Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > General > Audio encoding

Closed Thread
 
Thread Tools Search this Thread Display Modes
Old 28th May 2004, 14:38   #501  |  Link
Eye of Horus
Banned
 
Eye of Horus's Avatar
 
Join Date: Jun 2002
Location: The Netherlands
Posts: 326
Quote:
Originally posted by ursamtl
[B]Oops, sorry. I posted just the text without using the http:// button on the forum interface because I didn't want to repost a link to a file that might potentially be corrupted. I didn't notice that the forum software automatically converted it to a link.
No need to apologize : perhaps it gives some extra visitors :-)


Quote:
Well, did you try the dipole without moving your speakers? I found the results were better with my speakers at +-25°. At +-10° there was very little sense of width to the sound at all.
Yep ! I try out every bidule :-)

Quote:

I look forward to the new bidules. As for the 32 bit conversion, yes, I have done this a couple of time for comparison sake and it does provide a marginally cleaner sound especially in high-frequency details such as cymbals, etc., but I didn't notice any appreciable difference in terms of the Ambisonic effect. I'm lucky in that, for one thing, I can monitor my mixes in real time on my PC so I don't always burn DTS CDs. In addition, my DVD player reads CDRWs so when I do write burn DTS CDs, they're temporary tests to see how the sound is transferring to my 5.1 system. Once I settle upon an upmix method that gives me the results I seek, I'll do some proper full CDs and start with 32-bit files. I still haven't found a satisfactory answer concerning the whole question of dithering. It's generally agreed that going from 32-bit to 16-bit with no dithering reduces sound quality, but so far the Surcode documentation hasn't mentioned whether it downsamples to 16-bit and if it does so, does it use dithering. All it says it that the program accepts 16- or 32-bit files for input. The CD redbook spec is 16-bit, but it's not clear that a DVD player recognizing a CD but playing a DD or DTS wave file does so only if the file is 16-bit.

By the way, yes, your earlier post did clear things up between us. I'm at work and don't have time for a detailed reply, but yes, there are surely cultural nuances to our communication that can cause misunderstandings. For example, I once worked in an office with some Swiss and German personnel who were working on a contract for the Canadian govt. here. They would get into discussions that seemed to me like they were ready to haul out weapons and kill each other, yet walk out of the office at the end of the day smiling at each other! Here I was the poor Canadian fellow shocked into thinking I was about to witness murder! These folks told me this was simply their European temperament and nothing I should be concerned about. Perhaps we Canucks are just too polite! Plus, having worked with them for awhile I found I was adopting some of their attitude.

Have a great weekend!
Ursa

LOL !!! How recognizable !!!

Even some of my friends in the USA (for over 8 years now) still have problems in a discussion with me :-) They always think I'm angry or too sharp.

Have a good (productive !) weekend too !

kind regards,

EoH
Eye of Horus is offline  
Old 28th May 2004, 15:19   #502  |  Link
kempfand
Registered User
 
kempfand's Avatar
 
Join Date: Oct 2001
Location: Switzerland
Posts: 225
Ursa: Regarding the Stereo Dipole method I was posting here, I wrote it is a prototype, as the 2 "only" speakers for the Ambisonics (placed in the back) are probably not enough to make it a full AmbioPhonics by the books. Do do things correctly, there are at least 4 speakers (square) or even more, as well as convolution with real space impulse responses. Also, the dipole spreads the image in the front onyl (ideally over 180 deg), but there are methods which use a 2nd dipole in the back.

I have also read from two sources (Farina being one of them) that a frontal dipole works best if it comes from above (i.e. two closely spaced speakers at the ceiling to produce a dipole). I can't really report any personal experience on this, as my other half would probably get a nervous breakdown if I changed the living room too much ... ... but maybe one day ...

On the 32 bit conversion of the feeding Stereo-in, I quote here a Farina post from April 2002 from SurSound, which you/all might find interesting.
Quote:
Now CEP2 instead defaults to the most usual mode.
In any case, being a float number, no dither is required (nor any special trick) when rescaling. Most availablòe libraries of loops and sound effects are actually 32-bits waveform. All known software smoothly converts betweeen 32 and 16 bit waveforms, so probably You are already using a lot of 32-bits WAV files, and You do not know this.... CEP, as any other program natively supporting 32-bits files (such as Audio Mulch, Cubase, Sound Forge, etc.) plays without problems a 32-bits file over a 16-bits sound board.

This does not pose any conversion problem, because the sampling rate is unaffected, and the bit depth can always be increased without any particular trick.

This is advantageous also when the original sampling was also at 16 bit. A complete mixing chain, which start at 16 bits, makes all intermediate computation and fading at 32 bits-float, and switch back to 16 bits only at the end, gives a sound quality which is definitely much better than a 16-bits integer processing chain.

Apart from the storage space on the HD (which is very cheap now), all the computations made in float-32bits are nowadays very efficiently performed in the FPU of the computer, which can outperform the most powerful DSP chips by a factor 16 or more, if the computer is properly programmed (making extensive use of the SIMD instructions, for example). This means that, in practice, che computational load is lower working in 32-bits float than in 16-bits integer!
Regards,
Andreas
kempfand is offline  
Old 29th May 2004, 00:10   #503  |  Link
ursamtl
Registered User
 
Join Date: May 2004
Location: Montreal
Posts: 729
SIR delay compensation groups added to FTP

Hi,

I've just added a zip file containing three SIR delay compensation groups to the FTP server and to my web page at www.geocities.com/ursamtl.

