>> v4: ReaSurround

[jm duchenne]

(duplicated)

[jm duchenne]

Originally Posted by dawhead

this is how ambisonics works. and yes, it makes sense, and yes, it works.

you encode the desired position of the source with the source signal (this type of audio signal is called b-format). the b-format signals all do go an ambisonics decoder, which knows about the actual speaker arrangement, and does a bunch of math to put each source in the right place in the soundfield. ambisonics has some issues, but it also does a LOT of things right.

Yes... but you cannot do FX on the ambisonic encoded signal, you have to decode it before to process the channels, according to what speaker arrangement you have choosen in the decoder...
Ambisonics can do a lot of things, as long as you don't have to process the sound.

[jm duchenne]

The more I play with ReaSurround, the more I find the way that the height parameter has been implemented clever. Perfectible, but clever !

A suggestion/request : could we use the mouse wheel as an alternative to the Alt+Drag combination to change the height value ?
We could change it while moving the blobs...

[yerbouti]

Originally Posted by jm duchenne

Yes... but you cannot do FX on the ambisonic encoded signal, you have to decode it before to process the channels, according to what speaker arrangement you have choosen in the decoder...
Ambisonics can do a lot of things, as long as you don't have to process the sound.

So, is Reasurround using an ambisonic or standard amplitude panning algorithm (or VBAP)? I doubt it uses ambisonic.

Could the developers give us info on that please?

[schwa]

Originally Posted by yerbouti

So, is Reasurround using an ambisonic or standard amplitude panning algorithm (or VBAP)?

It is a mix of standard panning (aka geometry) and VBAP. It's not extra-fancy. If it sounds good, it's probably because of how the parameters are smoothed while sound sources are in motion.

[griz lee]

Originally Posted by jm duchenne

The more I play with ReaSurround, the more I find the way that the height parameter has been implemented clever. Perfectible, but clever !

A suggestion/request : could we use the mouse wheel as an alternative to the Alt+Drag combination to change the height value ?
We could change it while moving the blobs...

I agree Jim. It's damn clever.

I really doubt it's ambisonics. Either that or it's the best ambisonics i've ever heard. It sounds like good old amplitude panning, done correctly with good scaling to me.

I particularly like the way you can zoom the speaker dimensions in/out. Really great for 'oh, it's good, but i need a little more/less separation'. The whole thing is total genius. You'd normally have to backtrack on each channel for that.

These guys, and schwa in particular, deserve a medal for awesomeness.

--
Edit: ... ah now we know..

[musicbynumbers]

I think what might be really good for this is a actions to assign 3 midi CCs to each of the 3 rotor dials so that we can select a reasurround, choose our 3 parameters, click on the input "node" to be affected and then start moving your "p5 midi glove" in the air!

EDIT: in fact we can probably do this now already by assigning CCs to it in the normal way.. off to try

[griz lee]

... i just went bonkers with parameter modulation on the XY locations and diffusion settings and got all kinds of wacky auto-pan effects that should absolutely never be consigned to a production.

Super fun though...

[jm duchenne]

Originally Posted by griz lee

I agree Jim. It's damn clever.
I particularly like the way you can zoom the speaker dimensions in/out. Really great for 'oh, it's good, but i need a little more/less separation'. The whole thing is total genius. You'd normally have to backtrack on each channel for that.

Yes, but a special parameter to diminish the "areas" size more is needed when we have more than 8 speakers.
I have tried on my 17.1 system, and there is no way to separate the FCL and FCR from the others. 5 in a line is too much for the current area value when zoom is at minimum value, and the speakers are the farthest possible.
Or we need a largest zoom view, because when the x width and y depth values are at max, the speakers are not visible anymore...

[musicbynumbers]

..ok tried it and they're not there yet

and I know they are infinite encoders and we kind of want to be able to assign them to standard CCs for fast/absolute changes so hopefully we can find a way to do this

[jm duchenne]

Originally Posted by musicbynumbers

I know they are infinite encoders and we kind of want to be able to assign them to standard CCs for fast/absolute changes so hopefully we can find a way to do this

I have searched for too.
Imagine : inputs groups buttons + MIDI CC control of the edit parameters ?
Wouaouh...

