Reducing Headphone Treble Response with Paper Filters in Front of the Driver
Takato14 is a freak for old headphones and regularly sends in boxes of vintage cans for measurement. (Thanks, mate!) Evidently, he's also a sucker for a lost cause and decided to try to get rid of the nasty 6kHz peak in the Sennheiser HD700 response.
His thesis is the pads were to blame and proceeded to perform a significant modification, which did net good results. He felt they could be further improved with modifications to the acoustic exit hole behind the driver, and small disks of filter paper in front of the driver. The filter paper was intended to reduce some of the treble response as he felt them a bit too hot.
I did some initial measurements on the mod as he sent them in, but there were some odd artifacts in the bass distortion. When I sent him the results, he asked if I could remove the filters and measure the cans without. I did so and found the distortion to be significantly reduced but, of course, the treble was back to being too elevated.
Well, now I was curious. Where did the distortion come from? He included four filters for each earpiece, what would the response be with only one or two? And...there you go, a nice little rabbit hole to dive into. I proceeded to measure the headphones with a variety of damping to see what it would look like.
In addition to the filters Takato14 included, I also tried some toilet paper as a replacement for the Tyvek filters as they seemed to be rather impermeable to air. Here's a combined frequency response plot for all the headphones.
I think it's probably best not to read too much into any changes in bass response as that is somewhat sensitive to positional changes on the measurement head. For a number of these measurements I used my IEM testing routine, which only takes one measurement of frequency response and allowed me to complete the survey much more quickly. But it also means that the bass response is not averaged over a number of measurements, which means it may be inaccurate. For the same reason, treble response noise may be a bit greater on the unaveraged measurements.
It doesn't much matter though as I was more interested in the gross trends in treble response and how distortion would change as the amount of filtering increased.
In the legend of the above plot, the headphones are arranged in order of increasing amounts of filtering from top to bottom. I also made the color of the plots move from yellow through the color wheel to green so as to make the progression of change a little more easily apparent.
What is fairly readily apparent is that the treble does indeed reduce as more damping is added. (Duh.) But it's nice for the DIYer to know the magnitude of change as the filters are added, which is as much as 10dB at around 8kHz. That's a significant amount of tuning ability!
Unfortunately, there's a down side. Apparently, if you add too much filtering you end up creating a problem in that at some point distortion starts to markedly increase in the bass. My guess is this might come from two possible sources:
1) The filter papers (especially the Tyvek ones) might be mechanically moving along with the driver. When the driver pushes out, the filters will move away from the driver lowering the acoustic impedance they present. When the driver pulls inward, it may suck the filters inward increasing the acoustic impedance they present. This asymmetry of positive and negative going pressure would create even order distortion.
2) Because the filter papers are just resting between the driver and dust cover, they can move around. When they do, it's possible that that they make some small amount of noise on their own. Sort of softly clapping into each other. I thought this might have been the case with the Tyvek filters, but it turns out toilet paper in place of the Tyvek also causes a similar distortion.
Here's a .pdf booklet of all the measurements, again put in order of less filtering towards more.
One thing to observe there is the initial transient edge of the impulse response when using filters 1,2, and 3 (the Tyvek filter) is is reduced relative to the measurement of filters 1, 2, and two plies of toilet paper. However, bass distortion is worse with the toilet paper. This indicates to me that the Tyvek is slowing impulse response, possibly because the air isn't moving through it as much as driving the Tyvek back and forth acting as a second series diaphragm, which would delay fast transients.
At any rate, my take is that the very thin damping materials Takato14 used (scavanged from surgical breath masks) do provide provide roughly about 1-2dB of treble reduction per layer above 5kHz. A handy little rule of thumb for DIYers, I reckon. And secondarily, once you get to about 3-5 layers of that (or 1-2 plies of toilet paper) you bigin to get the possibility of distortion from the acoustic impedance of the paper. Of course, different headphones will respond differently, but I would think the technique is still broadly aplicable if you want to knock the treble down a few dB.
If you're interested in the full story on these cans, see this superbestaudiofriends.org thread. Thanks for the lost weekend, Tak!