This zip file contains three Plogue Bidule groups I've created as alternatives to loading dummy instances of SIR in your bidules to compensate for its sample delay of 8960 (in SIR v.1005). The savings in CPU usage can be significant, ranging from 5-10%


SIRDelComp2 provides 2 channels in and out with a 8960-sample delay on each. This replaces one dummy instance of SIR
SIRDelComp4 provides 4 channels in and out with a 8960-sample delay on each. This replaces two dummy instances of SIR
SIRDelComp4+2 also provides 4 channels in and out with a 8960-sample delay on each, plus 2 channels directly through (pins 5 & 6) for those who want 5.1 channels on one neat bar. For example, L, R, C, LFE, might have delay compensation with channels 5&6 connected to one instance of SIR elsewhere in the bidule.


When I conducted tests on my PC (Athlon Thunderbird 1.1GHz), the results were as follows:
Average CPU usage
-----------------
1 SIR with stereo impulse + 1 dummy SIR: 28%
1 SIR with stereo impulse + 2 dummy SIRs: 32%
1 SIR with stereo impulse + SIRDelComp2: 20%
1 SIR with stereo impulse + SIRDelComp4: 22%
1 SIR with stereo impulse + SIRDelComp4+2: 22%
2 SIRs with stereo impulse + SIRDelComp2: 27%
2 SIRs with stereo impulse + 1 dummy SIR: 33%


Although I did no testing of memory usage, obviously there should be some saving given that SIR is 1.6MB and the groups are a few KB!

Enjoy,
UrsaMtl
ursamtl is offline  
Old 29th May 2004, 00:50   #504  |  Link
Shayne
Registered User
 
Shayne's Avatar
 
Join Date: Jan 2002
Location: northern canada
Posts: 215
Quote:
Originally posted by kempfand


........So you must set 'never' for the WAVE Format Extensible (with Preference)........

Regards,

Andreas
Sorry for the late reply but thank you Sir. All is fine now and my computer no longer needs to live in the past.

I have checked pin outs and this release seems to be consistent with besweet (all is well). I did find myself dropping the gains on lfe down to -6.5 db but we are batching again.

Weird being a forced beta tester. But free is good.

Peace
Shayne is offline  
Old 30th May 2004, 23:24   #505  |  Link
kempfand
Registered User
 
kempfand's Avatar
 
Join Date: Oct 2001
Location: Switzerland
Posts: 225
Quote:
Originally posted by ursamtl
I have yet to discover convincing impulses for giving me that "being there" feeling. For example, at work I can listen to music through a nice set of headphones, but the plugins or impulses suggested when I search on the net for "out of head" sounds just don't seem to work. In theory, with the right impulses I should be able to sit there and have the music sound as if I had a PA sitting in front of me, but it hasn't worked so far. But, I digress...
Check the samples at the bottom of The FIReverb Suite audio demonstration for beautiful "out of head" binaural listening experience.

The FIReverb Suite is a superbe package IMHO (as well as the small version CATT-Acoustic

Cheers,
Andreas
kempfand is offline  
Old 30th May 2004, 23:31   #506  |  Link
kempfand
Registered User
 
kempfand's Avatar
 
Join Date: Oct 2001
Location: Switzerland
Posts: 225
Quote:
Originally posted by Shayne Weird being a forced beta tester. But free is good.[/B]
Plus it's one of the rare examples where beta is labelled as beta. Far too often, betas is sold for $. Also, I've never seen an example where the company (Plogue) was so responsive to implement enhancements, and help for bug fixing.

I'm not a fan of advertising, but I'll be in what will hopefully be a line when the free-version changes to "fee-version".

Cheers and peace,

Andreas
kempfand is offline  
Old 31st May 2004, 00:46   #507  |  Link
Shayne
Registered User
 
Shayne's Avatar
 
Join Date: Jan 2002
Location: northern canada
Posts: 215
Well you would think a batch process routine would need to be implemented to make it truly useful that there would be any sort of line.

But they need not listen.

Peace
Shayne is offline  
Old 31st May 2004, 02:43   #508  |  Link
Umma
Registered User
 
Join Date: May 2003
Posts: 76
Quote:
Well you would think a batch process routine would need to be implemented to make it truly useful that there would be any sort of line.
I wanted a batch routine, too, when I first started with it, but Plogue crashes on me after 5 or so straight uses without closing it down and restarting it. I just use one big wav file now.

Plogue wasn't even developed with this sort of music manipulation in mind. I may be wrong, but I think it was mostly meant for midi or homegrown electronic music. It would be nice if they took out the midi part and sold a "mini-Plogue" to those of us who like to fool with music conversions such as those we are doing over here.
Umma is offline  
Old 1st June 2004, 21:12   #509  |  Link
kempfand
Registered User
 
kempfand's Avatar
 
Join Date: Oct 2001
Location: Switzerland
Posts: 225
Quote:
Originally posted by specise_8472
Have just uploaded to the FTP server some new toys.
Daphy will put on Web server when he can.

It is a very good Mono to stereo conversion utility. And feeding the result into one of the above 2 to 5 methods produces astounding results. Better than I had hoped for. I myself personally use my Allinone process.

