jeffrey ni.: Building IEMs

My dad came back from China yesterday with parts to make an IEM. He brought back putty, SE535 shells, ED-29689 (well the China OEM edition) balanced drivers, MMCX connectors, litz wire, one pair of filters, and a headphone cable.

I noticed that the sound filters are too small to snugly fit into the output nozzle of the SE535 shell, so I decided to first try making the earbuds without filters and with the drivers just directly wired to the connector. I held everything in place with putty (the driver is not vented, so it is ok).

I was surprised at how good they sounded. The biggest part was the isolation: with foam tips outside noise is seriously attenuated.

One interesting thing I found was how easily the sound balance could be affected by the distance of the driver to the nozzle. At first, one side's driver was about half a millimeter farther than the other side to the output, and the imbalance was instantly obvious. After shoving the driver in closer so both sides matched, the volume was nicely balanced. I'm not sure if it is an amplitude effect, or if it is a brain processing effect caused by phase differences.

I felt that these sounded really nice for vocals and piano/orchestral works. Even though balanced armature drivers are generally thought of as having bad bass reproduction, I thought that the low end was fairly well-represented (but not punchy). Doing some basic frequency sweep tests, I could start hearing bass somewhere around 30 to 40 Hz.

I next tried adding a filter to the setup. I took apart the IEMs and added a bit of acoustic tubing (from DIYearphone) to couple the driver to a green filter (well I ordered a green one but they look white, so I'm not sure).

Took this photo during disassembly (accidentally broke the MMCX connector)

I really wasn't able to tell if there was any difference in sound between the filtered and unfiltered driver, but this might also be because I wasn't able to test them back to back. There as fairly large gap in time between when I heard the unfiltered and filtered ones (since I had to add the filter!). I might try to thicken the filters so they fit snugly in the output nozzle so it is easier to compare different filters.

The problem with both of these setups is that the driver is directly connected to a device's output. The drivers have a relatively low impedance (~10 ohms @ 1kHz according to the Knowles datasheet), and the volume level needs to be really low for them to be safe to listen to. As a result, relative to the music's amplitude, the noise floor is incredibly high. Any time my computer or phone's output turns on, noise is immediately heard. Not to mention I have barely any control over the volume of the music (on Android, one level above muted is loud already!)

To remedy this, I decided to put a 18 ohm resistor in series with the driver. This should effectively increase the output impedance of the amplifier, and decreasing the voltage drop across the drivers. At the same time, this will change the characteristics of the RL circuit formed by the resistors and inductor in the driver. I'm not too sure how it effect the sound output, but I'm not worried.

I didn't have any SMD resistors, so I just used the through hole ones I had on hand. I measured their resistances and they were within 0.5% of each other (not bad for 5% tolerance resistors). It's incredibly important to get matching resistors or else each side will sound different.

After adding the resistors, I didn't have room for the filters so I just decided to leave them out (and I couldn't tell any difference at the time).

I also broke the headphone cable when disconnecting one side (no wonder the cable was so cheap) and had to fix them. Even with more expensive cable it's probably good to always disconnect the IEM by pulling on the connector, not the wire. Woops.