[dr cornelius]

Curious about this:

https://auralid.genelec.com

 

[audiohobbit]

This is a service to generate your personal head related transfer function. Normally this would require you in an anechoic room with a sphere consisting of many loudspeakers around your head, in-ear microphones in your ears and a lot of time.

Genelec will offer to do this by simulations, with so called BEM software (Boundary Element Method. That's similar to the Finite Element Method with which I'm working myself professionally in automotive engineering. BEM is more useful and more often used for acoustic calculations).
There is a finnish software company that develops such software: http://www.kuava.fi/software-solutions/waveller-fbem/ (scroll down a bit and you can see a video with a head-model for HRTF generation by simulation.
I would not be surprised if Genelec, which is also a finnish company, is working with this software and/or company. I also found some papers on this topic and on the Waveller software by finnish universities.

The somewhat new and special thing about the Genelec service is that they will generate a 3D model of your head and ears that will be used for the simulation only by using a video of your head and ears that you can take yourself with your smartphone for example.

I've been talking to a guy from Genelec at the recent High End show here in Munich. He said that in June there will be more info on this service online on their web-pages, and that the HRTF generation should cost "only" 500 euros. If this is correct (no guarantee!) it would be not much for this kind of service. (Since I'm working myself in simulation I can roghly guess how much work and cpu time this needs).

You already can find prices online for such a service, also at a finnish company (coincidence...? I doubt it): https://holvi.com/shop/waveller-hrtf-service/
And this is without the 3D modelling of your head and ears.

What you finally get is a file in the so called SOFA-format: https://www.sofaconventions.org/mediawiki/index.php/SOFA_(Spatially_Oriented_Format_for_Acoustics)

This is used mainly by professional software and the service is mainly aimed at professionals.
Therefore you can at least hope that the whole process is accurate enough compared to a real HRTF generation in an anechoic room.

Compared to processes like the Smyth Realiser for example (I own the A8 and wait for the A16), the SOFA file contains no information about room and loudspeakers, so also no reverb at all.
Listening to music with this would sound very very dry and lifeless (just like in an anechoic room). At least I think so, I can only speculate because I have no experience with these SOFA files.
For stereo or immersive movie sounds the software has to add some sort of reverb/room impuls response I think.

 

[audiohobbit]

This is a service to generate your personal head related transfer function. Normally this would require you in an anechoic room with a sphere consisting of many loudspeakers around your head, in-ear microphones in your ears and a lot of time.

Genelec will offer to do this by simulations, with so called BEM software (Boundary Element Method. That's similar to the Finite Element Method with which I'm working myself professionally in automotive engineering. BEM is more useful and more often used for acoustic calculations).
There is a finnish software company that develops such software: http://www.kuava.fi/software-solutions/waveller-fbem/ (scroll down a bit and you can see a video with a head-model for HRTF generation by simulation.
I would not be surprised if Genelec, which is also a finnish company, is working with this software and/or company. I also found some papers on this topic and on the Waveller software by finnish universities.

The somewhat new and special thing about the Genelec service is that they will generate a 3D model of your head and ears that will be used for the simulation only by using a video of your head and ears that you can take yourself with your smartphone for example.

I've been talking to a guy from Genelec at the recent High End show here in Munich. He said that in June there will be more info on this service online on their web-pages, and that the HRTF generation should cost "only" 500 euros. If this is correct (no guarantee!) it would be not much for this kind of service. (Since I'm working myself in simulation I can roghly guess how much work and cpu time this needs).

You already can find prices online for such a service, also at a finnish company (coincidence...? I doubt it): https://holvi.com/shop/waveller-hrtf-service/
And this is without the 3D modelling of your head and ears.

What you finally get is a file in the so called SOFA-format: https://www.sofaconventions.org/mediawiki/index.php/SOFA_(Spatially_Oriented_Format_for_Acoustics)

This is used mainly by professional software and the service is mainly aimed at professionals.
Therefore you can at least hope that the whole process is accurate enough compared to a real HRTF generation in an anechoic room.

Compared to processes like the Smyth Realiser for example (I own the A8 and wait for the A16), the SOFA file contains no information about room and loudspeakers, so also no reverb at all.
Listening to music with this would sound very very dry and lifeless (just like in an anechoic room). At least I think so, I can only speculate because I have no experience with these SOFA files.
For stereo or immersive movie sounds the software has to add some sort of reverb/room impuls response I think.

Edit: Experts say that simulated HRTFs are even more accurate than measured HRTFs, read here: https://www.slideshare.net/TomiHuttunen1/ownsurround-hrtf-service-for-professionals

“… virtualisation by using the simulated HRTFs is preferred over virtualisation by the measured HRTFs …”

That is in principle correct for many types of simulation but is highly dependent on the accuracy of input data. in this case the accuracy of the 3d models:

The simulation needs a triangulated surface mesh for the head and torso • Obtained using 3d scanning: • Structured light scanning • Photogrammetry • Laser scanning • Well-known requirements for resolution*: • <2mm on pinna • <5mm elsewhere

I think photogrammetry is the keyword here, that's what Genelec is going to do. Here a video with a comparison between structured light scanning and photogrammetry with a smartphone camera:
(I have to watch it in full length with sound at home)

 

[bearcatsandor]

Compared to processes like the Smyth Realiser for example (I own the A8 and wait for the A16), the SOFA file contains no information about room and loudspeakers, so also no reverb at all.
Listening to music with this would sound very very dry and lifeless (just like in an anechoic room). At least I think so, I can only speculate because I have no experience with these SOFA files.
For stereo or immersive movie sounds the software has to add some sort of reverb/room impuls response I think.

