time-assignment speech interpolation

Time-assignment speech interpolation, or TASI, is a transmission technique that detects voice activity in real time and uses the silent gaps between words and pauses to carry speech from other channels on the same circuit. On a single analog telephone trunk, TASI can nearly double capacity by recognizing that human speech is discontinuous; a speaker occupies the line for only about 40 percent of the time during a conversation. Rather than allocate fixed time slots, TASI assigns transmission time dynamically to whoever is actually talking.

The system works through a bank of voice detectors that monitor incoming channels continuously. When a detector identifies speech above a threshold level, it triggers an assignment gate that routes that speaker's signal to the next available output slot on the shared trunk. The receiving end maintains a parallel detector and gate system to route incoming speech back to the correct channel. This handoff must occur in milliseconds; delays longer than about 50 milliseconds become audible to users as echoes or strange timing artifacts.

Implementation and limitations

TASI was most widely deployed on transatlantic and transpacific submarine cables and long terrestrial microwave links in the 1960s and 1970s, where transmission capacity was extremely expensive and scarce. A four-channel voice trunk could typically support five or six simultaneous conversations. However, the technique carries a real penalty: occasionally, during periods of high activity, all available slots fill and new speech is blocked or clipped. This blockage rate had to be engineered into trunk design; a well-engineered TASI trunk aimed for less than one percent of speech lost to blocking during peak hours.

The system also introduced audible artifacts. Gaps between words could be compressed or expanded slightly as the assignment gates switched, and the noise floor during silent periods sometimes changed noticeably when a speaker regained an open slot. Users also noticed that interruptions became more frequent: a speaker might be cut off briefly if their assigned slot was reassigned during a pause they expected to fill.

TASI became obsolete with the rise of digital switching and fiber optic transmission in the 1980s. Modern voice compression and packet-based protocols make continuous bandwidth assignment unnecessary. However, the fundamental principle survives in modern voice coding standards that exploit speech discontinuity; variable-rate codecs use the same underlying observation that silence is cheaper than speech.