NEW CODEC! 128 kbps and less for 5.1 surround encoding for MPEG4 (Xvid/Divx) movies

[3dsnar]

Alright. Thanx for your ideas :)

[3dsnar]

Scharfis. I will try to change Aud-X HGHQ into DPLII compatible, during this weekend,
because I think that your idea is GREAT!
----
What weight do you suggest for LFE?
(After few thoughts, I know how I can also mix it in and separate it nearly perfectly in case of Aud-X decoding)

[Rockaria]

As you see, AC3filter has both DS/DPL and DPL II encoder(speaker set), ideally can be attached at the aud-x out pin(5.1ch PCM).

However, having these DPL encoder inside of the aud-x dsfilter would make a good comparison & option.

[bobcat56458]

I second the Dolby ProLogic II addition, also there have been posts here on this forum asking for a way to do this in real time. I’m speaking of having a stereo TV tuner card, or line input on the PC, output Dolby Prologic II to the speakers, this would be a great additional feature. I would also like to see a way to temporarily disable, or unregister, the audxdshow.ax filter\codec because with my P4 1.8GHz processor when I have Aud-x running, and play a uncompressed, or DV AVI video file it sometimes makes the video run not so smoothly. Also when I’m editing stereo audio\video I want to hear the sound output in stereo, not 5.1 surround. I like the 5.1 Built-in Pseudo Surround module for playback of my MP3 files though. I have an Audigy I soundcard connected to a 5.1 amp by way of its analogue inputs\outputs. Thanks for your continued development of this fine application\codec.

[3dsnar]

Bobcat,
thanx for this suggestion. This is exactly what I mean.
A realt time DPLII encoder (similarly to already embedded AC3 encoder).

[Rockaria]

This post is moved from other thread because of going too deep on the aud-x only issue. Any aud-x issue will be discussed here.

We seem to have got some clearer data than the white paper in terms of the codec mechanism :
1. FV : the core technology of the aud-x codec which decides the fidelity
2. mp3 : asynchronous storage of core & meta(FV) data, affects the core(base) quality

Supposing upper scenario as the base architecture, there might be some more inquiries like below :
ENCODING:
1) Obtaining energy and phase relations between channels for future reconstruction from the merged signal (performed by Aud-X). Such information is stored in a feature vector - FV - (for each frame such a vector is produced)
FV : a meta storage for energy & phase relations between the channels on each frame
What is the frame size(nSamples) ? depending on the quality criteria?
2) Merging all channels
Core-Stream : 1ch or 2ch to be encoded in the mp3 body
3) Encoding with LAME
Encodes core only...
4) Inserting the Aud-X feature vector to mp3 stream (there is a space which can be utilized for any kind of information)
Where exactly : header, body or trailer?
Using own mp3 file access routine?
How much space is available in the meta data area for Lame & mp3?
Also isn't it(meta space) dependant on the core codec?

DECODING:
1) Decoding mp3 frame to obtain base audio stream
2) Getting Aud-X FV and reproducing the complete 5.1
Is it possible for streaming purpose when you get the FV set after the stream(body)?
What if the meta data area is altered by any tag programs or truncations ?

My impression of the asynchronous stream access(core & meta) is that it has some potential problems.(streaming, truncation...)
Like WavPack's hybrid format, a seperate meta file(if not exists, it decodes as lossy) can be a good idea for the synchronous & safe access.

Meanwhile, the core blackbox(FV manipulation)'s spatial image abstraction(relative to the core?) into energy-phase sets sounds fantastic but surely is beyond my touch, and now agreed totally different from the phase-shift technology.

Many things I described could be wrong. But that is my best analysis based on the given data, requiring some clarifications, for whoever is interested in aud-x.

[Rockaria]

OK. This is not my area of expertise. But I roughly know the answers to your questions (frankly, my area of expertiese are all the phase and energy manipulations, which are not so interesting for you ;) )
Our software expert is not very easy to reach...until March.
-
But I will try to answer, based on what I know.
If you have a let say 128 kbps mp3 stream, it is possible to change it
into for example 192 kbps, by manipulating some header information. And this makes some room free for additional information (FVs in our case).
So available space is not a problem, since it is possible to control the average kbps available.
---
There are no problems with streaming. This is just a matter of a delay to get
the main audio and FVs related to that. In practice, when you reproduce Aud-X sound, it is in fact streamed from the disk to the DirectShow filter. And jumping from one point of the stream to the other is possible...

