nobody publishes where a phone speaker stops

i wanted one number. the frequency a phone speaker stops at, so i could stop guessing and mix against it.

it doesn’t exist.

no standards body, no cross-brand dataset, nothing i could reach. and the measurements that do exist can’t be added together: one review quotes the point where output is 3 dB down, the next 6 dB, the next prints a range with no tolerance at all. three numbers, three different questions. pooling them into “phones roll off at X” is making something up with a citation stapled to it.

so the honest version is that nobody knows where your listener’s speaker gives up, including your listener.

the one measurement i did find

a 2015 13-inch macbook pro, asked for 125 hz, returns 75.5% total harmonic distortion. a yamaha monitor in the same table gives 1.4%. good headphones, 0.3%.

that number needs its condition attached or it becomes a lie. the researchers say plainly that the sound pressure over that range was very low, so it’s a big percentage of a small thing. the laptop is barely playing down there, and most of the little it manages is harmonics rather than the note you asked for.

your brain builds the note anyway

this part has been solid for sixty years. play the harmonics of a note without the note, and people hear the note. 300, 400 and 500 hz gets heard as 100 hz, the number they all divide by.

which is where the advice comes from. saturate the bass, make harmonics the small speaker can actually produce, and the listener assembles the low end you couldn’t send.

and here is what i could not find

the listening tests are real. fifteen listeners, twenty-four double-blind paired comparisons, every enhanced version ranked above the unprocessed one.

but read what they measured. bass quality preference. that’s the endpoint. none of those studies measured whether anyone heard a deeper pitch, or identified the missing fundamental, or found it louder. the mechanism is well established, the tests never tested it, and the two have been quietly glued together for twenty years.

then the trade. one study ran the same comparison twice, once for bass impression and once for audio quality. sixteen subjects. the processing won on impression and lost on quality, against a hidden reference. so it works, and it costs you.

what i would actually do

saturate the bass. it’s right. just do it knowing you are buying an impression and paying in cleanliness, not restoring anything.

keep the band you feed it narrow. past about three octaves the intermodulation becomes clearly audible, which is the authors’ wording, not mine. theirs fed 20 to 70 hz in and kept 70 to 140 out. that’s their setup, not a law.

then check it on something small. jbl, sony and bose all ship bass processing in firmware and sony’s is on out of the box, so the thing you send isn’t the thing that plays.

it keeps happening

the other film does it from the other end. the tuning arithmetic is closed and has been for centuries: stack twelve pure fifths and you land 23.5 cents above where seven octaves put you. you can’t remove that gap, only move it, and every tuning system is an opinion about where.

then somewhere in the middle of checking, the “vallotti” preset in almost every synth turned out to be named after the wrong person. vallotti’s own printed text specifies a fifth of 27:40. the preset uses a different comma, which barbour attributes to thomas young in 1800. not everyone agrees, di veroli prints the preset’s version and calls it vallotti’s.

two films, same shape. go and check a number, and it’s either not published or it’s filed under someone else’s name.

the spiral that will not close is up now, and the script that computes every figure in it is free here. the phone one is up today.