Head-Fi.org › Forums › Equipment Forums › Sound Science › HRTF and binaural measurements of Sennheiser HD650, HD700, AKG K550
New Posts  All Forums:Forum Nav:

HRTF and binaural measurements of Sennheiser HD650, HD700, AKG K550 - Page 3

post #31 of 54

Isone is just a fancy looking crossfeed with EQ and room reverb. The built-in EQ curves (including the "HRTF cues") did anything but work for me.

Imo, you get better results if you do the equalization by hand.

post #32 of 54
Quote:
Originally Posted by JMS View Post

Let me add some more explanation as to what I'm trying to accomplish, which I really should have done in the initial post.

 
The motivation is to establish an ideal frequency response for headphones, in terms of subjective preference. For loudspeakers, we already have good research on what this response should be: when measured in-room, it's a flat line sloped downwards (see Floyd Toole's book on loudspeakers). Now, is there an equivalent statement that we can make for a headphone's frequency response?
 
Given the loudspeaker research, we might guess that the ideal headphone response should also be a downwards sloping line like the in-room loudspeaker response. However, we need to account for modifications to the in-room response by the listener's HRTF. To measure the listener's HRTF, we place a dummy head (mine) in the listening seat, and measure how the in-room response with the microphones mounted in-ear differs from the response with the microphones free standing. These were my measurements (a) and (b). Our candidate for the ideal headphone response would be a flat line sloped downwards but modified by this difference. This is why I said the original loudspeakers' and room's frequency response are largely irrelevant: we are only interested in the difference between two measurements.
 
Then we measure our headphones using the same in-ear microphones. These were my measurements (c). The ideal headphone should have a measurement (c) such that:
 
(c) = "flat line sloped downwards" + ((b) - (a))
 
Note that if the room and speakers were ideal, then (a) = "flat line sloped downwards", so we'd want (c) = (b).
 
Hope this clarifies the motivation behind these measurements.


Sorry there's one thing I don't really get. I'd assume you haven't subtracted the natural crossfeed happening with speaker listening? With (a) and (b) the mic on one ear will record both the sound traveling directly to it and part of the sound of the other mic while that doesn't happen with (c). Unless you want something as an approximate frequency response, that equation for determining (c) doesn't look correct to me.

 

Btw very cool experiment. I thought about grabbing some mic like the one you used but not sure which brand is good.

post #33 of 54
Thread Starter 
Quote:
Originally Posted by tranhieu View Post


Sorry there's one thing I don't really get. I'd assume you haven't subtracted the natural crossfeed happening with speaker listening? With (a) and (b) the mic on one ear will record both the sound traveling directly to it and part of the sound of the other mic while that doesn't happen with (c). Unless you want something as an approximate frequency response, that equation for determining (c) doesn't look correct to me.

 

Btw very cool experiment. I thought about grabbing some mic like the one you used but not sure which brand is good.

 

Only with (b) is the mic mounted in one ear. With (a) it's mounted on a tripod stand. In both cases a stereo sine sweep is played, so sounds from both loudspeakers are picked up by the mic. The difference between (a) and (b) is explained by what the head and torso do to the frequency response.

post #34 of 54
Thread Starter 
Quote:
Originally Posted by firev1 View Post

You are not bothered by the 6k peak and dip at 2-3k? Cool stuff here anyways. Consider me a fan :)

 

I've only just started to play with equalization, so I started addressing the bigger issues first, which to me was the broad lower midrange bump. I haven't played with the 6k peak yet, but because it's high-Q (looks like 1/6 octave), it might be relatively benign. I've seen some CSD (waterfall) plots of the HD700's having lots of ringing right around that frequency, but I'm not sure it's as bad a problem as it looks. The research (ref: Toole) tells us that we are sensitive to faults in the frequency domain rather than the time domain, and resonances are less audible in headphones than in loudspeakers anyway.

 

As I just came back from vacation, I'll update after I've had time to play with it more.


Edited by JMS - 12/30/12 at 1:43am
post #35 of 54
Quote:
Originally Posted by JMS View Post

 

Only with (b) is the mic mounted in one ear. With (a) it's mounted on a tripod stand. In both cases a stereo sine sweep is played, so sounds from both loudspeakers are picked up by the mic. The difference between (a) and (b) is explained by what the head and torso do to the frequency response.

So you are saying there're only 1 mic recording in both (a) & (b)?

post #36 of 54

Disregard, I failed at reading comprehension.


Edited by ComfyGrados - 1/3/13 at 12:50pm
post #37 of 54

Lol, disregard.


Edited by ComfyGrados - 1/3/13 at 12:45pm
post #38 of 54
Thread Starter 
Quote:
Originally Posted by ComfyGrados View Post

It doesn't appear as if the difference between the speaker measurements with the microphone in and out of the ear reflect anything close to a HRTF when it comes to your ears' interaction with headphones... because it doesn't appear your in-ear microphones are measuring things quite like an eardrum hears things.

 

First of all there's no peak at 2-3 kHz on any of your in-ear measurements, whether they're of speakers or headphones.

 

Second... your in-ear measurements seem to peak at different places depending on whether they're measuring headphones or speakers. When measuring speakers they appear to peak the most between 1-2 kHz, around 4 kHz and around 12 kHz. However, when measuring headphones, due to the way headphones and ears interact, you have a distinct peak in your in-ear measurements at 5-6 kHz and a distinct dip at 7-8 kHz.

 

I do not believe you can achieve your goal with your current in-ear microphones as it doesn't appear they're measuring anything close to what your eardrums actually hear, which would be a peak at 2-3 kHz and probably a peak around 7.5 kHz due to the acoustic impedance mismatch between transducer, ear canal and ear drum causing a half wavelength resonance in the canal.

