A better Dolby Pro Logic II downmix

[Fr4nz]

Originally posted by sycho
no matter how it is encoded the Dolby Surround Prologic II soundtrack will sound bad in prologic just becase the way the matrix was designed, buying a profession encoder will just yeild better panning from front to back or vice versa, this tread was started to try to tell the people of doom9 about the inproper downmix values and now it has turned into a debait about playing Dolby Surround Prologic II soundtrack on a Dolby Surround Prologic system. It all comes down to this, if you are using Dolby Surround or Dolby Surround Prologic use the "surround" downmix and nothing but, if you are using Dolby Surround Prologic II use only the "surround2" downmix and nothing else.

Hey Sycho calm down, we were only discussing various aspects of DPL2...
Take a camomile ;)

[nuked]

ok... thnx frnz. That answers my question then. I am still trying to figure though frank what you keep meaning by this concept of sound "cancellation" you keep mentioning and implying in passing. Sound from speakers really doesn't cancel. Even mathematically it only cancles in certain places in the room and most of those places depend on frequency. In practice, with minor diference in speakers and reverb off walls and having 2 ears in diferenent locations and I don't what... but i know if you play sounds of oposite polarity out of 2 speakers they just don't cancel. Decoders may actively cancel the sound some to some extent. Is this what you mean? Anyway... not that important to the discussion, just I get the impression sometimes that people seem to get an idea that by downmixing the sound in oposite polarity you are somehow hiding it and making it inaudible or something, not that I'm saying you think that, but it's a myth worth refuting once and awhile on the record.

[nuked]

by the way.. when I hear that LFE can lock the steering to center it makes it harder to be overly impressesed with the engineering in these decoders. Seems like there should be a high pass filter on the input to the steering amps. Bass obviously doesn't need to be steered. Seems like a big oversight but then maybe there are technical difculties in doing that.. I can't imagine why though.

[Sycho]

if you have problems of the LFE locking the sound to the centre channel speaker than don't include it in the downmix, a real downmix to Dolby specs don't include a LFE

[nuked]

oh of course.. it should not be included. But that's silly. It's not just LFE though, any deep bass effects, and there's plenty even outside of LFE, don't really need to be interfering with the steering.

edit: oh yeah.. to say something on topic... I do think it's pretty cool that bleo did the work to find these new matrix values.

[frank]

As sycho said:...It all comes down to this, if you are using Dolby Surround or Dolby Surround Prologic use the "surround" downmix and nothing but, if you are using Dolby Surround Prologic II use only the "surround2" downmix and nothing else.
Discussion closed.

I encode DPL2 only, and... I will buy a new receiver, DPL2 built-in - at some time.

[nuked]

peace man... no need to be defensive. Just asking a question.. was this claim based in theroy or tests? Fr4nz answered it. Thanks.

edit: oh and thanks for pointing out that the discusion was finished... in case anyone missed my previous post where I thanked Fr4nz for his answer. Of course if anyone else wants to comment I think they are free too until a moderated actually closes the thread.

Cheers
-nuked

[Fr4nz]

Did I say something wrong?

Sycho seemed a little aggressive in his sentences so I asked him to calm down.

IMHO there's no need to close the thread, we were only discussing/improving our theories about DPL2 encoding, and this thread has to stay open.

[nuked]

Well maybe it was me that came off as aggressive since both frank and sycho seemed a little defensive. If so I apologize. I certainly wasn't meaning to attack anyone's credibility or anything like that... nor could I attack franks as I think his credibility is probably reasonably well established in this subject.

-nuked.

[Midas]

I'm about to release a new version of Azid; version 1.9.

One of the things I've done is to update the DPLII matrix, according to this thread.

1) What would be the most correct matrix if the input source only contains one surround channel, e.g. 3/1 in DPLII. Will it be correct to sum the surround channels into one like this:

Lt = L + 0.7071 C - 1.366 S
Rt = R + 0.7071 C + 1.366 S

because

1.366 S = 0.5 S + 0.866 S

It gives you very much relative engergy from the surround channel in the output... Can someone please confirm this?


