superjawes wrote:If you want some interesting reading on the subject of personal preference, Harman International is doing research on what constitutes a "good" sound for headphones. I've been reading this through Inner Fidelity (most recent writeup linked).
It's interesting because there is already a pair of headphones (NAD VISO HP50) that was designed based on this research, but they also found that the right frequency balance for "a good speaker in a good room" isn't necessarily a good response for "head-mounted speakers".
It's clear that a flat frequency response isn't the ideal, but what is? The referenced research was supposed to figure that out (and it looks like they did a pretty good job), but the ideal changes depending on who's doing the listening and in what situation. I tried to come up with a way to create an FR band using my own ears as the mics, so I had data that worked well for me, and I think it turned out very nicely.
I used Audacity to generate sine waves (there are better suited tools, but this was what I had handy). I started at 125 Hz (I forget why I didn't test bass, this was months ago) and an amplitude of 1, and then adjusted a 175 Hz signal's amplitude until it sounded the same volume to my ears as the 125. When that's done, compare the 175 to a 250, and on down the line. Recording all of the resulting amplitudes and applying some corrections gives an FR band. This is what I got for my heavily modded Grado SR60i's:
... and the SR60i graph from the Inner Fidelity link, for comparison:
Mine is tilted CCW a tad, has more bass at a slightly lower resonant frequency with higher Q (though it's not really visible here), is very nice 2.5K to 10K, clearly shows the drop at 12.8K, and my hearing doesn't quite make it to 16K (other testing shows my ears are good to 14 or 15K). Correcting for scale, the 2K peak is roughly the same height on each graph.
The Harman correction and my own graph agree much better than most corrected FR graphs I've seen. Most of the differences between them can be explained by my mods, and I think the only difference that's perceptual is the CCW tilt. The tricky bit is the 2K resonance - my mods reduced it greatly, but it shows up about the same.
Thing is, peaks on a FR graph are often also points where transients are bad - these Grados take a lot longer to start or stop making a sound at 2K than they do at other frequencies. A quick scan, such as that done for most FR graph generation, doesn't give the resonance time to reach anywhere near full volume. Switching on and off a 2K sound with other frequencies nearby for comparison makes it very clear there's a problem, in a way that sweep-ish testing doesn't.
Overall, I learned a lot more about the behavior of my headphones from this process than I would have any other way. I only wish I'd thought of this before the mod process; seeing the effects of mods in detail would have been great fun.