A note on the frequency graphs: As I keep adding more graphs, keep in mind that, since the artificial ear I use is made of blu-tack, it will slightly change its shape as the weeks go by. That means there will be variations in the response graphs as a function of time, f.e. they may develop a dip somewhere in the treble that wasn't present in earlier measurements. For this reason, if I really need to compare a new graph to and old one, I re-measure the old one as well. When you compare older graphs to newer ones, keep in mind to not look too closely into the details in the graph, like individual peaks or dips, but rather pay attention only to the overall shape of the graph (relationship between bass, mids and treble, for instance).
Another note: Frequency graphs of various kinds are scattered here and there in this thread. The graphs in this particular post represent only a selected few of those. You can browse the thread for more graphs if you feel adventurous; and if not, it's possible that you may not be missing anything too important.
5. Frequency graphs (Update: 11 Mar 2012)
Alright, peeps (anyone still use "peeps" for people?). As I outlined in my post above, I've done some experiments and figuring-outs and so on on how to record better measurements.
I'll go over the details of how my measuring setup has changed, including the mic calibrations and that, but first I'll just unload the graphs. If you're interested in the background stuff, you can read the whole post; if not, you can take in the graphs and leave.
Each graph is immediately followed by a description of what I heard in sine sweeps for that particular headphone.
Measurements were taken with an artificial ear setup. Three measurements were taken of each headphone and averaged into the result (1: ear in the center of the cup; 2: ear slightly upwards from the center; 3: ear slightly downwards from the center). The black line shows the result with 1/3 octave smoothing (given the somewhat tentative nature of my recording setup), and the light grey line has the unsmoothed result.
(Note that there is a tiny peak at 50 Hz in all graphs – ambient interference I assume, but feel free to offer guesses or knowledge on that as well.)
AKG K 250
In my post here you can read my pre-measurement description of a K 250 sine sweep. For ease of browsing, I quote it here entirely: "Sound starts around 30 Hz, climbs up until it stabilizes around 150-170 Hz, then stays reasonably flat (remember, this is by ear) from 170 to 2000 Hz. A broad hill (I didn't note the relative amplitude change, but it wasn't a massive one) between 2300 and 4000 Hz, having its peak somewhere around 3400 Hz and ending at a very noticeable, narrow notch between 4100 and 4300 Hz (it's quite deep, and I don't think I've heard anything like it on other phones). Reasonably flat again between 4300 and 9000 Hz. Peak at 9700 Hz; peak at 11500 Hz; starts seriously falling after 12000 Hz; tiny peak around 14000 Hz, <my hearing presumably fades out somewhere around 16 kHz>."
The description matches the graph surprisingly well, considering that the treble area is the most difficult to measure correctly (and, to a degree, it can't be measured to fully match how you would hear it).
This is what I heard when I did some sine sweeps after taking the measurement: Possibly hear a slight hike up in volume at 170 Hz – a slight notch at ~380 Hz, which is also visible on the graph, though it could be something else as well – curves down slightly, starting at 1.8 kHz, then leads into a peak at 3 kHz – peak takes a dip at ~4 k Hz and quickly goes up after that – comes down at 6 kHz, has another peak at 7 kHz, goes down after that – peaks at 10 kHz, then a dip, then peaks at ~11.5 kHz.
I love the graph; the K 250 has to be the ultimate in Grado-like AKG.
I think the microphone is slightly exaggerating the area around 5 kHz, but I can't be sure and don't want to make further calibrations willy-nilly, so I'll leave it at that until I have more measurements under my belt to either confirm or negate that suspicion.
AKG K 340
You can find measurements by Tyll (a guy with an expensive measuring box) of the K 340 here. My graph is both similar and dissimilar to what Tyll measured. Mine doesn't have the big treble peak at 9 kHz, although rhythmdevils (whose pair it is that Tyll measured) says the phones are bright in the treble, which my pair definitely isn't. My graph doesn't have the same valley at 5-7 kHz or so either, but that may be due to the fact that I've taken out all the cotton from my pair. My graph and Tyll's both have the dip at around 1.3 kHz, which Tyll makes a specific note about, and there's a small ledge leading into it in both graphs. The relationship between the extremes of the mids seem similar in both graphs. My K 340 seem to have better-extended bass, so they might just be the bass heavy version, if there is such a thing. Given that I've taken out the cotton, you might well see a bigger curve in the bass if I hadn't.
