A question about frequency response, particularly treble.

Are you talking about headphones? if so, I believe it has to do with resonance and other acustic properties of external ears. 

In my experience that weird notching seems to correspond to driver construction qualities rather than your subjective sonic experience (... you know what I mean by subjective, I hope, we can't digest eachothers' dinner but there's nothing magic going on there). I've noticed that a highly positive treble notch seems to correlate with a quicker sounding driver - though I lack the tools to measure it. Wish I had the necessary gear to do some waterfall plots.

 

Edit: Also, I really hate logarithmic graphs for headphones. I would prefer that the regions which have the least information be sussed out more. They're very sensible for speakers, because the goal with speakers is literal fidelity to the input signal, but with headphones an "ideal" reproduction is a little more up in the air. Couple some speakers to the side of your head and things change a bit. The frequencies that get scrunched together could do with some unpacking, in my opinion, we'd have better data for comparison.

Edited by NotJeffBuckley - 8/2/11 at 8:15am

Err, the logarithmic scales makes perfect sense, it's how our ears work.

The freaking huge 10,000 Hz band from 10 kHz to 20 kHz? Just a single octave, like the octave from 20 Hz to 40 Hz.

For speakers it's great, tells you all the information you could possibly want, but for headphones some more precision in the ranges that get especially squashed would help to illustrate correlative relationships between particular treble profiles and bass/midbass profiles and driver construction. Could be useful data.

Frequencies above 10kHz are not treble. The last octave of audible sound is primarily just harmonics. Other than triangles and cymbals, musical instruments don't go that high. It is much more important to get the frequencies below 10kHz flat than the stuff up at the edge of human hearing.

I'm way less interested in the actual frequency response relative to what one hears and way more interested in seeing how different peaks and valleys might correlate to different objective measurements of behavior, e.g. waterfall plots showing quick decay on cans with spiked treble versus slower distorted decay on cans with dipped treble (If that were the result). My feeling is that it's some driver property rather than them just ignoring everything above 10khz. Beyond that point it's virtually impossible to distinguish between a sine and a square wave, that's not the point, I just want to know if those spikes or dips relate to the impressions people give regarding "fast" and "slow" headphones. Seems like there could be data there, useful...

 

Although just to be cantankerous I'll suggest you load up your favorite DAW and a song you know well, run a linear phase lowpass and carve out everything above 10K really aggressively and see if it still sounds like it did

I've done it. Have you?

Yes, sounds fairly muffled. One of the fair complaints against earlier or poorly encoded mp3s is that they aggressively lowpass the content to shave off data.

 

Edit: I should probably note that at 24 and having taken care of my hearing, I'm not as biologically predisposed toward being unable to sense a difference with regard to high frequency content as older men - I don't presume to make a statement regarding your hearing, I don't know anything about your hearing, etc., don't think that please. Since 18 I've gone down from 22khz top end (which I was proud of at the time) to more like 19khz top end. A shame, even taking care of your ears won't change biology...

 

The presence or lack of high frequency content altered without regard to phase in a recording can have a substantially measurable effect in the form of FM modulation/intermodulation distortion; whether it's audible is a different question, I guess, but for the purposes of the simple test I should think the effects of a strong (24dB/octave) lowpass at 10khz ought to be dramatically audible provided your hearing is sufficiently undamaged or non-declined. I forget where hearing settles in adult males, but my brain keeps tossing out 14-16khz for the majority of our adult lives, so I would think there's audible content there at least which would be immediately noticed as having been altered.

Edited by NotJeffBuckley - 8/3/11 at 8:56am

If it sounded muffled, you *definitely* were filtering below 10kHz or you're speaking entirely out of your head and haven't really done the test. (I'm betting on the latter.)

10kHz-20kHz is one little octave. It sounds like a lot of numbers on paper, but in practice, it's very little. It's audible, but it doesn't make the sound muffled because it doesn't touch the fundamentals. If you were familiar with equalizers, you'd know that. You'd also know that rolling off above 14-16kHz has almost no impact on music. Barely audible only with direct A/B comparison... And even then only with music that happens to have triangles and cymbals.

