In-channel multi-frequency signaling design

In-channel multi-frequency (MF) signaling sends address digits as combinations of voice-band tones over the same channel that later carries speech. Its design must make the signals fast and reliable while never being confused with speech.

Design considerations

1. Choice of frequencies: tones must lie inside the 300–3400 Hz voice band but be spread so filters can separate them. Example: CCITT R1/No. 5 uses 700, 900, 1100, 1300, 1500, 1700 Hz (200 Hz spacing); R2 uses 1380–1980 Hz forward and 1140–540 Hz backward (120 Hz spacing).
2. Code structure (2-out-of-N): each digit is two tones out of six, giving $\binom{6}{2} = 15$ codes. Exactly two tones must be present, so any missing or extra tone is detected (self-checking).
3. Protection against speech imitation (talk-off): speech rarely contains two pure tones of exactly the right level and duration, so detectors require precise frequency, level and minimum duration; MF receivers are usually connected only during set-up.
4. Signal level and twist: tones must be loud enough above noise but below overload; the level difference between the two tones (twist) must be limited.
5. Frequency tolerance: allow for carrier-system frequency shift (a few Hz) while still rejecting adjacent tones.
6. Timing: minimum tone duration and inter-digit pause (e.g. about 68 ms tone in R1) set the signaling speed.
7. Compelled or non-compelled operation: compelled (R2) — each forward signal stays on until a backward acknowledgement arrives, giving reliability on unknown links; non-compelled (R1) is faster but needs good links.
8. Start/stop signals: KP (start-of-pulsing) and ST (end-of-pulsing) codes mark the digit string.
9. Separation from line signaling: supervisory signals (seize, answer, clear) are handled by a different method (e.g. 2600 Hz or out-band), so the MF set is reserved for addresses.