Can't you just capture and add room/reverb to the resulting SOFA file?

 

[Maestroso]

Has anyone tried it yet?

 

[edwardsean]

Has anyone tried it yet?

ditto

 

[SuperEB]

Just placed my order for Aural ID. Will report back.

 

[audiohobbit]

Cool. How do you plan to use the SOFA-data?

 

[SuperEB]

I have two setups - I'll start with my recording studio DAW and use the Sparta Binauralizer plugin which can load sofa data, playing music through the DAW. Once I'm happy with the the settings, possibly adding some artificial ambiance as well as a headphone EQ curve (just switched from SR225's to Sundaras!), I'll created a stereo impulse response file which will be the sum of all of the above processing. My "portable" setup (linux laptop + Asgard 3 w/DAC) will have an impulse deconvolver plugin loaded using Carla between the media player and the output going to the DAC.

 

[SuperEB]

@audiohobbit

I see you have SMYTH Realiser as your icon. How do you like it?

EDIT: Never mind - just re-read this thread. I read it a couple of weeks ago but joined today.

 

[Maestroso]

@audiohobbit

I see you have SMYTH Realiser as your icon. How do you like it?

EDIT: Never mind - just re-read this thread. I read it a couple of weeks ago but joined today.

hi superEB,

Have you had a chance to try AuralID already?

btw, I own an A16 and it’s great!

 

[SuperEB]

I just got my Aural ID this morning, but only got to listen for a few minutes. Definitely does what it says, but it is like listening in an anechoic environment. I'm fine with that because my desire for years has been to be able to listen to speakers in a room without any colorization from the room, as I get enjoyment out of listening very analytically to how recordings were made. What you are getting with aural ID is the ability to place sound sources at any angle relative to your head, but since there are no early reflections, there is little sense of distance. It is definitely outside of your head, but is it close? Is it far? It's hard to tell.

HRIR's are used in angular perception, but that our ears/brain need to hear reflections off of nearby surfaces to discern the distance from the sound source. I plan on experimenting using BRIR's and/or artificial reflections for the fun of it,

 

[Maestroso]

Thanks! How is elevation? If you find the time, it would be cool to know how well mixes translate between systems. I’m especially interested in the spectral accuracy.

As for ER/tail: You might want to try IRCAM SPAT with your SOFA file. Ircam forumnet offers the max patch for free (alternatively, you can try the Flux standalone).

 

[SuperEB]

Elevation seems to be accurate. I wasn't at first, due to the perceived frequency response created by the combination of my ears and headphone drivers. Everything seemed high by 15-20° and some sounds were shifted slightly right. The left-right difference started at around 2kHz, gently sloping to a max difference of 0.6dB at 13kHz. I created a 1/12 band EQ in Reaper, played pink noise one band at a time and then adjusted the pan for each track (frequency band). I was able to use this data to apply a frequency-left-right-balance compensation curve. I pulled the rtings measurement data out of AutoEQ, smoothed it to 1/3 octave and had REW create a "rough" correction filter. In time I will create a more accurate curve to match my headphones and ears. All of this put things back in their correct place in azimuth and elevation.

I also replicated the first ceiling, lateral reflections and rear wall reflections with the correct delay, angle and attenuation for a treated studio control room. The effect is subtle but definitely improves localization across the stereo field. Everything is a bit more focused in its position. After setting this up, I was listening with my eyes closed. It was spooky to open my eyes facing my studio monitors and have the sound coming out of them even though they weren't turned on!

 

[hazenut]

@SuperEB What kind of headphone do you use with aural id? I'm sort of want to test it out. I know a couple of headphones are diffuse field tuned, however, even with er4sr, I have perceived quite a bit of uneveness in the top octave FR, the latter in fact serves no small amount of relevance in pinpointing the exact location of things (various data give that 5k-15k range is extremely important ). Those distortions are undesirable. Circumaural headphones like HD600 might be good(haven't got one to try, have more expensive and better cans but simply don't work for virtual surround sound), but the ideal ones would be those with very smooth and consistent FR so they can be easily EQed to a diffuse field. What's your experience on this?
Another thing is genelec has not specified how many datapoint do they collect. It seems it's 836, which correspond to points on a sphere of fixed distance from the center of the head. I have seen other services(e.g. Waveller HRTF Service) provide more than one distance data. Kind of interests me, because I always itch to do some virtual sound field creation (like higher order ambisonic stuff). With accurate headphone virtualization, that's finally realizable. I'm also interested in tools dedicated for this kind of creation. Currently there are too many format conversion, encoding/decoding tools. They have limited ability of positioning and tweaking sound objects and handling reverbs, perhaps with their main focus on studio mixing and fixed multichannel format in mind. What I imagine is tools that enables fully editable, time varying positioning in 3 dimensions(including depth), sound transmission parameters(air absorption, damping, etc), reverb settings not limited to particular type of scenarios, such as hall, room, or regular shape, but fully customizable, placing objects in a virtual scene that alter the sound in tangible way. I'm sure such thing exist, or some workflow to get similar result, otherwise there would be no 3d audio in games or movies. Putting that thing to work in a standalone way to create art is extremely tantalizing/
@Maestroso I've heard of SPAT long ago, tremendously advanced tool with staggering price. Don't know there is a free version (even so that's max and still need to get max/msp, perhaps?), will check it out.