[esimms86]

IMO the video is highly deceiving. It doesn't say anything ABOUT Mesh2HRTF, just what you possibly could do with the results. You don't get the slightest idea of this video how complicated the numerical simulation of an HRTF actually is!

I had a quick look at the project pages: https://www.oeaw.ac.at/isf/das-institut/software/mesh2hrtf and https://mesh2hrtf.sourceforge.io/
My professional background as a mechanical engineer working over 18 years in numerical (FEM - finite element method) simulation tells me that this is far from easy and will have a very very steep learning curve!
In numerical simulation you'll need a few years of experience just to get the basics right.
Yes, FEM simulation is now even included in most CAD programs and you can create a mesh with the press of a button and get some results, but you don't have any clue what you're really doing and where the many many pitfalls are...

And I suppose with the BEM (boundary element method) used for acoustical simulations it will be the no different.
I'm experienced in the FEM simulation but not at all in BEM. We occasionally do some acoustical simulations with FEM but normally not airborne noise for high frequencies (because in FEM you have to mesh the actual air volume with a mesh seed many times finer than the shortest wavelength of interest, so it's basically used for low frequencies, in my case e.g. car interiors).



Simply said: No.
Not at all. Why should they?
If one poses this question it easily shows that one has not understood the difference between a (simulated) HRTF and the (very easy in comparison) PRIR-capturing method.
No offense!

No. Just the HRTF in SOFA format as far as I understand.
You'll need a loudspeaker-room simulation to "feed" the HRTF, otherwise you'd hear a perfect sound source, but in an anechoic chamber, and that I think even for surround movies would be far too "dry" sounding.

Yes, and by far the easiest way.


I don't think so. The ear canal goes very deep and has some curves. You would need a complete 3D model of the ear canal down to the ear drum, which imo is not possible with an external 3D scanning device.

Which leads directly to the first main problem of simulation of an HRTF: You have to very accurately capture the 3D shape of your ears and head.
See here: https://sourceforge.net/p/mesh2hrtf/wiki/Basic_tutorial_3d_scanning/

And my understanding is that they don't describe a reliable tested way to do this but mainly ideas etc.

Then you need to get these data into Blender freeware 3D software. So I think you have to learn one or two things about Blender first.
They have some plugins for Blender obviously. Then you have to create the surface mesh for the BEM elements in Blender.
Meshing normally is THE crucial part in numerical simulation. As I said you easily get a mesh with the press of the button but, at least in mechanical (and fluid dynamics) simulations the quality of the mesh (mesh size, geometry of the elements etc.) is what makes good results, and with complex geometries normally you have to make (lots of) manual corrections to the mesh. And this needs some experience.
As I said I have no experience in BEM and their mesh requirements, but I suppose there will also be many.

They have a tutorial, but as far as I can see it doesn't deal with highly detailled geometry data:
https://sourceforge.net/p/mesh2hrtf/wiki/Tutorials/

You'll need a mesh of the ears that may look like this: https://mesh2hrtf.sourceforge.io/images/Pinna.jpg i suppose.

Which leads to the next question: Performance
Numerical simulations (at least with FEM) at that detailled mesh level need strong multiprocessor machines with lots of memory and even then a simulation could take days.
Again: I have no idea about the performance of the BEM and their implentation of the solver.

If you want to get a complete HRTF you'll need a lot of point sources on a sphere around the head with the full freq response up to 20 kHz which I think will take lots of computation time.

Same is true if you want to measure a real HRTF (as shown in the video).
And this is where Smyth comes in and they say we don't need a full HRTF, but just for those spots where the loudspeakers are (and the room response comes as an add-on, normally not wanted when measuring a pure HRTF, but for a PRIR measurement it is what you want to capture) so the PRIR capturing is by far easier than capturing a real HRTF.

Finally with Mesh2HRTF it seems you need Matlab (which is no freeware and not so easy to use either) for postprocessing (but I'm not sure if really needed to just get the SOFA file).

Then you will have a SOFA file with your HRTF. Still no loudspeakers and room. And no headtracking either. (yes there's some talk about headtracking in the video, yes with a complete HRTF all the data needed will be there but how should the head movement be tracked? the video talks a lot about possibilities but no real solutions as far as I understand).


So I would say better to pay the couple of hundred euros (which is absolutely cheap for a numerical simulation...) to Genelec and let them do the work...



just for reference ( searching for the 3 authors Ziegelwanger H.; Kreuzer W.; Majdak P. )
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5321476/
(from 2016, saying s.th. about "Calculations were run on a Linux cluster consisting of eight PCs with Intel i7-3820 processors running at 3.6 GHz. Each PC was equipped with 64 GB of RAM. In total, more than hundred HRTF sets were calculated in this study and the calculations lasted several days.")

