Ken Stox <stox@balr.com> (09/23/90)
Recently, we have been seeing a bit of discussion of the cost (to the
operating company) of a data call versus a voice call. All the
statements I have seen, so far, seem to agree that a data call costs
the operating company the same as a voice call. TTBOMK (To The Best Of
My Knowledge), this is not true for the following reasons:
First of all, I should state this is the case for a DIGITAL
phone system. If everything were still analog, many would be false. In
fact, this is where the problem lies! It seems that everyone is using
the analog case.
1) Although your connection is analog in nature, it will only
be that way until it reaches the C.O.
2) Once digitized at the C.O., the digital data from your
phone call is blocked into packets of data which are routed through
the phone network.
3) Human speech contains a great deal of dead air/silence.
When you are pausing in a word/sentence/etc., you are no longer
sending data. The phone company can now send more packets of data over
that trunk line while you are pausing between word/sentences/etc.
4) Modems don't pause, they will use every available packet
for that data path. In other words, a modem conversation will not
allow any other packets through.
So, we can now understand why the RBOC's get so blustered
about data traffic. The service that they expected you to use 50% of,
you are using 100% of. I am sure we all feel a great deal of pity for
that poor accountant, who, at this very moment is writhing in agony
over uncollected potential revenue. No doubt, in the not so distant
future, the RBOC's will figure out how to bill you on a packet by
packet basis. This may be the beginning of a much more equitable
method of billing ( right, when hell freezes over :-> ) by which the
customer purchases X number of packets at a given routing grade. Well,
someday, maybe ISDN.
Ken Stox internet: stox@balr.com
BALR Corporation uucp: {uunet|att|attmail}!balr!stox
600 Enterprise Drive voice: (708) 575-8200benyukhi@uunet.uu.net (Ed Benyukhis) (09/24/90)
In article <12490@accuvax.nwu.edu>, stox@balr.com (Ken Stox) writes: > 2) Once digitized at the C.O., the digital data from your > phone call is blocked into packets of data which are routed through > the phone network. Digitization of speach does not imply packet switching. > 3) Human speech contains a great deal of dead air/silence. > When you are pausing in a word/sentence/etc., you are no longer > sending data. The phone company can now send more packets of data over > that trunk line while you are pausing between word/sentences/etc. Speech interpolation techniques are not prevelent in the land networks yet. Edward Benyukhis
crocker@uunet.uu.net (Ronald T. Crocker) (09/24/90)
In article <12490@accuvax.nwu.edu> stox@balr.com (Ken Stox) writes: > 3) Human speech contains a great deal of dead air/silence. >When you are pausing in a word/sentence/etc., you are no longer >sending data. The phone company can now send more packets of data over >that trunk line while you are pausing between word/sentences/etc. > 4) Modems don't pause, they will use every available packet >for that data path. In other words, a modem conversation will not >allow any other packets through. From my experience (former Bell Labs), the type of multiplexing that you describe above (item 3) is not typical of any switches (digital or analog) that I am familiar with. Most telephony connections are "circuit-switched", i.e. equivalent to hooking a pair of wires between the two parties. The only "packet-switched" connections that I know of are those for ISDN packet data (B or D channel), and these are handled as "special cases," at least in the 5E. Voice is not packet data. It is not treated in a packet manner. Whatever happens to be on the voice channel is digitized (PCM), transmitted across digital carrier facilities (T1) to another switch, decoded to the equivalent analog signal, and played out of the receiver in the handset. No where in this loop is anything trying to figure out if the digitized voice signal represents "quiet". T1 is simply a multiplexed digital version of 24 analog trunks. Voice-grade lines are 64Kbps, T1 channels are [nominally] 64Kbps. Maybe if there were some compression done the case would be different, but I don't know of any of that either. Ron Crocker Motorola Radio-Telephone Systems Group, Cellular Infrastructure Division (708) 632-4752 [FAX: (708) 632-4430] ...!uunet!motcid!crocker
coleman@cs.ucla.edu (Michael Coleman) (09/25/90)
stox@balr.com (Ken Stox) writes: > 4) Modems don't pause, they will use every available packet >for that data path. In other words, a modem conversation will not >allow any other packets through. This may be kind of a dumb solution, but why can't the phone company detect dead modem carrier and compress it the way they do with silence. I realize that modem silence isn't as simple as people silence, but there must be some way to do this. How about it?