[griz lee]

Originally Posted by jm duchenne

Yes, but a special parameter to diminish the "areas" size more is needed when we have more than 8 speakers.
I have tried on my 17.1 system, and there is no way to separate the FCL and FCR from the others. 5 in a line is too much for the current area value when zoom is at minimum value, and the speakers are the farthest possible.
Or we need a largest zoom view, because when the x width and y depth values are at max, the speakers are not visible anymore...

That sounds very important.

I've 'only' got 5.1 in this studio, but have done loads of large area stuff in the past.

The zones need to scale correctly in some proportion to the number of speakers. I guess what's here is fine for 5.1, and i've had some great results with it already, but as the system is so flexible it would be a shame if something like this prevented the big layouts from happening.

[Kihoalu]

.

I did octaphonic work many years ago using 8 speakers for playback (one in each corner of a cubical space), since then I have not heard anything that was true 3D (and of course, 5.1 and 7.1 are not). Ambisonics can do 3D but this is a term that was invented after I did my 3D work, and I have not heard any actual "Ambisonic" materal.

The "panners" involved in real 3D mixdowns must be able to pan left, right, back, front AND up, down (Hard to do with mechanical panners) but easy with a computer 3D layout visualization, will have to see how this new Reaper implementation works.

Here is a link to some ambisonic work done with Reaper where some 3D examples are mentioned:

http://www.slideshare.net/DrWig/futu...ing-ambisonics

(corrected post new info found after posting)

.

[jm duchenne]

Originally Posted by schwa

It is a mix of standard panning (aka geometry) and VBAP. It's not extra-fancy. If it sounds good, it's probably because of how the parameters are smoothed while sound sources are in motion.

It's not extra fancy, and it is what makes it so powerful
Geometry + VBAP + some well choosen parameters = you can do what you want (and even more !)...
Nice work indeed.

[jm duchenne]

Originally Posted by blakflag

Related to the surround capabilities, this bug is something I posted over a year ago:

http://forum.cockos.com/showthread.php?t=36271

Reaper is lacking some sort of VST capability that some surround plugs (Uhbik) require to be able to know what arrangement the channels are in. ("speaker arrangement"). Seems like a part of the VST spec that Reaper might want to (PLLLLEEEEEEASE) implement given this new enhanced surround capability.

PLEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEASE.

I have described the problem on U-he forum at KVR, and here is a part of his answer :
That said, I would consider all hosts that allow for multichannel processing and that do not propagate the channel layout flawed. (In Audio Units specs this information is mandatory, as is in RTAS and I think also in VST3)
However, as this seems to be a major problem with many VST hosts, I'll think about a way to unlock multichannel support no matter what.

Good news, isn't is ?

[martin.leese]

Originally Posted by Kihoalu

I did octaphonic work many years ago using 8 speakers for playback (one in each corner of a cubical space), since then I have not heard anything that was true 3D (and of course, 5.1 and 7.1 are not). Ambisonics can do 3D but this is a term that was invented after I did my 3D work, and I have not heard any actual "Ambisonic" materal.

If you want to hear Ambisonics then there are over 200 pieces available for free download from Ambisonia.com. Almost all of these are 3D (full-sphere). If you want 3D then you will need to download the B-Format versions and to use a decoder. Ad-hoc decoders are available (and are listed here).

Regards,
Martin

[malcolmj]

Originally Posted by jm duchenne

I have tried on my 17.1 system ...

I would love to see a photo of your control room!

[schwa]

Originally Posted by jm duchenne

I have tried on my 17.1 system

Originally Posted by malcolmj

I would love to see a photo of your control room!

They are just special headphones.

[Kihoalu]

.

Originally Posted by martin.leese

If you want to hear Ambisonics then there are over 200 pieces available for free download from Ambisonia.com. Almost all of these are 3D (full-sphere). If you want 3D then you will need to download the B-Format versions and to use a decoder. Ad-hoc decoders are available (and are listed here).

Regards,
Martin

Thanks for the Ref! - will check. I am familiar with making suitable decoders using Op-Amps and faders. There should be a Reaper plug (or several) about now.

.

[nicholas]

Originally Posted by jm duchenne

Hope that it may help ;-)

Sure does, thanks a lot!

[martin.leese]

Originally Posted by Kihoalu

Thanks for the Ref! - will check. I am familiar with making suitable decoders using Op-Amps and faders. There should be a Reaper plug (or several) about now.

