[rsaavedra]

Quote:

Originally Posted by Tyll Hertsens /img/forum/go_quote.gif I'm hoping that just doing 5 measurements with slight movements in between and averaging the five

I wonder if the "slight movements" part will be applied systematically and in identical manner to all headphones. What will those slight movements be like? Is it something like the following?
Measurement 1: Center of driver right on the center of ear canal.
Measurement 2: Center of driver half an inch below the center of ear canal.
Measurement 3: Center of driver half an inch above the center of ear canal.
Measurement 4: Center of driver half an inch in front the center of ear canal.
Measurement 5: Center of driver half an inch behind the center of ear canal.

There should be some way to make sure they can be replicated with sufficient accuracy.

 

[SP Wild]

Quote:

Originally Posted by xnor /img/forum/go_quote.gif I'm sorry, I just noticed that the head's curve in the overlay is totally off. No real similarities after all. :/ I will remove the image to avoid confusion. I'm not so sure about this. I assume that the head is built with an average characteristic, probably from doing tests with hundreds of people, in mind. Correcting the raw measurement data coming from the head with the smoothed correction curves from the manufacturer would just decrease accuracy of the results, imho. Noooo, (too much) smoothing is evil! Imo, not only does it get rid of the spikes that could be accounted for driver resonances (admittedly CSD would be more useful here) or construction imperfections ... but the smoothing could make any headphone look good. (and maybe sell well? That's not what this should be about, right?) Looking forward to those measurements. Thank you for doing this.

"Assuming" doesn't sit well with me, and "average" when things can be so diverse?

I also think graphs must be made as simple to interpret as possible. This is so that dummies like myself can actually make any sense of them.

 

[Tyll Hertsens]

Quote:

Originally Posted by rsaavedra /img/forum/go_quote.gif I wonder if the "slight movements" part will be applied systematically and in identical manner to all headphones. ...snip... There should be some way to make sure they can be replicated with sufficient accuracy.

I'm gonna make a little video of what I do when placing the headphones, but all I can say is that I'll try to make it as consistent from headphone to headphones as possible.

For example, some supra-aural sealed headphones are nearly impossible to move at all without breaking the seal, and some full-size headphones fit very close around the ear while others have a lot of potential movement.

Quote:

Originally Posted by SP Wild /img/forum/go_quote.gif "Assuming" doesn't sit well with me, and "average" when things can be so diverse?

Well, his assumption is correct, the head has an average but fully detailed ear. If I smooth the data but not the compensation (which does have detail from the individual characteristic of the average ear) we'll get an erroneous reading.

But here's the problem: A headphone is not the same acoustic signal as the sound source used in the DF or ID compensation. There should be a special compensation curve used for headphones, but there isn't one. (Maybe one result of what we're doing here will be such a thing.) I tend to agree with xnor that two speakers 30deg off axis to either side should be the sound field that headphones try to simulate. I'm not sure if it should be done in an anechoic chamber, however --- not that I wouldn't want to see the measurements.

As an annoying side note: It's a bit problematic that the head is an average of both men and women because it makes the head smaller than the average male by quite a bit. Since most headphone listeners are male, most headphone makers tend to make headphones that are a little too big for the head. This causes a situation where it's sometimes harder to seal the cans on the head properly. It would be hard to overstate what a pain in the butt this is; I'm constantly having to figit with cans to get them to seal. I've asked about this and those in the know indeed acknowledge the problem, but there's nothing that can be done about it. The spec for the head writen into the standard can't discriminate against half the population, so they'll never make a male average head.

Anyway, the head is carefully engineered to be avarage in the sense xnor means.

Quote:

I also think graphs must be made as simple to interpret as possible. This is so that dummies like myself can actually make any sense of them.

I am VERY focused on this idea. However, it's important to derive the simplified curves from accurate detailed curves. So it's accurate data gathering first; accurate compensations next; and then simplification of data last. But I'm with you 1000%, the end result of this entire measurement activity is a better understanding of the audio quality of headphones for as broad an audience as possible. That means getting to simple is crucial.

Having accurate data, albeit noisy with detail, will be fun for us geeks though.

 

[xnor]

*thumbs up* Tyll, I agree to everything you said in the last two posts, especially the nice and slow and fun for us geeks part.

About the 30deg measurement, I don't know either if an anechoic chamber would be the right thing, guess not. After all I have no idea what the right thing is but we're about to find it out, aren't we

But I don't know anyone who seriously listens/enjoys music in such a chamber.
Actually I was more thinking of an audio engineers treated (mastering) room with 2 monitors. Some people say that too much reflection absorbing material might kill reflections that are necessary for nice imaging, too little will cause deep nulls etc. in the FR.

 

[Tyll Hertsens]

Quote:

Originally Posted by xnor /img/forum/go_quote.gif About the 30deg measurement, I don't know either if an anechoic chamber would be the right thing, ..... Some people say that too much reflection absorbing material might kill reflections that are necessary for nice imaging, too little will cause deep nulls etc. in the FR.

The issue, I think, is that the sound is coming from a single source, at one angle if incidence, producing significantly deep and sharp nulls and peak indicative of the pinna reflections at that angle. With headphones the sound is not hitting your ear like a normal wavefront would as it's more curved 'cuz the source is closer, and at low to mid frequencies the length of the halfwave is longer than the longest internal dimentions and the sound is more 'coupled to' rather than 'propagated at' the ear.

So, mimicking two speakers in space is still not quite right.

 

[SP Wild]

Hope you guys work out the finer details and all the best with your endeavors. I'll be keen to follow the progress and will do so in the sidelines as I think I am just a little out of my depth. I have a feeling that this will be revolutionary and you guys will achieve something of absolute importance in this field. Cheers

 

[xnor]

Quote:

Originally Posted by Tyll Hertsens /img/forum/go_quote.gif The issue, I think, is that the sound is coming from a single source, at one angle if incidence, producing significantly deep and sharp nulls and peak indicative of the pinna reflections at that angle.

