LTT Lab is going Brüel & Kjær

[sir_frigorifero]

my bad. i thought there wasn't a thread since i did not find it in the HP subforum

 

[MayaTlab]

IEM bass response varied wildly from user to user

Sorry for being a bit of a pedantic arsehole, but I believe that in the LTT video the IEM bass response variation is the on-fixture variation, not the variation between individuals.
In-ear measurements with IEMs might be very, very difficult to do. The only "in concha" headphones I'm personally half-confident extracting some data from in-ear measurements are the Airpods 3 given their leakage tolerance and their feedback mechanism.

 

[thewas]

my bad. i thought there wasn't a thread since i did not find it in the HP subforum

I don't think its your bad as possibly it belongs better in the HP subforum.

 

[AdamG]

I don't think its your bad as possibly it belongs better in the HP subforum.

Multiple threads combined here as requested.

 

[Haint]

Sorry for being a bit of a pedantic arsehole, but I believe that in the LTT video the IEM bass response variation is the on-fixture variation, not the variation between individuals.
In-ear measurements with IEMs might be very, very difficult to do. The only "in concha" headphones I'm personally half-confident extracting some data from in-ear measurements are the Airpods 3 given their leakage tolerance and their feedback mechanism.

Yeah I was aware, but we can safely assume the human to human variance on IEMs is 100x worse than even closed backs. This is why even $5 ones come with 10 different tip sizes.

 

[Vacceo]

I think that would have been more interesting as there are so many more speakers out there to test, but the niche of computer/tech/gaming users are heavily biased to using primarily headphones so it makes more sense for them. That being said, if they are only planning on releasing their findings in review/video format, there is no way they are ever going to be able to pump out enough content compared to other sources of this head/earphone data. The good thing is that such a mainstream source will be using objective data, which will further encourage manufacturers to make innovations now that a wider lens is cast on them.

What could be interesting is an objective approach to spatial audio on headphones. I have no idea if that´s even possible and I guess the whole idea of "spatial on headphones" works on a lot of psychoacoustics, but still, that´d be an interesting thing to explain with measures.

Gaming on headphones, to me, is like comparing a McDonalds hamburguer to a grilled angus ribeye: sure, both contain beef and they can be broken down to aminoacids, but that´s it.

 

[MayaTlab]

What could be interesting is an objective approach to spatial audio on headphones.

I think that you first need to know where the sound is meant to be located in a virtual space, and find methods to know whether or not individuals can accurately locate it without any other cues than sound.
Skimming over a few articles, quite a few of which related to VR, it seems that quite often it involves something akin to a game. Sounds are placed within a virtual space, and test subjects asked to point to it, sometimes even move towards it. Their results are scored.

 

[Lambda]

Oh

"spatial on headphones" works on a lot of psychoacoustics, but still, that´d be an interesting thing to explain with measures.

is it psychoacoustics or more physioacoustics?
first the game engine needs to have good model of your ear and head.

Head-related transfer function - Wikipedia

en.wikipedia.org

So i don’t know you might need to get your ear scanned at some point.
Maye in the future we can use CT/MRI scans.
But then the headphone is changing this of cause. so ideal this needs to be modeled as well.

If you don’t wan’t do mess to mush with complex models i would guess headphones with good "linear" group delay in the Mids - high frequencys and linear flat response are what you want.

 

[Vacceo]

Oh

is it psychoacoustics or more physioacoustics?
first the game engine needs to have good model of your ear and head.

Head-related transfer function - Wikipedia

en.wikipedia.org

So i don’t know you might need to get your ear scanned at some point.
Maye in the future we can use CT/MRI scans.
But then the headphone is changing this of cause. so ideal this needs to be modeled as well.

If you don’t wan’t do mess to mush with complex models i would guess headphones with good "linear" group delay in the Mids - high frequencys and linear flat response are what you want.

Those´d be the same principles applied to conventional stereo on headphones. Worst case, at least you have a good baseline sound and if the spatial configuration does not sound well, you can always turn it off.

 

[LearningToSmile]

Data driven approach to audio already showing up in their videos, still from the latest ShortCircuit(part of the Linus Media Group family of channels) video about Volvo S90 Recharge and its Bower & Wilkins sound system:

 

[amirm]

One of the challenges with B&K 5128 is that by itself it introduces variability in measurements. I found it very hard to seat IEMs for example. And the curvature of the "head" makes sealing headphones harder. You may argue that our head is also rounder than a flat plate but that roundness is variable. In my view it is better to get standardized measurements that are more repeatable than not.

I think the only reason to get 5128 is to brag about its cost and hence commitment to the testing which is what was done in the video. It is short lived though because as soon as the measurements come out and don't match with tests by others, riots start in streets and gets them nothing but grief. Maybe they have a big enough fan base to overcome this but in my view, they are going the wrong direction here.

 

[Lambda]

I think the only reason to get 5128 is to brag about its cost and hence commitment to the testing which is what was done in the video. It is short lived though because as soon as the measurements come out and don't match with tests by others, riots start in streets and gets them nothing but grief. Maybe they have a big enough fan base to overcome this but in my view, they are going the wrong direction here.