Plug-ins are listed separately here. Don't know if any will run under Reaper.

Regards,
Martin

[jm duchenne]

Originally Posted by schwa

They are just special headphones.

It would certainly be easier to find a home...

[jm duchenne]

+ ReaSurround: use standard surround speaker placements

(I know it looks the same as the old one, but this one's better. No, really. It is.)

Sorry, but no, this one is not better at all.
I will try to explain why with my terrible english... (I hope that you will understand nevertheless).

I think there is a confusion between real speakers positions and symbolic positions in such a spatialization tool.

If ReaSurround was based on ambisonic coding, it would be different, since the decoding of the WXYZ components would require to know the exact positions of the loudspeakers.
In a polar representation system, like in Sonar or Logic, where speakers are inevitably in a circle, it would be also different.

As a tool based on geometry and vbap, what is important is the radius of the "level control area" (sorry, I have no better words), which is symbolized in ReaSurround by the yellow circles.
In Samplitude, the default 5.1 position of the 3 front speakers is this one :
It is the opposite of the real speakers position, but very effective to do a front panning in cartesian coordinates, like in ReaSurround.

The representation of the yellow circles is also rather confusing, since the higher level value is not at its center (like in Samplitude), where is the speaker picture, but at the opposite of the plugin space center.
So, to place an input only on the left speaker for example, we need to move it at the extreme end (top left) of its yellow circle.

Without something that shows the level value (in dB or in %) it is hard to know it.
(EDIT)
It is a rather strange choice, that can result in very risky results when we move several inputs at once with edit knobs : sound can disappear and reapper abrupltly each time a blob leaves or enters a circle.
Why don't you put the max level at the center of the area, where the picture of the speaker is ?
And by the way, we really need a "pan law" for this amplitude control. The current one is very straight.
See the Soundfield editor in Samplitude.

Since the transition between sound/silence is very straight at the "external" edge of the circles, it makes it very difficult to play on precise positions.
Suggestion : a surrounding line that connect all the maximum level points could symbolize this maximum value perimeter will help to control the real level (but I think that a centered system would be preferable).
(EDIT)Another correlated possible mistake, is that when the inputs blobs are at the same position than the speakers pictures, the sound level is about 9 dB lower than it must be (and is when when they are at the edge).

It is a shame because with the Diffusion parameter we can "oversize" the speakers easily, but here they are already so "interlaced" that there is no use of the diffusion parameter.

It is not new with 0.7a, but the new default positions with less distant and irregular makes it harder to use.
Of course, free positionning is always possible...
But if the size of the speakers areas cannot be changed, possibly for each one, then the more the speaker symbols are equidistant, the more the users can control the spatialization.

Thanks !

[jm duchenne]

Originally Posted by musicbynumbers

I think what might be really good for this is a actions to assign 3 midi CCs to each of the 3 rotor dials so that we can select a reasurround, choose our 3 parameters, click on the input "node" to be affected and then start moving your "p5 midi glove" in the air!

EDIT: in fact we can probably do this now already by assigning CCs to it in the normal way.. off to try

Not yet. But soon ?

Hey, you have a P5 Glove too
Yes, iy will be fun...

[griz lee]

Originally Posted by jm duchenne

Sorry, but no, this one is not better at all.

It is a shame because with the Diffusion parameter we can "oversize" the speakers easily, but here they are already so "interlaced" that there is no use of the diffusion parameter.

It is not new with 0.7a, but the new default positions with less distant and irregular makes it harder to use.
Of course, free positionning is always possible...
But if the size of the speakers areas cannot be changed, possibly for each one, then the more the speaker symbols are equidistant, the more the users can control the spatialization.

Thanks !

Schwa - i think we're going to need some control over the size of speaker coverage area to make this fully configurable.

[jm duchenne]

Originally Posted by malcolmj

I would love to see a photo of your control room!

My previous studio, a few years ago, 32 channels compatible with 22.2 :





(this is not me who put the wallpaper on the ceiling !)

[malcolmj]

Originally Posted by jm duchenne

My previous studio, a few years ago, 32 channels compatible with 22.2

Nice!

It's been years since I last mixed in surround. Testing this plugin has made me remember just how much fun it is!

Cheers,

Malcolm.

[schwa]

Originally Posted by jm duchenne

the more the speaker symbols are equidistant, the more the users can control the spatialization.

