[AutumnCrown]

When looking at FR response, I find it difficult to estimate how much of the total energy of a headphone a given band (e.g. 7-10 khz) represents. For example, the SR009 and LCD3's FR's look fairly similar overall, when in fact one is rolled off a bit on top, and the other is a bit bright.

Would it be easy to represent the FR response as a table where the percent of energy for a given band is given? EG:

20-60 hz 12%
61-120 hz: 10%
121-200 hz:12%
201-400 hz: 10%
...and so on to 20 hz or so

Perhaps this could be normalized against the harman target response, so for a headphone that met the response perfectly it would read

20-60 hz +0%
61-120 hz: +0%
121-200 hz: +0%

For a bassy headphone, it would read:
20-60 hz +5%
61-120 hz +5%
121-200 hz +5%

It would be surprised if this hasn't been thought of before. But it seems like it would make reading FR responses easier. Sort of like having the "area under the curve" in statistics be in a chart, because it's a lot of information to remember. And that's for a single curve. It would also help in comparing graphs from different measurement systems and with different compensations.

 

[castleofargh]

so all in all how many values do you figure this thing would have? 10? 15? you expect somebody who's lost by 2 FR to make sense of that easily? I just don't see it.

IMO everybody is much better served with the basic tool of placing 2 headphone curves on the same graph. maybe matching the levels at 1khz could be adjusted on demand as sometimes headphones could have a big area with similar response but the difference at 1khz shifts everything and may give the wrong subjective idea. but it's still IMO the most explicit way of showing differences in FR.

 

[SilverEars]

When looking at FR response, I find it difficult to estimate how much of the total energy of a headphone a given band (e.g. 7-10 khz) represents. For example, the SR009 and LCD3's FR's look fairly similar overall, when in fact one is rolled off a bit on top, and the other is a bit bright.

Would it be easy to represent the FR response as a table where the percent of energy for a given band is given? EG:

20-60 hz 12%
61-120 hz: 10%
121-200 hz:12%
201-400 hz: 10%
...and so on to 20 hz or so

Perhaps this could be normalized against the harman target response, so for a headphone that met the response perfectly it would read

20-60 hz +0%
61-120 hz: +0%
121-200 hz: +0%

For a bassy headphone, it would read:
20-60 hz +5%
61-120 hz +5%
121-200 hz +5%

It would be surprised if this hasn't been thought of before. But it seems like it would make reading FR responses easier. Sort of like having the "area under the curve" in statistics be in a chart, because it's a lot of information to remember. And that's for a single curve. It would also help in comparing graphs from different measurement systems and with different compensations.

You are right they do look similar except that STAX bass lower in the subs. Interesting thing is, they don't sound similar. I find STAX do not sound like how the measurement are represented. Treble sounds more prominent than the Audeze, and Audeze seems more representative of the response. You will find that the bass response is different as well. Stats tend to not have such a low-end body.

Despite the graphs, expect different sounding headphones.

 

[71 dB]

It depends on the signal itself how much "energy" you get. Bass-heavy music gives you more "energy" at low frequencies.

Energy is the integral of power over time. One second of one watt [W] power equals 1 Ws or one joule [J] of energy. So if you listen to a record 10 minutes instead of 1 minute you get probably 10 times more (assuming the power of the music is pretty constant) energy blasted at your ears.

Mathematically frequence response it derived from impulse response (taking the magnitude value of the complex-valued impulse response). Impulse response describes a linear time-invariant (LTI) system completely. Frequence response loses the phase information of the impulse response and does not describe a LTI system completely. Two different systems can have identical frequence responses, but completely different phase responses so that the systems sound different.

Linear system means that the following are true:

f(x+y) = f(x) + f(y)
f(a*x) = a*f(x)

Time-invariance means that the system doesn't change in time. In practice everything changes, but usually slow enough for the assumption to hold. In other words, your headphone sounds today the same it did yesterday, but maybe not 10 years from now. 100 % linear systems hardly exists, but audio products are often linear-enough to appear completely linear to our ears.

