What I wanted: a 2-input stereo mixer consisting of 50¢ of opamps, 2$ of pots and jacks, and 2$ of power supply, marked up to 10$.
What the internet wants to sell me: 35$ of ~mystery mixer~ that, according to the reviews, bleeds signal back to the input devices and has bizarre clipping issues. Available from 10+ identical listings!
I swear I'm going to have to spin a whole PCB just to get a device that works
Everyone knows and loves¹ the well-known, decades-old 3.5mm tip-ring-sleeve (TRS) audio jack. Pure analog, cheap as heck, present in everything from cars to laptops, the ultimate in backward compatibility. You might assume there is some benevolent committee of gentle old fellows, who for the last 50 years, have gently steered this ecosystem, publishing little books of numbers and tolerances describing the electrical and mechanical parameters and inspecting devices for compliance. You would be wrong.
There is no standard. There are simply a lot of physical artifacts that happen to mostly, somehow, work with each other. The tip-and-sleeve 1/4" jack actually originates with telephone switchboards, having been patented in 1902 and commercially available with 3 conductors as Western Electric 1907 catalog plug no. 85, used for the operator's headset. The 1/8" version, also called 3.5mm (even though that's about 10% bigger than 1/8"), was introduced with transistor radios.
These transistor radios (see above, schematic of Sony EFM-117J) used strange-by-our-standards transformer-coupled drive schemes, with pairs of complementary transistors driving a center-tapped winding, coupling to the output winding. These would have effectively produced whatever voltage necessary to drive the output, sort of like a flyback transformer, operating as essentially a fixed-power driver. Great design, makes good sense why they did it that way, the volume knob just sets the bias voltage for the entire output stage and therefore how much energy gets coupled into that transformer. Unfortunately, I don't have any funky transformers, and I desire to make this whole thing cheap and small, so I won't be going out and acquiring any².
Modern devices do not incorporate transformers in their audio output drivers, presumably because of size, weight, and cost constraints. Instead, it's just transistors, either monolithic or integrated into the output stage of a high-output op-amp. This does not, however, generally produce a fixed-power driver, but a fixed-voltage or fixed-current driver. This brings us to the least standard part of this non-standard: impedance. It is genuinely impossible to google anything about headphone impedance, because your results will be full of articles for audiophiles. These articles will try to confidently tell you a large number of conflicting and wrong things about headphones.³
Impedance is just alternating-current resistance. It uses the same units (ohms) and works the same way: you have a component with a specific impedance, and the voltage across that component is the impedance times the current (Ohm's law). So, if your driver works so that 100% volume is 1V, you will drive as much amperage as necessary to get a full volt across the output. Since power is current times voltage, this means that a lower impedance will pull more current to hit your fixed voltage, and thus more power. Too little power, and your headphones are too quiet; too much and they might catch fire. This means we need to decide what our maximum and minimum expected impedances are. The only wrinkle is that impedance is dependent on frequency, so your headphones don't have a single impedance, but rather a whole impedance-frequency curve (Audiophiles generally ignore this fact) -- we'll take this into account when picking a minimum and maximum.
To start, the Android Device Specification⁴, which dictates what devices can be sold as android-compatible (in theory, anyway), specifies that headphones must not have an impedance less than 16Ω, and ideally should be between 32Ω and 300Ω. Next, because that's deeply vague, I found two headphones and tested them with my LCR meter: my beloved ATH-M30s, which are inexpensive studio monitor headphones; and a pair of old no-name earbuds. The monitors were between 76Ω and 71Ω, and the earbuds were between 33Ω and 37Ω. A fixed-voltage driver would put twice as much power into the earbuds! It would also deafen the earbud user unless they adjusted their volume, and hence voltage, down.⁵
Anyhow, definitely going to just build this mixer, and probably open-source it when I'm done. If this piques anyone's interest, feel free to ask questions!
Bottom line for someone trying to make a headphone driver: assume headphone impedance between 30Ω and 80Ω, make sure your device won't catch fire if someone plugs in 16Ω headphones, and there is no standard to design to.
¹ I hope reading this post has, at least somewhat, cured you of this unearned feeling.
² I do now have a horrible urge to discover if you can use a ethernet transformer for this, but I do actually want an audio mixer, not a research project. Insofar as this is not already a research project, anyway.
³ These dreadful articles inspired me to write this rant!
⁴ https://source.android.com/docs/core/interaction/accessories/headset/plug-headset-spec
⁵ It is traditional to describe the subjective audio quality in articles about headphone impedance. After adding a series resistance to the earbuds to make the volume comparable, I'd describe the quality of the 1kHz test signal as "physically painful" in both the low- and high-impedance headphones.