[travis-bickle]

WavesNX, Creative SFXI, JVC xp-ext1 and a bunch of others offer personalisation of HTRF.

But none of them allows interchanging of sound relevant parts within the hrtf afaik

 

[GeorgeA]

I’ve started a thread for the Smyth Research Realiser A16 Speaker Edition. You’re welcome to write your opinions.
https://www.head-fi.org/threads/smyth-research-realiser-a16-speaker-edition.960912/

 

[Thomasphoenix98]

Any customization like Adding head circumference and intra-aural distance with replace the info in the HTRF. I find the patent generic. And even with Atmos decoding on the Airpods Max something like Audeze Mobius with only upto 7.1ch sounds superior. I do think they will benefit from customization
.

 

[sander99]

I bet what they are doing/will do is just an automatic manLoud process, so basically just EQ, which is not really magic.

I don't think so. First of all the ManLoud requires your input about what you subjectively think sounds equally loud to you, and what Smyth wants to do is an automated process. Second the ManLoud is effectively changing the HPEQ, applying some additional EQ to the two headphone channels only. The processing of the PRIRs will involve changes to the actual room impulse responses, per virtual speaker, per lookangle, per ear. It would boil down to applying additional EQ per virtual speaker, per lookangle, per ear. (And hence, if it works, has much more potential than a ManLoud). I suspect they will try to isolate the direct sound portion of your personal "low-quality" PRIR, by looking at what happens first in the impulses (direct sound will arrive first) because that will not be influenced by the room - for higher frequencies at least - only by the frequency response of the speaker (maybe the use of their specific tweeter is required of which they know the frequency response).

 

[VandyMan]

Regarding personalisation of an HTRF... On the A-16, I think there are two ways Smyth can potentially customize a generic PRIR for a specific user.

The most effective would be to have the user create an HPEQ against a know reference (ie a specific model of headphones that has low sample deviation). They would then create an HPEQ with the "same" headphones on the dummy head that was used to create the PRIR. The difference between the two HPEQ would give you the custom result that you could apply to any PRIR created with that dummy head.

The second way would be very similar, except you measure a lot of users against the know reference and then create an averaged HPEQ. This would obviously be less accurate, although maybe not by much for most users. One could also use the averaged HPEQ data to make a custom PRIR into a generic PRIR.

 

[sander99]

The most effective would be to have the user create an HPEQ against a know reference (ie a specific model of headphones that has low sample deviation). They would then create an HPEQ with the "same" headphones on the dummy head that was used to create the PRIR. The difference between the two HPEQ would give you the custom result that you could apply to any PRIR created with that dummy head.

Sorry, no way this works.

Regarding personalisation of an HTRF... On the A-16, I think there are two ways Smyth can potentially customize a generic PRIR for a specific user.

The most effective would be to have the user create an HPEQ against a know reference (ie a specific model of headphones that has low sample deviation). They would then create an HPEQ with the "same" headphones on the dummy head that was used to create the PRIR. The difference between the two HPEQ would give you the custom result that you could apply to any PRIR created with that dummy head.

The second way would be very similar, except you measure a lot of users against the know reference and then create an averaged HPEQ. This would obviously be less accurate, although maybe not by much for most users. One could also use the averaged HPEQ data to make a custom PRIR into a generic PRIR.

Sorry, even less way this works.

Changing the HPEQ in whatever way is never guaranteed to give a good result with someone else's or a dummy head PRIR.
It might sometimes give some improvement in localisation, more often some improvement in timbre, often only for a subset of the speakers and at the cost of degrading some of the other speakers. A real correction is only possible by independently correcting the impulses per ear, per speaker, per lookangle. This is not possible by appying EQ to the 2 headphone output channels. And the information needed for this can never be obtained from a HPEQ measurement or from comparing two HPEQ measurements. A HPEQ can not contain any information related to how a sound is affected by bending over the top of your head for example. Imagine the following thought experiment: take two people who have identically shaped and sized ears, with also the part of the head close to the ears (as far as covered by the headphone cups) identical shaped and sized. But with different distance between the left and right ears, and the rest of the head completely different shaped and sized. Their HPEQs would be indentical (assuming identical mic positions when measured), but their HRTFs and PRIRs would be different!
It is for this very same reason that a manloud can not solve all hrtf mismatch problems in all cases.

