Website banner

This page is a summary of my research, done to determine the effects of various design parameters on the sound of circum-aural headphones using Dayton Audio CE38MB-32 drivers. This info was then used to tune, or voice (as it’s commonly called), the sound of the Homebrew Headphones.

As the MiniDSP EARS system was not available at the time, all the test results presented here are from my flat plate test rig with locating pins and holes for the headphone. This rig provided insane repeatability. I tore the headphone down and rebuilt it with almost no change to the before and after curves, same for removing it from the rig and replacing it on the rig. The MiniDSP system is not near as good. Info on my test equipment can be found here.

As I couldn’t find any publicly available practical information on tuning headphone sound, I’ve had to figure it all out from scratch.

From my research, here are the main factors that affect the sound of closed back headphones, roughly ordered from most effect to least:

Holes On Rear Of Driver

The rear of the driver has a series of holes covered in a cloth-like material. The density of this will affect the frequency response of the driver. While we can’t easily change its density, blocking the holes with plasticine will have the same effect – reducing low frequency response.

Driver holes blocked and un-blocked Driver-holes---all-blocked

Initially the Homebrew Headphones design lacked brightness and top end (ie high frequency, treble response). It took me a while to realise this (and it was almost an accident that I did!), but if you need to increase the top end response of your headphones, the easiest way is to reduce the low end, by blocking these holes.

The black line below shows what I believe to be close to the optimal ‘tuning’ of the Dayton CE38MB-32 in the Homebrew Headphones design: all except one hole blocked. Graph for rear vent blocking

Leaks In the Rear Cavity

The frequency response will be different if the cavity behind the driver is sealed to the surrounding air, or vented. This is similar to a loudspeaker, albeit the effect seems to be a lot smaller.

Venting the rear cavity ads some low frequency response over a completely sealed configuration. In reality, controlling the degree of venting would make the build much more difficult, so I pretty much dismissed this as a variable when doing the design.

Graph for rear cavity leaks

Rear Cavity Volume

Reducing the rear cavity volume reduces bass, albeit the effect is not huge. In the test below I roughly halved the volume.

Graph of reduced rear cavity volume

Seal Between Front and Rear Cavity

If there is a leak between the front and rear cavity, such as poor sealing around the perimeter of the driver mounting, you’ll loose some low/mid frequencies.

As the Dayton Audio CE38MB-32 lacks high-end in the Homebrew Headphones housing, I recommend spot gluing the driver in.

Front to rear leak graph

Leaks In Ear Cavity

I don’t consider sealing of the ear cavity (ie sealing of the cushions around the ear) a design variable, but more a problem that can occur – ideally you want this to be sealed.

This seal can break down due to hair running under the cushions, incorrect fitment of the headphones or mis-adjustment of the headband. If the angle of the headband is not correct for your hear, I recommend heating it up with a heat gun, and bending it until you get a good fit.

The graph below shows what happens if these leaks occur:

Ear cavity leaks graph

The hair leak was simulated by putting a pencil under the cushion. This is probably more extreme than most hair leaks, but shows the effect nevertheless. I sealed the cushion to headphone housing interface in one test (green line) as I wasn’t 100% sure this was sealed properly. It looks like it was!

Other Factors

I have tested a few other factors, but they didn’t have any noticeable impact:

  • Foam in the rear cavity (as is often done in loudspeaker building)
  • Foam and cloth lining in the ear cavity

I have noticed that on some commercial designs the driver is angled approx 5° rearwards, ie ‘into the ear’. I tried to test this, but without a replica of a human ear on my rig, I couldn’t measure the effects of varying this angle.

Next: Testing Your Design