Physical vs. digital modding (convolution) again.
Yamaha HP-50S. Measurements are raw; compare only to each other and not to measurements outside this post.
Right channel, stock foam (but without the white paper disc):
Left channel, modded (two nylon discs):
Right channel, stock foam + a convolution filter applied to the test tone:
The convolution filter was aimed at turning the stock right channel response to a response identical to that of the modded left channel. Which it did. The two responses are identical, given some variation in positioning, except for the increased peak around 15 kHz or so. I listened to the phones with the left side modded and the right side convolved - they sounded more or less the same.
No artificial ear was used in the above measurements. Below are measurements with an artificial ear to test whether the convolution filter was ear shape-dependent or not. The filter was lowpassed at 12 kHz for experimental reasons.
Left, modded (no convolution):
Given some variation in ear placement (more than previously), the graphs remain similar. Based on this one test, the filter seems to have but a weak dependence on ear anatomy at worst (which is good).
I also noticed that there are a few small ridges - around 5 and 6 kHz - that occur in the same place in all of the measurements. This often means it's an ear-independent feature and thus something that's easy to fix with EQ. To that end, I made a crude convolution filter based on the earless measurements from above aimed at eliminating the two small peaks in those areas of ringing, then applied that filter to the impulse response from the modded left channel (which was measured with an ear):
The two spots of ringing are gone. Since the filter was created from earless measurements and nonetheless worked for an eared measurement as well, there's a good chance that it's not dependent on the ear's anatomy, and would thus work for my own ears too. (This was somewhat of a theoretical test since I filtered an already measured impulse response rather than filtering the test tone and measuring the result off the phones themselves - but usually the end result from these two actions seems much the same.)
Edited by vid - 10/22/12 at 4:01pm