Excellent point. The key observation is that the usefulness of a spatialization tool is how easy it is for the user to separate sound sources, not how well it represents the geometry of the listening space.

I can think of two possible ways to handle this:

1. Don't represent standard surround placements, and simply place the speakers visually around a box (as it was prior to pre7).

2. Do represent standard surround placements, but adjust the size of the default influence area for each speaker. So in a 5.1 surround setup, the front 3 speakers will each have a small influence area, and the rear speakers will have a larger influence area.

Regardless, we will eventually add the ability to control each speaker's influence area as well, and also change to the more standard behavior where placing a sound source directly on a speaker (with no source diffusion) causes it to be heard only from that speaker (vs the current behavior where you place the sound source so that the speaker is between the listener and the source).

[jm duchenne]

Originally Posted by schwa

Excellent point. The key observation is that the usefulness of a spatialization tool is how easy it is for the user to separate sound sources, not how well it represents the geometry of the listening space.

Exactly ! May I quote you if I need it ?

I can think of two possible ways to handle this:

1. Don't represent standard surround placements, and simply place the speakers visually around a box (as it was prior to pre7).

2. Do represent standard surround placements, but adjust the size of the default influence area for each speaker. So in a 5.1 surround setup, the front 3 speakers will each have a small influence area, and the rear speakers will have a larger influence area.

Yes...
Why not both ? I mean to series of Presets ?
Most DAWs offer two or more view types, because none of them can be perfect for everybody or every purpose.

Regardless, we will eventually add the ability to control each speaker's influence area as well, and also change to the more standard behavior where placing a sound source directly on a speaker (with no source diffusion) causes it to be heard only from that speaker (vs the current behavior where you place the sound source so that the speaker is between the listener and the source).

It is perhaps only a matter of drawing things.
What seems sure to me, is that the speaker picture at the center of the influence area lets suppose that the maximum level is at the center and that it decreases in all directions equally.
The current design works only as expected when the speaker is at the center of the space, like in Imax 6.1 or in 17.1 : the "fade out" area is equal in all directions !

And also : why is the source position restricted on the edges ? It makes actually impossible to have a correctly working 5 front channels in SDDS.

(the "1" blob cannot go higher than that)

[jm duchenne]

Originally Posted by schwa

(vs the current behavior where you place the sound source so that the speaker is between the listener and the source).

So, if there would be a drawing that shows how the level changes according to the input position (a little like in Samplitude), it would be something like this (for a speaker placed at the left side) ?

[schwa]

Originally Posted by jm duchenne

So, if there would be a drawing that shows how the level changes according to the input position (a little like in Samplitude), it would be something like this (for a speaker placed at the left side)

Not quite. The current design is more like a Venn diagram. If a source is within a speaker's influence area, the source's contribution to that speaker is its proximity to that speaker, relative to its proximity to all other speakers whose influence area includes the source.

Within a speaker's influence area, it is similar to VBAP in that placing the source on the same vector as the listener and the speaker maximizes the source's contribution to the speaker, but unlike VBAP in that even a point source will contribute to other speakers if it is also within their influence area.

In other words, a speaker's influence area (the gray circle) is only a fade zone relative to other speakers' influence areas.

This is all open to discussion though. It would be possible to change this design to straight VBAP, while still using the source diffusion and speaker influence area settings.

[griz lee]

Some kind of influence area control would be the thing I think.

I like this approach though schwa, it produces some stable images.

If a set of presets could contain the influence area information too, then i think you've cracked both problems.

Could we have a choice of alogrithm perhaps - vbap mode, or your amplitude/vbap mode?

[jm duchenne]

Originally Posted by schwa

Not quite. The current design is more like a Venn diagram. If a source is within a speaker's influence area, the source's contribution to that speaker is its proximity to that speaker, relative to its proximity to all other speakers whose influence area includes the source.

Within a speaker's influence area, it is similar to VBAP in that placing the source on the same vector as the listener and the speaker maximizes the source's contribution to the speaker, but unlike VBAP in that even a point source will contribute to other speakers if it is also within their influence area.

In other words, a speaker's influence area (the gray circle) is only a fade zone relative to other speakers' influence areas.

This is all open to discussion though. It would be possible to change this design to straight VBAP, while still using the source diffusion and speaker influence area settings.