/// snip ///

For those interested - it is basically feeding the mono signal through two banks of cascaded filters. Each bank is a 19 pole Allpass filter. Then feeding the resultant signals through my x-talk filter to produce better channel seperation.
Species: I tested it (both the original one, as well as the 44.1 & 48 ones). Setup: Mono Signal -> Mono_To_Stereo -> Audio File Recorder. The levels of the recorded L and R are extremely different. Same results when EoH runs this.

Guess it's just a small wrong routing somewhere, and would appreciate if you could re-check.

Kind regards,
Andreas
kempfand is offline  
Old 6th June 2004, 06:09   #510  |  Link
Tantulus
Registered User
 
Join Date: Sep 2003
Location: Massachusetts
Posts: 34
SAD5.1 with impulse response

Well,finally I got around to experimenting with Bidules. I decided to add the "ambient hall" impulse from NoiseVault to SIRs on the SAD5.1 bidule. At that time I also decided to spring for a better center speaker and a powered subwoofer. This gave me a chance to try out my "SIR" SAD5.1 adaptation in a decent listening room. Honestly, I was blown away! My cramped living room never did justice to surround sound. Anyway, I felt the soudscape fill the room and the Impulse Response gave enough ambience that when I shut my eyes I could actually imagine being in a concert hall. My impression is that at least with orchestral music one can get a sense of "broader-than-room-size" with a decent speaker setup. Although, at times I felt that I was just bathed in the music rather than sensing the music orginated from the front as in a concert hall, I was very satisfied. Just as mentioned earlier in the thread, the salesman wanted to know how this was done and called others in the store to hear the disc.

Also, as Kempfand noted, there are discs that capture the ambience of the enviroment where they are recorded in stereo and I found with these recordings I get an added benefit of nice localization of instruments. Ambience was augmented without the need for using impulse responses. I guess the results of encoding into 5.1 ultimately depends on the orginal recording process, the speaker setup, the room acoustics and the bidule used. Once again, hats off Kempfand; my receiver has DPLII, DTS, and Neo6 settings. None of them come close to the bidules produced in this forum. At best I get some broading of music and tighter vocals in the center channel with TV or CD's.

Anyway, that's my two cents!

Finally, please, please, could somebody explain what the AmpIn and AmpOut settings in the HNM filter does? I looked through the whole thread I can only find references to settings but not their meaning.

Also, some recordings tend to be too bassy and no amount of changing the low pass control on the subwoofer corrects this. I know I can add a gain control but can this be adjusted with the HNM filters connected to LFE?

A less timid but greatful,
Scott

Last edited by Tantulus; 6th June 2004 at 06:31.
Tantulus is offline  
Old 7th June 2004, 01:35   #511  |  Link
ursamtl
Registered User
 
Join Date: May 2004
Location: Montreal
Posts: 729
Ursa's UpMix Studio

Learning to like Ambisonics
This forum has several guides or suggestions for converting 2-channel stereo to some variation of surround sound, usually 5.1. Each method is fun to experiment with, and each presents something different. I’ve been playing around with them and trying different combinations for a few weeks now. What follows is the result of my experiments so far. I’m quite pleased with the sound it provides and I hope you’ll enjoy it too.

Basically, I took the equations on Angelo Farina’s web site and created a bidule group that converts 2-channel stereo (UHJ) to B format (WXY, the Z is null because no height info is available). This group is a basic Ambisonic encoder with no controls for azimuth, elevation, distance, or anything. It just creates the basic signals.

Then I used different combinations of the equations on Richard Furse’s web site to put together a similar basic bidule group to serve as an Ambisonic decoder. I tried the calculations for the “Surround” speaker layout, but the sound wasn’t that impressive, and Furse’s site points out that this results in symmetry distortion. I then considered the Pentagon suggested by EoH and kempfand, but when I looked at Furse’s suggested equations, I discovered that they do not form a truly equilateral pentagon. Thus, rotating this by 36° to provide three front channels and two back results in a sort of lopsided listening field. To lay out one's speakers to match the calculated points of the rotated Pentagon, one would have to place the center speaker roughly three degrees to the left of center, the left speaker at 74° from center and right speaker at 71° from center. Even then, the center of the rotation would not be centered between the left and right speakers.



I got better results from the Octagon 2 layout. Since it calls for 8 speakers, and there are only five available in my target 5.1 system, I created phantom speakers by splitting the signals for each of the side left, side right, and center surround speakers and feeding them through the two speakers on either side of each phantom.



I tried the equations for both the strict spherical harmonics layouts and the controller opposites. As Furse points out, the latter results in a more diffuse image. I found the spherical harmonics layout sounded better.

When I put this all together and tested it, the sound was quite good, much more expansive than any other Ambisonics I’d heard to date, yet still providing that sometimes uncanny sense of realism that simple Dolby Surround upmixes just can't convey.

I then played around with various enhancements to the input stereo signal, such stereo widening, brightening the stereo edges with EQ, etc. These provided some interesting results, but it got complicated and CPU intensive.

Then I thought about the nature of the signals, if the Y signal consists of the spherical harmonics representing the width of the soundfield, then boosting it with respect to the others should spread the sound out. I tried it, and it worked! More importantly, nothing else in the sound suffered. Stereo wide effects generally kill tight, accurate bass response, but this widened Ambisonic image did not. I also tried the same trick on the X channel to see if a similar improvement would occur in the depth of the soundfield, but it didn’t provide such a spectacular result. However, applying reverb to the X channel--even simple mono reverb--does provide some amazing results.