---

As far as I remember, the mechanizm of insterting the FVs is also used by mp3surround and was used earlier by mp3pro.
So actually knowing this, we started to investigate this issue.
I guess when you read carefully MPEG Layer3 Bitstream Syntax andDecoding.doc you can find all the answers to your questions there
(as our colleague did).
If you do not have this document, I can send it to you.
---
@3dsnar, you can delete the content from other thread.
Thanks for the clarification on the aud-x architecture.
Maybe the major success key seems to lie on the PV technology, looks like more suitted for the game 3d effect area, IMO.

[3dsnar]

OK. Deleted. I will report
DPLII related changes in our codec.
Thanks for all the valuable suggestions.

[3dsnar]

Aud-X v.1.1 Dolby Pro Logic II compatible is available for downloading.
Please read more in a separate thread devoted to the new release.

[bond]

@3dsnar: I have a idea for ultimative compatibility to standard MP3 (or other codecs) decoding:
- while Aud-X encoding downmix the 5.1 signal into two channels using a DPL2 matrix
- then build a 5.1 separation bitstream according to this downmixing.hm if compatiblity to mpeg is a goal, why not use the mpeg surround standard right away? :confused:

3dsnar, would it be hard to make your encoder compatible to this?

[3dsnar]

Well, now you can decode the stream by a DPLII decoder,
which is not ofcourse as good as a digital decoding methods.
--
But it is still better than stereo, when you use an external DVD player and DPLII decoder.

If you can use a PC with 5.1 speakers, or hook it up to an external
Dolby Digital decoder, it is better to use our PC based DirectShow decoder,
since it producess surround sound of much better quality than DPLII analog decoder.

===

Do you mean mp3surround?

- it is not DPLII compatible (AFAIK),
- it has poor channel separation quality, with significant phase errors.
- it is NOT FREE,
- It is not VirtualDubMod compatible,
- There is no directshow filter (to decode it) available
- It does not have transcoding and decoding tools available.

[bond]

Do you mean mp3surround?no, i dont mean mp3surround, which is imho just another proprieatary surround method

i mean the official surround standard mpeg released
http://codingtechnologies.com/products/mpgsrnd.htm
http://www.hydrogenaudio.org/forums/index.php?showtopic=36355
http://forum.doom9.org/showthread.php?p=699471#post699471

[3dsnar]

Aud-X is mp3 backwards compatible.
It can be used with AVI and is CBR.
With encoding bitrates similar to AAC 5.1
(i.e. 192, 128 and 80 kbps).
----
Now it is also DPLII compatible.
And this is useful for external home theatre equipment.
----
So as you can see it might be an interesting alternative,
to mpeg-4 5.1 format.

[bond]

so you cant/dont want to make your surround implementation compatible with the official mpeg surround standard?

why not?

[3dsnar]

I have nothing against the standard, and we would like
to see in the future if our spatial information encoding could be used with
mpeg-4, as an extention to Aud-X (so it will cover both mp3 and mp4).
-------
W tried to make something compatible to mp3 and AVI
due to their huge popularity. That is all.
And it turned out to be comparable (in terms of quality and bitrate)
to existing mpeg-4 based surround encoding technologies.

[Gabriel_Bouvigne]

Aud-X is mp3 backwards compatible.
It can be used with AVI and is CBR.
With encoding bitrates similar to AAC 5.1
(i.e. 192, 128 and 80 kbps).

I do not really understand how you can add ancillary data to mp3 and still have a cbr stream, without:
*modifying Lame's bit allocation
or
*waste a lot of space by setting the original mp3 cbr stream 1 cbr step higher.

As an example, PlusV was also inserting ancillary data into the mp3 stream from Lame. But to do this they had a frame repacker, and the result was abr, not cbr.

[bond]

I have nothing against the standard, and we would like
to see in the future if our spatial information encoding could be used with
mpeg-4, as an extention to Aud-X (so it will cover both mp3 and mp4).
-------
W tried to make something compatible to mp3 and AVI
due to their huge popularity. That is all.
And it turned out to be comparable (in terms of quality and bitrate)
to existing mpeg-4 based surround encoding technologies.hm, i think you are misunderstanding the "mpeg surround" standard! its not only for mpeg-4, it also works with mp3!