The condition of your ears has nothing to do with it. I'm sure you can hear quite nicely. But there's very little up there to hear, and it certainly can't be described as "treble".
Edited by bigshot - 8/3/11 at 11:30am

Would you like a screenshot of a wav, my DAW, my preferred linear phase EQ, and a pair of files to compare the two? The frequency range we're speaking of has everything to do with the details that distinguish cymbals with rivets from those without, or being played with a brush vs. a mallet vs. sticks; and the harmonic AND intermodulation distortion products of the mix as a whole below it. If above 10K didn't matter, we wouldn't insist on it in recordings, would we? No, I don't traffic in race-to-the-bottom activities, I'd like the high frequencies retained, if it's all the same to you.

 

What point are you trying to prove, by the way? Remember that the genesis of this discussion was my thought that there are disparities in driver "types" that are reflected by certain artifacts in the very high frequencies, and unpacking the compressed logarithm might yield useful data to establish some firmer correlation. I remember you from years back, I got kicked out of the no-double-blind thread thanks to too much consorting with you, haha. You'll bite to the bone to win an argument. I'm not as hungry, but do recall what I was originally wanting out of the information rather than whatever it is you're trying to demonstrate by dismissing everything in the 10khz to 20khz range and insulting my intelligence and character by suggesting I'd lie to argue as small a point as this.

I'm afraid I don't know what you're talking about. I was simply pointing out that above 10kHz is not what we consider treble and although filtering it off is noticeable, but it doesn't impact the fundamental frequencies of music- only cymbals and triangles.

There was a study where they presented people with two recordings- one 20 to 20 and one 20 to 10. Listeners indicated that neither sounded significantly better to them.

It has been years. But to the point, what I'm talking about is my conjecture that there's a commonality between quicker drivers and slower drivers that seems to manifest as similar high frequency profiles as shown on a logarithmic frequency response graph. I've noticed a similar correlation with THD products, but it's not appropriate to go from correlation to causation, obviously... With more data to go on it could however potentially be shown to be an accurate or inaccurate conjecture. If I had the necessary equipment I'd love to do some waterfall plots on headphones' responses to short pulses of varying frequency.

 

I guess I'm shorthanding 10khz+ as "treble" for convenience' sake, I just mean "high frequency." The appropriateness of graphing it logarithmically for a representation of frequency response across octaves isn't what I'm calling into question, rather I wish there were more data there; the headphones people qualitatively refer to as "quick" versus "slow" seem to have some things in common, but all I have now is "seems" and it doesn't necessarily seem to correlate with driver mass alone, nor does driver mass seem to correlate with the high frequency profile (which has the appearance, to my eyes, at least, of being prone to artifacts in measurement in the first place, but what can you do?)

I'm always happy to look at numbers on paper, but I judge withhuman ears. The buck stops there. How things *sound* is what really matters. It's always surprising to me when gnat hairs are precisely measured and quantified and basic descriptions of how things sound are exaggerated, described in vague flowery terms or completely misstated.

I'm always happy to look at numbers on paper, but I judge with human ears. The buck stops there. How things *sound* is what really matters.

Treble starts somewhere around 4-5kHz. Above 10kHz its importance to how music sounds drops off very fast. The range of human hearing covers about ten octaves. The top octave is the absolute least important to sound quality. Because of the principle of masking, even the midrange balance is more important to whether something sounds muffled than the topmost end.

Some audiophiles have the idea that upper frequency extension is what makes one stereo better than another. That isn't true. Some audiophiles also take pride in having golden ears that hear beyond the range of normal hearing. That's nothing to be proud of. There's more noise and squeal from electronics and flourescent lights up in that range than there is music. Normal hearing is much better for music than having a bat's ears.
Edited by bigshot - 8/3/11 at 1:38pm