 

[cukis350]

well it's another day, another preorder suckers asking the same question "will any of us that preordered an A16 will ever received one?" Has all the kickstarter backers gotten theirs? TIA

 

[batfier]

well it's another day, another preorder suckers asking the same question "will any of us that preordered an A16 will ever received one?" Has all the kickstarter backers gotten theirs? TIA

I would expect ... as usual (unfortunately) ... Nope and Nope

By the way, actually as preorderer we should have a valid contract. As a kickstarter you take some risk, loosing your money, but get a big discount. This does not apply to preorders as far as i know. Smyth's even provided an inital delivery date and promised several other Before kept quite completelly.

Simply frustrating...

On the other hand, wasting time with buggy hardware and firmware doesnt sound better

In 5 years or less, at least if the smyths continue in this speed or continue at all, this technology maid be obsolete.

If apple or other big players with deep pockets take on this topic, there will be a smartphone solution for it, may be with some additional hardware for the measurement or a more professional solution in an apple store or so.

The window is closing, wake up Smyth's guys.


p.s. sorry, must be the gray weather outside...

 

[VandyMan]

Sorry, no way this works.

Sorry, even less way this works.

I don't want to argue it here, but I disagree. It is just math. Feel free to msg me if you want to discuss.

 

[illram]

https://www.patentlyapple.com/paten...-related-transfer-function-hrtf-paramete.html
Seems Apple found a way to decouple the user specific hrtf parts for mixing with room measurements not done by yourself

Apple suing them out of business with a questionable patent before I get my pre order would probably be the most comical way for this whole thing to end.

 

[castleofargh]

I don't want to argue it here, but I disagree. It is just math. Feel free to msg me if you want to discuss.

He's right. You need to consider what the captured PRIRs are. They give information about a given sound source from a given position, and how that incoming sound is altered by the shape of your body+head when you look toward a given direction. Each created impulse has it's own information about one direction and one head looking somewhere specific. Even if the interaural delays were the same for you and the dummy head, with one EQ correction, you can expect to improve one direction. and with your idea, the direction you'd improve is where the headphone is. which is the last place we're interested in. The all point of HRTF processing is to stop feeling like the sound comes from the headphone. The final signal will still have the cues for given directions altered by the dummy instead of you.
This would work if from the start you had a HRTF that is very close to the dummy's.

Trying to remove deviations caused by the headphone is important, but it's an issue that comes on top of you having your own HRFT. Solving one doesn't solve the other.



As for averaging with your second idea, that's already what a dummy head is. Just listen to some recordings done using a dummy head instead of just 2 parallel mics, and decide how that works for you. For me it's been a solid disappointment anytime I tried, but my body isn't very average in the first place, so it's a sad but predictable result. Some people seem to enjoy those same binaural recordings very much. again, it's all about having the right body.

 

[arnaud]

Talking about buying a startup, rather than smyth, I’d see rather apple look into audio XD, who’s behind the speaker edition of the realiser.

I’ve listened to the french forum podcast where they discuss the implementation and, while still vague in the descriptions, I realized (pun intended), they weren’t merely replacing the headphones / hpeq with loudspeakers / xtalk cancellation but actually virtualizing the surround and ceiling channels from mathematical models derived by machine learning processes.

In essence, they appear to be making both magnitude and phase compensations to the surround channels not from personalized hrtfs but instead from heading specific models that represent « common denominator across population of hrtfs », extracted by means of machine learning.

There’s a bit of marketing gimmick and, I am dubious by principle since it’s well known human hearing uses idiosyncratic nature of head/ear shape to localize sources above and avoid cone of confusion etc. However, all who’ve given it a try were apparently looking for the speakers behind them…

What is fascinating here to me is that AI/machine learning opens doors for extraction of patterns from complex data that is an HRTF for distillation of generic features that appear to require little brain adaptation to make us believe in the validity of the virtual source location.

That is so much ahead than the mere convolution work that the realiser does from individual data, this realiser speaker edition almost feels like a trojan horse with the future being is in these math models (as we’re probably not too far away from software bsed realtime convolution rather than relying on dsp)…

cheers,
arnaud

 

[weijiangnan]

Does anyone know the current status of smyth? Last November said the first quarter of this year at the end of the delivery, but now no movement ah? I am a pre-order user.

 

[Thomasphoenix98]

Does anyone know the current status of smyth? Last November said the first quarter of this year at the end of the delivery, but now no movement ah? I am a pre-order user.

from the events that have unfolded in the past couple of years, pandemic, Chip factory fire, Chip shortage etc. I think trying to convert the order to a market price, seems to be the only realistic option to get the A16 in the short term.

 

[infernix]

Is there any way to adjust the first bands on a manLOUD (the two yellow bars) individually for the L and R channel? I can adjust the gain but only for L+R together. Am I missing something? An unlock toggle somewhere perhaps?