What am I hearing?
The call starts normally: dial tone, a number, ringing. The ringing comes from a distant office, over a long-distance trunk. On an R1 trunk an idle circuit carries 2600 Hz in both directions. A seized circuit carries no tone forward, and until the call is answered the far end keeps its tone on in the backward direction (ITU-T Q.311, Table 1). The near office filters that backward tone out of your earpiece (Q.313 §2.3.4).
So when 2600 Hz arrives from the caller's end for at least 300 ms, the far office concludes the caller has hung up (Q.313 §2.3.3). It drops the call, the ringing stops, and its end of the trunk goes back to idle. That release is the "kerchunk".
When the tone stops, the far office sees a new call arriving on an idle trunk. It answers with a "wink": it drops its own tone for at least 140 ms while it attaches a register, then puts it back (Q.311, Table 1). The register now waits for digits in multifrequency (MF) code: two of six tones between 700 and 1700 Hz per digit. KP (start of pulsing) opens the sequence and ST (end of pulsing) closes it. Each digit lasts 68 ms, with 68 ms gaps (Q.320, Q.322). The far office then routes the call as if another office had asked, and the network answers with ringing, a busy signal, or an announcement's tones.
The international system of the CCITT (now ITU-T), System No. 5, did the same with two frequencies, 2400 and 2600 Hz. It uses a "compelled" handshake, where each signal is held until the other end acknowledges it (Q.141). The MF digits follow at 55 ms each (Q.151, Q.153).
Who used this, and how did it end?
The Bell System published its signaling frequencies in the open technical literature, including the MF code in the Bell System Technical Journal of November 1960 (C. Breen and C. A. Dahlbom, "Signaling Systems for Control of Telephone Switching"). A toy whistle given away with Cap'n Crunch cereal could produce 2600 Hz, and John Draper became known as "Captain Crunch" for it. The whistle could only send 2600 Hz, though. To send the MF digits as well you needed an electronic tone generator, called a blue box, and these were already circulating in the 1960s. Ron Rosenbaum's 1971 Esquire article "Secrets of the Little Blue Box" made them famous, and Steve Wozniak and Steve Jobs built and sold some before they founded Apple.
The fix was to take signaling out of the voice channel. AT&T began introducing Common Channel Interoffice Signaling (CCIS) in 1976, which carried call setup on a separate data link. Its successor, Signalling System No. 7 (SS7), took over most of the world's networks over the following decades.
What did a payphone send?
Payphones on the Automated Coin Toll Service (ACTS) told the toll office what you had paid with short bursts of 1700 + 2200 Hz, a different pattern for each coin. These are historical tones, and the timings I use are the widely published ones: I have not seen the Bell System Practice, so they are marked approximate. Listen only; there is no download for these.
Sources and accuracy
Frequencies, levels, signal durations and the order of the line signals follow the ITU-T texts for System R1 (Q.310–Q.332) and System No. 5 (Q.140–Q.164), and the address sequence follows Q.107. The exact recognition and response delays inside the allowed ranges are my choices, as are the "kerchunk" click and how much leaked through to the caller. They are listed in the open questions. A separate test suite, written from the ITU texts alone, decodes the digits and the line signals from the audio.
This page makes sound files for listening and teaching. It does not place calls, and these tones have no effect on any working phone network.
Related: dial, ringback and busy tones by country · special information tones · a short history of the modem.