Sowing how much variance measurements can have depending on how they are taken is by itself valuable!
(but lets not go there again...)

Till now they did a good job in explaining measurement variances and averaging there results form relative large sample sets.
Lets see how this plays out.

 

[amirm]

Sowing how much variance measurements can have depending on how they are taken is by itself valuable!

Such variance cannot be quantified in any meaningful way as it highly depends on headphone type, how the headphone was put on the fixture, issues with the fixture itself, etc.

You have to step back from this and look at the high level picture and only then do headphone measurements make sense. A literal measurement and trying to defend the data, using methods like averaging and such, just doesn't work. It will result in nothing but arguments especially among people who have seen measurements different with other fixtures.

Cooking up your own solution can also be risky. Look at RTING's effort for example:

 

[SamV]

Yeah I was aware, but we can safely assume the human to human variance on IEMs is 100x worse than even closed backs. This is why even $5 ones come with 10 different tip sizes.

We wanted to measure the bass of IEMs on humans as well, and even got the equipment for it. But we couldn't get around building the rig during the rental period. We might still do it in the future.

 

[SamV]

I think that you first need to know where the sound is meant to be located in a virtual space, and find methods to know whether or not individuals can accurately locate it without any other cues than sound.
Skimming over a few articles, quite a few of which related to VR, it seems that quite often it involves something akin to a game. Sounds are placed within a virtual space, and test subjects asked to point to it, sometimes even move towards it. Their results are scored.

Finding a reliable way to do this objectively is on our agenda, but it may not be that easy. It's also not the highest priority test for the beginning.

 

[SamV]

Data driven approach to audio already showing up in their videos, still from the latest ShortCircuit (part of the Linus Media Group family of channels) video about Volvo S90 Recharge and its Bower & Wilkins sound system:

That's the plan. As mentioned in the notes, this was just a taster, we're working on improving the methodology.

 

[SamV]

One of the challenges with B&K 5128 is that by itself it introduces variability in measurements. I found it very hard to seat IEMs for example. And the curvature of the "head" makes sealing headphones harder. You may argue that our head is also rounder than a flat plate but that roundness is variable. In my view it is better to get standardized measurements that are more repeatable than not.

I think the only reason to get 5128 is to brag about its cost and hence commitment to the testing which is what was done in the video. It is short lived though because as soon as the measurements come out and don't match with tests by others, riots start in streets and gets them nothing but grief. Maybe they have a big enough fan base to overcome this but in my view, they are going the wrong direction here.

I don't think there's a way to get around the issue of variability in measurements, especially since the variabilities happen on humans too and can be non-linear. I agree with you on the IEM seal issue, but that problem exists on the 711-based fixtures too, although to a lesser extent.

We did consider getting a flat plate fixture, and we still might in the future, but getting a fixture that better represents human hearing seemed like the better move for the start. That was our takeaway from the human bass measurement survey we did. The 5128 does a decent job of representing the average bass, while Kemar would be too leaky and the 45CA would not be leaky enough. Nevertheless, for over/on-ear headphones, we are considering doing measurements on humans to supplement our HATS measurements.

Our general philosophy was to start from the top, and do the best job we can in every step down the chain. The first step was to get a HATS and we saw tradeoffs in either option: getting more representative data vs. being more compatible with Harman research. In the end, we felt more confident in being able to make our results be compatible with Harman research than trying to make sense of less-representative data.

 

[SamV]

Sowing how much variance measurements can have depending on how they are taken is by itself valuable!
(but lets not go there again...)

Till now they did a good job in explaining measurement variances and averaging there results form relative large sample sets.
Lets see how this plays out.

We haven't been able to find a way out of the measurement variance issue yet. It seems unless one does personalized measurements, headphone measurement results are always going to represent the average human, not any individual.

We are also not concerned with our measurements not matching what's already out there on the internet. If we can show that our measurements actually represent humans better, we should be good.

 

[fieldcar]

I'm really curious how you will approach this. It seems like @crinacle and MRS/super reviews (https://squig.link/) rely on a canal simulator and align the depth of the IEM to result in a 8.0KHz coupler resonance. I've liked this since it makes IEM graphs between most reviewer's somewhat comparable.
Source: https://crinacle.com/2020/04/08/graphs-101-how-to-read-headphone-measurements/

Whether that's correct is debatable, but you have to start somewhere. I feel like you'll have to do a bit of borrowing of methodology.

I'm sure you will get a lot of people riled up no matter how you approach it. Don't let it stress you out and good luck with the new and exciting position at LTT!

 

[Lambda]

We haven't been able to find a way out of the measurement variance issue yet. It seems unless one does personalized measurements, headphone measurement results are always going to represent the average human, not any individual.

Is it possible to buy or make replacement ears that are anatomically totally different and fit them to the head?
Sure every ear is different and it is good to have some "average" ear but as viewer i wonder how mush of a difference it makes
Old Mann Ears vs female kid ears for example.

Maybe its possible (and it would make a fun video) to make some silicone ear castings from the LTT staff

or just put some putty on the head