benyukhi@uunet.uu.net (Ed Benyukhis) (09/26/90)
In article <12545@accuvax.nwu.edu>, motcid!crocker@uunet.uu.net (Ronald T. Crocker) writes: > From my experience (former Bell Labs), the type of multiplexing that > you describe above (item 3) is not typical of any switches (digital or > analog) that I am familiar with. Most telephony connections are > "circuit-switched", i.e. equivalent to hooking a pair of wires between > the two parties. The only "packet-switched" connections that I know > of are those for ISDN packet data (B or D channel), and these are > handled as "special cases," at least in the 5E. ^^^^^^^^^^^^^ I agree that most connections are circuit-switched and that most calls are POTS type calls. But what is so special about B/D channel packet switching??? > Voice is not packet data. It is not treated in a packet manner. ^^^^^^^^^^^^^^^^^^^^^^^^ It could be. VSCS (FAA) at Bell Labs is implementing just that i.e. Packatizing voice for air traffic controllers communications. Voice packatezation perhaps warrants some discussion/explanation by someone more familiar with the process. How about it Pat???? > Whatever happens to be on the voice channel is digitized (PCM), > transmitted across digital carrier facilities (T1) to another switch, > decoded to the equivalent analog signal, and played out of the > receiver in the handset. No where in this loop is anything trying to > figure out if the digitized voice signal represents "quiet". T1 is > simply a multiplexed digital version of 24 analog trunks. Voice-grade ^^^^^^^^^^^ > lines are 64Kbps, T1 channels are [nominally] 64Kbps. Maybe if there ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ When you say Voice Grade Lines, are you referring to th subscriber loops?? And if you are, than, how about BRI connection piping 144Kbps over a wire pair. And even this can be increased by playing tricks with the loading coils. Also, depending on the Super Frame format, you might not even get a 64Kbps clear channel on the T1 either. In general, these are long and, at times, complex subject matters. Regards, Ed Benyukhis, Motorola, CID. (708)632-4658
csowden@compulink.co.uk (Chris Sowden) (09/26/90)
There is a CCITT recommendation describing Digital Circuit Multiplication Equipment (DCME) which improves the efficiency of international trunks. DCME works by encoding the speech data from 64kbit/s down to a lower rate such as 32kbit/s (ADPCM) and by only connecting the speech channel to a trunk when it is carrying a burst of speech (Digital Speech Interpolation). DCME, while not quite fitting the packet model, does allow a number of 64kbit/s speech channels to be carried on a smaller number of transmission channels. Voice band data, OTOH, is not suitable for compression by DSI. Also, ADPCM can cause problems with high speed voice band data. Digital data (ISDN) requires a whole channel to itself. DCME can only be cost effective where the circuit costs are high (long trunks). If DCME is used, data calls will require more bandwidth on these trunks than voice calls. Therefore, data calls will cost the operator more to carry. Chris Sowden
tnixon@uunet.uu.net (Toby Nixon) (09/27/90)
> This may be kind of a dumb solution, but why can't the phone company detect > dead modem carrier and compress it the way they do with silence. I realize > that modem silence isn't as simple as people silence, but there must be some > way to do this. How about it? CCITT Study Group XV is currently leading a study (along with Study Groups VIII and XVII) on compression of fax and modem traffic on digital trunks. It would involve demodulation of the signal, transmission of the original bits, and remodulation when the signal reaches the other end. This way, instead of wasting an entire 64kHz DS0, the network can use only 2.4KHz for a V.22bis connection, 9.6KHz for V.32, etc -- and pack them into DS0s to save bandwidth. In the case of fax, the half-duplex nature of the link allows the network to insert this mechanism at any line turnaround. Also, the fax T.30 handshake always indicates what kind of modulation scheme is going to be used next, so the demodulator/remodulators can track the protocol and keep the fax machines happy, without introducing any appreciable delay. With data, it's a bit trickier, because you don't know for sure what the modulation scheme is going to be, and in the case of full-duplex modems there's no natural place (like a line-turnaround) where the network can switch in this mechanism. It's an interesting project. Toby Nixon, Principal Engineer Fax: +1-404-441-1213 Telex: 6502670805 Hayes Microcomputer Products Inc. Voice: +1-404-449-8791 CIS: 70271,404 Norcross, Georgia, USA BBS: +1-404-446-6336 MCI: TNIXON Telemail: T.NIXON/HAYES AT&T: !tnixon UUCP: ...!uunet!hayes!tnixon Internet: hayes!tnixon@uunet.uu.net MHS: C=US / AD=ATTMAIL / PN=TOBY_L_NIXON / DD=TNIXON
laird@slum.mv.com (Laird Heal) (09/27/90)
>Recently, we have been seeing a bit of discussion of the cost (to the >operating company) of a data call versus a voice call. > 1) Although your connection is analog in nature, it will only >be that way until it reaches the C.O. Forgive my ignorance, but it seems to me that with mature technology the voice call should be more costly than a data call. This is because of the simple nature of the modem tones as compared to the less predictable modulations of human voices. Furthermore, correct me if I am mistaken but the back-channel of the current listener still sends any background noise, from breathing to, in the case of U. S. Sprint, pins dropping. The exception listed was for cable transmission where the number of circuits was strictly limited: so that's what the modem 'guard tones' are for? A modem's tones should be much more predictable than a human's voice, and generally any compression algorithm or sampling could clock much lower for data transmission. The most preferable from the Telephone Company's standpoint would be direct digital transmission point-to-point because they could bypass analog devices, and measure data throughput. It does seem another case of paying the utility more for service that costs them less; can someone prove to me that touch-tone service actually does cost the phone company more today than ol' reliable? Laird Heal laird@slum.MV.COM (Salem, NH) +1 603 898 1406
