[Br777]

Quote:

My M50 settings are getting crazily complex! It's always small tweaks for the better though. I've gotten all the small precision detailed changes, now it's all just broad tonal shifts for the ultimate goal of perfecting the sound to my own ears.
 
 

 
after using and reccomending electri-q for quite some time, i ultimately switched to easy-q
 
mainly b/c it has a precut function so that none of your added peaks cause distortion
also its very easy to use.. just as easy if not even easier than electri-q. 
 
you may want to give it a whirl

 

[AstralStorm]

SA6 cable was so horrible, the isolation hardened and shrank when exposed to body heat - then broke.
 
Those peaks in M50 are pretty typical of lower-end headphones - it's very hard to tune that region, it's mostly case resonances.
(From my experience with speakers, the only thing you can do is to reshape the case - and the results are pretty unpredictable. Or of course replace the tweeter.)
 
However these seem not as lousy as these old Beyerdynamic DT231 I had to dust off and wear till I get my new IEMs.
Beyerdynamic sure likes to boost their highs.
 
Oh, about the boost causing distortion - put a VST preamp before the filter. I'm working with -10 dB one here. Most VST hosts have one built in.
 

 
---
Quote:

You can not go wrong by improving the drivers housing. You remove/reduce resonances and distorting backwave reflections, things that would require very precise EQing, even though some things just can not be done by frequency adjustment. When you equalize by ear all your work is dependant on the exact gear you used to EQ. Change cable/amp/pads and all your work is wasted. Stiffen by adding mass/replace with stiffer material. Dampen by putting soft, breathing, yet dense material (cosmetic pads are just fine if you want to try) behind the driver, exact location depends on type and construction.

 
No, you can go wrong, shift resonance frequencies to the wrong place, or reduce those that actually were linearizing the response. It's pretty impossible to do it right without lots of trial and error or an excellent mathematical model of your headphone's construction. (incl. material response, which is hard data to get) That's part of the reason why good gear is expensive.
Changing amp shouldn't do anything to frequency response unless your amp is colored, not strong enough (usually cuts extension) or has lots of harmonic distortion. (in other words, bad)
Changing pads can improve or worsen things, but at least those are typically easy to restore to original. Shouldn't hurt to experiment if you have spares.
Changing cables? Now you're really pushing your credibility. I've never seen a cable do anything important to the sound. Really bad ones can maybe cut highest end very slightly. Cables with large capacitance can change sound, but using these for audio is insane.
The only things you cannot improve with eq are harmonic distortion and square wave response - first depends on case and the driver, also on the amp, the second depends on the driver and coil inductance.
Square wave response can be changed by increasing or decreasing the mass and magnet of the driver, stiffening or loosening it with proper coating, adding capacitance (such as with a cheating cables!) - however these treatments might have an unknown effect on linear frequency response.

 

[AstralStorm]

Oh, hello again.
While equalizing, make sure you check with real music or pink noise too, or skip a bunch of octaves and not just balance nearest neighboring frequencies. It's too easy to forget the set point and end up with overly bright or dark signature if you're not careful, e.g. the above chart vs the really definitely final one below.
The former one had the Sennheiser veil feel to it, nice and non-fatiquing but not linear actually.
 

 

[jerg]

Quote:

Oh, about the boost causing distortion - put a VST preamp before the filter. I'm working with -10 dB one here. Most VST hosts have one built in.
 
---

Where might I find the VST Preamp option? I'm sorry, all of this is really hard for me to figure out, it was a lot of trial and error just for it to save the current EQ settings I made lol.

 

[AstralStorm]

It's not in this plugin. But you know you can chain VST plugins, right?
Sample one is FreeG: http://www.kvraudio.com/get/2506.html
 
Also, most hosts have such a volume control. For example, VSTHost, which I'm using for system sound (in conjunction with Virtual Audio Cable), has a preamp control on input and output in addition to nice VU and RMS meters.
 
My EQing method starts with finding the max excursion, which is typically at the ends of the range (bandwidth limit compensation).
Then you should preamp so this max stays at 0 dB, adjust analog volume until 1 kHz tone at, say, -6 dB full scale sounds slightly annoying. This should be around 50-60 phon.

 

[WTFGoD]

the first link for pink noise/wave sweep is dead can someone help me out with a link?

 

[MaciekN]

AstralStorm, you are both wrong and right in what you wrote in response to my previous post. You are right because everything that you have written is true but it is pure theory.
 
