jackson@ttidcb.UUCP (Dick Jackson) (05/08/85)
We are thinking about sending voice through a packet data network (our own). I would much appreciate references to prior work in this area - I know that (at least) NAC,as it then was ,id a study for the Government a while back, but I don't have chapter and verse. Thanks in advance, Dick Jackson Citicorp - TTI
taylor@ecsvax.UUCP (Steven Taylor) (05/09/85)
There is a short article about packetized voice in the February '85 issue of Data Communications (p.45). Technically, there is really no good reason the voice can't be packetized, the primary questions are simply questions of bandwidth and economics. The typical data rates for 'digitized voice', which would have to be used, are on the order of 16 to 32 kbps. Most packet switches available today have maximum data rates available for inter-nodal transfer at 56 kbps. Since packet switching is based on the probability that a large number of users can simultaneously share a circuit if there is a large enough group and the composite rates are sufficiently high, it is very probable that the data rate of the packet switch will have to be on the order of 1.5 mbps (T-1) for packetized voice to work well. Otherwise, the inherent delay caused by the sharing of the facilities will be intolerable. Evidently, there is also some degree of reluctance to push this development as fast as possible because of the questionable economics of building such a fast voice switch when enormous bandwidth is available from sources such as fiber optic cable.
jcp@brl-tgr.ARPA (Joe Pistritto <jcp>) (05/10/85)
The neatest packetized voice experiment I know of is the one that BBN (Bolt Berenek & Newman, the people who brought you IMPS), is doing under DARPA contract. Basically, they have a multi-processor 68000 machine, called a Voice Funnel, a Butterfly Gateway, or a Butterfly Supercomputer depending on which contract # you ask about, which takes voice and spills it onto a 1.544Mbps Wideband Satnet channel. I don't think they've tested it via the Satellite yet, but they have tested it over ground microwave links. Really slick. They are also developing an interface to subscriber telephone lines for the thing. The person to contact is Bob Hinden of BBN if you want more info. -JCP-
jbn@wdl1.UUCP (05/15/85)
Packet voice seems to be one of those futuristic ideas, like monorails, whose time has passed. Xerox PARC played around with it a few years ago, building a rather expensive unit that allowed voice transmission over Ethernet. But it was never made into a product, and wasn't that useful even at PARC. Digital voice I/O gear is now easy to come by; the Texas Instruments PC is available with voice I/O; they provide an answering machine program as a demo. Unfortunately, the TI machine lacks DMA, so you can't run most Ethernet cards. BBN has been playing around with packet voice for years now, with no really interesting results; you can send voice over the ARPANET but the net has to be gimmicked to give voice data extra priority for it to work. The only real justification for packet voice seems to be for high-reliability jam-resistant military systems. And even there there are other approaches. John Nagle
dgary@ecsvax.UUCP (D Gary Grady) (05/17/85)
> Packet voice seems to be one of those futuristic ideas, like monorails, > whose time has passed. > ... > The only > real justification for packet voice seems to be for high-reliability > jam-resistant military systems. > > John Nagle I was under the impression that packet voice was used in cellular telephone systems that are going up in some major cities. My understanding (which may well be way wrong) is that small, low power radio exchanges serve little "cells" of a city, with voice being packetized over the air to allow relatively narrow use of the frequency spectrum and time-base multiplexing. Does anybody have any information on cellular radio phone systems that would confirm or contradict this? -- D Gary Grady Duke U Comp Center, Durham, NC 27706 (919) 684-3695 USENET: {seismo,decvax,ihnp4,akgua,etc.}!mcnc!ecsvax!dgary
lauren@vortex.UUCP (Lauren Weinstein) (05/18/85)
Cellular radio is not packetized voice. It's just plain, old 800 Mhz transmissions, which you can receive on an old UHF TV if you feel like wasting your time that way. The "cells" simply handle particular regions and switch the voice connection from cell to cell as necessary for a moving vehicle. In the future, there may be digitized/scrambled voice channels for cellular radio--but that hasn't happened yet. --Lauren--
jst@wucs.UUCP (Jon Turner) (05/20/85)
John Nagle writes > Packet voice seems to be one of those futuristic ideas, like monorails, > whose time has passed. > ... > The only > real justification for packet voice seems to be for high-reliability > jam-resistant military systems. > Not true. Packet voice offers two main advantages over circuit-switched voice. First, it requires less than 40% of the bandwidth, since there is no need to tie up the channel during silent periods. Second, it integrates voice and data in a single communication network. The primary beneficiary of this integration is data, since the size of telephone networks offers economies of scale that aren't available in pure data packet switches. The expectation is that an integrated packet network can bring the cost of high speed data communication down to an affordable level for residential customers. When I was at Bell Labs, I worked on a project (called the Fast Packet Network or FPN) whose purpose was to demonstrate the feasibility of packet voice on a large scale (ie. replace the current circuit switched voice network with a packet network). Many of the ideas behind this project were not new. BBN has been in this game for a long time as John pointed out. Anyway, the project was successful in demonstrating that packet voice is not only feasible, but cost-competitive with circuit-switching (something that almost no one in the telephone world would have believed at the time). Since leaving the Labs about two years ago, I've lost touch with the project and have no specific knowledge of their plans, but they are continuing to work on it, and rumor has it that they are running a field trial in California. On the whole, they've been fairly