kabra437@pallas.athenanet.com (Ken Abrams) (09/27/90)
In article <12490@accuvax.nwu.edu> stox@balr.com (Ken Stox) writes: > 2) Once digitized at the C.O., the digital data from your >phone call is blocked into packets of data which are routed through >the phone network. This has been explained in some detail in earlier posts but I think a quick recap is in order. While what Mr. Stox says might be true (to some degree) sometime in the not too distant future, it is NOT true today. As far as I know, there are no telco owned switches in service today that use packet switching for voice. At the present time, a path through a digital switch used for voice grade communications is not pre-emptable. Time division multiplexing is NOT the same as packet switching. A voice grade call gets a "full time" channel regardless of what is transported, voice, data or even silence. Ken Abrams uunet!pallas!kabra437 Illinois Bell kabra437@athenanet.com Springfield (voice) 217-753-7965
goldstein@carafe.enet.dec.com (Fred R. Goldstein) (09/28/90)
In article <12623@accuvax.nwu.edu>, motcid!benyukhi@uunet.uu.net (Ed Benyukhis) writes... >> Voice is not packet data. It is not treated in a packet manner. ^^^^^^^^^^^ >It could be. VSCS (FAA) at Bell Labs is implementing just that i.e. >Packatizing voice for air traffic controllers communications. Voice >packatezation perhaps warrants some discussion/explanation by someone >more familiar with the process. How about it Pat???? [Moderator's Note: Fred Goldstein comes to mind. PAT] This discussion sounds like a rerun of one we had last year, around TASI, but I'll jump in anyway. I've even sent a contribution to ANSI T1Y1 for their next meeeting explaining why Digital voted NO on a proposed packet voice standard (syntactic and checksum matters), but that's beyond the scope of this thread. Packet voice does occur on the public switched telephone network, but it's not common. Old-fashioned Time Assignment Speech Interpolation using analog gear has gone the way of the FDM open wire carrier. But newer digital interpolation gear does exist, mostly in private nets and in international calls. It's not worth the effort for domestic calls, since raw transmission is cheap enough and any packetization adds delay, making echo cancellation (not so cheap) necessary. AT&T uses a device of their own manufacture called IACS (Integrated Access & Cross-Connect Switch, if I recall) to compress international telephone calls. (Undersea cables aren't cheap!) An effect of the fax explosion, which they reported at a T1S1 meeting, was that they've had to reduce the number of derived channels from each physical pipe, since there are no gaps in fax modem tones. So yes, modems do add a little to the cost of calls, but only overseas. IACS uses a technique called Embedded Adaptive Delta Pulse Code Modulation. This is like ADPCM except that the low-order three bits of a five-bit sample are not used for predicting the next sample. So if the network is particularly busy, it can throw away the lowest-order bit or two from each speech sample, by truncating the last 32 bytes in a frame which carries speech samples arranged by bit significance. Thus the tail end of the frame is all low-order bits. On average you may get 30 kbps or so, but during the busy hour it may drop and still sound okay (just not quite as good) and at real off-hours you may get over 32 kbps and sound better than normal. BTW, the ISDN service definitions for "telephony" and "3.1 kHz audio" differ in that the former permits the use of speech processing, TASI, etc., while the latter requires that the network listen for the 2100 Hz disable tone that modem calls begin with. ISDN interworks with the analog network using the 3.1 kHz audio service. But again, for the bulk of domestic toll and essentially all intra-LATA and local calling, you get raw circuit mode and the network doesn't care one whit about whether you have a modem or microphone. Fred R. Goldstein goldstein@carafe.enet.dec.com or goldstein@delni.enet.dec.com voice: +1 508 486 7388 opinions are mine alone; sharing requires permission
grayt@uunet.uu.net (Tom Gray) (09/28/90)
In article <12664@accuvax.nwu.edu> Ken Abrams <pallas!kabra437@uunet. uu.net> writes: >X-Telecom-Digest: Volume 10, Issue 684, Message 3 of 11 >In article <12490@accuvax.nwu.edu> stox@balr.com (Ken Stox) writes: >> 2) Once digitized at the C.O., the digital data from your >>phone call is blocked into packets of data which are routed through >>the phone network. >quick recap is in order. While what Mr. Stox says might be true (to >some degree) sometime in the not too distant future, it is NOT true >today. As far as I know, there are no telco owned switches in service >today that use packet switching for voice. At the present time, a The ATT IACS (Integrated Access and Control) system uses compression to carry voice data. It is a fast packet system using 384kbit chunks of T1 channels. A whole industry of networking companies is selling compression equipment for private networks - ATT sells its IACS to industry to lower their networking costs.
