mxf too. The templates offered are probably in the "sweet spot" region of the engine.

More correctly, the templates match the tape standards that these codecs are meant to digitize.

So...
With regards to AVCHD, it appears the recommendation is to get it into an intermediate format when editing such as MXF or Cineform for least trouble?

I'm trying a trial version of Cineform Neo Scene right now....

I'd like to get back to the point where I can go straight from the timeline to finished product like I can with the standard def stuff without Vegas choking.

"With regards to AVCHD, it appears the recommendation is to get it into an intermediate format when editing such as MXF or Cineform for least trouble?"

I have very few issues with avchd directly on the time line. Are you sure it's "AVCHD" that you're dealing with? (I ask simply because there seems to be confusion on the exact meaning of "avchd")

I render out my work (usually a 3 hour time line) as avc (or sometimes a M2TS with no audio) with a separate ac3 file then import it all to DVDa. I have no issues (or recompressing in DVDa) and I turn out perfect Blu rays.

If you really wish to get into intermediates though I would choose MXF and saty away from cineform... too much trouble.

I understand what you're saying that each type of source media has an optimum capture codec and data rate, like DV. and HDV have different bitrates coming out of the camera.

I read in the manual that one way to deal with this is to capture via raw DVI directly to an AVC recorder, so you don't need any other codecs except a proxy for AVC.

I'm going to take that advice. I like the way MXF "feels". I dunno, but it just feels efficient.


=================================================
If you really wish to get into intermediates though I would choose MXF and saty away from cineform... too much trouble.
Message last edited on 1/1/2010 9:13:16 PM, by Rob Franks.

"each type of source media has an optimum capture codec and data rate, like DV. and HDV have different bitrates coming out of the camera."

BudWzr, what are you talking about?

I *think* my trouble has been the combination of AVCHD (and I'm sure that's what it is), and the other media I'm using.

I'm not sure I would be having trouble if AVCHD was all that I had on the timeline.

I appreciate the information here. I've been working with Neo 2 Booster to get some of this type of work done on a trial basis, and although it really kicks butt in the AVCHD arena, I just don't care for it on most everything else, compared to Vegas.

A little late, but remember that vegas pass 4:4:4 between tracks. So if mxf looks great on the vegas timeline, it will render out from 4:4:4 back to 4:2:2 or 4:2:0 where any losses occurs from 4:4:4, not mxf org. There are "tricks" to controlling motion loss. The simplest is appending small renders together without rendering them again. Look around, there many good apps for this. Smart rendering was going to make this easy, I can not state this is true in all cases or any unique workflow I've seem yet---- But I don't use vegas enough to state positively either way.

You're talking about *color* resolution (subsampling). That's all well and good, but the bigger loss is in the luminance.

And smart render doesn't work if you apply ANY effect to the entire track or timeline.

"And smart render doesn't work if you apply ANY effect to the entire track or timeline. "

Are you speaking of avchd?? IF that's the case then it is untrue. Granted it's not perfect... it takes (my time lines anyway) roughly anywhere from 100 to 180 frames for the "no re-compress" to kick back in again after a transition/effect

That sounds fine except in the process you may create a non compliant file. Not saying this will cause problems down the track but it may.
Smart rendered HDV will not PTT. Smart rendered mpeg-2 causes issues with some DVD players. I don't know if this will cause problems with BD players or not.

Bob.

Re-read what I said. I am not talking about transitions. I said any effect which controls the ENTIRE track/timeline. Oh wait.. I think I see the trouble.

If you apply an effect that runs the entire length of your project, then you will not be able to smart render. If you add an effect to a single clip, or to just a transition, then it will smart render apart from that.

"If you apply an effect that runs the entire length of your project, then you will not be able to smart render. If you add an effect to a single clip, or to just a transition, then it will smart render apart from that."

In a nutshell... that's correct. The object of a smart render system is to "pass through" those parts of your time line that have not been changed or altered in any way. If you alter a single event then only that event will need to be uncompressed... each frame re-written... then re-compressed. The rest has not changed so there is no reason for it to be uncompressed/re-compressed... it just passes through (or essentially gets copied). If however you introduce some kind of change that affects your entire track.... well... the WHOLE thing will get uncompressed... each frame re-written... then re-compressed.

So "smart render" implies no actually smartness or optimizations per say, it's just smart enough not to do a "dumb render"?