So, I put it all together in a bidule consisting of nested groups, and learned how to put together some controls in a bidule. This resulted in the following controls, which are available from a control panel when you double-click on the UpMix Studio group in the center of the bidule:



Input Level A standard bidule Gain. The standard setting is at 0dB.
LFE? Checking this box causes all really deep bass centered around 32Hz or so to be redirected to the LFE channel. If the box is unchecked, the channel will be null. If you have monitoring capabilities with a subwoofer on your PC, try disconnecting all other outputs but the fourth from the left, and you'll hear exactly what's going through the LFE channel. By the way, the LFE channel is not affected by the Bass Boost controls below.

Enable Soundfield and Bass Boost? I put this control there for the purists. If you don’t want any width control or bass boost, and wish to just upmix to 5.1 using the basic Ambisonic implementation in this bidule, uncheck this box. If you do want the enhancements, check it!

Soundfield Width Gives you the chance to boost the Y or width channel, thus spreading out the sound. This also tends to make the highs a little crisper, but there is no EQ involved. This one control provides an incredibly spacious yet accurate soundfield. Of course, it all depends on the source file. Try, some Floyd or Roxy Music's Avalon.

Bass Boost Level Adjusts how much boost there is to the bass frequencies below the crossover frequency.

Bass Boost Crossover Frequency Moves the upper limit of the bass boost. If you’d like a more solid bottom but without much more bass, Move this all the way to zero and use the Bass Boost Level to boost the deepest bottom end a bit. This puts a nice bottom on drums without overpowering anything.

And if you really like the sound of the grand canyon…..
If you have a fairly dry piece of music wish to add some reverb, try adding it only to the X channel. I was astounded by the huge, professional sound image I got when I did this. I also tried running both the X and the Y channels through the two channels on a SIR VST. It was pretty massive, but almost too much so. Have fun, play a bit.



Monitoring
To get the most out of this system, it’s best to monitor the sound while you make the adjustments. If you don’t have a surround sound system on your PC, just hook up the standard 2-channel output. I’ve found that this gives you a really clear idea of what the sound of the final mix will be. Do not hook up the surrounds, the center or the LFE to this, however, as the sound will be inaccurate. It amazes me just how nice even a 2-channel mix sounds through this bidule. I think I'll use it to do some new mexies for listening in the car.

Note: You’ll have to remove this monitoring sound output device if you want to process in Plogue Bidule’s offline mode.



Of course, if you have a surround system and a multichannel ASIO setup, you can monitor the whole mix.



So, have fun. As I said, this provided me with some very nice surround mixes. Unless you add some reverb, the natural ambience of the original recordings is all you have, and most of the time, that’s all you need.

<b>You can download the resulting bidule from

My website on Geocities, another link

here

or an offline processing mode version from the Projects\Ursa's UpMix Studio folder on the NeedfulThings server.
Thanks to daphy for doing the mod for offline processing and linking the player and recorder!

As for what to do with the multichannel wave file, that’s been covered extensively in this thread and others. If you’re encoding to AC3 using SoftEncode, just load the file in and launch the encoder. Otherwise, use Besweet to split the 6-channel file and then encode using your preferred AC3 or DTS encoder. By the way, as mentioned earlier in this thread, converting your input wave files to 32 bits does indeed make a difference. The sound difference may be subtle, but you will notice more detailed high frequencies, etc.

For more reading, check out the following:

This whole fascinating thread!
Thanks to EoH, kempfand, kpex, species, daphy, andy and all the others, you’ve started something very, very cool.

Conversion between UHJ and B-format by Angelo Farina.

First and Second Order Ambisonic Decoding Equations by Richard Furse.


UrsaMtl, June 6, 2004

Moderator Edit: Fixed Image links per request of ursamtl

Last edited by KpeX; 14th May 2005 at 14:14.
ursamtl is offline  
Old 7th June 2004, 22:07   #512  |  Link
kempfand
Registered User
 
kempfand's Avatar
 
Join Date: Oct 2001
Location: Switzerland
Posts: 225
ursamtl: Thanks for sharing your work. It's always good to see what others do, and playing with the various possible schemes in Bidule is so much fun, as you also mention. A few remarks:

1) In the current uploaded bidule (June 6), I believe that the decoding maatrix for C is broken (in "B-Octo2SH-to-5.1"). It should be C = 0.1768xW + 0.2500xY (but currently is C = 0.1768 x W + 0.1768 x X). The "0.2500"-box is there but not linked, so it probably is just broken.

2) It is true that "Spherical Harmonics" decodes result in a (slightly) loss of directional information, but it should be added that "Controlled Opposites" produce a larger listening area, and guarantee that speakers will never generate an `anti-phase' signal.

I did many tests some time back and personally prefer the "controlled opposites", but it's a question of taste (not right or wrong).

3) I analyzed your modified Octagon2 decode, and it clearly produces a slightly more narrow soundfield as compared to the Pentagon decode. This is also expected, as the different speakers receive different W's, i.e. you would expect some 'distortions' for the soundfield (in the case of the modified Octagon2 some kind of squeezing along the x-axis, especially in the FL and FR areas of the soundfield).