[3dsnar]

I understand that you are talking about Binaural Cue Coding?

[3dsnar]

OK. Yes, you are referring to BCC (based on which mpeg surround extention
has been created).
Well, I actually think that this approach has several limitations related to phase preservation. Therefore I carried out my own research on reconstructing the surround panorama (which resulted in Aud-X concept).
---

[Rockaria]

Well, if it is backward-compatible with existing core codecs, shouldn't the SAC standard also be forward-compatible(pluggable) with the future spatial codecs(with own ancillary & core data), as far as they meet the requirements?

The DPL II incorporation surely has some benefits than AC3 encoding-passthrough if the fidelity is not a big issue. But the bandwith & channels between the PC and external digital decoder shouldn't be of any constraint on DPL II & AC3 stream.
. The s/w AC3 encoding seems to have more than 500ms of delay, of course depending on the CPU, compared to the 32ms(soundstorm DICE) of h/w encoder, which asks for the a/v syncronization setting in the players or DSFilters.
. The external AC3 decoders are known to have some delay(as much as 1000ms) depending on the models.

As seen with existing established DSFilters such as FFDShow or AC3Filter, all the 'plus alpha' in the aud-x(plus lots of other DSPs) are already provided if it simply allows the filter chains(instead of the redundancy).
What is questioned and to be verified/implemented with aud-x(FV technology) seems to be the quality & standardization for general purpose audio codec use(not specific to movie or game use), not the 'plus alpha' at the moment, IMHO.:)

[3dsnar]

I also started thinking of a proper DPLII decoder in our DSfilter.
I.e. since it is unknow how the signal is in fact decoded, I will have
to be a bit creative... I know how to separate the center channel
in the frequency domain (by analyzing energy, phase relations
and differences between imaginary and real parts of the channels).
I also have a rough idea how to create a stereophonic subwoofer stream,
so
I think that I will include the decoder in our DSfilter within
next couple of weeks.

[tebasuna51]

I also started thinking of a proper DPLII decoder in our DSfilter.
...
I also have a rough idea how to create a stereophonic subwoofer stream
Stereophonic subwoofer stream?
I think a proper DPLII decoder must be a upmix to 5.0 (without LFE), because:
1) Dolby recommendations.
2) Don't mistake LFE channel with subwoofer.
3) Low frequencies don't need be directional.
4) The hardware multichannel audio player always redirect the low frequencies no matter in which channel are present.
5) This redirection must be a user choice. For instance, I have a 5.1 audio equipment but only 5.0 speakers, then I redirect all low frequencies to front speakers (with enough capability to play low frequencies). The normal choice is redirect to subwoofer.

Then, for a DPLII upmix, I think the LFE channel is unnecessary.

[3dsnar]

oops.
I am sorry. what I wrote is rubbish...
I meant stereophonic surround channels,
and I wrote subwoofer by mistake.

---
Regarding the LFE in a DPLII downmix.
Well, I decided to include it, because people seem
to like it. I also did some tests with and without it
with my DPLII standalone decoder, and I think that
downmixing the LFE sounds better...

But I guess this is a matter of taste.

[JnZ]

Hi 3dsnar,
your codec looks very well. So I decide to try it. I'm encoding now 137min long 6ch WAVe and waiting for result.

BTW: I found small error in cmd encoder. Instead percentage progress it shows only 1.#J% ;) This happend only with this long wavefile...

Thx for Aud-X.

[tebasuna51]

Regarding the LFE in a DPLII downmix.
Well, I decided to include it, because people seem
to like it.
I have not problem with include the LFE channel in DPLII downmix (disregarding Dolby recommendations), there are a info and can be preserved. Only in the upmix I think is unnecessary because the criterium to select the signal is only the frequency range, and this job is always maked by the hardware audio player.

[3dsnar]

JnZ:
Thank you :)
I hope you know that you can use our release of VirtualDubMod to
encode the complete movie with Aud-X (you can find the release of VDM
on our web site).
------------------
So you take the input movie with AC3 inside (AVI, VOB, etc) and directly
(using the VDM) you generate Xvid video (or whatever you use) with mixed in Aud-X surround
sound.

=============

Tebasuna51:
100% agreed

[JnZ]

My 137min Terminator 2 test WAV is OK. So give it to my frend to test it on 5.1 speakers (I have only stereo :( ).
Coding via VDM Aud-X is OK.

