[jerg]

Quote:

 
But the big question everybody probably wants to know is, what's it sound like EQ'ed up? Some of that special airiness that the DT880 is known for disappears, but with it goes most of the harshness and sibilance, and the overall balance is much more natural to my ears. Detail, of course, is somewhat lessened but not to a terrible extent. As far as I'm concerned, that detail was never there to begin with if it needed the extra treble to bring it out. Overall, these are still airy, quick, and resolving with the EQ, except now they don't bite my ears off anymore. The biggest revelation, actually, is the midrange--it really perks up (which it would have to, because, as I've argued before, it's not "recessed" between the treble and the somewhat light bass on these, but rather overshadowed by the high end).

 
Truth is, most "sound signatures" (good or bad) of headphones are just deviations from the fully neutral sound. Things like airiness / sibilance / lushness / bloating / rolloff. This whole headphone equalization business is trying to remove all "signatures" from the sounds they produce and unify them into theoretically near-perfect sound productions, disregarding technical inadequacies.

 

[scruffsaudio]

I see a different graph for the 880 250 ohm model on Headroom.
I see this one:
 
 
 
 
 
Quote:

Just popping in here again. A long while I posted about doing this with my DT880. Since then I've refined my methodology (or, more precisely, did what the tutorial actually said and used the pink noise instead of just sine-sweeping) and the results were quite different. Basically what I did wrong before was lower the "baseline" along with trying to suss out the peaks. Therefore, though I finally got the peaks sorted, everything around them was too low as well. This, I suppose, is why the tutorial recommends the pink noise, so as you can make the notches just as wide and deep as they need to be to eliminate the peaks and nothing else.
 
Some folks wanted to see my results before, so here's the updated curve:
 

 
 Filter 1: 5827.6 Hz | -6.0 dB | 0.3 Oct BW
*Filter 2: 7577.8 Hz | -6.0 dB | 0.5 Oct BW
 Filter 3: 10920.0 Hz | -6.0 dB | 0.5 Oct BW
 Filter 4: 13879.0 Hz | -5.0 dB | 0.3 Oct BW
 
*This peak will almost certainly occur in a different place for different people, as I believe it's related to my own ear resonance.
 
What's really interesting is that it matches up very well with the following, even though I used nothing but my ears to make my curve:
 

 
These are the raw frequency response curves for the DT880 and the DT990. My curve lines up almost perfectly with the one for the DT990, leading me to believe that HeadRoom's plots for the DT880 are off, or else my sample is more DT990-like in its treble. Maybe this is why so many people have different impressions of the DT880--could there perhaps be some DT990-esque DT880s floating around out there? (FTR, I have the 250 ohm version, and both these plots are for the 250 ohm variants of their respective models).
 
Anyway, I trust my ears, not the HeadRoom plot, and my ears say that there are four ~6 dB peaks of varying widths. The second peak in my plot, which occurs above the 5-6kHz one (and which is absent from the HR graph), seems to exist in every headphone I test, so I think it's some sort of resonance peak (it's usually around 7.4kHz, which is double the 3.7kHz presence peak). Also, my third peak is centered higher up than in the above plot. Looking through some of the DT880 curves others have posted, I see I'm not alone in my final result. Many folks have four peaks, usually of similar amplitude and centered in similar places.
 
But the big question everybody probably wants to know is, what's it sound like EQ'ed up? Some of that special airiness that the DT880 is known for disappears, but with it goes most of the harshness and sibilance, and the overall balance is much more natural to my ears. Detail, of course, is somewhat lessened but not to a terrible extent. As far as I'm concerned, that detail was never there to begin with if it needed the extra treble to bring it out. Overall, these are still airy, quick, and resolving with the EQ, except now they don't bite my ears off anymore. The biggest revelation, actually, is the midrange--it really perks up (which it would have to, because, as I've argued before, it's not "recessed" between the treble and the somewhat light bass on these, but rather overshadowed by the high end).
 
Overall, I highly recommend following this method. I offer my curve just to illustrate that there's not a whole lot that needs to be done to a DT880, apparently, so it's a good candidate for this type of adjustment. Don't copy my result verbatim; your pair of DT880s is likely different (more like the HeadRoom ones perhaps?), and so are your ears.

 

 

[stv014]

Quote:

 
I see a different graph for the 880 250 ohm model on Headroom.

 
That is because it is an equalized graph, while the one in the post you quoted is raw (i.e. exactly what was recorded from the dummy head without any correction).
 

