PAL pulldown

[Jan Marijniszoon]

If you convert a Blu-Ray to a PAL DVD, you have to speed up the audio and speed up the frame rate to 25 fps.

Now I was wondering, why not pulldown a 23.976 fps source to 25 fps (with 50 fields by pulldown).

DGPulldown is able to do this, although this is not standard offcourse.

Can someone explain to me why this is never used? Why do people always choose the speedup-method?

[2Bdecided]

It is sometimes used (commercially, and by people here), but the hic-ups in movement due to the repeated frames or fields (1 repeated frame per second is more common than 2 separate repeated fields - at least in commercial use) are objectionable. Smooth movement + 4% speed up is generally preferred.

Cheers,
David.

[Jan Marijniszoon]

Yeah but I just do not understand...

3:2 pulldown adds 6 frames via fields now don't it?

Then why should it be a problem to add 1 frame per second with 2 fields then? This way you can benefit from the higher resolution (720 x 576) and you can benefit from the unaltered audio track.

[Guest]

The judder is just more visible because it is not nicely spread out.

3:2 adds 2 fields for every 8 fields.

[Ghitulescu]

Yeah but I just do not understand...

3:2 pulldown adds 6 frames via fields now don't it?

Then why should it be a problem to add 1 frame per second with 2 fields then? This way you can benefit from the higher resolution (720 x 576) and you can benefit from the unaltered audio track.

The audio track speed-up I think is now a matter of history. There are software (and hardware - almost every DJ stuff has this, it's called MasterTempo) that compress the time and keeps the pitch.

[Jan Marijniszoon]

The judder is just more visible because it is not nicely spread out.

3:2 adds 2 fields for every 8 fields.

Thanks, now I understand!

[Jan Marijniszoon]

The audio track speed-up I think is now a matter of history. There are software (and hardware - almost every DJ stuff has this, it's called MasterTempo) that compress the time and keeps the pitch.

And how does eac3to do the job according to your opinion?

[Rumbah]

I guess eac3to just resamples the audio and therefore changes the pitch.

You can do a better change with Avisynth and
TimeStretch(tempo = (100.0*25.0*1001.0)/24000.0).

[Ulf]

I guess eac3to just resamples the audio and therefore changes the pitch.

That is my finding as well.

For changing the frame rate of the video (while preserving the pitch of the audio) from 23.976 or 24 fps to 25 fps I use the following Avisynth code:

freq = Audiorate()
TimeStretch(pitch = (100.0*Framerate())/25.0)
AssumeFPS(25,true)
SSRC(freq)

[Rumbah]

But don't forget
ColorMatrix(mode="Rec.709->Rec.601", threads=0)
if you aim for SD resolutions.

[2Bdecided]

I didn't realise TimeStretch in AVIsynth (from the SoundTouch library) was that good - but it is.

It's not perfect, but I think it's the lesser to two (three!) evils.

Cheers,
David.

[Jan Marijniszoon]

But don't forget
ColorMatrix(mode="Rec.709->Rec.601", threads=0)
if you aim for SD resolutions.

Is it better to apply this before downscaling or after?

I also asked this question in this thread: http://forum.doom9.org/showthread.php?p=1296069#post1296069
maybe it is better to answer it there

[Jan Marijniszoon]

That is my finding as well.

For changing the frame rate of the video (while preserving the pitch of the audio) from 23.976 or 24 fps to 25 fps I use the following Avisynth code:

freq = Audiorate()
TimeStretch(pitch = (100.0*Framerate())/25.0)
AssumeFPS(25,true)
SSRC(freq)


Thanks for this tip!

[Jan Marijniszoon]

What is the difference between the following codes:

TimeStretch(tempo = (100.0*25.0*1001.0)/24000.0)


and

freq = Audiorate()
TimeStretch(pitch = (100.0*Framerate())/25.0)
AssumeFPS(25,true)
SSRC(freq)

[Jan Marijniszoon]

In the meantime I tried this methode:

freq = Audiorate()
TimeStretch(pitch = (100.0*Framerate())/25.0)
AssumeFPS(25,true)
SSRC(freq)

The waveforms look pratically the same as the result from eac3to. And eac3to uses SSRC as well. So the question here is, is this really a better method than eac3to uses?

[Ulf]

What is the difference between the following codes:

TimeStretch(tempo = (100.0*25.0*1001.0)/24000.0)


and

freq = Audiorate()
TimeStretch(pitch = (100.0*Framerate())/25.0)
AssumeFPS(25,true)
SSRC(freq)

The first piece of code you use if you only import the audio in Avisynth.

The second one you use if you have imported both video and audio. TimeStretch(pitch = (100.0*Framerate())/25.0) decreases the pitch, while AssumeFPS(25,true) increases the pitch to the original value (and speeds up the video to 25 fps).

The end result for the audio is the same in both cases.