   In real life cup resonances are usually not controlled at all, contributing to colouration rather than linearity. Consumer amplifiers more often than not have some design flaws that contribute to sound colouration. As for cables, I totally agree that conductivity and inductivity are important but stock cables, especially on low and mid-fi gear, are simply too thin to physically allow enough flow of current, thus limiting lowest bass extension. This may also cause bass to veil mids, as it can not extended well it concentrates in mid and upper bass.
 
But my point was that every part of our rigs has it's impact on what's at the end. When you change it you have to go through all the EQing again. What I suggest is to do the dirty job first and then, when you have a better base to work on, do the EQing to fix what's left. From my experience, there are things which you just can not fix with EQ.
 
If you check out some popular threads about modding headphones you will find that what I wrote about damping, replacing stock materials etc is often essential for achieving better performance, even with expensive gear like top Denons to give an example.
 
 
 
 
 
 

 

[Arksun]

You have it the wrong way around surely. The whole point of doing an HRTF is to show how a persons ear would actually percieve it!. So that should be the graph you go by that gives the closest idea of what you'll hear, not the raw initial response graph.
 
Of course tolerances/differences in headphone manufacturing, size of head, ear flap shape etc all play their part too if trying to use an eq to flatten things further, the problem then is, due to the steepness of some of those bell curves and their sizeable depth too in linear phase, they introduce their own pre and post ringing artifacts. Very few linear phase eq's on the planet can really reduce these artifacts down to a minimum, only one that springs to mind really is PlPar EQ on its highest quality level 7 setting. This munches on the cpu and also creates a big latency delay to the signal.
 
Quote:

This graph is the raw frequency response data from HeadRoom's own dummy head, before the dummy head's HRTF (Head-Related transfer function) data has been subtracted from it. While it is not at all indicative of the headphone's actual frequency response, it is indicative of the dummy head's perceived response, that is, what the dummy head actually heard at the time of testing, which is what matters. And the dummy head's perceived response is very interesting indeed. Look at these graphs:

 

 

[sh4w]

Heres what I came up with after playing around with sinegen for about an hour and finding all the major peaks on my K701.

2503,5 Hz -7dB BW: 0,2
6100,0 Hz -7dB BW: 0,3
7432,0 Hz -5dB BW: 0,2
12110 Hz -6dB BW: 0,2

 

[jerg]

My M50's setting has changed a lot again.
 
This time I added an even increase in mids and bass, and suddenly music sounds so much fuller.
 

 

[AstralStorm]

Quote:

You have it the wrong way around surely. The whole point of doing an HRTF is to show how a persons ear would actually percieve it!. So that should be the graph you go by that gives the closest idea of what you'll hear, not the raw initial response graph.
 
Of course tolerances/differences in headphone manufacturing, size of head, ear flap shape etc all play their part too if trying to use an eq to flatten things further, the problem then is, due to the steepness of some of those bell curves and their sizeable depth too in linear phase, they introduce their own pre and post ringing artifacts. Very few linear phase eq's on the planet can really reduce these artifacts down to a minimum, only one that springs to mind really is PlPar EQ on its highest quality level 7 setting. This munches on the cpu and also creates a big latency delay to the signal.

Specifically, linear part of the HRTF would be the difference between raw measurably flat response and corrected-by-ear response - you can apply this to every corrected device. (equalized to measurable linearity with the same rig, possibly automatically)
However, there are other parts of HRTF that are non-linear, so equalization doesn't cover everything. Crossfeed and some slight reverb help a lot here.
Of course typical crossfeed uses a sphere model of the head - it's an open engineering problem to produce proper crossfeed (esp. with correct frequency response of copied sound) given e.g. 3D face mesh - and do it with reasonable performance.
For basic crossfeed, try Bauer Stereophonic to Binaural (BS2b), for more advanced, maybe Ear-Fit.
For reverb... well, there's plenty around. SIR is excellent if you can find a good impulse response. TAP reverb (LADSPA) comes with a few excellent impulses.
 