I would suggest you still give the Pentagon decode a try, as it produces a perfectly equilateral pentagon (72 deg between each speaker). You can use 2 options:

i) Using BFprocEdit (rotate the soundfield by 36 deg) and then Emigrator (decode to the Pentagon), as on the guide on the 1st page of this threat or

ii) Modify the decoding matrices and use a bidule scheme similar to the one you did:

For Spherical Harmonics:
speaker________________Weights
__________x_______y_________W_______X_______Y

C ___1.0000__0.0000____0.2828__0.4000__0.0000
FL___0.3090__0.9511____0.2828__0.1236__0.3804
SL__-0.8090__0.5878____0.2828_-0.3236__0.2351
SR__-0.8090_-0.5878____0.2828_-0.3236_-0.2351
FR___0.3090_-0.9511____0.2828__0.1236_-0.3804


For Controlled Opposites:
speaker________________Weights
__________x_______y_________W_______X_______Y

C____1.0000__0.0000____0.2828__0.2000__0.0000
FL___0.3090__0.9511____0.2828__0.0618__0.1902
SL__-0.8090__0.5878____0.2828_-0.1618__0.1176
FR___0.3090_-0.9511____0.2828__0.0618_-0.1902
SR__-0.8090_-0.5878____0.2828_-0.1618_-0.1176


In summary, I think you did a great job, and I'm sure it was great fun an a good deal of deep understanding of how Ambisonics works, and what it does and does not.

Kind regards,

Andreas
kempfand is offline  
Old 7th June 2004, 22:18   #513  |  Link
kempfand
Registered User
 
kempfand's Avatar
 
Join Date: Oct 2001
Location: Switzerland
Posts: 225
Re: SAD5.1 with impulse response

Quote:
Originally posted by Tantulus I guess the results of encoding into 5.1 ultimately depends on the orginal recording process, the speaker setup, the room acoustics and the bidule used.
Tantulus: Think you made a key statement here, so I am re-quoting it. A big deal depends in fact on the original recording and how it was mixed . As a matter of fact, I know some repurposing experts, who use LP's as source, just because they contain the 'right' mix for what they want to do. Now I'm not suggesting to use LP's as source, but I think it makes the point.

As to the AmpIn and AmpOut settings with the HNM filter, I am sorry I cannot really help, as I never use an LFE, and if it is part of a bidules, I just rely on the settings published. I simply think there is too much danger of "things going wrong", unless of course one knows exactly what one does.

To quote from recent posting from the QuadrophonicQuad forum, from a guy I highly respect for his knowledge:

Quote:
The biggest problem with nearly all of the DTS-CD's is that they are home cooked, and essentially qualify as bootlegs.
If the person who did the original transfer was working on, say, PC speakers or HiFi speakers then they used an inaccurate monitoring environment, and as the studio saying goes "If you can't hear it you can't mix it".
Another big area for error is the extensive use of the LFE channel as a sub channel, when it is by definition a completely different thing.
The Sub channel exists to extend the bass response in systems that do not have 5 full range monitors/speakers, and NOT as a subwoofer.
When you start encoding music with an LFE channel, you run the following risks.

A/. The end user has not set things up correctly, and that pumpimg bass just isn't there.

B/. You can never know all the crossover frequencies that end user sub/sattelite systems use, so phase cancellation is a big problem. There is also the reverse of this, and what is piped to the LFE may well add to what is already in the bass end of the main speakers, giving the "bass heavy" problem you describe

C/. (The most common) the system for playback is simply not properly calibrated, the sub is too loud, the balances between all the other 5 channels are phase delayed & comb filtered and the result is an unpredictable bloody mess.

D/. There is a bass management setting on the AV amp, and it is conflicting with what is piped to the LFE Channel.

E/. Incorrect crossover frequency set in the DTS conversion. In an ideal world, you do not use the LFE but instead use an LFE splitter to emulate the effects of Bass Management. Set the X-Over to 80Hz and the slope to 24dB/Octave, or even steeper if possible. Never, ever encode with the LFE active unless you know that your monitoring environment is 100% accurate & properly calibrated. And only then with extreme caution.

In short, do not use the LFE channel unless you are doing film scores or the 1812 overture, and trust to the Bass Management in the AV amps to sort everything out. It is what it is there for.
Kind regards,

Andreas
kempfand is offline  
Old 8th June 2004, 00:01   #514  |  Link
ursamtl
Registered User
 
Join Date: May 2004
Location: Montreal
Posts: 729
@kempfand,

Thanks for the feedback. Just a couple of brief comments for now as I have some balcony gardening to take care of.

Quote:
Originally posted by kempfand
1) In the current uploaded bidule (June 6), I believe that the decoding maatrix for C is broken (in "B-Octo2SH-to-5.1"). It should be C = 0.1768xW + 0.2500xY (but currently is C = 0.1768 x W + 0.1768 x X). The "0.2500"-box is there but not linked, so it probably is just broken.
Actually the numbers I chose are intentional. I did forget to delete the unused 0.2500 constant. You see the Octagon is designed for a soundfield where the center speaker is at the same distance from the listening point as the others. Since most surround systems situate the center speaker on the same plane as the fronts, I felt it would be more accurate to attenuate the center somewhat. After doing a scale drawing in Illustrator (which I will share once I resolve some problems linking to graphics), I determined that since the center speaker on the same plane is 0.7071 the distance (interesting how that number keeps coming up , I attenuated the center by that amount. As for there being different values going to different speakers, of course there are. That's how the phantom speakers are created. For example, the front left speaker receives 0.1768 for it's own signal, plus 884 (half of 1768) for the phantom left side speaker. This totals 0.2652. For the rear left speaker, I took this amount representing the rear left plus half the rear side, plus an additional 884 for half of the rear center phantom speaker. In fact, all the virtual speakers in my implementation of Furse's Octagon 2 receive the same amount of sound pressure with the exception of the center, the reason for which I just explained.


