TMPGenc rate control question

[MJH]

Hi there!


:) Disclaimer: I've looked through the guides and did a forum search, but I didn't find much user-feedback (as opposed to the program manual's advice) to the following question:

:confused: Question: If you're going to move video from DVD to DVD-R but are *not* particularly concerned with file size (within reason), what is the best option + option settings for rate control?

Most every discussion of rate control understandably focuses on adjusting bit-rates to keep the size within the DVD-R capacity. However, I'm not moving an entire film, but only a scene or two, so I've got much more space available.

The best option *seems* to be the CQ-constant quality option (and so says the TMPGenc manual), but I'm wondering what settings have been proven best to veteran users. I've read in one post that the b-spoilage needs to be kept within 0-5 rather than the default 20 (may not be true).

Anyway, thanks for your help in advance, and my apologies if this issue has been addressed to death. I'm conducting my own experiments on this, but I wouldn't say 'no' to the sage advice of you TMPGenc experts. :D

MJH

[MJH]

Found what seems to be a very good guide for this sort of thing:

http://dvd-hq.info/Compression.html

MJH

[r6d2]

That's a good guide indeed. Anyway, if you're really not concerned with space, disable B pictures, use closed GOPs and go for CQ 100 with spoilage -100 for P pictures, and High Quality (Slow) motion detection.

According to the TMPGEnc manual, and some tests I've done, this will give you the highest quality and biggest BR needed for each frame, provided you are a millionaire in space.

[MJH]

Thanks r6d2!

I followed your advice and the image quality is excellent! However I do have a follow-up question/issue:

In the constant quality settings, what should I specify as being the max bitrate? The guide above suggests a range between 6000 and 8000, but I really have no idea myself.

I set the max rate at 6000 and used the *highest* quality setting under motion search detection, and the only drawback to the video is a slight blurring when people move (more so than you regularly see on a DVD). Could this be the result of a too-low or too-high max bitrate, my choosing the 'highest quality' setting over the 'high quality' one for motion detection, or something else entirely? Is interlacing / non-interlacing / 3:2 pulldown an issue here?

I realize that the final product will not be as perfect as the DVD from which it was ripped, but I'm trying to get as close as possible under the circumstances detailed in my post above (e.g. only select material is ripped, so I have lots of space available).

Thanks!

MJH

[r6d2]

Originally posted by MJH
In the constant quality settings, what should I specify as being the max bitrate? The guide above suggests a range between 6000 and 8000, but I really have no idea myself.
You should specify the highest number hich will remain between specs. i.e., 9800-Audio for a DVD.
I [...] used the *highest* quality setting under motion search detection, and the only drawback to the video is a slight blurring when people move (more so than you regularly see on a DVD).
I could not tell you the reason whithout looking at it myself. But my advice is to use High (slow) and not Highest (slowest) as motion estimation, since it has been reported that sometimes TMPGEnc may find motion were there is none, resulting in actually lower quality at certain spots.

If you do this, I guess it will be as good as it gets.

With regards to interlacing, that's a whole different topic, but generally if interlacing is a problem it shows up very clearly. Try the guide below to check your source if in doubt:

http://www.doom9.org/ivtc-tut.htm

[dvd_maniac]

Why use ANY compression on the video if you have so much space?
If you plan is to put it onto a dvd-r why not just rip the source dvd using stream proccessing or only rip the chapter and/or cells you want? You could go even farther with tempgenc mpeg tools to cut out any unwanted scenes from that chapter>>>

[r6d2]

Originally posted by dvd_maniac
Why use ANY compression on the video if you have so much space?
@dvd_maniac, you have a point here. :)

Why didn't I think of that?

[dvd_maniac]

I do this all the time. I overthink a problem so much that I believe the solution must be a difficult one, when the easiest way is sometimes the best way.

[r6d2]

Originally posted by dvd_maniac
I do this all the time. I overthink a problem so much that I believe the solution must be a difficult one, when the easiest way is sometimes the best way.
The old, plain and simple KISS concept. :D

[dvd_maniac]

Don't think I know the KISS concept.
What is it?

[r6d2]

Originally posted by dvd_maniac
Don't think I know the KISS concept.
What is it?
Use the search button :)

Keep It Simple, Stupid.

