Hi Dietz,

congratulation for this great product. I recently finished testing and have bought a license.

But there is an issue that should be further analysed. It is about how the Mir engine handles the dry signal. My understanding is, the Mir virtual mic polar patterns are applied to the dry signal as well. Checking all the virtual mic factory presets for me they feel like they do a spatial distortion of the spatial image that creates tension. Feels like a lens error of an image looks like.

How did I solve it? By claiming that the law of energy conservation is also valid for sound. That means, that for every angle of the mic polar patterns the amplitude square sum of left and right channel has to be constant. So I created a small python script for a simple cardioid setup with variable mic angle and calculated the StdDev from the equal power law for several mic angles.

Result: mic angle for cardioids (50% directivity, only 2 capsules for L/R, no center capsule) which leads to minimum error of energy conservation is about 67°.

Application in Mir: Concerning spatial image distortion this simple 2-capsules-setup is by far the best I have tested so far. I expect that this can be further improved by a more complex capsule design.

If someone is interested in the subject, I can provide the python script and diagrams.

Hi Vokars,

great to hear that you liked MIR 3D! Hope you enjoy your new tool. :-)

You are absolutely right that the dry signal gets ENcoded to Ambisonics before it gets DEcoded together with the newly created, positional impulse responses. The underlying idea is to match the virtual position of the source as exactly as possible with the position of the originally recorded IR (which doesn't include any direct signal for exactly this reason).

The decoding process depends on a multitude of parameters, like stage geometry, position of source and virtual mics and their angular relations, the directivity of the source signal, the number and shape of virtual capsules or the spherical coefficients. It is also important to understand that a stereo-signal is always represented as two (linked) mono-sources in MIR. Consequently, there will always be, or better: there _has_ to be some kind of imaging distortion - actually this is one of MIR's basic paradigm's. :-)

If you want to preserve the stereo image of your source by all means you should use MIR 3D as some kind of "glorified AUX send reverb": Use an AUX send from your source track to feed an AUX bus with a MIR 3D instance that's set to "100% wet". That way you will lose MIR's abilities to take care for the dry signal's positioning without sacrificing the wet signal component.

HTH,

Hi Dietz,

thank you very much for your fast response. The idea to use Mir with 100% wet setting and mix it outside Mir with the wet signal is a good solution in many situations. Thank you for this advice!

Did you give my proposed capsule setup a try? Very simple: +-67° direction, 50% directivity, distance minimum 15cm, I use 35-70cm. Instrument width should be adjusted to the original dry signal at the recording distance. How does it sound to you? Do you think, it makes sense to regard energy conservation in general as a constraint for some factory mic presets? I have to admit due to my education (physicist) I can't image anything else especially because theory and experiment are so perfect in line with each other.

By the way: The Blumlein crossed Fig. of 8 also fulfills the energy conservation, because with Fig. of 8 the math equation sin^2 + cos^2 = 1 is always true for the power sum of left and right channel. But: It contracts the -90°/+90° listening angle interval to -45°/+45°!

Also that means: It could be a good extension for the Mir capsule designer? A button for optimize a certain capsule design for minimal image distortion by fitting one paramter (e.g. mic angle, ..) with a state of the art standard minimizer using channel energy sum StDev as cost function.

Hi Vokars,

perhaps it would be a good idea to explain what you are actually trying to achieve. As I wrote above, MIR is a multivariate system. There are so many influencing factors that your results will seem arbitrary quickly in another context.

... could it be that you're actually reverse-engineering Ambisonics, inadvertently ...? ;-)

Kind regards,

Hi Dietz,

I try to explain, that mic capsule presets, that are not in line with the laws of physics, are experienced as some kind of artefact, that do not sound naturally. That means: The fact, that a user can design any capsule design he wants, means that in most cases these designs doesn't sound well as long as they don't fulfill basic physical requirements / constraints.