But today I test small M$ test 6ch WAVe, and results are very interesting.
(http://download.microsoft.com/download/winmediatech40/Utility/1.0/W98NT42KMeXP/EN-US/6channel.exe)
This WAVe must be converted to real WAVe, because it have strange ID.

Audio have little higher pitch, than original (test yourself). I test it on 128, and both 192kbps profiles. And subwoofer have less expressive.

[Rockaria]

I also started thinking of a proper DPLII decoder in our DSfilter.
I thought that the aud-x core(FV-technology) was a lot improved than the 10 year old technology as for the quality and bit efficiency?
IIRC, the DPL II requires the same amount of the bit rate on each channel but just volume-reduced(check the matrices) and shift-phase(+-90 degree, rear) merged in the two fronts. There's no abstraction layer to reduce the size, making it(encoding/decoding) straightforward and sounds with equal fidelity on each channels if decoded correctly(yeah I know center!).

Any recent: advanced Spatial Audio codecs are expected to sound(reconstruct) a lot better than DPL II, given the same amount of the bit rate in total with its bit-efficiency and improved algorithm. It made me believe the incorporation of the DPL II with aud-x should be at the end of the aud-x decoder(speaker set or mixer DSP) like other approaches. It's because of less restriction of the bit rate between the external decoder, also because not all the receivers have the mp3 decoder equipped.

Another established DSF FFDShow also has lots of useful mixer options including the DPL II which, I tested, seperates the channels with good quality when applied the correct DPL II matrices. Maybe aud-x is gonna include all the FFDShow features...

Also any combination is mathematically possible for each encoder & decoder set, but requires a lot of art...
The pair DPL II encode->FV-decode makes little sense as well as the pair FV-encode->DPL II-decode.
But the chain FV-encode->FV-decode->DPL II-encode(mpeg/PCM)->external DPL II-decode is expected to have some advantages, provided & proved the FV-technology is more efficient & effective than DPL II in the quality.

Good Luck.

[3dsnar]

JnZ, the wave is the older type wave format.
The pitch change is caused by the fact that originally it is 48000 Hz.

So you just changed the header and also the information of the sampling rate
was changed to 44100 Hz.

The pitch change also results in less expressive LFE reproduction.

---
You can play the wave (produced by the Aud-X frontend) with FFDshow for example.
Or maybe you can manage in the conversion that you performed
to preserve the 48 kHz sampling
frequency.

[3dsnar]

I thought that the aud-x core(FV-technology) was a lot improved than the 10 year old technology as for the quality and bit efficiency?
IIRC, the DPL II requires the same amount of the bit rate on each channel but just volume-reduced(check the matrices) and shift-phase(+-90 degree, rear) merged in the two fronts. There's no abstraction layer to reduce the size, making it(encoding/decoding) straightforward and sounds with equal fidelity on each channels if decoded correctly(yeah I know center!).

Yeah. I know all of this. I have read some interesting discussions on Doom9
regarding this subject :)
http://forum.doom9.org/showthread.php?s=&threadid=57988

So now HGHQ contains a proper DPLII information.
I did a test (connected my PC to my home theatre system via SPDIF).
So when I reproduce the HGHQ stream and route it via SPDIF with Aud-X DirectShow filter, I can hear surround produced by Aud-X. When I played the sound with DS incompatible player, the stereo PCM was sent via SPDiF which resulted in DPLII decoding (my amplituner switches to DPLII automatically if the incoming stream is PCM).
The surround was also decoded...but the difference is significant...
And it is true that DPLII is not much comparable to digital encoding techniques.


Any recent: advanced Spatial Audio codecs are expected to sound(reconstruct) a lot better than DPL II, given the same amount of the bit rate in total with its bit-efficiency and improved algorithm. It made me believe the incorporation of the DPL II with aud-x should be at the end of the aud-x decoder(speaker set or mixer DSP) like other approaches. It's because of less restriction of the bit rate between the external decoder, also because not all the receivers have the mp3 decoder equipped.

Well, in fact Aud-X does not use the DPLII at all. But since the downmixing
is a linear channels merging (which I've learned during discussions on this forum - I thought that DPLII downmixing has some nonlinear DSP operations) I realized that this should not affect the decoding concept, but the way the spatial information is gathered must be modified. So I did the modifications during this weekend and this resulted in the latest release of our codec.

