gould@pilot.njin.net (Brian Jay Gould) (05/10/89)
I have seen numerous comparisons by carriers, of their noise levels (dropping pins and such). The modem manufacturers specify data rates based upon noise free lines. So... has anyone ever attempted to chart data rate versus carrier? (for several modem types) If not, I may have some funding for someone interested in doing some research. -- - Brian Jay Gould :: INTERNET gould@pilot.njin.net BITNET gould@jvncc - - UUCP rutgers!njin!gould Telephone (201) 329-9616 -
dave@rutgers.edu (Dave Levenson) (05/13/89)
In article <telecom-v09i0160m03@vector.dallas.tx.us>, gould@pilot.njin.net (Brian Jay Gould) writes: > I have seen numerous comparisons by carriers, of their noise levels (dropping > pins and such). The modem manufacturers specify data rates based upon > noise free lines. > > So... has anyone ever attempted to chart data rate versus carrier? > (for several modem types) We have used three different LD carriers in recent years. Our experience is that each of these (SBS Skyline, MCI, and AT&T) shows significant variation between different calls to the same place, using the same carrier. The variation in data transmission quality varies even more among calls placed over the same carrier to different destinations. The variation between these carriers was no more than the variation between different calls using the same carrier. We also find plenty of noisey intra-lata calls. If the local carrier cannot deliver a clean dial-up circuit across town, how can they be expected to provide a uniformly clean interface to any of the inter-lata carriers? It doesn't matter much how good the toll carrier is if the local carrier is bad. A given circuit is as good as its weakest link, isn't it? -- Dave Levenson {uunet | rutgers | att}!westmark!dave ...the man in the mooney
INTERMAIL@a.isi.edu (Robert Gutierrez / MCI ID: 367-9829) (05/17/89)
From: westmark!dave@rutgers.edu (Dave Levenson) >In article <telecom-v09i0160m03@vector.dallas.tx.us>, gould@pilot.njin.net >(Brian Jay Gould) writes: > I have seen numerous comparisons by carriers, of their noise levels >(dropping pins and such). The modem manufacturers specify data rates based >upon noise free lines. > So... has anyone ever attempted to chart data rate versus carrier? > (for several modem types) >We have used three different LD carriers in recent years. Our >experience is that each of these (SBS Skyline, MCI, and AT&T) shows >significant variation between different calls to the same place, >using the same carrier. The variation in data transmission quality >varies even more among calls placed over the same carrier to >different destinations. The variation between these carriers was no >more than the variation between different calls using the same >carrier. I could not agree more. Even at MCI, the circut quality could even depend on the time of day you call. Scenaro: You call from San Francisco to New York. If you called at 9pm, your call would travel over our backbone fiber that runs between San Francisco and Washington DC, then via digital radio to New York (over-simplification here), but it is possible that you try at 9AM instead, and all trunks are busy on that fiber route, then we have to find an alternate route for you call to terminate in New York. So, lets see, we send you to Los Angeles first, then Phoenix, Arizona, then Dallas, Texas, then St. Louis, Missouri, then......you get the picture. Then there is unexpected problems that can contribute, like "backhoe fade"....:-) >We also find plenty of noisey intra-lata calls. If the local >carrier cannot deliver a clean dial-up circuit across town, how can >they be expected to provide a uniformly clean interface to any of >the inter-lata carriers? It doesn't matter much how good the toll >carrier is if the local carrier is bad. A given circuit is as good >as its weakest link, isn't it? Ahhh yes, the local carrier lines. I have a problem in that regard. I am on Portal (cup.portal.com) in Cupertino, California. I call from Hayward, California, so, it is intralata. I am calling from a 1AESS to a DMS-100, distance: 25 miles. I would hope to think that there are directs between HYWRCA11CG0 (Hayward C.O.) and SNJSCA12CG0 (San Jose C.O. serving Cupertino). When I call Cupertino at 2400 baud, I always get a bunch of curly characters at the beginning of the transmission, usually going away after repeated <cr>'s (or, during retraining/equalization). This usually points towards frame errors/slips on a T-1 carrier, a common problem on the DMS-100, but I have also gotten this while calling through MCI which we have direct circuts to that C.O., but not from work, where we route through San Francisco instead (I think it's San Francisco 10 or 12 were on). I searched my own calls from work, and they terminate directly into SNJS/12, but I have no way of searching my intralata calls to see if I hit the tandem on their end (Santa Clara) or my end (Oakland), and see if one of the tandems is screwing this up. Of course, trying to explain all this to repair (611), and they usually say "huh?". Let us know if you do proceed with the modem experiments, and the results and problems you encounter. Robert Gutierrez MCI Telecommuncations Western Region Trouble Management Center Hayward, California.