Afaik, speaker measurements in an anechoic chamber don't show such nulls so I think the main problems are reflections from walls/objects. I can only guess that this would also hold true for the same measurement done with a dummy head.
Ethan Winer wrote some stuff about this ("comb filtering is the root of all evil"). Did you see his AES presentation video that was posted some weeks ago?

 

[Tyll Hertsens]

Quote:

Originally Posted by xnor /img/forum/go_quote.gif Afaik, speaker measurements in an anechoic chamber don't show such nulls so I think the main problems are reflections from walls/objects.

That's correct the peaks and nulls come from reflections.

In an anechoic chamber with a single point source of sound, the ear on the dummy will have it's own reflections as it listens. Because it's a point source of sound, it will set up the particular HRTF from a single angle of incidence, which will have very particular and deep notches and peaks due to the shape of the ear.

When you have sound arriving from many angles of incidence (diffuse field) then the ear produces many differing types of HRFT simultaneously which tends to smooth out the peaks and valleys as the ear responds to the various angles of incidence simultaneously.

For example the way they take the DF response of the head is to put it into a very "live" chamber (concrete walls, no acoustic absorption) and pump it up with pink noise from multiple speakers in many positions around the room. The ear is then presented with sound coming from all directions at once, averaging out all the deep notching from the from the ear reflections when sounds come in from discrete angles.

Here's an HRTF measurement set for FR at various angles horizontally.

Sound coming from one angle only will have a particular FR.

Again, making the compensation for a particular case, will make it less usable broadly.


I don't think I did see the video, got a linky?

 

[xnor]

Here it is: YouTube - Audio Myths Workshop
A bit like myth busters minus the big explosions.

Ethan wrote that having two ears already smoothes out the peaks a bit ("A peak or deep null in one ear is likely not present in the other ear, and vice versa.") and moving around your head, even if its just a few millimeters will also cause some "smoothing".

Quote:

Originally Posted by Tyll Hertsens /img/forum/go_quote.gif When you have sound arriving from many angles of incidence (diffuse field) then the ear produces many differing types of HRFT simultaneously which tends to smooth out the peaks and valleys as the ear responds to the various angles of incidence simultaneously.

Yeah but then you get that "huh, the sound is coming out of my head?" feeling I guess.

Quote:

Again, making the compensation for a particular case, will make it less usable broadly.

True. :/

Even if such a 30deg dummy head measurement wouldn't be the right thing I'd still love to see one.

 

[Tyll Hertsens]

Quote:

Originally Posted by xnor /img/forum/go_quote.gif Here it is: YouTube - Audio Myths Workshop A bit like myth busters minus the big explosions.

Yeah, I did see that one. Good stuff.

Quote:

Even if such a 30deg dummy head measurement wouldn't be the right thing I'd still love to see one.

Me too. But the one I'd really like to see is this:

Put the head in an anechoic chamber.

Beyond the wall in front of the head is another room which is a "live" room with hard concrete walls in which many speakers are playing pink noise.

Then you put a large opening between the two rooms such that the center of the opening is at 30 deg off axis to one side of the head, and the opening is maybe 15 degrees wide and tall relative to the head. This would present a large but diffuse sound source with some angular information.

One subtlety might be that you actually need two sound sources on either side of the head, because when you listen to speakers both ears hear both speakers, so you may still need to have the "far side" sound source to complete the normal EQ you hear from your HRTF.

Fun to think about ... not likely to happen.

 

[Tyll Hertsens]

Alrighty then, finished a bunch of measurements and they look pretty good I'd say.

You can download the PDF here.

Any page that has the word "Box" at the end of the title is on the new system, others are previously gathered data at HeadRoom.

I'm quite please the measurements are as consistent as they are.

 

[rsaavedra]

Gee the raw data in the HD650 - Box measurement suggests a right driver hugely different from the left one. The right one hardly had any treble right below 20 kHz, while the left one is up there between -10 and -15dB.

Something similar with the first Beyer DT880, and with the second Beyer DT880 32 Ohm.

Will big differences between left-right drivers prompt repeating the measurements with another pair of headphones of the same model?

PS. Why are older Headroom measurements being mixed with your new ones?

 

[Feanor]

I'm suprised at the left/right differences too. Any thoughts on why it occurs?

 

[Tyll Hertsens]

Quote:

Originally Posted by rsaavedra /img/forum/go_quote.gif Gee the raw data in the HD650 - Box measurement suggests a right driver hugely different from the left one. Will big differences between left-right drivers prompt repeating the measurements with another pair of headphones of the same model?

I think (because it's the subject of the next test that will determine a little more clearly) that what you're seeing is the reflections from the pinna creating a comb filter notch at that frequency.

Here's a set of raw data from the same set of HD800s in slighlty different (~5mm) positions around the ear.

You can see the variations in the highs can be extreme due to small positional changes. It leads me to believe that almost all the fine details in the curve above 8k have little to do with the output from the cans themself and more to do with position changes.

Again, this is exactly why I'll be taking 5 FR measurements and averaging them: to see if the noise in the highs can be reduced to give an accurate measure of response up there.

Quote:

PS. Why are older Headroom measurements being mixed with your new ones?

So I can see if the measurements are repeatable from previous measurements --- which they are for the most part.

 

[xnor]

Quote:

Originally Posted by Tyll Hertsens /img/forum/go_quote.gif Here's a set of raw data from the same set of HD800s in slighlty different (~5mm) positions around the ear.

That could explain some of the day/night differences that people hear with cables.