or Connect
Head-Fi.org › Forums › Equipment Forums › Sound Science › AES 2012 paper: "Relationship between Perception and Measurement of Headphone Sound Quality"
New Posts  All Forums:Forum Nav:

AES 2012 paper: "Relationship between Perception and Measurement of Headphone Sound Quality" - Page 9

post #121 of 135

Vocals will carry both melody and the usual semantic content that goes with language - which of those you prefer emphasized in your music in subtle ways can to my mind be a similar question of preference as whether you like your bass boosted or not - if indeed more subtle. Though I can also see that the mind has been made up in this thread by and large that all headphones should be neutral (won't happen, this would put most manufacturers out of business, would it not). I for sure prefer neutrality as well, but not to the point of dictating it quite so harshly.

 

Having said that, the floor is still likewise open for discussion on the effects of this left/right divide on the study presented in the Harman paper, a study in which this variable was not controlled for apparently. Would people have a preference to hear vocals primarily in either the left or the right ear? I'll also note that the S & W article I've been referring to was published in The Lancet; I'm sure they put more than a cursory thought into it.


Edited by vid - 7/16/13 at 4:59am
post #122 of 135
What I mean is: you can't boost the melody so the left ear hears it better, or whatever you were thinking could be done. Ab Initio explains he facts of the matter very well. The only thing I would change about what he wrote is "perfectly flat" to "perceived Flat relative to a calibrated stereo" which is sort of what this thread is about(or was rather). We still don't know that curve after making headphones for decades. Of course calibrated stereos haven't been around as long, but B&K were able to derive a preference curve that's similar to todays a long ti e ago. We got derailed on to brain lateralization and how it effects what we hear--which is a great topic BTW. It's just the wrong thread for it.
post #123 of 135

It's exactly the right thread for it since the issue has the potential to affect the results in the Harman study. Or do you have an argument against that?

 

As far as "boosting" the melody, I don't think we're on the same page. The way I understand it anyway is that it's a question of whether the melody or the grammar is processed, given priority in processing, or whatever. You don't need to extract the melody and then feed it into the ear.

post #124 of 135
There are a million things that have potential to effect the Harman study.... I actually completely understand your second paragraph and it's part of the reason that I think brain lateralization is for a different thread.

Anyway,
If you look at the HP1, the most preferred headphone SQwise in the first Harman paper, it measures most similar to the dtmblabber calibrated room measured curve. https://dl.dropboxusercontent.com/u/16343460/Relationship%20between%20Perception%20and%20Measurement%20of%20Headphone%20Sound%20Quality.key.pdf
The Mad Dog measures very similarly to the HP1 when you look here: http://dtmblabber.blogspot.com/search?updated-max=2013-07-14T18:11:00-07:00&max-results=1
The Mad Dog is known to measure similar to the LCD2(the HP1). It was found in the later Harman studies to "The unequalized Audeze LCD-2 was described as dull and lacking presence or energy around 1-2kHz". http://www.innerfidelity.com/content/harman-researchers-make-important-headway-understanding-headphone-response Not too far from what the measured curve would suggest.
post #125 of 135
Vid, I am with Ronion and most posts here. Your topic does not seem to be the key focus of the Harman study (it may affect their results, but there are lots of other factors there that can affect the results).

If our goal is to have Dr. Olive provide responses, then please note that Dr. Olive has not chimed in since page 7. I imagine he might not want to comment on things which he is not interested in or doesn't have an answer for.
post #126 of 135
Quote:
Originally Posted by vid View Post

You don't need to extract the melody and then feed it into the ear.
This is exactly why you shouldn't color the left and right speakers differently.

I agree, this is the wrong thread. i think you should start a new thread to discuss left brain/right brain processing as it is certainly an interesting topic smily_headphones1.gif
Why dont you start the thread and copy your original post on the matter (or link to it) and include a link to the paper?
Cheers
post #127 of 135

I suspect scientists would be more prone to hang around here had we more scientific rigor on offer rather than an attempt to have shortcuts and to accept results regardless of objections raised.

post #128 of 135
Quote:
Originally Posted by vid View Post

Having glanced through the slides and the blog post, I wonder whether the channel balance of the headphones was considered in tandem with the type of music played, and further, whether listeners might have a preference for a perfect channel balance or one where certain frequencies are emphasized on one but not the other channel? This I wonder after having looked over the short piece Why do mothers cradle babies on their left? by Sieratzki and Woll, where they suggest via reference that the left ear is more adept at recognizing melody while the right grammar.
Channel balance has been shown to be Attribute of preferred headphones. Maybe people like grammar just as much as melody. There could be several reason for that as I think has been well covered. wink.gif

I'm not sure if any genre research has been done. I wouldn't be shocked if Harman has done it though. They have done research on reliable recordings to judge spectral balance. Ideally the desired FR should be built into the recording--the playback aystem would ideally be neutral. I've seen papers done on the general frequency content of various genres, if playback was standardized, there should be no real need for different frequency curves for different genres. That said, playback is far from standardized. There may well be just as much difference between any two recordings, or recordings from any 2 studios as there is for genres. Who knows? (Hopefully someone). If a particular genre fan like a particular curve over a neutral one then you'd think the whole genre of studio mixing/mastering engineers had playback systems skewed in a similar manor. I wouldn't be shocked.
post #129 of 135

Should the channel balance have been normalized in this study? If so, why, and if not, why?

post #130 of 135
I'd say that depends on how it's measured. Straight off the driver? Yes. Why: otherwise the imbalance could skew the result b/c we are looking for an ideal curve. If the ear is involved in the measurement: no, we live with those differences everyday. If we corrected for them, it would skew the results again.