quiet about it, but occasionally some tidbits leak out. Some of the basic FPN patents have been issued recently, which has taken the lid of the publication restraints that Bell had imposed previously. I've recently written a paper describing the basic technology and giving the arguments in favor of an integrated packet network as an alternative to the much heralded, Integrated Services Digital Network. I'll be happy to send copies to anyone interested. GTE has been doing similar work using a different switching technology, which they call burst-switching. This has been described in a couple of articles in the IEEE Communications Society magazine. Burst-switching is essentially a fast circuit switching scheme in which a circuit is established and taken down for each burst of information (voice, data or whatever). It's more geared towards voice than the packet switched approach and is in my opinion less flexible, but the objectives of this work and the FPN work are clearly similar. In summary, I think it's premature to write off packet voice as idea whose time is past. Jon Turner Washington University in St. Louis 314-889-6193 UUCP: jst@wucs.UUCP or ..!{ihnp4,seismo}!wucs!jst ARPANET: wucs!jst@seismo.ARPA CSNET: wucs!jst@seismo.ARPA%csnet-relay -- Jon Turner Washington University in St. Louis 314-889-6193 UUCP: jst@wucs.UUCP or ..!{ihnp4,seismo}!wucs!jst ARPANET: wucs!jst@seismo.ARPA CSNET: wucs!jst@seismo.ARPA%csnet-relay
ron@brl-tgr.ARPA (Ron Natalie <ron>) (05/20/85)
> I was under the impression that packet voice was used in cellular > telephone systems that are going up in some major cities. My > understanding (which may well be way wrong) is that small, low power > radio exchanges serve little "cells" of a city, with voice being > packetized over the air to allow relatively narrow use of the frequency > spectrum and time-base multiplexing. Does anybody have any information > on cellular radio phone systems that would confirm or contradict this? True, except for packeting. Just uses conventional audio. The neat part is the electronics for selecting and switching off between the 666 channels and from cell to cell. You can listen to Cellular telephone calls with a conventional FM receiver tuned the appropriate frequency (or as some one else on the net pointed out, you can use certain video tape recorders). -Ron
jbn@wdl1.UUCP (05/24/85)
> Not true. Packet voice offers two main advantages over circuit-switched voice. > First, it requires less than 40% of the bandwidth, since there is no need > to tie up the channel during silent periods. Second, it integrates voice and > data in a single communication network. The TASI subchannel-assignment system used on transoceanic cables for decades avoids tying up the channel during silent periods, and achieves roughly similar bandwidth economy. And most newer switching systems, especially PBXes, are digital circuit switches and handle both voice and data in a similar way. You don't need packet switching to achieve these gains. It's interesting to hear that Bell Labs tried to build a big packet network. I would appreciate references, if available. To date, all packet networks have been tiny by telephony standards, and no one has convincingly demonstrated that packet switching technology can be scaled up to networks with millions of subscribers. Most of the X.25-based systems are circuit switches underneath. John Nagle
jst@wucs.UUCP (Jon Turner) (05/27/85)
In article <438@wdl1.UUCP> jbn@wdl1.UUCP writes: > > The TASI subchannel-assignment system used on transoceanic cables >for decades avoids tying up the channel during silent periods, and achieves >roughly similar bandwidth economy. And most newer switching systems, >especially PBXes, are digital circuit switches and handle both voice and data >in a similar way. You don't need packet switching to achieve these gains. > TASI can give you the same bandwidth gains as packet switching, but it's purely a transmission system. As soon as you reach a switching system the signals are separated back out to the full 64 Kbs. Consequently the compression must be repeated for each hop. There are switching techniques that can carry the compressed signals through switching offices. None of course, have been implemented on a large scale. It is also true that you can get integrated voice and data communication in digitial telephone switches, but only if you're happy with a 64 Kbs circuit switched channel. The burstiness of most data applications makes this a less than ideal option (yes I know, it sure beats 1200 baud modems). The kind of system I referred to in my earlier note can provide channels in any size from 100 bits per second to over a megabit per second, and can handle bursty signals, charging the user only for what is used, not for the silent periods. > It's interesting to hear that Bell Labs tried to build a big packet >network. I would appreciate references, if available. To date, all packet >networks have been tiny by telephony standards, and no one has convincingly >demonstrated that packet switching technology can be scaled up to networks >with millions of subscribers. I never said that Bell Labs tried to build a big packet network, only that they have done (and still are doing) research trying to establish that it is feasible. Not much has been published to date. You can find two papers in the Proceedings of Globecom 83. One is authored by Len Wyatt and myself, the other by Bill Hoberecht. There is a paper by John Kulzer and Warren Montgomery in the Proceedings of the International Switching Symposium (84) and there are two papers in the December 83 issue of the IEEE Journal on Selected Areas in Communications, one by Warren Montgomery and the other Y. C. Jenq. There are also several patents which issued recently. I don't have the patent numbers here (I'm typing this from home), but will be happy to post them if there is enough interest. >Most of the X.25-based systems are circuit switches underneath. Huh? The ones I'm familiar with certainly are not circuit switches underneath. What do you mean by circuit switching? -- Jon Turner Washington University in St. Louis 314-889-6193 UUCP: jst@wucs.UUCP or ..!{ihnp4,seismo}!wucs!jst ARPANET: wucs!jst@seismo.ARPA CSNET: wucs!jst@seismo.ARPA%csnet-relay