Head-Fi.org › Forums › Equipment Forums › Sound Science › Equalise to get a flat frequency (WHY NOT?)
New Posts  All Forums:Forum Nav:

Equalise to get a flat frequency (WHY NOT?) - Page 4

post #46 of 119
Quote:
Originally Posted by barleyguy View Post


So by EQing, you are trading soundstage coherency for frequency balance.  One the primary parameters that determines audiophile sound quality is phase, often called "group delay" in specs.  It's what can make one headphone have great soundstage and another one collapse everything into left-right-center.

 

The short version is that by EQing, you are mangling your soundstage in the pursuit of a flat frequency response.  Genre of music is relevant though.  If the music you listen to is already massively EQed by the recording engineer, a little more ain't gonna hurt.  If you are listening to something that was recorded well though, you're much better off buying headphones that already have the frequency response you want and not damaging the signal any more than necessary.

 

This has not been my experience. But I also usually use a crossfeed plugin which vastly changes soundstage as well (for the better). So it could be that I'm fixing any phase issues already.

 

I use some of my headphones for mastering, and need them to match my mastering monitors. Using EQ to get their responses closer together (there are none that would match precisely) is a very useful and workable solution. 

 

It is just a bonus that it makes them even better for casual listening.

 

*shrug*
 

 

post #47 of 119
Quote:
Originally Posted by jcx View Post

anyone hear the SVS Realizer with a personal calibration?  - it been demoed at a national meet, there are threads here with owners apparently happy with the result, a Stereophile review discussing minutiae of the reproduced "soundfield"

 

the "lesson" is that (really advanced  customized to the user, room) EQ can make  a pair of headphones sound like This real set of loudspeakers in This real room

 

that experience will leave you laughing at posts in this thread about how EQ can't change "sound signature", "fix" soundstage, imaging...

 

http://smyth-research.com/index.html


Apples and oranges.  The personal calibration is performed in-ear, various digital sound processing is performed, and ESL is probably used to keep distortion characteristics of the headphones themselves low in-spite of potentially large equalization.  The Smyth Realizer is currently a one box solution that can't be described as "just" an EQ, which by itself does not inherently fix potential issues with headphones.

post #48 of 119

Like Shike said, the Realiser is far more than just an EQ.

 

It does show what you can do with DSPs if you're clever enough.  Music really can be reduced to numbers and manipulated mathematically but a lot of people don't seem to like the idea.

post #49 of 119
Quote:
Originally Posted by barleyguy View Post

I'm more in the purist camp.  You're not going to make an imperfect headphone "more perfect" by using EQ.  EQ by definition changes the phase of the signal, and phase is one of primary things that determines soundstage.  So by EQing, you are trading soundstage coherency for frequency balance.  One the primary parameters that determines audiophile sound quality is phase, often called "group delay" in specs.  It's what can make one headphone have great soundstage and another one collapse everything into left-right-center.

 

The short version is that by EQing, you are mangling your soundstage in the pursuit of a flat frequency response.  Genre of music is relevant though.  If the music you listen to is already massively EQed by the recording engineer, a little more ain't gonna hurt.  If you are listening to something that was recorded well though, you're much better off buying headphones that already have the frequency response you want and not damaging the signal any more than necessary.

 

My two cents...


Inaccurate, linear phase EQ don't change the phase response at all. Regular EQs, unless you go crazy with it change the phase response by 20 degrees or so, it;s far less messy that what a multi-way speaker does.

 

 

post #50 of 119
Quote:
Originally Posted by liamstrain View Post



I disagree. No headphone is perfect. But I guess it all comes down to preference and need. 

of course, but finding one near perfect for your own personal use is not that hard to achieve. not very hard to find an ''all-rounder''. most people are picky. i'm not. i like speakers or headphones that do their job and work best for me personally. everyone is different. i just can't own billion of different headphones personally like everyone else. i wouldn't know what to do with them.
post #51 of 119
Quote:
Originally Posted by khaos974 View Post


Inaccurate, linear phase EQ don't change the phase response at all. Regular EQs, unless you go crazy with it change the phase response by 20 degrees or so, it;s far less messy that what a multi-way speaker does.

 

 

Agreed about linear phase EQ.  But the vast majority of EQs on the market, including the ones built in to virtually every device, aren't linear phase.  They work through phase cancellation.