Another established DSF FFDShow also has lots of useful mixer options including the DPL II which, I tested, seperates the channels with good quality when applied the correct DPL II matrices. Maybe aud-x is gonna include all the FFDShow features...

I tried to make Aud-X compatible with FFDshow. So you can decode
the DPLII stream to 6 channels and it can be passed through Aud-X
(if you have both codecs installed).
Morover, when you are playing a stereophonic file (lets say mp3) the signal is decoded by FFDshow and passed to Aud-X, where it can be converted with the pseudosurround module and sent via SPDIF to the external amplifier.
etc...

Also any combination is mathematically possible for each encoder & decoder set, but requires a lot of art...
The pair DPL II encode->FV-decode makes little sense as well as the pair FV-encode->DPL II-decode.
But the chain FV-encode->FV-decode->DPL II-encode(mpeg)->external DPL II-decode is expected to have some advantages, provided & proved the FV-technology is more efficient & effective than DPL II in the quality.

Yeah, as I mentioned the DPLII matrix downmixing is a separate story.
I did it only to make the HGHQ stream more flexible in terms of applications,
i.e. so you can play it on an external DVD connected to the home theatre amplifier and have some sort of surround (the one that DPLII offers ;) )
But if you can use a PC, a proper surround, resembling the original AC3 is reproduced with the Aud-X decoder. To see the difference, please perform a test which I described in the upper part of this post.

[Rockaria]

Let me point out some contraries.

1. Your aud-x involves too many (contrary) stuffs : What do you mean by aud-x exactly. I believe you must exclude everything else besides the core FV technology, the quality of the Spatial Audio Codec we are questioning.

So now HGHQ contains a proper DPLII information.
Well, in fact Aud-X does not use the DPLII at all

2. possibly you can attach the aud-x at the end of other decoders such as ffdshow.
But have you ever tried to attach the ffdshow at the end of the aud-x at all, to utilize other excellent & proven DSPs?

I tried to make Aud-X compatible with FFDshow. So you can decode
the DPLII stream to 6 channels and it can be passed through Aud-X
(if you have both codecs installed)....

3. the discussion is two years old and actually nothing new as for the +-90 degree phase shift. : some say delay others say analog encoder... looks like identifying a 3d object reflected on the photo, grayed.

But since the downmixing
is a linear channels merging (which I've learned during discussions on this forum - I thought that DPLII downmixing has some nonlinear DSP operations) I realized that this should not affect the decoding concept, ...

4. the DPL II decoder can actually upmix any streams on 2ch(forced) to create any types of illusion. what matters is the reconstruction quality onto the originally DPL II downmixed contents from 5.1ch. If you want the compatibility, you must not add any ancillary data for later decodings.
But if you want to because of the better decoding, you are actually creating a new spatial audio codec combining the DPL II & FV.

I realized that this should not affect the decoding concept, but the way the spatial information is gathered must be modified. So I did the modifications during this weekend and this resulted in the latest release of our codec.

I have no trouble in my setup to evaluate any new codec. The only trouble I have experienced is that it's moving continiously before grasping(evaluating) anything... So I'll just have to wait and see. Thanks. :)

[edit]
OK. thanks for disclosing your philosophy clearly, I'll respect that.

[3dsnar]

By Aud-X
I mean the compression technology based on reconstructing the full 5.1 panorama from the spatial information (stored in the Feature Vectors) and the merged base audio signal. The carrier of the audio main stream is mp3 audio, while the spatial info is stored in the ancillary data of the mp3 stream.

However our DirectShow decoder is much more than the Aud-X stream decoder.
It also contains resampling algorithm, PseudoSurround algorithm, AC3encoder and DPLII downmix module.

Indeed, you can attach Aud-X only at the end of the filter chain.
However, since you can pass-through any incoming signal, there is no problem with it. I.e. you can apply any DSP algorithm of FFDshow or other filter (e.g. AC3filter). And the resulting stream will be fed to Aud-X and passed through it without any changes to the rendering device.
So I think there is no conflict.
This was the most reasonable compromise that I could have come up with...
That is why it is done in such a way.