I don’t see where the Genelec SOFA would work for consumers interested in listening to audio and/or watching videos for enjoyment. After all, most consumers don’t want to experience music/movies inside an anechoic chamber. Also, directional information is provided but not distance. Beyond that, there is no head tracking provided in the current iteration, with no promise that it will be provided in the future(though, in all honesty, in that regard they’re in good company with Smyth Research). Still, Genelec SOFA sounds like a remarkable achievement.

 

[arnaud]

And I suppose with the BEM (boundary element method) used for acoustical simulations it will be the no different.
I'm experienced in the FEM simulation but not at all in BEM. We occasionally do some acoustical simulations with FEM but normally not airborne noise for high frequencies (because in FEM you have to mesh the actual air volume with a mesh seed many times finer than the shortest wavelength of interest, so it's basically used for low frequencies, in my case e.g. car interiors).

I agree that regardless of the simulation method, pitfalls always exist. One here is that it’s not just about the head shape but also the way skin affects how soundwaves bounce off your head (just like how a thin carpet does absorb some sound, or a grass field especially near grazing incidence etc.).

You'll need a loudspeaker-room simulation to "feed" the HRTF, otherwise you'd hear a perfect sound source, but in an anechoic chamber, and that I think even for surround movies would be far too "dry" sounding.

I personally do prefer dry prirs and have a feeling hrtfs with an atmos feed may be good enough / preferable for me but indeed.

I don't think so. The ear canal goes very deep and has some curves. You would need a complete 3D model of the ear canal down to the ear drum, which imo is not possible with an external 3D scanning device.

Ditto.

Which leads directly to the first main problem of simulation of an HRTF: You have to very accurately capture the 3D shape of your ears and head.

It’s the same debate as ’can a prir from someone else sound good to me?’, except that there are more challenges starting from a simplified numerical model and benefits from lack of (intrusive) microphones to capture the response / lack of background noise / no loss of coherence due to head movements etc.

As I said I have no experience in BEM and their mesh requirements, but I suppose there will also be many.

Mesh quality and size could indeed affect quality of results to a small or large extent depending on how robust the bem solver is… The paper you linked shows that , in order to minimize localization errors, simulation should use the subject head geometry with sufficient resolution of the pinna.

Which leads to the next question: Performance
Numerical simulations (at least with FEM) at that detailled mesh level need strong multiprocessor machines with lots of memory and even then a simulation could take days.
Again: I have no idea about the performance of the BEM and their implentation of the solver.

There are flavors of BEM solvers, but indeed memory and cpu time will scale with upper frequency of interest and how clean/detailed you want the head/torso/pinna to be… The paper you linked is actually quite nice with both objective and subjective (loss of localization) evaluation of simulated hrtf performance and influence of mesh resolution.

If you want to get a complete HRTF you'll need a lot of point sources on a sphere around the head with the full freq response up to 20 kHz which I think will take lots of computation time.

I’d say it‘s not a humongous job these days unless of old / poor solver implementation. As noted in the paper, reciprocity is used so, you end up with 2 sources and many receiving ‘mic’ locations so that’s a lesser issue (actually, for mesh2hrtf, given timings below, you have no other option that using reciprocity given how slow it is).

just for reference ( searching for the 3 authors Ziegelwanger H.; Kreuzer W.; Majdak P. )
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5321476/
(from 2016, saying s.th. about "Calculations were run on a Linux cluster consisting of eight PCs with Intel i7-3820 processors running at 3.6 GHz. Each PC was equipped with 64 GB of RAM. In total, more than hundred HRTF sets were calculated in this study and the calculations lasted several days.")

That feels like actually pretty poor performance for 100-200 frequencies to 20kHz (1 to 3 hours per frequency). Mesh2hrtf is based on fast multipole bem method which, due to iterative nature of the solver, may suffer from poor convergence (especially with distorted meshes etc) at lower , higher frequencies or both. The advantage is that it’s free software .

 

[shmau]

Do you actually own an A16 AES, and are you using it with a DAC 8?

Yes.

 

[littlej0e]

2024 for the Smyth Realizer for the remaining kickstarters (there are still some of those) and the prepays. That's at the earliest, could even be 2025 if then.

Can confirm. I was able to (hopefully) get one within 10 working days by essentially paying full retail price. At this point, I would be quite surprised if remaining Kickstarter backers ever receive their units without doing the same.

 

[batfier]

No... they're just not sending out the lower cost orders probably because they'd be losing money. If you pay retail they can probably ship within a week or two. As has been discussed several times in the past few months.

offering a product, collecting money, no real intent to deliver it? sounds like fraud.

limited to no public communication after collecting the money, almost no answers to emails, ...obviously no interest in their customers!

do you recommend to invest more money in a company like this? this is only rational, if you already have your unit and hope to get support in case of problems.

interesting to see that we as their customers/community accept whatever they do or especially don't do. due to missing official communication we're looking for reasons to accept these kind of behavior and even defend this behavior!

kind of prisoner dilemma?