 

I'm a big fan of phase accurate speakers also.  Sitting in front of some Zu Druids right now.

 

post #52 of 119

"minimum phase" linear filters - like most analog EQ, and DSP IIR filters actually compensate for the minimum phase shift associated with amplitude variations in the transducer - EQ for flat frequency response and you improve the phase response

 

multi-driver transducers and the output above cone breakup at higher frequencies aren't "minimum phase" so linear EQ doesn't fully correct the "excess phase" part of the response

 

the "excess phase" response can in principle be fixed by more complex filters with all-pass phase EQ sections or in DSP FIR filters - usually you need automated measurement - our ears really aren't sensitive to phase above a few kHz

 

I do think the SVS Realizer is relevant to the discussion of what EQ can in principle do - even if you want to make distinctions between increasingly general "EQ"

 

1st order - a single frequency response adjustment curve - common to both R,L channels

 

next - custom per channel - even manufacturers bragging about matching their transducers only claim "1 dB" matching - clearly above ABX DBT detectable difference thresholds

 

cross-feed is simply "matrix EQ" - uses signals from both channels, cross-feed is the start of hrtf simulation, allows aproximating your hearing free field sounds with both ears

 

Dolby Headphone, some other SW adds simple idealized models of speaker separation, room reverb

 

the SVS system skips the modeling and measures the sound at your ears in the sweetspot of a real room and loudpseaker setup - including surround systems up to 7.1, adds angular response with the head tracking

and uses the in-ear mics to measure the headphones you're using, applying EQ for each channel - you can have separate headphone corrections in addition to multiple loudspeaker+room personal calibrations

 

 

for me all of the above are fundamentally related - we are multiplying the input channels by frequency/phase correcting filter transfer functions, possibly in in a full matrix and calculating the corrected 2-channel output to the headphones, only the head angle tracking of the SVS takes in any "new" information - and that just allows interpolation between transfer functions measured during the in room personal calibration


Edited by jcx - 1/14/12 at 8:43am
post #53 of 119

You can't forget various time domain related modifications that surely occur too with the SVS though, which would easily take it out of the "only EQ" territory very quickly.  I wouldn't call crossfeed an EQ, unless you want to call channel mixers EQs . . . and once again, crossfeed messes with time domain too.

 

The time domain modifications and crossfeed have much more to do with spacial cues and emulated room characteristics imo.  The EQ is a necessary part, but on its own wouldn't be sufficient without having key characteristics modified.  Moreover, it can not fix or improve every set of headphones which is predominately the point of this thread.  The answer is it can some, can't others, and will have some form of consequence in varying degrees.

 

Now if you said multiple stage EQs, digital processing, channel mixing, and time domain alteration (a lot of it may be redundant, sorry) can have a great impact on key characteristics of on an already good set of headphones then I doubt you'd have a single argument here.

post #54 of 119

Standard parametric equalization cannot correct the following issues:

1) case resonances - since those happen too late. You could squelch the driver, but that probably won't fix these, unless you go way below the pressure required to excite the resonant mode - which will probably cause the sound to be nonlinear.

2) complex "moving" driver resonances - you would need a time variant system for that and they're much harder to describe. These can be sometimes noticed in a correctly done CSD. A typical one in low end dynamic drivers is bass bleed into the midrange. (looks like an angled line on a CSD at the lowest end)

3) driver resonances not proportional in decay to peak size - eq will either overdamp (rare) or leave some of the resonance in. A FIR filter with adjustable phase/group delay can usually correct it.

4) poor THD behavior - "dirty" and sometimes "harsh" sound

 

Standard parametric EQ (IIR) sections do have relatively high phase delay, not exactly like a FIR minimum phase. It's somewhere in between. Do not confuse this with the common design approach of trying to achieve minimum phase as opposed to, say, maximum Q or smallest filter.

Anyway, static phase changes like those introduced by equalizers or most crossover circuits are very hard to hear, unlike the frequency response differences.

post #55 of 119
Quote:

Originally Posted by AstralStorm View Post

 

Standard parametric EQ (IIR) sections do have relatively high phase delay, not exactly like a FIR minimum phase.