Quote:
2) It is true that "Spherical Harmonics" decodes result in a (slightly) loss of directional information, but it should be added that "Controlled Opposites" produce a larger listening area, and guarantee that speakers will never generate an `anti-phase' signal..
Yes I read that too, but surprisingly, to me controlled opposites don't sound bigger. They actually remind me of the sound of audio that is out of phase, in that they produce an overly diffuse soundfield with nothing in focus. I personally found the spherical harmonics sounded much more appealing and more like the kind of professional 5.1 sound I so like.

Quote:
3) I analyzed your modified Octagon2 decode, and it clearly produces a slightly more narrow soundfield as compared to the Pentagon decode. This is also expected, as the different speakers receive different W's, i.e. you would expect some 'distortions' for the soundfield (in the case of the modified Octagon2 some kind of squeezing along the x-axis, especially in the FL and FR areas of the soundfield)..
Your statement really surprised me. In all the tests I've done on all the bidules over the past few weeks, the Pentagon decoded bidules sounded much more narrow to me. Until I actually tried the Octagon 2 layout (which I believe species 8472 came up with first in this thread. I simply added the phantom speakers). I dismissed Ambisonics and completely unacceptable for producing surround sound.

Quote:
I would suggest you still give the Pentagon decode a try, as it produces a perfectly equilateral pentagon (72 deg between each speaker)..
I did try a Pentagon layout and it sounded really compressed and lifeless compared to Octagon. Even a 4-channel cube layout is much better. Once took the calculations on Furse's web site and plotted them in Illustrator, I realized tht it does not seem to produce an equilateral pentagon (assuming the point of rotation is the geometrical center at -0.0489, ambisonic X). Check the diagram I just posted in my original message. Worse, once rotated 36° to the left, the soundfield is completely distorted. I already edited my message this morning once I had time to measure the angles in my drawing (which I'd left at work).

Even if the rotation is centered on the 0,0 listening position, there is still a problem with playing back a file with harmonics calculated for a wide soundfield, but on sound systems with real speakers placed differently. The ideal position for the front left speaker in the rotated Pentagon is at 74° from center, a full 29-44° from the suggested placement in a typical 5.1 system. the Right speaker is at 71°, or 26-41° from the typical position. Therefore, if you play back a file decoded through the Pentagon, the width of your soundfield is being compressed between 45-85°!! No wonder I found the Pentagon bidules lacking in width! I'm really surprised that you actually ended up making such a conclusion about the Octagon 2 layout. Sit down and draw it out, you'll see. I prepared some graphics for my post last night but there were some problems with linking to my geocities site. I'm going to place them on another site and edit my post. I think if you look at them and consider the geometry of it all, there's no way you can conclude that the Octagon 2 soundfield is compressed , certainly nowhere nearly as much as the rotated Pentagon.

Anyway, time to tackle the garden. By the way, did you try the effect of widening the soundfield with the slider I implemented? It sounds amazing!

Regards,
Steve.

Last edited by ursamtl; 8th June 2004 at 15:27.
ursamtl is offline  
Old 9th June 2004, 01:17   #515  |  Link
kempfand
Registered User
 
kempfand's Avatar
 
Join Date: Oct 2001
Location: Switzerland
Posts: 225
@ursamtl/Steve:

Thanks for the explanations. Few comments from my side.
Quote:
did you try the effect of widening the soundfield with the slider I implemented? It sounds amazing!
I tried, and the effect IS truely amazing. As indicated before, I also like the way you implemented this finding using bidule, as it makes testing so easy.
Quote:
I felt it would be more accurate to attenuate the center somewhat. After doing a scale drawing in Illustrator (which I will share once I resolve some problems linking to graphics), I determined that since the center speaker on the same plane is 0.7071 the distance (interesting how that number keeps coming up , I attenuated the center by that amount.
That makes it clear to me now.
Quote:
As for there being different values going to different speakers, of course there are. That's how the phantom speakers are created. For example, the front left speaker receives 0.1768 for it's own signal, plus 884 (half of 1768) for the phantom left side speaker. This totals 0.2652. For the rear left speaker, I took this amount representing the rear left plus half the rear side, plus an additional 884 for half of the rear center phantom speaker.
Understood. But:
Quote:
In fact, all the virtual speakers in my implementation of Furse's Octagon 2 receive the same amount of sound pressure with the exception of the center, the reason for which I just explained.
I disagree : Pressure (W) going into C: 0.1768, each of FL & FR: 0.2652, each of SL & SR: 0.3536.

Now: This is just a factual statement (i.e. observation) I made, i.e. not saying 'right' or 'wrong'. In fact, I have seen unpublished decodes from well respecetd experts in this area who also apply different pressures to the decodings to the speakers.

Quote:
Quote:
3) I analyzed your modified Octagon2 decode, and it clearly produces a slightly more narrow soundfield as compared to the Pentagon decode. This is also expected, as the different speakers receive different W's, i.e. you would expect some 'distortions' for the soundfield (in the case of the modified Octagon2 some kind of squeezing along the x-axis, especially in the FL and FR areas of the soundfield)..
Your statement really surprised me. In all the tests I've done on all the bidules over the past few weeks, the Pentagon decoded bidules sounded much more narrow to me. Until I actually tried the Octagon 2 layout (which I believe species 8472 came up with first in this thread. I simply added the phantom speakers). I dismissed Ambisonics and completely unacceptable for producing surround sound.
You clearly see the mentioned deviation when you plot your layout using the CATT-FIReVerb Suite, whereas there are no deviations for the rV and rE deviations when using a Pentagon decode.