 

[Argyris]

Quote:

 
Truth is, most "sound signatures" (good or bad) of headphones are just deviations from the fully neutral sound. Things like airiness / sibilance / lushness / bloating / rolloff. This whole headphone equalization business is trying to remove all "signatures" from the sounds they produce and unify them into theoretically near-perfect sound productions, disregarding technical inadequacies.

Agreed. I imagine that there are certain qualities of the presentation that come from aspects of the design apart from the frequency balance, but I agree that a lot of it comes from the signature (more so than many seem to believe). In the case of my DT880 I aimed to eliminate very specific, very noticeable flaws in the response, and not a "perfect" curve. I've noticed that when I try to compensate for the gradual rolloff in bass, for instance, the sound becomes even less characteristic (the "airiness" disappears completely). Technically, this is more accurate (though my own ears aren't accurate enough in this range for me to figure out the "proper" amount of compensation), but it's a change that gives me mixed impressions. It's a toss up as to which one I like better with some material, though the slightly mid-centric balance I get just by ironing out the treble preserves some of the euphonic elements of the original DT880 signature (the airiness, which I think comes from the relatively weak fundamentals) while still producing a relatively neutral response, and therefore I think it represents a good compromise.
 
My comments on the overall sound were made purely to describe the effects of my rather minimalist curve, which leaves everything below the treble well enough alone. Anybody who wants to go further can by all means keep going until the response is truly "flat" (or, more accurately, matches up perfectly with their own personal hearing curve), but that's beyond what the curve I posted will accomplish. As it is, with the treble smoothed out and the bass naturally a step back, the mids shine as they're the only thing left.

 

[Argyris]

Quote:

 
That is because it is an equalized graph, while the one in the post you quoted is raw (i.e. exactly what was recorded from the dummy head without any correction).
 

Exactly. I think the raw responses are far more useful for this kind of comparison. The bump around 3.7kHz is a modeling of the natural resonance of the human ear canal in this range and will appear, in some form, in every single headphone's measurement. It should not be EQ'ed down unless it's excessive (i.e. well over 10 dB or wider than it should be, like for the SRH440 with the SRH840 pads, which has a nasty resonance around 2.5kHz). Everything else, though, should theoretically correspond with the peaks you hear when you run the sine sweep program. I'll defer to anybody who has definitive knowledge one way or another, but I think there's a harmonic of the 3.7kHz resonance that appears in a lot of people's hearing, but doesn't seem to be modeled by the HR dummy head (or, apparently, in beyerdynamic's diffuse field equalization, or else they would have designed a notch there to compensate). For me it's in the same 7.4 - 7.5kHz range as for Piccolo (which is double 3.7kHz, suggesting a harmonic). I get the same peak in every headphone I test, and it never appears on any of the HR charts.
 
As for the smoothed out, compensated chart that everybody sees at HR, that basically subtracts the HRTF of the dummy from the response. This allows for a completely flat baseline against which to compare the output of the headphone, which is a lot easier for most people to understand than reading the raw data (which looks scary). Basically, the compensated charts compare the headphone's response to reality, leaving out the fact that reality is far from flat. A headphone that actually sounds "flat" from 20Hz to 20kHz would be unbelievably bass and treble heavy with incredibly recessed midrange and would be basically unlistenable.
 
BTW, if anybody wants to see the raw data for any headphone at HR, just subtract two from the chart's graphID number (e.g. one of the DT880 plots is 723, so its raw data is 721).

 

[scruffsaudio]

Quote:

 
BTW, if anybody wants to see the raw data for any headphone at HR, just subtract two from the chart's graphID number (e.g. one of the DT880 plots is 723, so its raw data is 721).

Whoaaaa..... Wildly changes the high end of the Q701.... what the heck.
 
EDIT - rest removed, read the above post carefully.
 