Here's what I heard in a few sine sweeps after the measurement: Bass sounds reasonably even from 40 Hz on – has a slight upwards movement around 150 Hz – maybe getting slightly louder at 300 Hz – starts to go down slightly around 580 Hz – goes up a bit after 1 kHz, then takes a dip at ~1.2 kHz – goes up and down after that – a slightly bigger peak compared the the other ones at 2.2 kHz – takes a dip after that, then peaks at 3.1 kHz – travels downwards until ends at a noticeable, narrow dip at 4.1 kHz – goes up from there to 4.5 kHz, but not quite as high as on the graph; maybe about the same level as the 2.2 kHz peak – reasonably stable until 6 kHz where takes a gradual dip that rolls down until 7 kHz – a notable peak at 9 kHz – another at 10.5 kHz – takes a nosedive at 13 kHz.
Beyerdynamic DT 990 (1980s)
(As a bonus, I've measured the 990s with both the original and the modern pads.)
I've no idea if anyone else has measured the vintage DT 990 (probably has), but if you know more, lemme know.
My notes from some sine sweeps after the measurement (with modern pads): Bass climbs up gently until about 110 Hz – a slight dip-let at 120 Hz – bass seems to start rolling off at about 250 Hz, but I can't be absolutely sure – seems to climb a bit around 500 Hz – peak at 3.5 kHz, comes down by 4 kHz – peak at 6.3 kHz – big peak at 9.3 kHz – takes a dip at 11.5 kHz – peak at 15 kHz.
------
The graphs below were added at later dates (see links to corresponding posts).
AKG K 240 DF
Philips SBC 3178 (presumed AKG K 260)
(Take the treble with a bit of salt, as I didn't have the original foam disc for the baffle; instead I used a bit of cloth over it.)
AKG K 241
K 250 measured in different pad/foam configurations
Comparison between K 241, K 250, and K 240 DF; measured roughly on the same day
-------
^ That's the end of the graphs. Now, about my measuring setup.
I still use the same blu-tack ear and fake ear canal, but I made some changes in the way I implement the (diffuse field) HRTF compensation. I also calibrated the microphone better (this is a problematic excercise, because I don't know the microphone's frequency response).
Below is the microphone calibration curve I used:
It consists of two things. First, a +5 dB hump between 1.5 and 2.5 kHz or so. This I felt was necessary because in earlier measurements that area was recessed way more than what I perceived, and indeed a valley was created where I heard none. The second thing is a smooth +3.5 dB or so boost to all frequencies above 500 Hz. If you look at my earlier graphs of the K 250, you'll note that it has about a 7 dB bass hump. But the problem is, there's no way a pair of headphones that a non-basshead (me) perceives as bass-lite could have a bass curve like that. Part of the problem was the presumably incorrect way I was applying the HRTF, but even withouth HRTF compensation the bass curve seemed off. I didn't know how much off it was, so I compromised by giving higher frequencies the 3.5 dB boost to lift them closer to the bass. This seems to have resulted in a more accurate graph based on what I hear.
As I said, I also changed the way I apply HRTF compensation. I used to EQ it in, but while looking at Tyll's graphs (http://www.innerfidelity.com), I noticed that his HRTF compensation seems to target specific frequencies while leaving others untouched. So, for instance, his bass curve doesn't seem to get tilted upwards in the compensated graph like it would if you EQ'd the HRTF in. It seems that applying the HRTF is simply a straightforward subtraction, which is what I've done in the graphs in this post.
I also did, as I said at the beginning, three separate measurements and averaged them together ((a + b + c) / 3) into the final graph. The official recommendation seems to be five measurements, but it gets a bit tedious, so I've cut corners.
As far as the response curve of the K 250 goes, it seems really flat, and basically that's how I hear it up until 2 kHz or so. There seems to be no "ideal tilt" to it, i.e. a graph that's slightly higher in the bass region and generally travels down a slight curve downwards to the treble. I don't quite understand why AKG had the 250 emphasize the treble region so much, either, as they seem quite neutral otherwise.
If someone has an idea of where the K 250 sort of curve might be useful (some specific studio job, etc.), let us know. Thanks.
Edited by vid - 4/14/12 at 5:18pm
) + having the ear canal tube would get you reasonably close for applying that particular HRTF, no?