But same remark as above: Just an observation As a side-remark, it's interesting to see how people have have & are putting patent protection to 'their' decoding specs (IMhO one of the reasons why Ambisonics never really made it to the masses). My personal opinion is summarized well in Anegelo Farina's presentation from May 2004 (at the Berlin AES convention): "We are still learning what is the best way to render ... over a standard 5.0 (or 5.1) setup".

Quote:
I did try a Pentagon layout and it sounded really compressed and lifeless compared to Octagon. Even a 4-channel cube layout is much better. Once took the calculations on Furse's web site and plotted them in Illustrator, I realized tht it does not seem to produce an equilateral pentagon (assuming the point of rotation is the geometrical center at -0.0489, ambisonic X). Check the diagram I just posted in my original message. Worse, once rotated 36° to the left, the soundfield is completely distorted. I already edited my message this morning once I had time to measure the angles in my drawing (which I'd left at work).
This is actually the point I am most keen to understand better. Maybe I just have 'bricks' in front of my eyes, but plotting the Pentagon (rotated by 36 deg) gives a perfect equilateral Pentagon in my plots (see the x,y coordinates I gave above). In fact, I was and am using this to construct signed filters to decode B-format to 5 speakers (using 15 filters in total, 3 for each of the Pentagon-rig speakers, using 2 instances of Prinstine Space to do the 16 concurrent convolutions).

Quote:
I think if you look at them and consider the geometry of it all, there's no way you can conclude that the Octagon 2 soundfield is compressed , certainly nowhere nearly as much as the rotated Pentagon.
In addition to my remark above (i.e. Octagon2_MOD is compressed a bit on FL & FR, as per CATT-FIReVerb analysis), I think a 'complication' comes by the fact that there are 2 aspects:
a) decoding for a specific speaker configurarion. If this was the only criteria, we should all go for the Surround-decoce (= ITU-5.1), which is what we actually do not.
b) actual speaker-set-up in our listening room, which delivers the created output.
I other words: A decode for a specific speaker configuration (be it Pentagon or Octagon2_MOD or xyz) CAN sound give good results, as a function of listener's preference and actual speaker set-up (which most often is not the standard ITU-5.1).

Hope the gardeing was a succes. Time to get some sleep here.

Regards,

Andreas

Last edited by kempfand; 9th June 2004 at 08:35.
kempfand is offline  
Old 9th June 2004, 19:24   #516  |  Link
Tantulus
Registered User
 
Join Date: Sep 2003
Location: Massachusetts
Posts: 34
LFE & speaker placement

Thanks Kemfand for your reply:

Part of my problem with the LFE channel is that I have too damn many controls on my AV receiver (Onkyo TX-DS797) not to mention my subwoofer. Since I use my system for DVD's and music I decided to take the bidules as is and do the adjustments on the receiver so that I can switch from one to the other.

Quote:
B/. You can never know all the crossover frequencies that end user sub/sattelite systems use, so phase cancellation is a big problem. There is also the reverse of this, and what is piped to the LFE may well add to what is already in the bass end of the main speakers, giving the "bass heavy" problem you describe
I have a switch on the subwoofer that allows my to adjust the phase which seems to be set correctly for my ears.


I have a dial that allows my to adjust the low pass from 200htz to 60. The manual for the subwoofer recommends setting the low pass at 200hts because I'm using the decoder's SUB OUT jack. However, this takes on to much of the bass and results in the bass heavy problem. Although its a pain to keep changing the Lowpass it does allow me the flexibity of how much bass I want to capture from the other speakers depending on the format.

Quote:
C/. (The most common) the system for playback is simply not properly calibrated, the sub is too loud, the balances between all the other 5 channels are phase delayed & comb filtered and the result is an unpredictable bloody mess.
I went to "Radio Shaft" and bought a SPL meter to set the levels of all the speakers to 75 decibel sound pressure. However, again by rereading the manual I found that I can vary the levels while monitoring the music to get the reponse I'm looking for.

Quote:
Set the X-Over to 80Hz and the slope to 24dB/Octave, or even steeper if possible
I'll check out this setting on the subwoofer.

I feel I do need the LFE. It brings out the lower register instuments such as cellos, bass's and bassons. I'm curious why you do not use the LFE. Do you have speakers with good bass response or do you use the receiver?

Anyway, your respones has spurred my on to find the ideal settings. Thanks again.

Now to Ursmtl:

I noticed in your diagramns that you have the listener placed in a central location relative to the speakers. However the manuals suggest that the surround speakers should be placed in line with the listener about 3 feet above the ears. Are you implying there is a sweet spot for ambisonic 5.1? Also, I recently purchased surround speakers that can switch from dipole to monople to bipole. Would any of these settings be better for the surround effect or do I need back surround speakers?

Thanks everyone for your assistance!