 

Is anyone aware of in-ear microphones which produce more accurate measurements like an eardrum would actually hear?

 

 

If we ignore the measurements because it violates one of our preconceived notions about a 1-2khz peak or whatnot, then whats the point of measuring? In fact, we are testing the very notion whether there should be a 1-2khz peak. Recent research has suggested that existing assumptions (such as IEC calibration) are not subjectively ideal.

 

My measurements are made with the mic at the entrance of the ear canal, so that effects due to head, torso, and pinna should already be accounted for. Sound at this point should be the same whether it comes from headphones or speakers, minus differences in timing and direction due to reflection. Those are certainly important effects, but we have to approach things one at a time, so I'm limiting to frequency response first.

 

As previously discussed in this thread, my measurements of the KSC75, HD650, and HD700 match up well with at least one other set of published measurements (changstar). They don't match Headroom's or Innerfidelity's well, but nor do these two match up with each other, so I have more reason to believe in mine, especially since those use a dummy head instead of a real head (mine).

 

It is only recently, with the AES2012 papers discussed in this forum, that we have begun subjectively verifying what previously was theorized on paper. My guess is that the old speculation is not going to hold up under subjective testing. An interesting anecdote is that PSB M2U headphones has just won product of the year at Sound & Vision magazine. They measure very flat and sound like possibly the most neutral of all headphones to me.

post #39 of 54
Quote:
Originally Posted by tranhieu View Post

So you are saying there're only 1 mic recording in both (a) & (b)?


Ok I think I finally got where I was wrong. It doesn't matter whether 1 or 2 ears are being used here since it's the SPL that is being measured to derive the frequency response (for EQ-ing later?). The effect of time and phase delay is not being considered and thus unaccounted for.

post #40 of 54
Quote:
Originally Posted by JMS View Post

 

 

My measurements are made with the mic at the entrance of the ear canal

 

Disregard everything I said. Sorry.

 

I don't know why, but I thought you were using mics at the eardrum. I brain farted and failed at reading comprehension. Are you measuring with a blocked or open ear canal? If blocked, how are you blocking your canal?

 

Does anybody have a target diffuse field EQ curve for blocked meatus measurements at the entrance of the ear canal? I can't seem to find one. I'd be interested in seeing what peaks "are supposed to be there" and comparing them to your measurements if you're measuring with a blocked meatus. I've read that blocked meatus measurements are more repeatable than open. Looking at your mics, it looks like you could plug your ear canals behind the mics rather easily, and those mics appear to put the mic elements in what I believe is the correct place for blocked meatus measurements, close to the tragus.

 

Google says the target curve can be found at:

 

"This will help, Fig.5 on pg.225 & Fig.7 on pg.226:
H. Møller, C.B. Jensen, D. Hammershøi and M.F. Sørensen, “Design Criteria for Headphones” J. Audio Eng. Soc. 43(4), 218-232 (1995)
"

 

If you could find a target curve and your measurement technique produces repeatable results which are showing results similar to the target curves (similar but different peaks in similar but different places)... then you might be able to produce some target curve-compensated measurements.

 

Cheers.


Edited by ComfyGrados - 1/3/13 at 3:42pm
post #41 of 54

oh god how did this get here i am not good with computer why did i hit quote instead of edit why can't i delete my own posts


Edited by ComfyGrados - 1/3/13 at 1:06pm
post #42 of 54

Apologies if this question sounds naive, as I haven't really studied much about this area of audio (even though I hope to eventually), but would the results be different if this experiment was being conducted with a sub-woofer in the setup?  Intuitively, it seems that would make a difference, and if so, is there anybody currently conducting this type of experiment with such a setup? I only ask cause if this is being conducted with some standard studio nearfields, even a lot of the good ones don't tend to have much of a response in the bass region... Granted, it's not easy to get a good nearfield+subwoofer pair in a studio and have everything will come out 'flat', but when it's done well, it is clearly better for doing audio work (from the limited experience I've had anyway).


Keep up the awesome work though either way. 

 

 

EDIT: nevermind, ignore me. Just looked up the speakers being used here and it seems they should provide good enough bottom end. I'm still curious as to how the characteristic frequency responses of the speakers used in these kinds of tests affect the results though...


Edited by oblique63 - 1/4/13 at 3:55pm
post #43 of 54

I wonder how much the HRTF varies from person to person. I would think it varies considerably. If so, this helps explain why people disagree on how "neutral" or "realistic" the voicing of a headphone is - and more generally its overall sound. And if so, any headphone's FR curve should be taken with a big grain of salt, depending on what HRTF was used to correct it, and whether that curve applies to any given listener.

post #44 of 54

I remember seeing that the standard deviation of the human FR using a microphone at the eardrum was 1 decibel. The variation that we see in listener descriptions has more to do with acclimation, taste, and training more than anything else, IMO. If Harman can get a bunch of testers to hear the same headphone similarly, I doubt that head differences are that big a deal.


Edited by briskly - 3/28/14 at 4:35pm
post #45 of 54

The HRTF is affected by many factors - head size, shape, density, ear-to-ear distance, maybe even the volume & shape of sinus cavities. Hard to imagine that with all those factors the STD across broad human populations is only 1 dB at various frequencies. That's virtually inaudible. Individual hearing response and acuity differ far more than that; why wouldn't the HRTF curves as well?

I need to read up on this topic more.

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › HRTF and binaural measurements of Sennheiser HD650, HD700, AKG K550