Even more EQ bands are necessary for many users

[ZoSo]

There's a growing community of headphone enthusiasts on the Internet right now. You can find these people at websites such as Head-Fi. A lot of these people, myself included, have become very interested in the pursuit of accuracy and neutrality through the use of equalization. Using tools such as sine wave generators or in-ear microphones as reference, we can equalize our headphones to neutrality with great precision.

The current 8 peak bands is not enough for users like myself. The headphones I'm using with my Sansa Clip+ right now require 12 peak bands to be neutral when using the parametric equalizer on my PC. My EQ settings for another pair of my headphones on my PC have a whopping 26 peak bands.

JdGordon says here http://forums.rockbox.org/index.php?topic=42724.0 that the absolute maximum possible is 32 bands. I feel there would be many Rockbox users who would benefit from having all 32 of these bands available; 30 peak bands, the hi shelf and the lo shelf. More and more headphone enthusiasts are discovering the art of parametric equalization every day, so more and more Rockbox users are going to be finding the current 10 bands to be limiting to their needs.

[wodz]

*IF* you need more then a few bands you totally miss the point of parametric eq, that is granted. Not speaking that it is a waste of cpu cycles which are rather limited in embedded systems. If you feel like needing such unusual setup feel free to modify the sources and recompile yourself.

[AlexP]

Ten is already too many really on these systems. As wodz says if you want more you will need to build your own version.

[saratoga]

I suspect if you're using 12 bands on a parametric EQ you may be doing something strange.  Its pretty common for even pro-level mastering hardware to have only 7 bands.  Can you post your EQ settings?  Maybe I can give you feedback.

[ZoSo]

*IF* you need more then a few bands you totally miss the point of parametric eq

I respectfully disagree. I've experienced a lot of success making headphones fairly neutral in the past by removing their deviations from neutrality with parametric equalization, all while without creating audible (to my ears) artifacts. If I am creating artifacts, I can't hear them, so I really don't care. The post-ringing of minimum phase equalizers is generally masked by the signal itself anyways.

Headphone responses can have a lot of sharp treble peaks which are unkind to the ears and fatiguing and this is compounded by the tendency of headphones to create ear canal resonances.

I know the "conventional wisdom" is to use only a few, broad adjustments with a parametric EQ, but there would be no progress if everyone conformed to "conventional wisdom." I've had tremendous success equalizing out very narrow peaks and valleys in headphone response without creating audible (to my ears) artifacts. I listen to a sine wave generator to estimate the location, magnitude and Q of these peaks and to verify I've nulled said peaks.

Not speaking that it is a waste of cpu cycles which are rather limited in embedded systems.

Well, obviously one would use less bands once they found it causing their music to lag/skip.

Can you post your EQ settings?

Here's the EQ settings I just came up with for my Koss KSC75, which use 12 bands. Pic attached are these settings turned upside down so as to represent a quasi frequency response measurement, i.e. these are what the headphones sound like to my ears. The large 7.5 kHz peak is an ear canal resonance; the other peaks and valleys are limitations inherent in the driver. I set the precut to -10 and the result is the bass response is completely flat like a Stax headphone or Etymotic earphone, and I can confirm this by measuring the headphones with a microphone. The somewhat harsh quality of the KSC75's treble is nullified by removing the sharp peaks in its treble response which result from driver deviations from neutrality and the large 7.5 kHz ear canal resonance which all headphones tend to produce with my ears.

My settings for my PC equalizer:

Filter  1: ON  PK       Fc     128 Hz  Gain  -3.3 dB  Q  1.00
Filter  2: ON  PK       Fc    20.0 Hz  Gain  10.0 dB  Q  0.80
Filter  3: ON  PK       Fc    60.0 Hz  Gain  -2.0 dB  Q  1.00
Filter  4: ON  PK       Fc   144.0 Hz  Gain  -1.0 dB  Q  0.20
Filter  5: ON  PK       Fc   3,829 Hz  Gain   6.0 dB  Q  5.00
Filter  6: ON  PK       Fc   1,968 Hz  Gain  -2.5 dB  Q  8.00
Filter  7: ON  PK       Fc   4,600 Hz  Gain  -2.5 dB  Q 10.00
Filter  8: ON  PK       Fc   5,400 Hz  Gain   1.5 dB  Q 10.00
Filter  9: ON  PK       Fc   7,500 Hz  Gain  -7.0 dB  Q  7.00
Filter 10: ON  PK       Fc   7,000 Hz  Gain  -2.5 dB  Q 10.00
Filter 11: ON  PK       Fc   8,500 Hz  Gain  -4.5 dB  Q  7.00
Filter 12: ON  PK       Fc  11,600 Hz  Gain  -4.0 dB  Q  7.00

I realize some of the really high Q value peaks are thought to be able to create artifacts, but again, I can't hear said artifacts. If the end user can't hear the artifacts, no harm done. I also feel the availability of higher Q values in Rockbox would be useful too, as the peaks in headphone response can be quite narrow at times.