 

[shmau]

offering a product, collecting money, no real intent to deliver it? sounds like fraud.

limited to no public communication after collecting the money, almost no answers to emails, ...obviously no interest in their customers!

do you recommend to invest more money in a company like this? this is only rational, if you already have your unit and hope to get support in case of problems.

interesting to see that we as their customers/community accept whatever they do or especially don't do. due to missing official communication we're looking for reasons to accept these kind of behavior and even defend this behavior!

kind of prisoner dilemma?

They have delivered most of the kickstarter units. Kickstarter backers know the risks when participating in projects and it is sadly common to not get anything back at all, so in that sense they've done better than many others.

 

[dsperber]

Yes.

I was unfamilar with the Okto DAC8, as well as its lesser stereo 2-channel output variant. So I did some online researching. Looks like quite the excellent device from its specs and glowing reviews.

Of course nobody (including Okto) is selling them (for almost a year it appears) because Okto is not taking orders, also because THEY CAN'T GET CHIPS!!! Sound familiar?

The Okto site doesn't have any info for ordering, or pricing or anything. You'd think they would have a simple disclaimer explaining that the product is currently not available for order or delivery, but I couldn't find that either.

Anyway, if the 7-month old listing is still active there is ONE of the stereo versions (which is perfectly suited for feeding a headphone amp) for sale on a Romanian 3rd-party marketplace named Audioweb. This particular item comes with a custom 12V high-output option (targeted for special amps I believe) so it might not be appropriate for every XLR-enabled headphone amp. The site is in Romanian but if you have Google translate in your browser you can quickly read it in English. Here is the URL for the translated-to-English version of the listing page. Unused, it is priced at 1130 euros, which apparently is around 100 euros over the original retail price. The listing includes links to two magazine reviews.

I was really just looking at its specs and reviews because I'd never heard of it until now. I don't actually need a replacement for my own superb 2-channel Audio-GD NFB9 DAC (Chinese made, stereo output) that I bought back in early 2012.

 

[shmau]

I was unfamilar with the Okto DAC8, as well as its lesser stereo 2-channel output variant. So I did some online researching. Looks like quite the excellent device from its specs and glowing reviews.

Of course nobody (including Okto) is selling them (for almost a year it appears) because Okto is not taking orders, also because THEY CAN'T GET CHIPS!!! Sound familiar?

The Okto site doesn't have any info for ordering, or pricing or anything. You'd think they would have a simple disclaimer explaining that the product is currently not available for order or delivery, but I couldn't find that either.

Anyway, if the 7-month old listing is still active there is ONE of the stereo versions (which is perfectly suited for feeding a headphone amp) for sale on a Romanian 3rd-party marketplace named Audioweb. This particular item comes with a custom 12V high-output option (targeted for special amps I believe) so it might not be appropriate for every XLR-enabled headphone amp. The site is in Romanian but if you have Google translate in your browser you can quickly read it in English. Here is the URL for the translated-to-English version of the listing page. Unused, it is priced at 1130 euros, which apparently is around 100 euros over the original retail price. The listing includes links to two magazine reviews.

I was really just looking at its specs and reviews because I'd never heard of it until now. I don't actually need a replacement for my own superb 2-channel Audio-GD NFB9 DAC (Chinese made, stereo output) that I bought back in early 2012.

They are selling them but they sell out in less than a day. I was able to purchase mine early this year. I think they are due for another batch soon. Seems like maybe around 1-2 months between batches currently.

 

[batfier]

They have delivered most of the kickstarter units. Kickstarter backers know the risks when participating in projects and it is sadly common to not get anything back at all, so in that sense they've done better than many others.

you are right regarding Kickstarter backers and most of them already received their unit.

the quoted response was for a pre order customer.

a pre-order is an order, it had a much higher price compared to the Kickstarter offering, but not such risks implied!

but now it seems the pre-order customers paid most of party here, with a much higher contribution and will receive NOTHING !!! ...not even a simple E-Mail...

but he, there is this "generous" offer to pay full price, which by the way increased over time as well... wow

if the money had been invested into stocks in the last 5 (five!!!) years, it would be almost full price equivalent.

 

[esimms86]

you are right regarding Kickstarter backers and most of them already received their unit.

the quoted response was for a pre order customer.

a pre-order is an order, it had a much higher price compared to the Kickstarter offering, but not such risks implied!

but now it seems the pre-order customers paid most of party here, with a much higher contribution and will receive NOTHING !!! ...not even a simple E-Mail...

but he, there is this "generous" offer to pay full price, which by the way increased over time as well... wow

if the money had been invested into stocks in the last 5 (five!!!) years, it would be almost full price equivalent.

