[khaos974]

You still haven't explained what good treble or clean treble is beyond that it is detailed.

 

[ac500]

 
Quote:

As to how to define good quality treble, I'd say a fast accurate transient response would fit the description.

 

 

[yifu]

Good treble is when cymbals sound real, not too sparkly, clean decay and fast transitions. 
 
Quote:

You still haven't explained what good treble or clean treble is beyond that it is detailed.

 

 

[khaos974]

Quote:

Quote:

As to how to define good quality treble, I'd say a fast accurate transient response would fit the description.

So, excellent impulse and step response define details?
Here's the impulse response of the LCD-3 and yet people can't stop describing how detailed it is.
 

 
 

 

[ac500]

Are you sure that was a correct measurement? That's certainly a strange impulse response.
 
That said, I see the first peak of the impulse begins at 0.0001 seconds. That's fast (twice as fast as the first peak of the SRH-940's impulse response). It just seems to take a strangely long time to settle.
 
We all know synthetic measurements generally aren't perfect. I personally don't understand why someone doesn't play a high quality FLAC song and measure using that, comparing to the source waveform and analyzing various characteristics of the deviation (maybe because it would probably require custom software).
 
(Oh, I guess it's because we all love listening to those square waves. And infinitely short duration impulses. And perfect single tone sinewaves swept across the frequency spectrum /sarcasm)

 

[R-Audiohead]

Quote:

 
We all know synthetic measurements generally aren't perfect. I personally don't understand why someone doesn't play a high quality FLAC song and measure using that, comparing to the source waveform and analyzing various characteristics of the deviation (maybe because it would probably require custom software).

EXACTLY
 
I've thought this for quite some time, actually.
 
I figured I was missing something

 

[khaos974]

Quote:

Are you sure that was a correct measurement? That's certainly a strange impulse response.
 
That said, I see the first peak of the impulse begins at 0.0001 seconds. That's fast (twice as fast as the first peak of the SRH-940's impulse response). It just seems to take a strangely long time to settle.
 
We all know synthetic measurements generally aren't perfect. I personally don't understand why someone doesn't play a high quality FLAC song and measure using that, comparing to the source waveform and analyzing various characteristics of the deviation (maybe because it would probably require custom software).
 
(Oh, I guess it's because we all love listening to those square waves. And infinitely short duration impulses. And perfect single tone sinewaves swept across the frequency spectrum /sarcasm)

Yes, that's the correct measurement, look at the pseudo period of the oscillation, it should be roughly 18/20 kHz, and of you look at the FR of theLCD-3 they were measured from, you'll see a huge frequency spike around 18 kHz, the pseudo oscillation in the impulse response is just another view for the 18 kHz frequency bump. And by the way, looking at the precise moment of when the 1st peak rises is pretty useless, depending on how your plot your graph, how you set the masuring aparatus with the different triger signa;, you'll end up with very different looking plots.
 
The other interesting point is that if you have a good enough sampling resolution and precision, with only the  frequency+ phase graphs you can plot the impulse response, the step response, the response to square wave just via the Fourier and Laplace transforms.

And finally, we are still going nowhere, details = good treble = fast accurate transient, fast and accurate are still vague imprecise and bulls***ting terms. It's like saying yo be accurate, a speaker must have a good frequency response, or a regular frequency response, it has little meaning, near zero content, insteast, saying that for a 50 to 20000 Hz the frenquency response must me contained within a ± 1.5 dB range averaged at 1/12th octave in a anechoic chamber is much more meaningful.
 
That's what everybody is interested in what is good , or fast accurate aren't good descriptive term,
 

 

[mikeaj]

Quote:

The other interesting point is that if you have a good enough sampling resolution and precision, with only the  frequency+ phase graphs you can plot the impulse response, the step response, the response to square wave just via the Fourier and Laplace transforms.

 
Are headphones linear enough for that?  Or do you just mean a good approximation?

 

[binson]

 
High frequencies have nothing to do with low frequency performance, nor are they needed for a detailed low frequency reproduction. This is naturally assuming that there are now major non-linearities in the system. However, in music there rarely is such material that would only have low frequencies. Let's take a bass line as an example. When the player plugs the string, at that very moment, there are also higher frequencies in the signal, and then you can't consider the bass line being just low frequency signal. 
 
I'd say the details come from the time domain performance. If you'd have a resonant free headphone with linear phase (flat group delay), and this naturally should be evaluated with a real head measurement, then you'd have a detailed sound. This, different phase responses, also (partially) explains why headphones with similar magnitude responses sound different. Continueing with the above bass line example, with time coherent headphones, plugging the string will sound detailed because all frequencies happen at right time. If you maintain a good phase response, you can even cut the high frequencies and it can still sound detailed, maybe darker, but detailed.
 