Our hearing is very complex and it isn't easy to come up with numbers that predict how bright a pair of headphone sounds.

 

[AutumnCrown]

It seems a serious problem to me that there are infinite different measurement systems and configurations. Ideally, if this system were implemented by measurers, it would make all of them comparable to each other without any superimposition or adjustment.

 

[castleofargh]

getting differences can't be helped so just forget about reasoning with that as a starting point. obviously we'd all want a complete perfect standard that's accurate within 0.5dB from 20hz to 20khz, but even the 2 drivers of the same headphone don't have that consistency, let alone measurements of them.

for starters transducers aren't very accurate(relatively to modern electronic!!!!!) and we're using 2 successively when measuring headphones. then just placement on a standard dummy head can create significant variations(several dB), change of seal and the low end immediately changes, a shift between the driver and the mic and the trebles might not look the same. just leave the headphone on the dummy head 30minutes and the pads will let the driver come closer or maybe even drop depending on what serves as cushion on the headband(and different headphones will behave differently because of weight, material...). then there is the issue of picking one dummy head with one mic, coupler/fake ear and decide that this will be the definitive standard. not only would it really mess the manufacturers of such products up, it would also mean to put a stop on research to find more accurate ways to make dummy heads. the ears used on the latest generation are not like the ones from 10 years ago, and that's a serious thing to consider when we wish to establish a solid standard. if it's solid for less than 2 years, then everybody has to move on to the newest gears, how many will have the will and the money to keep upgrading?

I believe that a protocol to measure a driver alone is something fairly stable and solid. but measuring headphones brings a lot of questions. do we remove the pads? do we let them on, and if so, how long before measuring? how many pairs of pads and headphones should we try to determine a final curve? how do we deal with placement when headphones are of different sizes and shapes and some will require a different placement if only to get a proper seal? it's not hard to arbitrarily decide on a standard, we do that all the time with everything and it helps a lot, but we have to set standards that people will be able to follow. so telling people to go setup about 90dB @1khz, that's relatively easy. pick a model of ear, why not. but how long should we decide to stick to that when clearly superior mics, ears, and whatever come out? and I'm just a noob having fun with cheap toys, so I'm probably missing several other serious questions where a single answer might not be as simple as we wish it to be. like, what's neutral on a headphone? nowadays we have the Harman curve as preferred subjective target(it wouldn't be too hard to jump from that to call it neutral, as usually we do have a preference for neutral signatures. but it's not what they claim to have done).

also I might be captain obvious, but us audiophiles getting FR graphs of headphones is mostly an amateur thing. the vast majority of the industry doesn't want us to have such data. so IMO, the handful of heroes sharing massive databases, become their own standard because of how few they are. it's clear to me that most amateurs with their cheap DIY rigs have at some point used Tyll's work(Innerfidelity) to set and calibrate their own system, even probably settling on positioning methods that seemed to come close to his results(despite how most amateurs measure a single position and Tyll averaged many of them).

 

[bigshot]

getting differences can't be helped so just forget about reasoning with that as a starting point. obviously we'd all want a complete perfect standard that's accurate within 0.5dB from 20hz to 20khz, but even the 2 drivers of the same headphone don't have that consistency, let alone measurements of them.

Thankfully, most people can't discern anywhere close to .5dB deviation listening to music, and most people who buy good headphones can't hear all the way up to 20kHz. There is such a thing as "good enough". I know absolutists try to push for specs that reach the bleeding edge of extreme situations, but when you are sitting in your living room listening to music, you don't need that. You only need that if you are sitting in an anechoic chamber listening to test tones... and I don't particularly find that sort of thing entertaining myself.

The specs of human hearing are more important than the specs of audio equipment. The impenetrable ceiling of audio quality exists more in our ears than it does high end equipment. If you work on just the specs you can hear and don't fret about what you can't, optimal sound is achievable.