 

[Joe Bloggs]

@Argyris: it looks to me like the difference between the peaks and troughs in your EQ curve is just too little.  You may want to go into Sinegen and try to find the quietest sounding frequencies in between your peaks (say, eg maybe 6500Hz in between your ~5900 and 7600Hz peaks) and find out the difference in dB between them and adjust the Q value of your adjustment points (probably smaller values) accordingly until the curve in your graph roughly matches what you hear in Sinegen.  Then you may find more of that detail you lost

 
When I did it right in Electri-Q I got this masking veil of high frequency static lifted out and the phones actually sounded MORE detailed than before

 

[Argyris]

Quote:

@Argyris: it looks to me like the difference between the peaks and troughs in your EQ curve is just too little.  You may want to go into Sinegen and try to find the quietest sounding frequencies in between your peaks (say, eg maybe 6500Hz in between your ~5900 and 7600Hz peaks) and find out the difference in dB between them and adjust the Q value of your adjustment points (probably smaller values) accordingly until the curve in your graph roughly matches what you hear in Sinegen.  Then you may find more of that detail you lost

 
When I did it right in Electri-Q I got this masking veil of high frequency static lifted out and the phones actually sounded MORE detailed than before

 
Thanks for the tip. You know, I've been looking at that myself and wondering how it could be right. I was imagining that that was just an artifact of the way the chart is presented. I'm fiddling with narrower bands now. So far it hasn't made a huge difference, but a noticeable one, which is probably right on line. In an A-B test the old version definitely seems a little more muffled than the revised one, which I suppose is because the troughs are more pronounced. Using the pink noise sample I determined that the thinner bands required a little more cut away to fully eliminate the noise (about 1.5 dB in each instance). The overall balance shouldn't really change, and it doesn't, but maybe I've found a little of that lost detail.

It's difficult to say, except that this curve is clearly more efficient/less destructive than the previous one. So far I haven't found any glaring drawbacks to the change, so I'll stick with it and refine it as necessary.
 
Just a general note to those using this method. The idea is to do as little as possible. That's why Joe's suggestion was a good one.
 
EDIT: I'm sure at least some of it is just that I'm paying more attention, but I swear this small adjustment to my curve has made a bigger difference than I initially thought. I had always thought with the previous curve (the one I posted) that strings sounded a bit dull and that everything was just slightly muffled, which may have contributed to a perceived lack of detail. That "veil" of sorts seems to have been lifted, and what's more, since my peaks are deeper but more focused, they seem more effective at eliminating some of the last vestiges of metallic timbre remaining in these cans. The result is smooth yet extended through the treble, detailed but not analytical or "dry" (whatever the hell that means, anyway). The balance didn't shift appreciably (which is good), and yet there's a definite improvement. How much is still difficult to say, but at this point I swear I'm going to lose sleep because I can't stop listening. I finally feel like I really got my money's worth from my Beyers.

 

[Argyris]

Quote:

What I'm doing now is running two copies of Electri-Q between the VAC input and the speaker output in VSTHost.  The first copy is loaded with the equal loudness curve preset.  With the first EQ alone, a pair of headphones with response at the ear equivalent to a loudspeaker with perfect response will theoretically sound as loud at all frequencies if you play tones at 60dB (loudness of a normal conversation)--if your perception of loudness is the same as the average ISO 226:2003 test subject.
 
With the second copy of Electri-Q, I will tweak my custom EQ preset until all frequencies do actually sound as loud to my ear with SineGen playing through the two EQs.  The EQ should then be the perfect when you remove the equal loudness curve EQ from the chain.  I will try this myself and report back on my findings.

That's an interesting idea. I never thought of something like that because I never bothered to figure out how to chain VSTs before. I can imagine that curve would be really harsh to work with, though--if I were to do something like that I'd definitely do it at reduced overall volume. 60dB is probably okay. Since my general philosophy has been to only attack obvious peaks in the 5-15 kHz range, I've never felt like I needed to compensate for the equal loudness contour. Using the pink noise allows me to pretty accurately hammer the peaks down to the level around them (I used to do this with pure sine sweeping and it was trial and error). I never made the mistake some here have made of actually trying to get the perceived response to be literally flat from 20-20,000 Hz (!), and I merely accepted that anything apart from sharp, obvious peaks in the treble was going to be difficult to accurately fix. Using your method might make that sort of thing a lot simpler.

 

[Joe Bloggs]

The other thing I do to keep detail while taming harshness and sibilance begs to be called a cheat.

