Stereo to surround extrapolation algorithms

[24-96 Mastering]

I'm aware of TC's Unwrap and Waves' UM225 algorithms. Are there currently any alternatives?

[reynaud]

in hardware there is the Z-Systems z-k6 stereo-to-surround processor and then in software the Arkamys Upmix which seems to still only be available for Pyramix systems.

Cycling '74 also had a VST and RTAS plugin called Upmix which they no longer seem to produce (not sure how much it shares with Arkamys' Upmix beyond the name).

cheers,
Reynaud

[tony666]

Build your own

I wouldn't be surpised if the Waves stuff (and the other systems perhaps too?)is somehow based on Gerzon's "surround sound from 2 channel stereo" article.

http://www.audiosignal.co.uk/Resourc..._stereo_A4.pdf

In the Gerzon article they're using a quadraphonic setup, but when you send L-R to the LS and RS speakers of a 5.1 system you would get the same i guess. You can shape the LFE channel however you want.

[bob katz]

Originally Posted by 24-96 Mastering

I'm aware of TC's Unwrap and Waves' UM225 algorithms. Are there currently any alternatives?

You bet! The K-Surround algorithm is available in the Z Sys hardware. We're, very slowly porting it to a plugin, but don't hold your breath, no one seems to have any time right now to do it. Here's a short synopsis:

If you're looking for very natural "unwrapping" of ambience and ambient material from stereo, then look towards the K-Surround. If you're into creating a surround where there was nothing to grab on before, then I think the TC Unwrap is the bees knees. So you gotta have both :-).


Best wishes,


Bob

[jayfrigo]

I've got one in for testing from a company that still hasn't made it public. The company was founded by one of the original MP3 team and a former New England Digital guy (Synclavier, direct-to-disk, 1st hard disk recording system). The process takes stereo to 5.1, 7.1, or even 9. I'm under NDA, so there's a limit to what I can say, but I could probably run a demo file through if you're interested.

It is fully patented (not just applied for), derived directly from the original mix (sums back perfectly unless you choose to further process/tweak), uses no steering, no delays or effects, and can actually place things discretely in each of 5 or 7 speakers, with no crosstalk, from a simple stereo source.

It works just as well for data-reduced files like MP3. It's one of those things I would have said probably could not be done before I had heard it. It's head and shoulders above other methods I've heard. Re: above, it's not based on Gerzon or any of the other methods out there. The patent granting would indicate to me that it's a new approach. It's DSP-intensive and not something you could simply put together with a fancy mix desk in a DAW.

[807Recordings]

^^ If you mean the Fraunhofer guys they already had a setup at Tresor with speakers all around the room. I am not sure if it is still setup there but it was around this time last year.


*just checked it was called IOSONO experience IOSONO **

[Key]

I believe he is talking about Penteo. But I thought that patent was pending.

[Tropicana]

On the IBC I heard a stereo>surround demo from orban.
It was still in beta but not bad at all.

[Jesse Graffam]

The industry standard.
LINEAR ACOUSTIC - Products - LA-5190 UPMAX:neo 5.1 Channel Surroundfield Synthesizer With Optional Analog Outputs, Remote Volume and Metadata Generation

Who's a select few of the people using Linear Acoustic stuff?
LINEAR ACOUSTIC - Testimonials

between the products these guys offer, they have a ridiculously huge market share of the DTV market, which is starting to be all there is anymore. if it's good enough for the Grammy's.... ehh....

the AERO.air lineup isn't half bad either... if it's good enough for the last 4 SuperBowls.... ehh.... but more of a broadcasting thing, unless you are handling a lot of Film -> DVD mastering.

[DrDeltaM]

There's also the Soundfield UPM-1 rackunit

[bob katz]

Originally Posted by jayfrigo

I've got one in for testing from a company that still hasn't made it public.

You're talking about a method which uses artificial intelligence and masking/unmasking techniques to separate out the sounds, using the techniques learned from lossy coding. I've heard a few of these in my room; proprietary and in beta or development and I've always detected artifacts. Many of them are good for demonstration and for forensic work, but when it comes to high fidelity, they suffer from the same kinds of artifacts (space monkeys) as your cell phone. There's always some artifact, you have to listen carefully. If there's enough material for masking, you're ok. But if you isolate one of the channels alone you can hear the space monkeys. Does that make it "broadcast quality"?

BK

[Sunbreak Music]

I didn't know Upmix from '74 was out. I much preferred it to Unwrap.

[24-96 Mastering]

Originally Posted by Sunbreak Music

I didn't know Upmix from '74 was out. I much preferred it to Unwrap.

It seems it's down again already on the website. Does anyone know how one might be able to obtain a copy / demo of it?

