dg@lakart.UUCP (David Goodenough) (07/26/89)
As is well recorded, when a pair of TB modems are PEP'ing at each other, only one end gets to talk at a time. Now, the modulation used by the TB uses many different tones, and presumably the demodulation does some real clever trickery with DSP to achieve the equivalent of a bandpass filter with a *REAL* high Q factor for each tone. Now, why is it not possible for one end to use some of the tones, and the other end to use some others, and achieve full duplex communicaions, at the cost of reduced transfer rates (i.e. 18000 in one direction, or 10000 in one direction, and 8000 in the other). Since the system is able to dynamically select a subset of the tones anyway, it would seem that most of the ability is in place. However, it can't be done, and I'll bet there is a good reason. What is that reason? -- dg@lakart.UUCP - David Goodenough +---+ IHS | +-+-+ ....... !harvard!xait!lakart!dg +-+-+ | AKA: dg%lakart.uucp@xait.xerox.com +---+
grr@cbmvax.UUCP (George Robbins) (07/28/89)
In article <630@lakart.UUCP> dg@lakart.UUCP (David Goodenough) writes: > As is well recorded, when a pair of TB modems are PEP'ing at each other, only > one end gets to talk at a time. Now, the modulation used by the TB uses many > different tones and presumably the demodulation does some real clever trickery > with DSP to achieve the equivalent of a bandpass filter with a *REAL* high > Q factor for each tone. Now, why is it not possible for one end to use some > of the tones, and the other end to use some others, and achieve full duplex > communicaions, at the cost of reduced transfer rates (i.e. 18000 in one > direction, or 10000 in one direction, and 8000 in the other). > > Since the system is able to dynamically select a subset of the tones anyway, > it would seem that most of the ability is in place. However, it can't be > done, and I'll bet there is a good reason. What is that reason? The technical issues are that running in a full-duplex modes requiers better filter/line hybrids and/or a guard band between frequencies to minimize the effect of the relativly strong transmitted back channel on the adjacent receive channels. Now it may be that Telebit felt that this hurt worse then the costs associated with running the packets in half-duplex style. You also have to decide whether to implement a fixed B/W back channel supported by conventional modem techniques or to do dynamic channel splitting based on thruput requirements. Trying to timeshare the DSP resource to handle simultaneous transmit/receive processsing could be a major effort. On the other hand, perhaps they simply weren't interested in this area and thought they had the reverse thruput issues covered adequately... -- George Robbins - now working for, uucp: {uunet|pyramid|rutgers}!cbmvax!grr but no way officially representing arpa: cbmvax!grr@uunet.uu.net Commodore, Engineering Department fone: 215-431-9255 (only by moonlite)
jon@shaffer.UUCP (Jon Doran) (07/29/89)
Telebit uses half duplex connections to avoid complicated echo cancellation. At each point where a modem is connected to a phone line, there is a small amount of echo. Thus a typical connection has two sources of echo. Now echo cancellation on a V32 connection is not THAT hard. V32 uses two frequencies near the center of the bandwidth, and the echo tends to shift in the band. But Telebit's PEP (not their V32!) does not use single character modulation, they use 512 frequencies distributed over the band. So the echo can appear to be valid traffic even after the frequency shift. The only easy solution at this point is to use half duplex connections, the echo is ignored completely. This does not cut into your bandwith; V32 used two frequencies for full duplex so Telebit would either need 1024 frequencies or would need to halve their transmission rate. BTW, a little T2500 trivia: Did you know that the T2500 is a 7.3 baud modem?? Jon Doran IBM AWD, Austin Tx
ch@maths.tcd.ie (Charles Bryant) (08/03/89)
In article <2486@shaffer.UUCP> jon@shaffer.UUCP (Jon Doran) writes: >Now echo cancellation on a V32 connection is not THAT hard. V32 uses two >frequencies near the center of the bandwidth, and the echo tends to shift in Nope. V.32 uses only 1800Hz in both directions. Each end subtracts out its outgoing signal from what it receives before examining the signal. This can mean that if one end hangs up, the other end does not notice. -- Charles Bryant. Working at Datacode Electronics Ltd. (Modem manufacturers)