Volume Control

After a quick test with the new IEMs, I decided to glue the shells together permanently with epoxy. The final result is rather nice looking:

Ooh.

However, they are still way to sensitive for most sources. They have a very low impedance and are too loud at low volume levels on my phone and computer. I did a bit of research and it turns out that in audio applications we want to do impedance bridging, or try to put the largest fraction of the voltage across a load by making the output impedance as low as possible (also explains why putting only one resistor in series with the IEM is bad). I needed an attenuator that would present a relatively large load (input impedance) on the output of the device, and provide a small output impedance for the IEMs.

The easiest way to create such an attenuator turns out is the simple voltage divider. By creating a voltage divider with a 33 ohm resistor and a 2.2 ohm resistor, we get an output voltage that is 16 times smaller. This lowers the output signal by 24 dB, so the sound is about 5.3 times quieter. The input impedance is about 35 ohms (with the IEM attached) and the output impedance is about 2 ohms.

Basic schematic.

I first tried out a commercial volume control from Radio Shack, which used two 290 ohm potentiometers to form the dividers for each channel. It sounded perfectly fine, just that the output impedance will change as a function of the volume setting and that the device adds another four feet of extension.

I only have carbon film resistor with a tolerance of 5%, so before prototyping, I decided to measure my resistors using a four point measurement of current through a resistor and the voltage across it at the same time. My several years old Radio Shack multimeters are probably not that accurate, but the measurement technique was something new to try out.

I wired up a simple constant current source with a LM317, and tested out the resistors to find pairs with the closest resistances.

Measurement setup.

Quick calculations.

The fact that my measurements are always off by a constant factor from the expected value makes me think that the meters are not that good (as I expected). But this is okay, as long as the error factor is always there, we can make still make relative comparisons.

I wired up the circuit for both channels, keeping the wire lengths as short as possible to make sure they didn't contribute much resistance to the circuit and cause signal leakage between channels.

It worked just as I expected. Only issues were that the prototype is rather awkwardly sized, there is a tiny bit of channel leak (barely audible), and that the connector I salvaged was not gold plated and had a bit of corrosion on it. I'll need to refine the package before using it on the go.