Bitrate drops when copying MPEG edits?

[Graham Jones]

I'm using KDDI's MPES PRO which was bundled with my JVC GR-PD1 camera.

I've tried other NLEs and even though they offer more in terms of options and flexibility - none are as stable.

I'm quite happy with MPES PRO as it suits my needs.

I have just one question!

I have learnt that it's better to keep timeline short, render, then dump rendered portion into new timeline - it's a way to avoid crashing and doesn't bother me as it's the way I work naturally anyway.

My question is: does this rendering - and potential re-rendering depending on how a given editing project develops - mean I am tampering with the picture quality?

I mean, hypothetically, if I were to render and re-render a scene say 6 times, in the course of editing the whole project, as I am moving stuff around and whatnot... is there a risk to the MPEG 2 quality?

As a test, I rendered a project and placed it back in the timeline, changed it slightly and re-rendered - I did this 6 times. After I created my 6th render I examined the quality.

I didn't notice any degradation on my 21" Ultrascan - I'm also using a 128 MB Radeon card and generally get nice picture...

However when I checked the 'property' of the file that had been rendered 6 times it's bitrate was different than the bitrate of normal, uncut shots in bin.

UNCUT, NORMAL SHOTS IN BIN:
* Stream Type : MPEG-2 TS
Bitrate : 19750973 bps


PORTION RENDERED 6 TIMES
* Stream Type : MPEG-2 TS
Bitrate : 19081265 bps


I'm not technical and for all I know this bitrate difference may be irrelevant to pic quality.

I know this might sound a bit finicky - but I am simply trying to establish whether rendering and rendering and re-rendering in the context of this NATIVE MPEG2 ONLINE NLE is a quality consideration or not a quality consideration at all...

Any help would be appreciated.

I hope it's okay that I am also posting this in HDV editing solution section - as I guess the users of the HD1 and HD10 may be most familiar with MPES PRO. Just thought it should go here too?

[Rob Lohman]

Since MPEG(2) and DV as well use a lossy compression you indeed
have a generational loss. The more times footage gets rendered
the worse it will get. The first couple of generations will be very
hard if not impossible to see with good quality software. At a
certain point it will start to get noticable compared to the first
generation footage.

[Graham Jones]

I have ascertained the bitrate has nothing to do with it...

thanks for that info rob... certainly at 6 generations there is no discernible difference, so I guess I'm safe as I only need it to surive 1 or 2 generations.. and it's a 'native' mpeg2 nle, so no processing unless I do effects which I don't need to do

[Rob Lohman]

Keep in mind that editing with MPEG2 more easily can result in
recompression than for example DV or MJPEG compressed
footage. Why? Because DV/MJPEG are frame compression codecs
and not intra-frame compression like MPEG2.

To give an over simplified example: suppose you have two 5
frame MPEG2 files with a 5 frame GOP structure where the first
frame is the full frame (F) and the following 4 frames are only
the changes (C); you get:

F C C C C (and that twice)

Now assume we want to paste these two simply together we
have no problems:

F1 C1 C1 C1 C1 F2 C2 C2 C2 C2

But, what if we want to have the second clip begin after just 2
frames of the first clip, you would get:

F1 C1 F2 C2 C2 C2 C2

Now this might work *if* you can switch to a variable GOP MPEG2
structure and the DVD format does support this. The question is
does the NLE. It would be even worse if you even decide to trim
the first couple of frames of the second file:

F1 C1 C2 C2 C2

Changes stored for the second file would be corrupt, so the
first would be re-encoded as:

F C C C C F C

and the second one as:

F C C C C

Notice how in both cases only the first two frames are original,
the rest of the frames already have a generational loss for being
re-encoded if the NLE wants to stick with one GOP structure.

p.s. most NLE's that support editing MPEG2/HDV simply re-encode
the whole thing (render) when exporting creating an extra
generation. Seldom are they true MPEG editors.

The editor you are using Graham is a true MPEG editor and it
should generate variable sized GOP structures to minimize
re-rendering and generational loss. Which besides the strange
GOP structures IS good news.

[Graham Jones]

yeah, I understand a few frames around cuts are re-created when neccessary

thanks for that break down, it makes it clearer