[Ulf]

In the meantime I tried this methode:

freq = Audiorate()
TimeStretch(pitch = (100.0*Framerate())/25.0)
AssumeFPS(25,true)
SSRC(freq)

The waveforms look pratically the same as the result from eac3to. And eac3to uses SSRC as well. So the question here is, is this really a better method than eac3to uses?

Yes, it is - it preserves the pitch. SSRC only changes the sample rate.

[mp3dom]

The majority of speedups (from 24 to 25) tends to change the pitch (only a bit of commercial titles have the pitch preserved). Timestretch is not perfect and is a bit risky with a 5.1 soundtrack where all the channels need a correct phase to allow a correct downmix to stereo. Resampling doesn't introduce these kind of errors and is, quality-wise, better. Also the timestretch algorythm depends on what source you have. There are good algorythm for voice and algorythm for polyphonic music, algorythms that doesn't introduce artefacts but doesn't keep the phase and/or the length and so on. In a film you have both voice and music so to have good results you need to apply the polyphonic algorythm to M/E and voice algorythm to the dubbing... keeping the correct phase (and this means that you need to edit often the channels in your DAW with high costs). If you apply the voice algorythm to all the 5.1 mix you'll end with drops or hiccup in the music; viceversa if you apply a polyphonic algorythm to a 5.1 mix you'll have voice with echoes or strong aliasing.
Good software like Prosoniq and others are not immune to these problems so at the end is a matter of taste: if you have a good sound system you'll hear the differences comparing the original audio to the speedup audio with preserved pitch.
Please consider also that the timestretch will fix the pitch but the film length will always be shorter by 4% compared to the original and sometimes directors (and often final users) doesn't like this.

[Ghitulescu]

Please consider also that the timestretch will fix the pitch but the film length will always be shorter by 4% compared to the original and sometimes directors (and often final users) doesn't like this.

Hopefully HDTV (24p) will put an end to this issue ;)

[Jan Marijniszoon]

I was wondering...

True film is 24 fps, but Blu-Ray still releases the films as 23.976.
Now do they 'alter' the audio source for this too? Wouldn't that be a shame?...why not switch over to 24 fps altogether?

[Rumbah]

I never tried to timestretch the original 5.1 sound, I just convert to stereo as it only gets played in a stereo environment (cheap dvd player and sound over TV). When I watch the movie, I use the original disc and full surround.

[mp3dom]

I was wondering...

True film is 24 fps, but Blu-Ray still releases the films as 23.976.
Now do they 'alter' the audio source for this too? Wouldn't that be a shame?...why not switch over to 24 fps altogether?

They slow down the film a little without timestretch (for the reason I've already explained). In this case the difference in pitch is inaudible since the slowdown is very low.
They are at 23.976 for practical (and economical) reasons. At 23.976 they can downscale to NTSC standard def. (23.976+3:2 pulldown) to obtain a 29.97fps without editing (they can made the bluray and the ntsc dvd at the same time. The masters are in HDCAM-SR and the HDCAM-SR VTR can output HD and downscaled SD at the same time). Also this is useful for international releases since the "old" dubbings can be used without editing the audio or the video.

[Jan Marijniszoon]

Thanks for the explanation so far...

But I still don't fully understand. How do they exactly 'alter' the audio from 24 to 23.976?
I mean; towards the end of the movie this would otherwise create an async of about 7 seconds. Do they resample it a bit?
This would still mean that they 'hurt' the original audiotrack from the cinema source.

[mp3dom]

They resample a little the audio track. The difference is anyway unnoticeable since with this kind of slowdown your movie will be shorter by some seconds.

[Ghitulescu]

Thanks for the explanation so far...

But I still don't fully understand. How do they exactly 'alter' the audio from 24 to 23.976?
I mean; towards the end of the movie this would otherwise create an async of about 7 seconds. Do they resample it a bit?
This would still mean that they 'hurt' the original audiotrack from the cinema source.

If people don't feel the difference from original WAVe to MP3 then they won't be hurt by resampling. ;)
On the other hand, very very few people really notice resampling artefacts (like those 48->44.1 or viceversa). In 96/192 world things are much simpler.
And every soundtrack is processed. What counts is the final product. If you like it ok, if not who cares? They won't do this for you, unless it happens that you own MGM or Columbia ;)

[Mug Funky]

additional to that, usually separate 23.976, 25 or 24 audio masters are output directly from the sound mix house, so there's no "tampering" with the masters, rather the mix is re-rendered from the source at different rates, using all the advantages of whatever nice resampling gear they have at their disposal. this also means that you can assume these mixes are correct and represent the intentions of the film makers.

(BTW, HDCAM-SR decks allegedly can do pitch correction, but in reality the results aren't really acceptable and it needs to be done in something a little nicer. also those decks are so stupidly laid out that who knows what they can and can't do, even with the manual at hand?)