[dvd_maniac]

Keep It Simple, Stupid.

yep, that's me alright

[chet]

I have a related question... I am quite new to the whole dvd->svcd game.

I'm using dvd2svcd with tmpgenc, and cannot work out what "cq test values" are! I've searched high and low and can't find a simple answer to this.

[r6d2]

Originally posted by chet
[B]I have a related question... I am quite new to the whole dvd->svcd game.
First, it's not a game... It's serious stuff! ;)

I'm using dvd2svcd with tmpgenc, and cannot work out what "cq test values" are! I've searched high and low and can't find a simple answer to this.
I guess you mean the values DVD2SVCD uses sequentially to determine the final CQ value...

Well, as CQ mode does not allow to constrain the size of the encode, you must find a CQ value for evry particular movie. This is done internally but DVD2SVCD, and you need not worry about it.

[chet]

as CQ mode does not allow to constrain the size of the encode
huh?

but anyway...no, R2D2, I do not really need to know.

I tried CQ overnight, and on an AMD 1700 it's taking waaaaaaaaay too long (i.e. more than 8 hours), and that's not including ripping the VOBs from the DVD.

I think I'm just going to try 2-pass VBR.

[r6d2]

Originally posted by chet
but anyway...no, R2D2, I do not really need to know.

I tried CQ overnight, and on an AMD 1700 it's taking waaaaaaaaay too long (i.e. more than 8 hours), and that's not including ripping the VOBs from the DVD.

I think I'm just going to try 2-pass VBR.
First, it's r6d2, not r2d2 ;)

Second, encoding speed in CQ is not slower than in VBR.

Third, 2 pass VBR will take twice the time CQ does (it's 2-pass, remember :)). That's 16 hours.

The prediction phase needed for CQ should not take very long, and is quite good.

Something smells bad in your PC. AMD 1700 should be in the 1x speed range with enough RAM (enough=256).

Check your IDE bus to see if DMA is enabled (search for a thread contaning "dmafix.exe" for details).

[chet]

re: r6d2, ooops I'm sorry.

thanks, found the thread, but the link to the file is down!

[r6d2]

Originally posted by chet
re: r6d2, ooops I'm sorry.

thanks, found the thread, but the link to the file is down!
Use the manual procedure. But first check if your IDE channels are using PIO or DMA.

[Kika]

Third, 2 pass VBR will take twice the time CQ does (it's 2-pass, remember ). That's 16 hours.

That's wrong. If caching is activated, the second pass can be much faster than the first one.

[r6d2]

Originally posted by Kika
That's wrong. If caching is activated, the second pass can be much faster than the first one.
@Kika, in your experience, how much is "much faster"?

[Kika]

Depands on the Video (and if you use Filters ans so on) how faster the second Pass works.
On my Machine (P4 1.9) and while Encoding to Half-D1 without any Filtering or Resizing, the second pass runs in Realtime, while the First pass needs double the Time of the Video-Lenght.

I think, pass Two is 33 to 50 % faster than pass one.

[r6d2]

Originally posted by Kika
On my Machine (P4 1.9) and while Encoding to Half-D1 without any Filtering or Resizing,
Half-D1 without resizing? :confused:
the second pass runs in Realtime, while the First pass needs double the Time of the Video-Lenght.
I think the difference from first to second pass may actually depend more on the motion estimation method you choose than on the filtering you made. Would you agree?

[Kika]

Half-D1 without resizing?

Yes, i use Half-D1 for my TV-Captures and i'm doing all filterings and resizings during the capture. So i can encode the Video directly without any other Filter or Frameserver.


I think the difference from first to second pass may actually depend more on the motion estimation method you choose than on the filtering you made.


No, TMPGEnc uses a caching-system for the results of the First pass. So the encoder must not do all the Work again. The second pass is always a lot faster than the first - except you are using heavy filtering whitch breaks down the Encoder-Speed.

[chet]

Use the manual procedure

how? I really am not so sure how to go about doing this.

[r6d2]

You have to delete the MasterIdCheckSum key from the following branch of the registry:

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Class\{4D36E96A-E325-11CE-BFC1-08002BE10318}

and then reboot. But see first if you channels are in PIO mode.

Read the thread, the detailed procedure is outlined there.