gould@pilot.njin.net (Brian Jay Gould) (05/19/89)
I agree that the quality of call to call will vary. Certainly, location choices will have a significant effect upon results. However, I have the feeling that some trends can be realized. We are still discussing whether to fund this research or not. We also need to find someone with the talent to lead the effort at a university if we are to proceed. -- - Brian Jay Gould :: INTERNET gould@pilot.njin.net BITNET gould@jvncc -
jimmy@seas.ucla.edu (Jim Gottlieb) (05/19/89)
In article <telecom-v09i0168m02@vector.dallas.tx.us> INTERMAIL@a.isi.edu (Robert Gutierrez / MCI ID: 367-9829) writes: >From: westmark!dave@rutgers.edu (Dave Levenson) >>In article <telecom-v09i0160m03@vector.dallas.tx.us>, >gould@pilot.njin.net >>(Brian Jay Gould) writes: >>using the same carrier. The variation in data transmission quality >>varies even more among calls placed over the same carrier to >>different destinations. The variation between these carriers was no >>more than the variation between different calls using the same >>carrier. >I could not agree more. Perhaps. But certain carriers are bad all of the time. Certain resellers use compression to squeeze more voice circuits onto a channel (i.e. DS1 or DS3). This tends to send data throughput down the drain. The company I am most familiar with is Telesphere (10555). I find that I consistently get about half the throughput I do with AT&T or Sprint. The volume on voice calls is also a bit low. When I confronted them about this, they said, "We are a voice service. We make no claims about our ability to carry data." With apologies to Mr. Gutierrez, I find that MCI still has the worst voice quality of the major carriers. They are still using way too many analog microwave hops with the stations farther apart than they should be (fog in the central California valleys kills connections from Southern to Northern California for instance). -- Jim Gottlieb E-Mail: <jimmy@denwa.uucp> or <jimmy@pic.ucla.edu> or <attmail!denwa!jimmy> V-Mail: (213) 551-7702 Fax: 478-3060 The-Real-Me: 824-5454
chip@vector.dallas.tx.us (Chip Rosenthal) (05/21/89)
gould@pilot.njin.net (Brian Jay Gould) writes: >X-TELECOM-Digest: volume 9, issue 160, message 3 of 8 >So... has anyone ever attempted to chart data rate versus carrier? The April 25 issue of |Data Communications| has an article "A quality comparison: Which carrier tests best?" They tested AT&T, MCI, and Sprint with 9600bps V.32 modems (specifically Codex 2264's). They summarized the results of the tests saying: Despite Sprint's new fiber network, AT&T was still the winner in terms of error-free data connections over all links. Not surprisingly, AT&T also won hands down in terms of call-setup time. [...] In terms of signal loss, Sprint was the winner, showing an average receive signal level of -10dBm [...] But when comparing link by link, there was no clear winner. Besides, all the carriers fell well within the acceptable limits for voice. In their tests, they compared: call-setup time, receive-signal level, and bit and block error rate. Their overall summary for connections with data errors (that is at least one error occured during the 1.5 minute BERT with a 511 pattern) was: AT&T ================ 15% SPRINT ==================== 19% MCI ============================== 29% |---------|---------|---------|---------| 0% 10% 20% 30% 40% "Percent of total tested connections with at least one block errored in transmission." However, as others have pointed out, the end-to-end connection is only as good as every link in the chain. -- Chip Rosenthal / chip@vector.Dallas.TX.US / Dallas Semiconductor / 214-450-5337
chip@vector.dallas.tx.us (Chip Rosenthal) (05/21/89)
Jim Gottlieb <ucla-an!denwa!jimmy@seas.ucla.edu> writes: >X-TELECOM-Digest: volume 9, issue 169, message 5 of 8 >Certain resellers use compression to squeeze more voice circuits onto a >channel (i.e. DS1 or DS3). This tends to send data throughput down the drain. The only tariffed service using compression is M44, which uses the 32K ADPCM speech compression algorithm. It is so named because the 24 8-bit channel bandwidth of a T1 line is allocated as 44 4-bit voice channels and 2 8-bit bundled signalling channels. The ADPCM algorithm has been engineered specifically to support modems. The Europeans had a G.721 ADPCM algorithm months before the ANSI T1Y1 committee approved theirs. What was the holdup? They wanted to get modem traffic right! Turns out that the original algorithm was found to have problems with V.22bis and some patches were made to account for this. The CCITT has since gone back and updated their standard with this same algorithm. There is no compression done at the DS3 level. In practice, DS3 is used in the public network only to multiplex several DS1's. I don't believe that ADPCM is commonly used by carriers for their normal public lines. Even if it were, most modem traffic would be unaffected. (I'm not sure about PEP or V.32.) There are lots of reasons for modem problems, but I don't think this is one of them. The assertion that compression is a problem is possibly qualified by mentioning "resellers". True, there is stuff out there which can make anybody sound like an old 78-rpm record which went through a car wash. But I would think you would have to go out of your way to get a super-cheap economy service to get this kind of stuff. -- Chip Rosenthal / chip@vector.Dallas.TX.US / Dallas Semiconductor / 214-450-5337
desnoyer@apple.com (Peter Desnoyers) (05/31/89)
In article <telecom-v09i0172m03@vector.dallas.tx.us> chip@vector.dallas.tx.us writes: >The only tariffed service using compression is M44, which uses the 32K >ADPCM speech compression algorithm. > ... >They wanted to get modem traffic right! People who build V.29 modems tell me that they didn't get it right enough. I've seen a paper on V.29 error performance over cascaded ADPCM lines, but I forget where. >-- >Chip Rosenthal / chip@vector.Dallas.TX.US / Dallas Semiconductor / 214-450-5337 Peter Desnoyers