It is not easy to have a mental representation, but, except for some cases, it works well.
I suppose that with some "area factor(s)", and perhaps a different way to show it, it will be OK.

[yerbouti]

Control of the influence area would really be great.

I can't really see advantages in using only straight VBAP.

Correct me if I'm wrong but I believe that the use of VBAP algorythm only occurs when working with height(3D)?

[schwa]

VBAP can be used in 2D (it's much simpler but the concept is the same).

[yerbouti]

Since we are on the subject,

We "officially" released the Octogris plugin yesterday. It is a free, open-source multi-channel sound spatialization tool (AU only, sorry).

You can get it here:
http://www.iact.umontreal.ca/site/?page_id=1091&lang=en

I guess with Reasurround it is now less relevant to Reaper users. But I'm sure there is a lot of creative way to combine those tools.

-Thanks the jm duchenne for helping us on that (I'm Raphaël from udm) and to the Cockos team for fixing the cocoa gui problem.

jm you could add it to your great website if you want.

[schwa]

Originally Posted by yerbouti

We "officially" released the Octogris plugin yesterday. It is a free, open-source multi-channel sound spatialization tool (AU only, sorry).

That looks like a fantastic tool, and it's extra fantastic that it's open source! Great work. All hail sheik yerbouti

[musicbynumbers]

Hear me out! (biggest benefit is for facilitating multi mix types in one session i.e. stereo and 5.1

I can see benefits to this for people wanting to send say 16 tracks all at once to a 16 imput bus with reasurround on it so they can go about automating the whole thing and having the volumes automation in the same place would be handy for the above way of working.

Not a biggy for me but thought I'd put it out there

[jm duchenne]

Originally Posted by yerbouti

Since we are on the subject,

We "officially" released the Octogris plugin yesterday. It is a free, open-source multi-channel sound spatialization tool (AU only, sorry).

You can get it here:
http://www.iact.umontreal.ca/site/?page_id=1091&lang=en

I guess with Reasurround it is now less relevant to Reaper users. But I'm sure there is a lot of creative way to combine those tools.

-Thanks the jm duchenne for helping us on that (I'm Raphaël from udm) and to the Cockos team for fixing the cocoa gui problem.

jm you could add it to your great website if you want.

Hi Raphaël !

I see that the GUI has evolved in a very nice one !

It is sure that with the ReaSurround (with some adjustments, like the speakers attenuation sliders on OctoGris , OctoGris and a lot of AcousModules will be of less interest for Reaper users !
Happily (?), there is still some people that use other DAWS ?...

[Herve]

Hello,

First, I do not often post on the forum but I am an avid reader. And often Jean-Marc give the response often before I understood the question ;-)

It is clear that Reaper is really taking the place of any other DAW. For my part with the recent arrival of ReaSurround I think to use Reaper only. This is a great chance to have Jean-Marc and the developer of Reaper very concerned with advanced surround sound. Many electroacoustic music composer électronic musicien or multimedia user on Mac or PC (including myself) will be very attentive to what comes out of Rea Surround. And we're very close to a truly unique result.

So... Some comments:

1°) About the graphic representation of the diffusion area

As a user old samplitude, I never found a equivalence for the clear representation of this value. The choice of the curve coupled with the variation of density representation of concentric circles in 2D space seems to be the best way. Another approach with Venn diagrams would be interesting, but seems a bit more difficult to understand directly. And a 3D approach was nice but maybe complicated to code and heavy for the CPU Usage, the Rea Surround plug-in must be considered as a tool that can be put on each track for a complicated mix so it must be very efficient.

2°) The input are displayed by Number and the output are displayed by letters maybe if we have the choice to rename the output or to use number for displayed output it's more practical. Because beyond 26 outputs we return to A.

3°) Display a grid (or not and choose the scale), choose the proportions of the room where we would put the speakers would be really helpful.

4°) Minor graphical bug
After you decide to replace the «*speaker D*» by the LFE and you go back, the name display of LFE in not replaced by «*speaker D*» again.

5°) The more than 16 input or output are not displayed


A simple general note for Reaper on the curve of volumes in the mix is really very boring. It seems like only linear had already explained Jean-Marc Duchenne in another post. A simple alternative in the preferences in logarithm scale would be really a plus. A decrease 6db only in half part of the of volume knob is really not suitable for sound design.

Thanks in advance

Best

Hervé