[reynaud]

demos are still available on the Cycling '74 site including the manuals. Latest version is 1.1.1

cheers,
Reynaud

[24-96 Mastering]

Originally Posted by jayfrigo

I've got one in for testing from a company that still hasn't made it public. The company was founded by one of the original MP3 team and a former New England Digital guy (Synclavier, direct-to-disk, 1st hard disk recording system). The process takes stereo to 5.1, 7.1, or even 9. I'm under NDA, so there's a limit to what I can say, but I could probably run a demo file through if you're interested.

Thanks for the offer, I'll get back to you on that, if I may, when I there's a challenging project up.


For now I need a working solution soon because there's work to be done and I sold my M6000 / unwrap a while ago. I was planning to just buy/use Waves um226 instead, but due to Waves being idiots (installing um226 would apparently uninstall my existing Waves bundles), that plan is now obsolete.

Thanks also for all other suggestions everyone, though I don't want to get any hardware box at this point for budget and workflow reasons.

So, software wise, there's:

- Waves UM266 (said to be good, but will probably require you to sell your soul to the WUP and pay for your existing plugins all over again)

- Cycling74 upmix (no longer sold, i.e. not an option, i guess)

- Starplugs 2-2-6 (dirt cheap... has anyone tried this?)

Are there any other software options I've overlooked?

[Sunbreak Music]

Originally Posted by 24-96 Mastering

It seems it's down again already on the website. Does anyone know how one might be able to obtain a copy / demo of it?

It came on a disc--I honestly can't remember if I demo'd it before purchasing.

I hope you can try it--there was some crossover in the development between Upmix and Unwrap. I just liked Upmix in both interface and sound.

There's a finite amount of control you have in this stuff anyway, considering what we're trying to do with it.

[jayfrigo]

Just to follow up, the one I'm talking about is not from the Fraunhofer guys, Penteo or Orban (which I think is just Penteo), or any of the others mentioned. And re: artifacts, some kind exist in pretty much anything that processes audio, whether it's a limiter with distortion, some kind of phasing or comb filtering sound, or the space monkeys. Even an analog compressor can pump and breathe, which is an artifact of its processing. EQs have group delay, and linear phase EQ has pre-echo (aka pre-ring). The best processes, including Cedar, Algorithmix reNovator, NoNoise etc. all will do funny things if pushed too hard. However, the good ones, like the previous examples, all sound great when used appropriately, Cielo included (process name, not company). But the laws of physics remain, no matter how pesky they are at times. Sorry to dodge the company name - just not my place to decide to go public for them.

[24-96 Mastering]

Sooo...

no more software options besides Starplugins 2-2-6 (doesn't cut it) / Cycling74 upmix (not sold anymore) / Waves UM226 (uninstalls other stuff you own) / Unwrap for Powercore (expensive & needs hardware)?

If anyone has any other suggestions, I'm dying to hear them

[Gutbucket]

You may take a look at the stereo>surround conversion techniques outlined here. I found that site while investigating ways of converting a number of 4 channel recordings I've made using a pair of stereo channels (Right/Left & Center/Back) to the standard five channel format. I haven't had time to dig into the site or try any of the techniques so unfortunately I'm unable offer any more direction or opinion on results.

In the more information that you probably care to hear about category:
My primary interest is finding ways to appropriately decorrelate that single back channel to two surround channels, but I also want the ability to optimize and re-balance things between the main and surround channels on occasion. Secondary applications for me are converting stereo recordings to 5 channel format and possibly rescuing a few of those 4 channel recordings in which I had problems with the L/R pair, rendering it unusable and leaving only the C/B pair. I'm thinking some M/S manipulation may be able to produce a pseudo L/R stereo result with the Center channel feeding the phantom center and the Back channel information spread to either side of the sound stage - or alternately - that some sort of upmix algorithm like those discussed in this thread may be able to produce a usable stereo output.

Hope that helps. Please post back to let me know what you end up doing if you can, thanks.

Lee

[24-96 Mastering]

Originally Posted by Gutbucket

You may take a look at the stereo>surround conversion techniques outlined here. I found that site while investigating ways of converting a number of 4 channel recordings I've made using a pair of stereo channels (Right/Left & Center/Back) to the standard five channel format. I haven't had time to dig into the site or try any of the techniques so unfortunately I'm unable offer any more direction or opinion on results.

Thanks for the suggestion, but as far as I can tell the first method described on that site is mirroring L/R on Ls/Rs, i.e. there doesn't seem to be any decorrelation algorithm / matrix at all, and all the others seem to be about demuxing multiplexed formats. Did I oversee something?

[tony666]

Originally Posted by 24-96 Mastering

Thanks for the suggestion, but as far as I can tell the first method described on that site is mirroring L/R on Ls/Rs, i.e. there doesn't seem to be any decorrelation algorithm / matrix at all, and all the others seem to be about demuxing multiplexed formats. Did I oversee something?