2) Since I am reviewing the downmix matrixes for the 2/0 outputs, are there any comments/updates/error with the standard DPL matrix:

Lt = L + 0.7071 C - 0.7071 BL - 0.7071 BR
RT = R + 0.7071 C + 0.0701 BL + 0.7071 BR

Edit: Does the same DPL matrix forumla apply when there is only one surround channel input, as well? I.e. the formula would then be Lt=L+0.707C-1.414S and vice versa. It this considered correct by the community?


PS! While we are at it: Does anyone have any ideas for *enhancements* of azid (restricted to audio-related issues)? Any known bugs (except for the occational click/pop-bug, that have been fixed now)?

-Midas

[bleo]

Hi Midas!
Love your work! :D

1) I don't think anything sums arithmetically. All the coefficients are squareroots because we are working on power.

If the input is 3/1, there's probably no need to double up the surround channel. Otherwise, it will sound twice as loud right? My preference would be:

Lt = L + 0.7071 C - 0.7071 S
Rt = R + 0.7071 C + 0.7071 S

This applies to both DPLII and DPL. I think both decoders will send half of the surround channel to each of the rear speakers. These will then sum to give you sound at the same level as the input.

2) The DPL downmix matrix is fine :)

Is there any chance you could add 32 bit integer wav output? ;)

bleo

[Midas]

Yes, my point exactly. You cant arithmetically sum two coeffesients like this because the DPL (II) matrix is designed for two surround channels, not one, so this is a special case.

Thanks bleo.

The next version of azid will have 32-bit integers as requested.

[Fr4nz]

Great work midas :D

[Midas]

@bleo

While talking about downmixes... I've recently aquired myself a 5.1 soundcard.

Do you have a good algoritm for DPLII 2.0->3.2 and DPL 2.0->3.1 upmix? (I fear that upmix isnt as simple as downmix, because of the VCA plates and the filters.)


Midas

[bleo]

I suppose you could use a simple Hafler matrix, but that wouldn't have all the fancy active steering that Pro Logic does. (I don't even fully understand how that works :p)

Or try the Ambisonics method.

[Sycho]

Originally posted by frank
The channel separation of DPL2 cannot reach the quality of AC3 tracks because of their analogue characteristics.

Frank what do you mean because of their "analogue charateristics", analog will always be better than digital. Do you mean, because of there "encoding charateristics"

[Fr4nz]

Hi all!

I've made some samples in DS2 with Besweet GUI + Besweet 1.5.b25 + azid 1.9 + Lame 3.90.3 and played back on my new Kenwood Dolby Digital decoder which is also capable of DPL2 decoding. Well the differentiation between front/rear is VERY good, really impressive, BUT generally you have the impression that the volume on the right rear speaker is a bit stronger than the left rear speaker (you can notice this especially when you should hear the sound in the rear center). Maybe the DPL2 matrix in Azid 1.9 has to be perfectioned?

Thanks for any reply!

Fr4nz

EDIT: "-s surround2" and "-s dplii" seems to call the same downmix matrix. In fact they produce identical files :(

Another person noticed this fact: http://forum.doom9.org/showthread.php?s=&threadid=70281

EDIT2: This is the log:

BeSweet v1.5b25 by DSPguru.
--------------------------
Using azid.dll v1.9 (b922) by Midas (midas@egon.gyaloglo.hu).
Using Shibatch.dll v0.24 by Naoki Shibata & DSPguru (shibatch.sourceforge.net).
Using lame_enc.dll v1.32 (8/8/2003), Engine 3.90 <http://www.mp3dev.org/>.

Logging start : 02/13/04 , 14:14:32.