The ancillary data, where the Aud-X spatial information is stored, is not "visible" for a standard (including hardware) mp3 decoder. And there is no compatibility problem. Simply, when you play the Aud-X mp3 stream, the main audio data will be reproduced. In case of HGHQ the audio data is stereophonic and downmixed from original 5.1 in a DPLII manner.
When the accillary data is accessed by the Aud-X decoder, it is used to reconstruct the 5.1 stream. The way of downmixing, however, is unimportant (assuming that any nonlinear processes were not involved). In other words, the created downmix does not have to be DPLII compatible, because it is not relevant for the decoding algorithm...

The only significant change in terms of the Aud-X spatial encoding algorithm was done once. During the last weekend, to consider the DPLII downmixing weights and procedure :)
Other upgrades are related only to functionality of the DirectShow decoder
(they are not a part of the coding/decoding methodology).

[elenhil]

Another established DSF FFDShow also has lots of useful mixer options including the DPL II which, I tested, seperates the channels with good quality when applied the correct DPL II matrices.

It did not play a single DPL-encoded stream of mine. Were they all encoded with incorrect matrices? Who cares? PowerDVD's DS filter did decode them properly, so should have any other. Which disqualifies FFSShow as a good free DS DPL decoder. And PowerDVD's not freeware (not to mention that it is poorly suited for things like playing avi's with more than one audio stream)

[3dsnar]

It did not play a single DPL-encoded stream of mine. Were they all encoded with incorrect matrices? Who cares? PowerDVD's DS filter did decode them properly, so should have any other. Which disqualifies FFSShow as a good free DS DPL decoder. And PowerDVD's not freeware (not to mention that it is poorly suited for things like playing avi's with more than one audio stream)

OK. If you are trying to play Aud-X stream, this means that it disqualifies FFDshow.
However, it makes no sens to decode such stream by DPLII direct show decoder - i.e. it is much better to use Aud-X decoder to obtain the 5.1 signal from Aud-X encoded audio (it is superior to DPLII decoded multichannel signal).
---
Normally for all other streams (not Aud-X) FFDshow and Aud-X DSfilter are connected in a filter chain. And the outcoming signal from FFDshow can be passed through Aud-X DSfilter without causing any problems.
---
Also, you can disable Aud-X by chosing in FFDshow output settings
"direct sound" output.
================
So normally any mp3 (excluding Aud-X stream), AC3, AAC, etc. can be played with all other filters.
================
Indeed, multiple audio strams cause problems. This bug will be fixed ASAP.

[Rockaria]

Who cares? ... Which disqualifies FFSShow as a good free DS DPL decoder. ..Perhaps me neither.. And it could be either yourself or FFDShow.
Were you able to find any DPL (II) decoders in the FFDShow mixer options?
I could find ENCODERS(mixers) only. Actually there is a 'dolby decode' on the top of the DSP lines. Try that one to DECODE.

[elenhil]

OK. If you are trying to play Aud-X stream, this means that it disqualifies FFDshow.
I did not refer to Aud-X streams specifically. I meant regular DPLII software-encoded (using BeSweet, etc.) streams used in 2ch AC3 DVD backup.

Say, will the new Aud-X DS filter decode DPL in mp3s only?

[elenhil]

Perhaps me neither.. And it could be either yourself or FFDShow.
Were you able to find any DPL (II) decoders in the FFDShow mixer options?
I could find ENCODERS(mixers) only. Actually there is a 'dolby decode' on the top of the DSP lines. Try that one to DECODE.
That's what I did. No separation at all.

BTW, neither does AC3Filter (with DPL/DPLII output presets) decode DPL(II) streams.

[Rockaria]

No, you are misunderstanding.
AC3Filter does not have any DPL (II) decoders(explicitly), it only has encoders as speaker set to be decoded on your external receiver.
Likewise, FFDShow has these encoders as mixer options.

Also if you are using any valid DPL II decoder, it will produce 5.1ch as far as the input stream is valid 2ch stereo(whether DPL II encoded or not), unless the input stream is messed up.

[elenhil]

No, you are misunderstanding.
AC3Filter does not have any DPL (II) decoders(explicitly), it only has encoders as speaker set to be decoded on your external receiver.
Likewise, FFDShow has these encoders as mixer options.

I see. Thanks. So, I'd like to see a universal software DPL _decoder_ capable of processing other formats besides ac3 or mp3 (it is OGG Vorbis that I have in mind, of course).

Also if you are using any valid DPL II decoder, it will produce 5.1ch as far as the input stream is valid 2ch stereo(whether DPL II encoded or not), unless the input stream is messed up.
But without proper positioning, of course.

