eds@solar.UUCP (10/25/83)
AT&T Communications (formerly Long Lines) certainly *does* do statistical multiplexing on voice trunks. The system is called TASI-E (Time Assignment Speech Interpolation), and I worked on its development during 1979-81 at AT&T Bell Laboratories. TASI-E is the third generation of TASIs, and the first with more than two systems installed (I think there are >20 in service now, with 240 trunks on 120 channels each). Currently, TASI-E is installed only on undersea cable circuits, where the cost of installing a new circuit far exceeds the cost of providing another virtual circuit via TASI. Better than a 2:1 compression of talkers to long-distance channels is achieved by exploiting natural gaps in conversation. [While I'm talking, you're listening, and even when I'm talking there are pauses.] Anyway, when a modem call comes into TASI-E, a circuit detects the echo-suppressor disable tone (2000 Hz for >194 ms.), and does two things: it removes the 50 ms. of delay from the modem circuit (delay is there to reduce the front-end clipping of speech bursts while they are being detected and switched), and it "pins" the trunk/channel connection as long as there is energy detected in either direction of transmission. This effectively reduces the number of long-distance channels available for connection to other trunks, so the average/maximum number of data calls present in a pool of 240 random trunks was an important consideration in the engineering of TASI-E. In 1978-79, Long Lines planned to install TASI-E on many domestic routes of longer than 1000 miles, but the plans were eventually dropped for a variety of reasons. I think the first big discussion of TASI was in a 1954 issue of the Bell System Technical Journal, if you're interested further. Ed Schulz, AT&T Information Systems Laboratories, Holmdel, N.J. (201) 834-3838 hou**!houca!solar!eds
preece@uicsl.UUCP (10/28/83)
#R:solar:-28100:uicsl:4100001:000:532 uicsl!preece Oct 27 09:59:00 1983 It seems foolish to mention this, given that digital transmission is not all that far off, but there's no reason they couldn't multiplex the modem channels by providing simulated carriers at the output end during periods when the carrier is unmodulated. The switching would have to pretty clean when a signal does appear, but I would think they could do it, if there were enough modem traffic to make the pinned channels a problem. Personally, I'd rather have true digital transmission... scott preece ihnp4!uiucdcs!uicsl!preece