If you take the image processing analogy, as ac500 gave earlier, it doesn't matter how high the resolution is, if the pixels are misplaced (phase error), or if the pixels keep moving (resonance). 
 
Actually, it's quite funny, that if you're judging the 'detailness', then the extend of the frequency range, or even the shape of the magnitude response, doesn't matter that much. It's not that crucial if some frequencies are missing, as long as they are not smearing the signal by being late or resonating. 
 
Naturally distortion has effect as well but that's fairly obvious, so no point arguing that here.
 
To the question "Good Treble = Detailed", I'd say those are two different topics. Sure, if you want to evaluate, if a cymbal hit sounds detailed through your headphones, you need to hear it first, thus you need treble. Whether 'good' means also 'detailed', is a matter of taste.
 

 

[ac500]

How do you define detail then, phase response? If that's the case, how would a phase shift on the order of .1ms matter to a bass waveform dealing with frequencies with periods on the order of 20ms?

 

[khaos974]

Quote:

Quote:

The other interesting point is that if you have a good enough sampling resolution and precision, with only the  frequency+ phase graphs you can plot the impulse response, the step response, the response to square wave just via the Fourier and Laplace transforms.

 
Are headphones linear enough for that?  Or do you just mean a good approximation?

Honestly I don't know if they are linear enough to do so, but another forumer, Soaa, recreated the square wave response and the impulse responses of multiple headphones based on the frequency response alone, and then compared it to Tyll's measured squared waves and impulse response. At firs glance, they were very close which would suggest a good deal of linearity for headphones.
 
 

 

[Maverickmonk]

Interesting topic, I'd love to see the mathematical proof, since math makes me feel all warm and bubbly inside, and this is pretty applicable to some things we've been doing in my physics class.

 

[binson]

Quote:

How do you define detail then, phase response? If that's the case, how would a phase shift on the order of .1ms matter to a bass waveform dealing with frequencies with periods on the order of 20ms?

 
That's a good question. There's a fair amount of literature about detectability of group delay, but in general the threshold varies from 1ms (around 2-4 kHz) to few ms (at lowest frequencies). And of course, 0.1ms doesn't make much different at lower frequencies, or at any frequency. I guess it for everybody's personal decision to say when you start to lose detail in the sound (as the phase errors grow).
 
Why not test it yourself? You seem to have tools for signal processing, so you could try taking a sound signal, and start playing with the phase, and see how it affects the details in the signal.
 
 
 
 

 

[ac500]

1ms for mids and a few ms for lows is perfectly reasonable to me. The debate isn't so much as to what the threshold is, but how much delay do headphones actually have? I don't know the actual physics of headphones in particular, but I'd be really surprised if it was more than 1ms (that seems absurdly slow). If the delay is indeed faster than what matters for bass, then we can then say that phase delay doesn't effect bass frequencies perceptibly. Therefore if we define "detailed" as transient response (including phase delay as you put it), we've just proven that "detailed" is a trait of headphones involving only the upper frequencies quality.
 
I think the reason people have such a hard time accepting this is like you said, a stum of a deep bass guitar cannot be thought of as simply low frequencies. A detailed and accurate representation involves frequencies across the spectrum, even though the upper frequencies are very small, they're still very important to conveying the detail of the sound. I think people don't realize that pure bass is really boring - simple smooth uninteresting and slow by definition. It's only when you incorporate higher frequencies that the detail can be heard.
 
Maybe an argument could be made that with some headphones, the actual transient response dynamically worsens when a bass frequency is being represented. In this case, it would be true that bass detail could be described as the ability to reproduce a bass signal without damping out all the higher frequencies, but arguably this is still a matter of reproducing quality treble under all cases (it's just disguised as a trait of bass for obvious reasons).

 

[Tilpo]

I think there are three problems with your hypothesis

1. From what I understand you say that details in bass are created by fast reaction of the driver. However the voltage almost always changes much quicker with high frequency tones. Most treble is at an about 10-20 times higher frequency which requires 10-20 times as high speed of the driver.

2. Perceived detail in bass might not always be detail in bass. The way we perceive sound is incredibly complicated and as a result I think what might be going on here is that better treble causes a preceived increase in bass detail. The bass may be reproduced just as accurately, but somehow the brain processes the sound into causing a perceived difference in detail.

3. This might sound cliché, but how can you be sure it's not a placebo? Unless either more people share your observations, or blind testes are performed there is no way to dismiss this possibility.
And don't go telling me you are unbiased, because that is simply speaking impossible.