Dial-up sound FAQ
Short answers. The explainer has the long ones, with players.
Why did dial-up modems make noise at all?
A phone line only carries sound, so modems had to talk to each other in sound. Before sending data, the two modems negotiated over the line: which standard to use, how fast, and how the line distorts each frequency. That negotiation is the handshake you hear. The tones are the actual data signals, not a sound effect.
Why could I hear it? Wasn't it meant for machines?
Modems had a small speaker so you could tell whether you reached a modem, a person, a busy line or a fax machine. The Hayes command ATM1, the usual default, kept the speaker on until the modems detected each other's carrier and then muted it. That is why the familiar clip stops just as the connection starts.
What is the first high tone after the ringing?
That is the answer tone: 2100 Hz from the answering modem. In V.8 modems it is ANSam, whose loudness sways ±20 % at 15 Hz and whose phase flips every 450 ms. The flips switch off echo cancellers in the telephone network (ITU-T V.25 and V.8 §7.2).
What is the 'chord' in the middle of a 33.6k or 56k handshake?
Line probing. Each modem sends 21 tones at once, every 150 Hz from 150 to 3750 Hz with four left out, so the other end can measure how the line attenuates and delays each frequency (ITU-T V.34 §10.1.2.4). The result decides the symbol rate and carrier for the call.
What is the long hiss?
Training. Scrambled data sounds like noise by design. The receiver uses it to train an adaptive equalizer, and the transmitter uses it to train its echo canceller. In V.34 it is signal TRN (§10.1.3.8).
What are the 'bongs' in a 56k connection?
Digital impairment learning (DIL) from ITU-T V.90 §8.4.1. The server modem plays a ladder of exact PCM codewords so the home modem can learn how the digital phone network alters each one. It only happens in V.90 and V.92 calls, which is why 33.6k calls sound different.
How accurate is this?
The handshake signals follow the ITU-T recommendations: frequencies, timings and bit layouts, with each segment citing its clause. An independent test suite decodes our audio (V.8 menus, INFO sequences with CRC, DTMF, line-probe spectrum, G.711 tables) and the 300 baud Bell 103 data decodes with minimodem. The data that follows CONNECT is an approximation, and so is V.92's phase 3. The full list is in our open questions.
Can I use the audio?
Yes. Everything is synthesized, not recorded, and you can download any handshake as a WAV file and use it as you like. A credit link is appreciated but not required.
Why does my country's dial tone sound different?
Every network chose its own call-progress tones. North America uses 350 + 440 Hz for dial tone; Germany and much of Europe use a single 425 Hz. We take each country's values from the ITU-T E.180 list of national tones.
Does this connect to anything?
No. The handshake is computed in JavaScript in your browser and played back. No audio, settings, or files are sent anywhere, and we never ask for the microphone. The site counts visits with Google Analytics 4 (see privacy).