H:\dvd\azid\BeSweet.exe -core( -input h:\output.ac3 -output h:\output.mp3 -logfilea H:\dvd\azid\BeSweet.log ) -azid( -s surround2 -d 2/0 -L -3db ) -ota( -hybridgain ) -ssrc( --rate 44100 ) -lame( --alt-preset standard )

[00:00:00:000] +------- BeSweet -----
[00:00:00:000] | Input : h:\output.ac3
[00:00:00:000] | Output: h:\output.mp3
[00:00:00:000] | Floating-Point Process: Yes
[00:00:00:000] | PostGain normalize to : 0.97
[00:00:00:000] +-------- AZID -------
[00:00:00:000] | Input Channels Mode: 3/2, Bitrate: 384kbps
[00:00:00:000] | Output Stereo mode: Dolby surround 2 compatible
[00:00:00:000] | Total Gain: 7.000dB, Compression: None
[00:00:00:000] | LFE levels: To LR -3.0dB, To LFE 0.0dB
[00:00:00:000] | Center mix level: BSI
[00:00:00:000] | Surround mix level: BSI
[00:00:00:000] | Dialog normalization: -4dB
[00:00:00:000] | Rear channels filtering: No
[00:00:00:000] +------ Shibatch -----
[00:00:00:000] | Source Sample-Rate: 48.0KHz
[00:00:00:000] | Dest. Sample-Rate: 44.1KHz
[00:00:00:000] | Attenuation : 0.0db
[00:00:00:000] +-------- LAME -------
[00:00:00:000] | 'alt standard' preset is used
[00:00:00:000] +---------------------
[00:02:22:560] Gain of 3.0dB had been asserted to file.
[00:02:22:560] Conversion Completed !
[00:02:22:560] Actual Avg. Bitrate : 186kbps
[00:00:34:000] <-- Transcoding Duration

Logging ends : 02/13/04 , 14:15:06.

[Fr4nz]

Okay. After testing some AC3 samples converted in DPLII I found that there's some leakage from R to SR. So maybe on software decoders that matrix is OK, but on hardware decoders the effect is this. Please someone answer!!

Fr4nz

EDIT: JUST FOUND that the leakage is audible also with Cyberlink PowerDVD 5.0.

[bleo]

er... surprisingly I haven't fully tried out the new dpl2 in Azid 1.9 yet... guess I've joined the 5.1 bandwagon...

But I'll try and suggest some things:
- Maybe we do need to do 90 degree phase shifts of the surround channels.
- Perhaps there is some rounding error in the downmix matrix coefficients, especially since they're normalized, that is enough to throw the dpl2 steering off balance. This may even be compounded by applying gain to the output and compressing to mp3 (?)
- My initial testing was all done with AC3Filter, and the coefficients were not normalized. I found this easier than normalizing and then applying gain to the output.
- My tests only had one channel active at one time. Having multiple channels active at the same time adds to the complexity of dpl2 that I am still trying to understand :p

[Fr4nz]

Bleo I've uploaded a small AC3 sample from Creative. It's very good because you can hear single voices coming from front, front left, front right, rear left and rear right. Very good for testing DPL2 downmix matrices!

The link is this: http://users.libero.it/i3ltt/Cazzate%20varie/prologe.AC3

[frank]

Hi, friends!
Sorry, I cannot make some tests in the next time because I am ill since months. (My arms are ill - too much worked)

Only some words to the tests I did with BeSweet and Azid 1.8.
1) For measuring I generated 6 sinus wav's (CoolEdit) for every channel and encoded to DPL2 with BeSweet.
2) Decoding the stream with GraphEdit and the latest DirectShow filters of PowerDVD / WinDVD back to 6 separate channel wav's I measured the amplitudes.

So you can verify the DPL2 downmix exactly because the DirectShow filters are certified by Dolby.

[Fr4nz]

Ok Frank care about yourself and come back soon :) We'll wait for you! I hope Midas come back also to see if we can do something for the leakage of L into RL.

[Sycho]

a simple fix for a 90° would be to place a small delay in the surround channels before the fold-down. I think that 5~10ms will do it, 1/200th 1/100th of a second respectfully. Be aware that this WILL cause unstable fold-down at certain frequency's

[Fr4nz]

We could compare the current matrix vs. the fixed matrix with my sample test (and if you want with other sample tests), and see what's the best.

[3dsnar]

This thread addresses 90 deg. phase shift vs 180 deg
phase shift in creating DPLII downmixes.
http://forum.doom9.org/showthread.php?p=837968&posted=1#post837968
I hope this
is usefull.
Cheers, 3d