 

That does not seem to be the case, I tested an IIR parametric equalizer and it is indeed minimum phase or at least very close to minimum phase. Taking the IR of the parametric equalizer at various settings and creating a minimum phase FIR filter out of it, the difference between white noise filtered with the original IIR parametric EQ and the minimum phase FIR filter is very small, and apparently mainly results from the limited length of the FIR filter (increasing it makes the difference smaller).


Edited by stv014 - 11/25/12 at 7:54am
post #56 of 119
Quote:

Originally Posted by AstralStorm View Post

 

2) complex "moving" driver resonances - you would need a time variant system for that and they're much harder to describe. These can be sometimes noticed in a correctly done CSD. A typical one in low end dynamic drivers is bass bleed into the midrange. (looks like an angled line on a CSD at the lowest end)

 

If it was a time variant system, then a single CSD (which is just a particular visualization of an impulse response) would not show it. Time variant implies that measuring the response repeatedly would result in different CSDs.


Edited by stv014 - 11/25/12 at 7:58am
post #57 of 119
I love how quickly these threads degenerate from useful tips on how to improve the sound of your system to scientific minutia that would discourage Einstein.

The best part is that instead of talking about how they've employed eq, most people are content to just give excuses why they've never used it. Nothing like real world experience!
Edited by bigshot - 11/25/12 at 10:58am
post #58 of 119

well i may not know too much about the science about it. but i know it works at least for me. (there are limits of course)

eq mostly fixes sound sig preferences imo. it doesn't really improve absolute performance. but it does compensate for the flaws (rolled of subbass for one) and improves percieved quality depending on user tuning skill and requirements needed to satisfy the user listening to a specific genre. certain curves like the dip in mids can push vocals away while raising it will get the vocals closer all over your face.

buying a better headphone will of course result you on having a better starting platform since it has less distortion,resonance problems and transient responce along with decay etc.
but a cheaper headphone with eq done right may sound better than a more expensive one only because the sound signature of the cheaper, tuned one satisfies the user so much that it outweights benefits of the extra detail,soundstage,etc of the more expensive headphone especially if the latter has a sound sig the user dislikes.

 

post #59 of 119

Quote:
Originally Posted by bigshot View Post

I love how quickly these threads degenerate from useful tips on how to improve the sound of your system to scientific minutia that would discourage Einstein.
The best part is that instead of talking about how they've employed eq, most people are content to just give excuses why they've never used it. Nothing like real world experience!

Equalization kicks butt and I don't know why some people don't use it. It strongly compensates for headphone deficiencies and improves on already great headphones.

 

An approach would be to use compensated FR plots from IF, ER, GE or other web site and try to compensate for the measured deficiencies. Then use our ears to fine tune to our preferences.

 

That said, the better and closer to our preferences the headphone, the less equalization and corrections needed. I'm happy with my HD558. Maybe later I'll upgrade. I have my KSC-75 around for running and my HD-202 for some odd reason. Maybe I'll sell them later since I don't collect headphones.

 

Quote:
Originally Posted by stv014 View Post

 

If it was a time variant system, then a single CSD (which is just a particular visualization of an impulse response) would not show it. Time variant implies that measuring the response repeatedly would result in different CSDs.

100% agree.

post #60 of 119
Quote:
Originally Posted by stv014 View Post

 

If it was a time variant system, then a single CSD (which is just a particular visualization of an impulse response) would not show it. Time variant implies that measuring the response repeatedly would result in different CSDs.

 

Yes, actually I've figured it out, what I'm talking about is a nonlinear effect, but still time-invariant. However, do remember that actual dynamic drivers are "time-variant" in the sense that they're dependent on conditions such as humidity and temperature. (affects stiffness and speed of sound)

What would you make of a time variant deterministic system? Suppose the driver cannot be excited by another beat for 0,01s after an excursion. That's an example of a time variant system, will not show in an impulse response, will definitely be repeatable. You'd need an impulse train to find this out. (or a maximum length sequence)

If the effect happens within decay, it will affect the CSD. (because the driver will respond to being hit by the pressure backwave)

 

Single impulse response will show some of nonlinear properties, but will not correctly describe them. Similarly, it can show some acausal properties.


Edited by AstralStorm - 11/26/12 at 1:30pm
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › Equalise to get a flat frequency (WHY NOT?)