mcmahan@ames.arc.nasa.gov (Dave Mc Mahan) (09/30/90)
In a previous article, Ken Abrams <pallas!kabra437@uunet.uu.net> writes: >In article <12490@accuvax.nwu.edu> stox@balr.com (Ken Stox) writes: >> 2) Once digitized at the C.O., the digital data from your >>phone call is blocked into packets of data which are routed through >>the phone network. >This has been explained in some detail in earlier posts but I think a >quick recap is in order. While what Mr. Stox says might be true (to >some degree) sometime in the not too distant future, it is NOT true >today. As far as I know, there are no telco owned switches in service >today that use packet switching for voice. At the present time, a >path through a digital switch used for voice grade communications is >not pre-emptable. Time division multiplexing is NOT the same as >packet switching. A voice grade call gets a "full time" channel >regardless of what is transported, voice, data or even silence. There is a company in Campbell, CA called StrataCom that I believe makes a packet switched voice network that is meant to run over a T1 line. I think they have had a product available for several years, but I'm not sure if any telco offices own it. They use all the standard tricks of not sending silence and re-generating a small amount of white noise for a listener so that he is pschycologically fooled into believing that he has a 100% connection. I think they are also offering products that let users insert data channels as well as voice into the T1 backbone. My phone book lists them as : Stratacom 3175 S. Winchester Bl. Campbell, CA 95130 (I think this is the right zip code, but am not sure) (408) 370-2333 dave
gd@dciem.uucp (Gord Deinstadt) (09/30/90)
hayes!tnixon@uunet.uu.net (Toby Nixon) writes: >CCITT Study Group XV is currently leading a study (along with Study >Groups VIII and XVII) on compression of fax and modem traffic on >digital trunks. It would involve demodulation of the signal, >transmission of the original bits, and remodulation when the signal >reaches the other end. This way, instead of wasting an entire 64kHz >DS0, the network can use only 2.4KHz for a V.22bis connection, 9.6KHz >for V.32, etc -- and pack them into DS0s to save bandwidth. We already *have* a system that does this; in Canada it's called Datapac. But instead of developing it into something worthwhile it's been allowed to rot. I wish someone would explain why they haven't added public-dial fax ports to Datapac. Or revised the tariffs so it was actually competitive with high-speed modems at LD rates. Gord Deinstadt gdeinstadt@geovision.UUCP
U5434122@uunet.uu.net (10/05/90)
In article <12893@accuvax.nwu.edu>, cognos!geovision!gd@dciem.uucp (Gord Deinstadt) writes: > hayes!tnixon@uunet.uu.net (Toby Nixon) writes: >>CCITT Study Group XV is currently leading a study (along with Study >>Groups VIII and XVII) on compression of fax and modem traffic on >>digital trunks. It would involve demodulation of the signal, >>transmission of the original bits, and remodulation when the signal >>reaches the other end. This way, instead of wasting an entire 64kHz >>DS0, the network can use only 2.4KHz for a V.22bis connection, 9.6KHz >>for V.32, etc -- and pack them into DS0s to save bandwidth. > We already *have* a system that does this; in Canada it's called > Datapac. But instead of developing it into something w Australia has a Faxstream service which supposedly demodulates the fax message, packetizes it, sends it through the network digitally and delivers it to the recipient when the recipient is not busy. Delayed delivery and broadcast are also available, but I don't know any details (Time to ring yet another 008 rep :-) ) I think you just have to subscribe your fax's phone line and have FaxStream intercept your outgoing calls on request. It is supposed to be cheaper than direct dial LD, but I don't know anyone who uses it. Danny