In order to see how many users such as myself are doing this kind of equalization to headphones these days one only needs to google "eq site:head-fi.org," or "equalize site:head-fi.org."

[AlexP]

I don't think those on head-fi are a particularly large section of the population. I do like that website though, it reads warmer and more open than most other websites, which seem muddy by comparison, especially at the mid page point.

That aside, if you feel it helps to have loads of bands then by all means go for it, just don't expect to see more bands appearing in the standard Rockbox builds.

[saratoga]

I respectfully disagree. I've experienced a lot of success making headphones fairly neutral in the past by removing their deviations from neutrality with parametric equalization, all while without creating audible (to my ears) artifacts. If I am creating artifacts, I can't hear them, so I really don't care. The post-ringing of minimum phase equalizers is generally masked by the signal itself anyways.

I don't doubt that this sounds ok to you, most of those adjustments have a minuscule effect on the actual signal as they are very small.  How are you generating these parameters though?  I'm quite skeptical for instance that you actually heard some of these resonances, as 1 or 2 dB adjustments can be hard to hear in fairly broad bands, let alone with very high Q.

[ZoSo]

How are you generating these parameters though?

0-1000 Hz by measuring the headphones with a microphone, past 1000 Hz by using a sine wave generator (Sinegen) to sweep through the frequencies while listening to the generated sine tones. I then mentally do my best to make what I hear match a smoothly ascending and descending diffuse field target curve. Pic related.

One of these days I'm going to build some in-ear microphones as described here: http://www.davidgriesinger.com/Binaural_hearing_and_headphones.ppt then use them to match the headphone response at my eardrum with the response of equalized-to-neutrality speakers at my eardrum. This type of thing is certainly going to require a lot more than 8 peak bands of EQ.

1 decibel differences are very difficult to perceive whilst listening to music... notsomuch whilst listening to test tones, though:

http://www.audiocheck.net/blindtests_level.php?lvl=1

I can reliably detect half-decibel differences using the above test.

I do like that website though, it reads warmer and more open than most other websites, which seem muddy by comparison, especially at the mid page point.

 

[saratoga]

How are you generating these parameters though?

0-1000 Hz by measuring the headphones with a microphone, past 1000 Hz by using a sine wave generator (Sinegen) to sweep through the frequencies while listening to the generated sine tones. I then mentally do my best to make what I hear match a smoothly ascending and descending diffuse field target curve. Pic related.

I think equalizing based on pink noise or whatever is a reasonable idea, but I am skeptical the EQ settings you have chosen are all that realistic.  I suspect that you could turn a few of them off and not even realize it.

1 decibel differences are very difficult to perceive whilst listening to music... notsomuch whilst listening to test tones, though:

That is not the same thing.  You're talking about a 1 or 2 dB change over a very narrow range of frequencies, not a volume change of the entire signal.  For example your 1.5 dB change at 5400 Hz with a Q of 10 affects  just  3% of the audible spectrum.  To have the same power as a Q=1 filter at 1 dB it would have to be about 10x as large.  Or to put it another way its like saying you have a Q=1 filter at 0.1 dB.  You might hear it on a pure tone if you listen carefully (and precisely align the tone with the center of the filter) but thats basically irrelevant to actual audio.

[ZoSo]

I suspect that you could turn a few of them off and not even realize it.

If I flip back and forth between the original EQ setting and one with the 1-2.5 decibel gain bands removed whilst listening to music, I can definitely hear the difference as I flip back and forth between them, and I can very easily hear those peaks and valleys in the unequalized response whilst sweeping a sine tone through the various frequencies and listening to it. You're just going to have to trust me on this one.

Yeah, they're small differences whilst listening to music, but once I know they're there from listening to the sine tones it's going to drive me nuts to not EQ them out just based on principle and all. I'm detail-oriented and like doing things right.

[toehser]

I would just suggest that the cost/benefit by upgrading from sub-$10 headphones would be saner than messing with 31-band EQ pink-noise analysis...

[ZoSo]

I would just suggest that the cost/benefit by upgrading from sub-$10 headphones would be saner than messing with 31-band EQ pink-noise analysis...

The only option if you want true, portable neutrality are Etymotic IEMs and those aren't practical for me because I produce above average earwax.

Really, the only truly neutral headphones or earphones out there right now are Etymotic ER-4B (diffuse field curve) and Stax Lambdas (free field curve, or diffuse field curve with Stax's diffuse field EQ box). Have any other headphones/earphones but those? You can benefit from EQ.

[toehser]

All I was implying is that if $10 headphones need 12-band-EQ, maybe $20 headphones only need 6-band-EQ...

[ZoSo]

All I was implying is that if $10 headphones need 12-band-EQ, maybe $20 headphones only need 6-band-EQ...

You'd think so, but it really doesn't work like that, unfortunately. Most headphones at any price point are quite far from neutral and typically a complete mess in the higher frequencies. Also, the KSC75 are better sounding headphones than most under $100, believe it or not.

[silvertree]

One has to wonder with such complex hearing if perhaps a dedicated audiophile music player wouldn't be a better option than any rockboxed unit?