Scott

Last edited by Tantulus; 9th June 2004 at 19:27.
Tantulus is offline  
Old 9th June 2004, 19:37   #517  |  Link
Tantulus
Registered User
 
Join Date: Sep 2003
Location: Massachusetts
Posts: 34
Batch Files

Sorry I forgot one more thing. For those of you with Windows 2000 Pro or XP. I'm trying to write a script using Windows Script Host to automate the encodeing process. However, it's going to take some time. Is anybody trying to do this?
Tantulus is offline  
Old 9th June 2004, 19:46   #518  |  Link
Eye of Horus
Banned
 
Eye of Horus's Avatar
 
Join Date: Jun 2002
Location: The Netherlands
Posts: 326
3 new exciting bidules !!

Hi all,

After 2 months of hard work, discussions all over, listening, rewriting, discussions again, adjustments, etc. etc.,
Kempfand and EoH proudly present our new bidules.
These bidules are completely different from everything published till now and in our opinion so are the results.
Our goal was to come as close as possible to the HardWare Repurpose guys and..... in our opinion on some tracks we even did better.
If you agree with us remains the question
We would love to hear your feedback !

NEEDED
--------

1. Besweet b28
2. MDA VST
3. Voxengo stereo VST (freeware !)
4. SSRC.exe
5. bidules and groups

DOWNLOAD
---------

The whole package can be downloaded here :
www.app.demon.nl/Kempfand-EoH-06-09-2004.rar

or on Needfull Things......

SETUP
------

1. Install Besweet, start.bat and SSRC.exe in the same folder
2. Install MDA and Voxengo Stereo into the VST folder of Plogue
3. Install the 3 bidules in the layout folder of Plogue
4. Start Plogue Bidule and rescan for plugins and groups
5. Make sure the filerecorder is saying : 16 bits !!!!

USAGE
------

There are 3 bidules which handle different kind of music.

1. Voice-Center.bidule
This bidule handles music with a singer in the center of the stereo. This is checked quite easily by putting your receiver into DPL mode and listen very carefully to the voice. If it's coming from the Center speaker only, then you should use this one.
After conversion, the voice comes out of the Center only.

2. Voice-Non Center_or_instrumental.bidule
This one is useful for songs where the voice in not (or not only) in the center and for instrumentals. After the conversion you will hear the voice from all speakers, but the accent is on the Center

3. Instrumental.bidule
Use this one or the previous for tracks without voice.

4. For a first run, don't change a thing in the bidules, but do make sure the filerecorder is using 16 bits. Of course your source need to be converted to 32 bits floating first for the best results !!

5. Save your output to the Besweet folder

6. Run start.bat, but first edit (right-click and chose "edit") it to suit your settings. Here is the batch routine :

BeSweet.exe -core( -input 01.wav -output f:\01- -6ch )

ssrc.exe --twopass --normalize f:\01-FL.wav f:\01-New-FL.wav
del f:\01-FL.wav

ssrc.exe --twopass --normalize f:\01-FR.wav f:\01-New-FR.wav
del f:\01-FR.wav

ssrc.exe --twopass --normalize f:\01-C.wav f:\01-New-C.wav
del f:\01-C.wav

ssrc.exe --twopass --normalize f:\01-SL.wav f:\01-New-SL.wav
del f:\01-SL.wav

ssrc.exe --twopass --normalize f:\01-SR.wav f:\01-New-SR.wav
del f:\01-SR.wav

copy f:\01-LFE.wav f:\01-New-LFE.wav
del f:\01-LFE.wav

As you see I use 01.wav as name for the 6channel WAV file in the Besweet folder (in my case on e and I do the processing to drive f:\ .
Change these values to the ones you wish to use. (Don't use spaces in the filename !) If you want the input WAV in the same directory, leave out the harddisc letter , but make sure it has another name after "-output" in the Besweet line !

In my case it will output 6 mono 16 bits 44.1 Khz WAV's to harddisc f: .

As usual you can use these to make a DTS cd......

EXTRA INFORMATION
--------------------

Voxengo stereo VST uses presets and we got a very good result with a preset of "pretty wide" . If you wish to experiment with these settings, it's very easy. Just doubleclick on the VST in Plogue and adjust to another preset or make your own. However : our results are all based on the "pretty wide" setting and we don't take responsibility for a worse conversion when using other settings

The same goes for the normalize routine. With these bidules we found that making all 5 mono files at the same peak level, gives the best result. It can be that you wish to have less sound in the rears. Just edit the batch and replace the setting for SSRC to your own preferences.

Regarding the messages about the LFE : all three routines don't have a separate LFE output ! The LFE file you see in the routines is empty !


Well, this is about it..... happy testing and please share your thoughts here !

kind regards,

Kempfand and Eye of Horus

Last edited by Eye of Horus; 9th June 2004 at 20:47.
Eye of Horus is offline  
Old 9th June 2004, 21:36   #519  |  Link
MaroonMike
Registered User
 
MaroonMike's Avatar
 
Join Date: Oct 2003
Location: Dixie USA
Posts: 23
EOH,

Are you suggesting that this new method is an improvement over the Ambisonics method? Can you explain how this method is different (from a sound perspective?)

Thanks for your work on this. Mike
MaroonMike is offline  
Old 9th June 2004, 21:38   #520  |  Link
AllTimeSToneD
HDTV Nerd
 
AllTimeSToneD's Avatar
 
Join Date: Dec 2001
Location: Beerland
Posts: 63
once i start the "start.bat" i get following error:

Error 58: Error : Unknown Input-File Format : "01.wav".
Quiting...

i named my wave file also 01.wav i also tryed to fix the wave file using besliced but still getting the same error
AllTimeSToneD is offline  
Closed Thread

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 15:32.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.