[Rockaria]

That's what I said by illusion.:)

You can try the ffdshow 'dolby decode' dsp.
Any formats(mp3,aac,vorbis...) decoded to 2ch PCM will be DPL(II) decoded(or upmixed) again if it is checked. not sure if it does DPL or DPL II though.

[elenhil]

You can try the ffdshow 'dolby decode' dsp.
Any formats(mp3,aac,vorbis...) decoded to 2ch PCM will be DPL(II) decoded(or upmixed) again if it is checked. not sure if it does DPL or DPL II though.
I have tried activating 'dolby decode' in ffdshow DS audio filter. Nothing even close to correct DPL decoding happened. Are you omitting some other crucial functions I must turn on to make ffdshow decode DPL?

[Rockaria]

Are you omitting some other crucial functions I must turn on to make ffdshow decode DPL?Instead, I guess you must turn off the mixer DSP(probably DPL II turned on?). What a side effect!:cool:

[Amnon82]

http://img443.imageshack.us/img443/3010/untitled39jr.png

Download XviD 5.1 Aud-X 128 Sample (19.44 MB | 2.28 Min) (http://www.megaupload.com/de/?d=XZB5X0N4)
Download Aud-X-Codec (http://www.aud-x.com)

Setup your system like this:

http://img442.imageshack.us/img442/7044/untitled45jv.png

[elenhil]

Instead, I guess you must turn off the mixer DSP(probably DPL II turned on?). What a side effect!:cool:
It was turned off.

[Rockaria]

I tested it again.;)
Uncheck the mixer itself, where is only(I believe) place to downmix the decoded(upmixed) multichannel by 'dolby decode'.

[3dsnar]

Amnon82, thanx for your example and using Aud-X
:thanks:

[Rockaria]

Finally, peace(well..). Must have been improved recently especially on mp3 HGHQ decoding.(I haven't done much of the quality test on the latest version)

I've not encoded the 128k mp3 myself yet but the movie sounds O.K.
But the SPBQ mp3(encoded before) sounded weird and aud-x decoded wav(also before) still has crosstalks on analog(disappeared by invert phase also).
Definitely need a total reencoding/test job but my impression is O.K.(busy...)

Now when the two-way compatibility(between the dsfilters) is ever enabled, I think I can perform some personal comparison tests on mp3 & m4a at the same decoding condition(DRC, DPL II, volume...).

[Summary] based on aud-x dsf & standalone compared to the original 6ch wav
STDQ.mp3 : artifacts
HGHQ.mp3 : clean, thin tone, wider seperation(artificial?), no crosstalk
SPBQ.mp3 : thicker than HGHQ but still thin tone, unbalanced, no crosstalk
HGHQ.wav, SPBQ.wav : all crosstalks on analog, removed by invert phase.
Done.

[3dsnar]

Rockaria, thanx for all your time and tests
:thanks:

Interestingly, IMHO SPBQ has better quality than HGHQ.

BTW. Artifact may be caused if the input channel order is incorrect in the 6 channel PCM.
(please check that out).

I do not know if you noticed, but the crosstalks happens mostly to phase-chaotic components (noisy components), while tonal ones are separated well. This results in preserving the spatial feeling, when listened to the decoded sound in 5.1 speaker environment environment (due to preserved phase of the tonal components). In the multichannel invironment it is (shoule be) impossible to hear the crosstalks.
I think subjective tests with headphones are not to much suited...for the purpose of 5.1 encoders evaluation ;)

[Rockaria]

fact is fact. If you want to see only what you'd like to see..why you listen.

What I've noticed is that aud-x only has made the crosstalks so far on this headphone.

Thanks for providing all this experiments.

[Sagittaire]

The core bitrates that we use
1) STRQ 80 kbps -> base mp3 is 56 kbps mono (24 kbps for spatial info)
2) STDQ 128 kbps -> base mp3 is 64 kbps mono (64 kbps for spatial info)
3) HGHQ 192 kbps -> base mp3 is 128 kbps stereo (64 kbps for spatial info)
4) SPBQ 192 kbps -> base mp3 is mono 64 kbps (128 kbps for spatial info)

Perhaps that VBR quality mode could be better ... if possible

VBR mode with lame
Quality mode VN -> base mp3 is VN lame join stereo (VN too for spatial info)
and
Quality mode VN -> base mp3 is VN lame mono (VN too for spatial info)