The guy makes a distiction between 2 methods/goals of getting surround out of a stereo track :
1. based on seperation (to place individual elements/instruments of the track somewhere in the surround field)
2. soundfield ( seems to be more based on the ambisonics theory, creating a more real surround I guess, using modern techniques )

Each of those methods has submethods with different algorithms. He's using freeware plugins and some none freeware to accomplish certain tasks. They're also using modules from the "plogue bidule" environment.

I'm not a surround expert, but I think you can actually create any kind of encoding algorithm /structure you want with the bidule environment, either using some plugins or making your own modules.

[24-96 Mastering]

Originally Posted by tony666

2. soundfield ( seems to be more based on the ambisonics theory, creating a more real surround I guess, using modern techniques )

Each of those methods has submethods with different algorithms. He's using freeware plugins and some none freeware to accomplish certain tasks. They're also using modules from the "plogue bidule" environment.

I'm not a surround expert, but I think you can actually create any kind of encoding algorithm /structure you want with the bidule environment, either using some plugins or making your own modules.

Thanks for your help. I understand the site gives detailed instructions for building a matrix in bidule. But unless I oversaw something relevant, the actual methods used to extrapolate surround content from stereo are all very, very basic (like simply extracting the side signal from L/R and putting it on the surrounds a la prologic) and the mentioned plugins that I tried don't really decorrelate to a meaningful degree either. If I missed a more elaborate upmixing algorithm in there, please do point it out.

[24-96 Mastering]

PS: I tried Fraunhofer SX (consumer) upmix algorithm today. How very, very awful!

[Gutbucket]

In my brief look at that dtsac3.com site I got the impression that he's using the plogue bidule environment for signal routing and various plugins for most of the manipulations, as Tony mentioned. Not having used it, the bidule environment looks like an interesting tool to go about doing some things I've wondered about. That's what primarily peaked my imagination.

When I came across the site I posted in the forum there mentioning what I was thinking about doing, to see if it was worth digging deeper at some later point. The site owner responded and gave me the distinct impression that his basic techniques are simple routings and Side information extraction like you mention, but the more involved techniques do decorellation and use other more advanced techniques, mostly based on the functions of the plugins used. The latter is what he suggested I take a look at for my needs. That's as far as I've gone with it.

[Key]

the dtsac3 site last time I checked - the site had an overhaul - mainly uses decoding based on QS Quadraphonic. But so does the most popular commercial decoder/upmixer DPLII.

I wouldn't get too caught up in complexity. It seems like you are assuming that complex = good. It really tells you nothing about how one of these decoders will perform/be percieved by two ears - for instance stereo isn't all that complex and there is no measurable separation from L and R when you utilize a phantom center image but this doesn't make stereo bad. Assuming this based on looking at the encoding and lack of decoding is to mistake theory for reality.

[24-96 Mastering]

Originally Posted by Key

the dtsac3 site last time I checked - the site had an overhaul - mainly uses decoding based on QS Quadraphonic. But so does the most popular commercial decoder/upmixer DPLII.

I wouldn't get too caught up in complexity. It seems like you are assuming that complex = good. It really tells you nothing about how one of these decoders will perform/be percieved by two ears - for instance stereo isn't all that complex and there is no measurable separation from L and R when you utilize a phantom center image but this doesn't make stereo bad. Assuming this based on looking at the encoding and lack of decoding is to mistake theory for reality.

No worries, I've been doing upmixing/extrapolation for a while, I'm not making any general assumptions about process complexity.
Simple M/S matrixing works great with some material. And with other material, it does not work at all. That's why I need tools that provide other decorrelation options.
I've worked with TC's Unwrap for a few years and found that to be pretty decent with very usable decorrelation alternatives, but I sold my M6000, so I'm looking for (ideally native plugin) alternatives now.

[24-96 Mastering]

Originally Posted by Gutbucket

but the more involved techniques do decorellation and use other more advanced techniques, mostly based on the functions of the plugins used.

Thanks, I probably missed that. The site has a lot of text, maybe I oversaw the relevant parts. I'll go have another look.

In the meantime, if anyone knows of any other extrapolation/upmixing plugins / methods, please tell

[bkuijt]

Record the output of your home theater amp with it's various settings.

But seriously I think you've sold the best thing for it... Unwrap is pretty good.
Z-systems are also used but I've had varying results.

[kosmokrator]

Götz-Michael Rieth of Eastside Mastering (former Emil Berliner/Dt. Grammophon) invented a process called AMSELII (IIRC).
I don't know if the product belongs to Emil Berliner, but just ask him:

east Side Mastering Studio - coming soon -

[Thor]

Hi Robin,

I recently did a survey of some of the people I know who work extensively in surround to find out exactly the same thing you're asking, and found that the current best of class is the TC Unwrap.

You can find a powercore used on eBay for cheap, and buy the TC Unwrap app for powercore. At least that's what I did, and it seems to run fine. Works well, although it's not as great as a real 5.1 mix. But you knew that already

Cheers,
Thor