That's the type of straight forward question I was looking for in my "me?" Response.
post #131 of 135

Is the  "in-room loudspeaker measurement for RR1_G and RR_G" compensation data available somewhere similar to the DF compensation referenced in ISO11904-2 ? 

From the Graph in the AES convention paper 8994 "Listener Preferences for In-Room Loudspeaker and Headphone Target Responses" as "reproduced"  by Tyll of innerfidelity.com ( http://www.innerfidelity.com/content/headphone-measurements-explained-frequency-response-part-two)
 
according to Tyll's analysis the data should roughly look like this
FR          dB

20          4    
60          4
200        0
1200       3
3000       12
10000     0
20000    -13

Edited by waxmaxmo - 10/28/15 at 4:41am
post #132 of 135

the conclusions we can make are limited to how the experiment was conducted. they went for a curve that is indeed close to most DF curves(for good reason), and the result could be interpreted as that curve tilted so that we end up with a shift of about 9db from bass to treble.

 

Tyll probably had more intel than I can get from the papers to draw his estimated harman curve, but from what I could get, we are unsure as to where that 9db-ish variation was measured. was it from a DF curve looking at the EQ change at 20hz and 20khz? or 40hz and 10khz? I'm not too sure.

 

the other significant unknown (at least for me), comes from how the test subjects could set the EQ. they had 2 knobs and that's it. maybe some would have liked to get the mids down but not the treble for all we know. or the opposite. and maybe the bass would have taken another shape if they had more than one knob to shape it?

 

 the overall conclusion I come to(maybe I'm missing something), is that the DF curve sounds too "bright" to most people(not a surprise for me at least), but is still a rather good general shape to start from.

post #133 of 135

All those questions and more have answers you can find on Sean Olive's blog, and this freely available summary of his more recent publications. The charts are taken directly from the publications.

 

The -9 dB slope is observed end-to-end deviation from the flat equalized speaker response. It wasn't explicitly about DF, but they noted the flat in-room response resembles diffuse-field response as measured by the GRAS 45CA with pinna simulators.

post #134 of 135
Quote:
Originally Posted by briskly View Post
 

All those questions and more have answers you can find on Sean Olive's blog, and this freely available summary of his more recent publications. The charts are taken directly from the publications.

 

The -9 dB slope is observed end-to-end deviation from the flat equalized speaker response. It wasn't explicitly about DF, but they noted the flat in-room response resembles diffuse-field response as measured by the GRAS 45CA with pinna simulators.


so 20hz to 20khz for sure? I've looked at the PDFs, the blog, the video, innerfidelity posts on the subject, even here and on other forums where we are usually 5 people to even care about those studies(which is strange for people supposedly passionate about having good sound), yet I'm still no sure Tyll's curve is what harman came up with, or if people wouldn't have preferred a different bass level if the Q value for the bass tuning had been different for example.

so if the answers are there, I certainly missed them.

post #135 of 135

This curve is ripped straight from the AES paper. I don't think the dashed curve (representing flat in-room at eardrum) is necessarily accurate, but errors introduced should somewhat cancel out in the solid black line.

 

 

About why the filters were set as such, they decided on that from previous investigations.

 

Quote:

The frequency (105 Hz) of the bass filter was chosen for several reasons. First, the majority of subwoofers are crossed over to the main speaker near (or slightly below) this frequency, so there are practical reasons to start from here. Secondly, variations in bass level due to acoustical interactions between loudspeakers and rooms occur near and below near this frequency. Thirdly, informal investigations by the authors found that extending the bass shelf frequency to above 105 Hz had an adverse affect on the timbre of vocals and other instruments whose fundamental pitches fall within this frequency region. On the other hand, boosting the bass below 150-200 Hz tends to enhance bass instruments without impacting the sound quality of the higher pitched instruments.

The treble filter frequency of 2.5 kHz was chosen because this is a common midrange-tweeter crossover frequency where the directivity of the loudspeaker begins to increase, and the in-room response of the loudspeaker begins to fall downwards. The exact amount of high frequency drop will depend on directivity of the loudspeaker, the ratio of direct-reflected sounds at the listening seat, and the absorption characteristics of the room.

 

As for variation between listeners, the curve represents an average result.

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › AES 2012 paper: "Relationship between Perception and Measurement of Headphone Sound Quality"