 
It's a multiband compressor.
www.head-fi.org/t/563120/multiband-compressor-just-the-thing-to-tame-bad-recordings-overbright-phones
 
This IMHO needs more love from the traditional audio guys even more so than the equalizer we're talking about here

 

[Argyris]

I've done that sort of thing before, though only as a last resort with tracks that are hideously sibilant on all my gear. I actually process them in a DAW, as opposed to running multiband dynamics as a plugin. People will have their sibilance in different places (and sometimes across multiple bands), and I like to be as precise as I can (difficult when presented with the final mix as opposed to the offending vocal track in isolation). It's not an ideal solution, which is why so far I've only done it rarely and for material I really like. For example, Genesis' A Trick of the Tail is one of my all-time favorite albums, and yet the sibilance on the edition I have (the 1994 remaster, which is the only good digital version) is unbearable. I have no idea if the original vinyl was that bad (probably not), and I won't even get started about the abomination that is the 2007 remaster (though they did fix the sibilance problems, at least). So my only choices were either to put up with it in stock form (unacceptable) or else try to be as minimally-destructive as possible with multiband EQ. The result was pretty decent, though the process has its own drawbacks (occasional artifacts that I can clearly hear, and probably a bunch of theoretical phase shift stuff that nobody can I can't hear), but they're preferable to my eardrums getting rammed repeatedly with the sonic equivalent of an ice pick every time I listen to my favorite song.

 

[Joe Bloggs]

My take on the multiband compressor is quite different.
 
Until quite recently, I listened almost exclusively at low volume, because my ears tired quickly from the high frequency components of transients quite quickly.  Using methods derived from this EQ tutorial to iron out the uneven treble response of my headphones was one part of the solution, the multiband compressor is the other half that is a constant part of my listening chain now.
 
To me it's like a HDR image processing plugin for audio.  It lets me see all parts of the audio scene more clearly without hurting my ears, just as with a HDR processed photo I can see a scene with the sun low in the sky without everything turning into sihouettes or blinding myself with the sun (or the clipped equivalent, the sky washing out).  It's allowing me to listen to music for extended periods at enjoyable volumes for the first time with an overall frequency balance that yields clarity but with the treble dynamics compressed so my ears don't tire.
 
(As far as I tell my current listening levels are still not exactly "loud"...)
 
The 5ORCERY plug-in claims to be completely transparent at rest state (when not compressing)... but I don't care if my settings are such that cymbals and hi-hats are being compressed along with sibilance... if the result is less fatigue and more enjoyable extended listening sessions then so be it.  I'm downloading a bunch of chart music (through a pay site) onto my phone for use as muzak while I'm doing stuff, for that purpose I'm applying heavy compression to the treble as well as my custom EQ while converting the files.  Even chart rubbish has its uses

But I probably couldn't listen to this stuff for 10 seconds without this processing

 
But all the other vocal genres I listen to (J-pop and anime) also feature heavy sibilance so I'm using compression for everything, just about.  I may remember to turn off the compressor if I'm putting the Planets or the 9th symphony on my phone

  But actually with the way I ride the volume knob all the time as I listen to music I may benefit from some kind of slow compression to all my music

 
But enough about me and my compressor...

 

[Agavehound]

Hey all,
     I just started playing around with SineGen and I've noticed something. When I slide through the different frequencies the sound is NOT equal on both sides of my HD598s. Could it be my hearing that is at fault? (50 yrs old and I've abused my ears over the years) Or do I need to make some adjustments to my system?
 
System:
 
Creative SB Titanium HD
Audio Console in Audio Creation Mode @ 44.1
EAX off
Equalizer Off
WMA Stereo 128 kbps
Foobar2k
 
Any help would be appreciated as I'm a complete noob at this sort of thing.'
 
Regards,
Chas
 

 

[Joe Bloggs]

Is the sound sort of biased to one side for all frequencies or does the bias change with frequency while staying centred for most frequencies?  In the first case the first thing to check is whether you somehow unchecked the "Linked" checkbox for the two channels and are setting left and right to output at different loudness.  In the second case, the usual suspects are the headphones--grab another pair of headphones and see if you get the same bias at the same frequency.  You probably won't.  But if you do, then I guess it's your ears after all

 

[Argyris]

If the bias seems to shift around a lot I'd almost bet money it's tracking differences between the two transducers. It's incredibly difficult (maybe even impossible?) to get a pair of drivers that will be perfectly matched across the entire spectrum. The result is not usually immediately obvious in normal use (though if it's excessive imaging might suffer), but this sort of test will spotlight it.
 
And if it's a static bias, like Joe said, check different headphones. If it occurs in all of them, then you probably do have a hearing imbalance. I'm 23 and I have a slight bias toward the left ear, so it's not necessarily an aging thing. From what I understand it's more common than you might think. And with headphones even a slight imbalance will be really obvious because of the way headphones isolate each channel on its respective ear.