A picture / sound says more than 1000 words: My propsed capsule design (only 2 capsules with optimized direction) is configured in below 1 minute. Just give it a try and compare it with the dry signal and then compare it to other capsule presets. That's done in below 5 minutes and you will understand, what I mean.

I am fine with my solution and I can use the capsule designer for my needs, but I think this is really an opportunity to make the Mir product even better and help many users out there to use the Mir engine in an optimal way by optimizing the capsule presets for minimal imaging distortions.

@Dietz said:

Just to be absolutely sure that I understand the issue correctly: Your old Cubase projects use (legacy) MIR Pro 32-bit? Then you have to make the transition to MIR 3D on Cubase 11 or earlier indeed. There's not much we can do about ...

Hi Dietz, 

apologize for my late response. I'm not 100% sure but I think so because Cubase tells me that 32 bit plugins are no longer supported. Anyways, I sent one project I mixed using the old MIR Pro to VSL so they can also test it on their computers. 

@falineleo said:

Just wish the 7th roompack - outdoor stage as like forest, lake...

Ohhh yes. The recording of forest IRs has been a dream of mine since the first conceptual studies of MIR were conducted in 2002 or 2003. We have thought a lot about how to realize this idea, but so far no one has come up with a solution that would ensure that the forest is absolutely quiet for at least 12 hours - no animals  and insects, no wind, no air traffic or cars, church bells, ... Any noise that occurs during recording will introduce strange-sounding artifacts into the final IR, and considering that it takes us about half an hour to record a single (!) position under ideal conditions, this would be a daunting endeavor. (For comparison, the MIR Venue we created from Synchron Stage Vienna contains nearly 80 source positions, each consisting of 8 directions ....)

If you happen to have an idea how to overcome these obstacles, I'm all ears! 😉

No idea of how to silence a forest 😊 but for me it would be great to have some "home" rooms and ambiences, like kitchen, garage, bedroom, car that could be used also for film postproduction purposes maybe

Hi Dietz,

actually 2 important dimension of this world: 1. wish and 2. truth

1. Wish: "But ... we _want_ these distortions, really! :-D Otherwise I wouldn't need MIR, would I?" ok, you try to explain to me, that bug could be feature .. Hm .. Some people may want that, other may not. I am one of these guys who wants the by far best reverb and imaging engine without having distortions. I expect not to be the only one. HOA of course is much of design and not just reproduction. But also here design can be natural e.g. when creating binaural sounds, where the HRTF has to be designed.

2. Truth: Assume there is a huge market for people, who want the best reverb but everything based on physics (we could call these people "conservative"): Does my proposed capsule preset would help all the people in this huge(!) market segment? I still can't wait for a feedback (I am aware that truth can hurt :( ) :)

Hi Vokars, 

I really try to grasp what you expect to achieve. You wrote,

@Another User said:

Did you give my proposed capsule setup a try? Very simple: +-67° direction, 50% directivity, distance minimum 15cm, I use 35-70cm.

... to begin with, I don't understand what you mean by "15 cm", or "35-75", sorry to say so. If you mean the distance-parameter in MIR 3D's Output Editor, then this is a purely virtual value that actually controls an artificial de-correlation of the late reverb tail. It doesn't affect the dry signal at all. Or do your refer to the grid on a Venue's stage? That would be indeed meters in reality, so again I don't know what to answer. 😕

Care to post a screenshot, please?

I see. Thanks for the clarification, Vokars.

If you look into the section "Factory Presets 02: Classic Stereo" you''ll find 1st Order settings called "Virtual ORTF" and "X/Y Wide" which should be pretty similar.

But again: This "Distance" parameter has no effect on the dry signal. Ambisonics is by definition a coincident microphone array, there is no distance between the capsules (apart from the physical limitations of the real world, of course).