I do share your doubt about EQs, however phase linearity isn't as important with IEMs and headphones unless your eq goes unstable (which can't happen with 99.(9)% of those, as they use 2nd order or lower filters), or the device uses multiple drivers. This is because there's far less phase interactions in the output. Also, most parametric eqs use a digital model of a 2nd order Butterworth bandpass filter, which has no ringing and nice phase behavior (-45 degrees at peak, smooth, proportional to frequency). FIR eqs can have ringing/preecho if the window size is too long and/or window shape is sinc, but have linear phase, best numeric and performance properties - most graphic EQs use these. The most reasonable compromise to me would be a 2nd order Linkwitz bandpass - this is almost linear phase and almost ideal eq, no ringing. Adds +1 dB when not bypassed - but this can be easily adjusted.
I suspect a few parametrics that boast linear phase actually use this one and not FIR.
 
I think that Electri-Q "Analog" filter is a convolution eq of okayish quality - it probably uses an impulse response of some real analog eq stage, thus copying harmonics. The "Digital" one seems to be a direct model using the bilinear transform (I don't have the code) and should have 0 preecho or harmonics in comparison, but depending on the exact specifics of the algorithm, it could be slower or less accurate.
The default Peak shape seems to be 2nd order Butterworth bandpass/band-reject. "Linear" filter in Electri-Q might be a FIR filter.
 
Numeric accuracy of the eq impacts final SNR. Here, FIR is king, because there are excellent ways to compute FFT and iFFT of the input accurately.

 

[sh4w]

After playing around for 1 more day with the sine sweep I came up with this:

2400 Hz | -10 dB | 0.6 | Peak Type I
2600 Hz | 2 dB | 0.2 | Peak Type I
5900 Hz | -10 dB | 0.4 | Peak Type I
6500 Hz | 3 dB | 0.2 | Peak Type I
8500 Hz | 12 dB | 0.4 | Peak Type I
10000 Hz | -3 dB | 0.1 | Peak Type I
11500 Hz |-12 dB | 0.4 | Peak Type I

I havent figured out how to select Peak type 2 yet, lol, cant find it anywhere :/

The sine sweep sounds pretty damn flat now. Its not 100% perfect, still a few peaks/dips left but nothing major.
The music sounds unbearable now after turning off the EQ, TOTALLY unrealistic mids, brain piercing highs and TONS of treble noise :blink:
And note that the dac/amp combo inside the MS40 is pretty damn dark, so its a very good fit for the K701 but imagine I had a brighter amp x_x

 

[plin]

Quote:

After playing around for 1 more day with the sine sweep I came up with this:

2400 Hz | -10 dB | 0.6 | Peak Type I
2600 Hz | 2 dB | 0.2 | Peak Type I
5900 Hz | -10 dB | 0.4 | Peak Type I
6500 Hz | 3 dB | 0.2 | Peak Type I
8500 Hz | 12 dB | 0.4 | Peak Type I
10000 Hz | -3 dB | 0.1 | Peak Type I
11500 Hz |-12 dB | 0.4 | Peak Type I

I havent figured out how to select Peak type 2 yet, lol, cant find it anywhere :/

The sine sweep sounds pretty damn flat now. Its not 100% perfect, still a few peaks/dips left but nothing major.
The music sounds unbearable now after turning off the EQ, TOTALLY unrealistic mids, brain piercing highs and TONS of treble noise

And note that the dac/amp combo inside the MS40 is pretty damn dark, so its a very good fit for the K701 but imagine I had a brighter amp x_x

I see that you totally 'flatten' the normal resonance around 3KHz that the ear is naturally producing. I don't think that this is wise. What do others think about this?
Should we completely wipe out this very important resonance? Should we tame it just a little (maybe 6dB only)?
 
 

 

[sh4w]

I see that you totally 'flatten' the normal resonance around 3KHz that the ear is naturally producing. I don't think that this is wise. What do others think about this?
Should we completely wipe out this very important resonance? Should we tame it just a little (maybe 6dB only)?
 
 

Yea I'm also pretty sceptical about this "ear canal bypassing" but so far it really doesnt sound bad. I'd really like to see opinions of people that are more educated in audio recording, though.

 

[AstralStorm]

This resonance is enhanced the closer the sound source is to the ear, so most IEMs have this frequency range boosted in effect.
Funny thing is that inserting certain IEMs deeper can improve this or make it completely unbearable.
Some people have pinna that attenuate this frequency range, while others have ones that boost it.
 
My latest simplified EQ for VSonic GR07 with decored Comply M foams shows this phenomenon pretty well, but different tips yield different eq curves.
(This eq is targetted for hardware implementation, thus simplified as much as possible.)