Not to get political but the stock market hasn’t been doing so great since the invasion of Ukraine. Also, the stock market seriously tanked at the start of the pandemic. Up until now, of course, it had recovered significantly.

I offer no excuses for Smyth Research’s lack of communication and I certainly am not an apologist for them. Also, in the spirit of full disclosure, I am Kickstarter back #174 so I have had my A16 since late January 2021. Having said that, I honestly believe that SR fully believed that they were on their way to fulfilling all of their Kickstarter orders when (1)their financial situation took a downturn for unexplained(to the public) reasons, and then (2)the unfortunate AKM fire happened. The even more unfortunate pandemic may or may not have impacted their supply chain. So now they are sitting on a reportedly limited stock completed A16s that they apparently are holding onto to make available, for the foreseeable future, only at full price. They started out following a business plan that hasn’t worked out for them or their customers so now everyone involved is paying the price to varying degrees.

People who made pre-orders are paying the most, as this group is currently out a significant outlay of money with no units to show for their investment in money and time. I honestly feel for you.

Kickstarter backers who have yet to receive their units are out less money but more time. Again, I feel for you.

And then there are people like me for whom the glass is more than half full, as we already have our A16s. I can legitimately complain that (1)the Realiser Exchange has yet to go online, (2)I have yet to receive my CanJam PRIR, (3)I have been offered DTS:X for a fee(I’ve since paid up and received it), though it was expected to be included at launch, and (4)Auro 3D is not visible on the horizon and could potentially also become available for a fee.

Compared to the previous mentioned groups of A16 investors, my complaints by far pale in comparison, and I consider myself extremely fortunate, with comparatively little to complain about.

I don’t believe that it was/is Smyth Research’s intent to act fraudulently, and I do believe that they have indisputably demonstrated poor management of their business. Their lack of communication re: an extended period of over a year with no public statements and rare responses to emails continues to be one of their biggest shortcomings, driving the frustration of so many on this forum and elsewhere.

 

[GeorgeA]

Actually, the French business partner of Smyth Research has been much more active both in promoting Realiser A16 and in providing new information about firmware revisions.

 

[mrtim6]

Not to get political but the stock market hasn’t been doing so great since the invasion of Ukraine. Also, the stock market seriously tanked at the start of the pandemic. Up until now, of course, it had recovered significantly.

I offer no excuses for Smyth Research’s lack of communication and I certainly am not an apologist for them. Also, in the spirit of full disclosure, I am Kickstarter back #174 so I have had my A16 since late January 2021. Having said that, I honestly believe that SR fully believed that they were on their way to fulfilling all of their Kickstarter orders when (1)their financial situation took a downturn for unexplained(to the public) reasons, and then (2)the unfortunate AKM fire happened. The even more unfortunate pandemic may or may not have impacted their supply chain. So now they are sitting on a reportedly limited stock completed A16s that they apparently are holding onto to make available, for the foreseeable future, only at full price. They started out following a business plan that hasn’t worked out for them or their customers so now everyone involved is paying the price to varying degrees.

People who made pre-orders are paying the most, as this group is currently out a significant outlay of money with no units to show for their investment in money and time. I honestly feel for you.

Kickstarter backers who have yet to receive their units are out less money but more time. Again, I feel for you.

And then there are people like me for whom the glass is more than half full, as we already have our A16s. I can legitimately complain that (1)the Realiser Exchange has yet to go online, (2)I have yet to receive my CanJam PRIR, (3)I have been offered DTS:X for a fee(I’ve since paid up and received it), though it was expected to be included at launch, and (4)Auro 3D is not visible on the horizon and could potentially also become available for a fee.

Compared to the previous mentioned groups of A16 investors, my complaints by far pale in comparison, and I consider myself extremely fortunate, with comparatively little to complain about.

I don’t believe that it was/is Smyth Research’s intent to act fraudulently, and I do believe that they have indisputably demonstrated poor management of their business. Their lack of communication re: an extended period of over a year with no public statements and rare responses to emails continues to be one of their biggest shortcomings, driving the frustration of so many on this forum and elsewhere.

As a pre-order customer it’s still surprising that no updates are forthcoming. They took our cash to build their business, I appreciate there have been some circumstances that are not Smyths fault, however their lack of customer care seems to be their trademark

 

[jk6661]

In my experience, the only sure way to get a response from James is to buy something. I sent him a few support questions weeks ago and still haven't gotten answers. In the meantime, I offered to pay up for DTS and got a response within a day. So there's essentially zero customer support unless you fork over money, even if (like me) you paid full price for the A16.

 

[sander99]

Hi hi, maybe we should offer to buy information from Smyth then.

 

[littlej0e]

Hi hi, maybe we should offer to buy information from Smyth then.

Ok, this made me lol, then cry because I realized it’s shockingly accurate. It was a rollercoaster of emotions…