-> Our Ambisonics Mics

Hi Dietz,

thank you for your feedback:

1. Yes ORTF is simliar.

2. Does it mean, you agree, that with this setup image distortion is low?

3. If yes: The critical issue now is: Of course the right angle could have been determinded by trial and error or by using a factory preset derived from tradition and experience. But in this case the correct angle was calculated simply from a physical law. And if I can do that, you can do that too. This example was just kept simple (2 capsules, stereo output), to explain it as simple as possible. But with this algorithm now 1. a machine can optimize physical correct 2. complex capsule presets and 3. multichannel output, where the trial and error method is not feasble as it is in this simple example with just one degree of freedom (z-direction angle).

I hope this helps.

Best regards

vokars

Dear Vokars,

As I said: There is no "absolute" value in MIR. As soon as you move away the source from the virtual microphones a meter or two, the decoder will already be narrowing the angle between left and right source track, thus "distorting" its imaging again - despite your carefully sculpted mic setup.

Thing is: We couldn't care less about it, because it was one of the main reasons MIR has been invented at all: To have the perceived width of a source change with its distance from the listener. This is what happens naturally in the real world, but it was always a bit cumbersome to achieve artificially with the tool-set of classical audio engineering. - MIR takes care for that all by itself, and we like it. 8-)

If I want my audio untouched, I don't feed it into MIR. :-)

Kind regards, 

Hi Dietz,

concerning: " As soon as you move away the source from the virtual microphones a meter or two, the decoder will already be narrowing the angle between left and right source track, thus "distorting" its imaging again - despite your carefully sculpted mic setup"

I solved that calibration issue by what I wrote above: Adjust the width of the dry signal with MIR switched on with the width, when MIR is switched off by listening (ignoring the width value shown in meters): At which width in MIR do both sound equal broad? This makes only sense if you position the instrument at the right distance from the instrument for this calibration process: Distance Mic to instrument in the MIR positioner during this calibration process should be the real estimated recording distance at which the dry signal was originally recorded. With that calibration procedure I am happy at the moment. Edit: That means, the positioner works fine with this calibration at least in my tests. It reliably adjusts width, dry/wet, and lowpass (if air absorption is switched on). I didn't check predelay (could measure some impulse responses with different distance settings). In my tests, no additional distortion occurs. With distance change other physical laws apply compared to a pure rotation that doesn't change energy: With larger distance the energy intensity (energy per area) of the dry signal approximately decreases with the area of a sphere (different dry wet ratio), high frequencies are absorbed more than lower ones, because this loss is velocity dependend and high frequncies have higher velocity (if air absorption switched on, this is usually modelled with a lowpass filter). Width is more complicated, because high frequncies behave approximately like beams, lower ones more like spheres (consequence of diffraction). Therefore my assumption: If the positioning makes problems then probabely there is a problem with the calibration process as described above.

Edit: And the EQing has to be done well. Otherwise there are distortions in the distance awareness e.g. higher pitch feels more distant. E.g. Piccolo and Eb clarinet high pitch sound too far away without the VSL character EQ but much better with the Teldex presets in Teldex venue. On the other hand I have found, that some optimizations could be done in some cases with the EQ presets. I am currently developing a model based matching EQ primarily for electric guitar und electric bass guitar. But I realized, that it also can find a critical 500Hz issue in the character EQ of the Eb clarinet (VI series). When published in a few months there will be a 1/2 octave freeware version that could save your team a lot of time when doing the EQing (1/3 octave version will be very affordable).

Best regards

vokars

I did a video review of MIR Pro 3D on virtualorchestration.it. Maybe it could be of interest to someone :)

Hi all,

just in case you missed the release of the most recent, quite unusual demo for MIR 3D RoomPack 7 - here is the link:

-> Stereo

-> Binaural 3D (headphones only!)

It was created for Karl Böhm Hall specifically by electronic music artist extra-ordinaire Dorian Concept, mixed in stereo as well as in 3D by Yours Truly. 8-)

Enjoy!