IEEE ComSocSCV Sept 8 meeting: High Speed Transmission on Twisted Pair in LANs and DSL
Our Sept 9th meeting features talks on high speed transmission on both LAN/data centers and DSL access networks. Details at
Several of your ComSocSCV officers spent yesterday afternoon and early evening at ATT Labs in San Ramon. We were surprised to learn of ATTs extensive use of VDSL2 in FTTN deployments of U-Verse (their triple play bundled servie that includes TV/VoD via IPTV, HIgh speed Internet access, and Voice (either POTS or VoIP). They are also using VDSL to reach subscribers in Multi Dwelling Units (MDUs). Separate from U-Verse, ADSL2 is being used for single point Internet access.
Our other talk will be the status of 10G Base T for LANs and data centers. It’s amazing that in 1993, High Speed Internet was only 100M bits/sec. Now 1G BaseT Ethernet is widely deployed and 10G Base T is coming along fast (the standard has been completed)
Here is a Brief History of Twisted pair based Ehternet and xDSL, based on prsonal observations in the 1980s and 1990s:
Twisted Pair based LANs
In the mid 1980s, ATT had a 1M b/s twisted pair transmission system named “STAR-LAN.” It never went anywhere as a cheaper version of coax based Ethernet (10Base2) was more popular. Then in the late 1980s Manchester coded 10BaseT became very popular.
Notes on nomenclature:
“BASE” is short for baseband, meaning that there is no frequency-division multiplexing (FDM) or other frequency shifting modulation in use; each signal has full control of wire, on a single frequency.
“T” designates twisted pair cable, where the pair of wires for each signal is twisted together to reduce radio frequency interference and crosstalk between pairs (FEXT and NEXT).
“UTP” = Unshielded twisted pair, as in UTP-3 (voice grade) and UTP-5 data grade) twisted pair.
“PMD” is the lowest sublayer in the IEEE 802.3 PHY layer. It stands for Physical Medium Dependent. Any coax, twisted pair, or fiber optic transmission system is the essence of the PMD sublayer.
Continuing the story……………….
Sometime in 1992, ANSI X3T9.5 began developing a standard for 100M b/sec FDDI on Twisted Pair. It was called “TP-PMD.” Discussion Group member (and former IEEE 802.3 Chair/ Vice-Chair) Geoff Thompson and I participated in that committee. It was chaired by a guy from DEC. In 1993 there was a performance test between the two competing twisted pair transmission technologies that were candidates for the TP-PMD standard. It was conducted at an independent test lab in New Hampshire. Crescendo’s technology (based on a 3 state Pulse Amplitude Modulation code called “MLT-3”) beat out National Semiconductor’s and was chosen as the TP-PMD standard.
Also during 1993, there was a “Fast Ethernet” standards war, with HP’s 100 VG AnyLAN (new MAC and PHY) battling 100BaseT (where the Ethernet MAC was not changed). 100BaseT had one version for UTP-3 and another for UTP-5. It was that latter version, known as 100 Base TX that dominated the market. Grand Junction seemed to be the ring leader of that camp, although Intel was a staunch supporter. 100Base TX used the PMD from TP-PMD without any changes. Ironically, both Crescendo and Grand Junction (as well as Kalpana) were all acquired by Cisco and that’s how Cisco came to dominate the LAN switching market.
Years later, 1G Base T and now 10G Base T became IEEE 802.3 PMD standards. I have not followed the market acceptance of those, but I’m sure our Sept 8 speakers from Teranetics will fill us in.
Digital Subscriber Loop (xDSL)
The first version of DSL was for Basic Rate ISDN U interface (between the Network Terminating Unit and the voice grade twisted pair access network. In North America, it was based on the 2B1Q line code (Pulse Amplitude Modulation) which was selected by the T1E1.4 committee in August 1986 as a compromise, because it couldn’t decide between 3 completely different transmission systems (I was actually at that meeting in Monterrey, CA and head the “dark horse” presentation by Andrew Siroka of Mitel Semiconductor. He claimed BT had done extensive tests that showed 2B1Q outperformed the other systems and that Mitel could make a transceiver with a significantly smaller die size (lower cost and power disipation) than the other proposed systems.
Bellcore’s Joe Lechleider – a member of the T1E1.4 committee (as was I), had suggested asymmetry would allow higher speeds than ISDN’s 160 kbps, perhaps as high as 1.5 Mbps. The theoretical results wound up being a lot more than 1.5 Mbps, depending on line length, bridge taps, condition of copper, etc. However, there was no standards group that seemed to be interested.
In the early 1990s, there was a new vision of telco TV- initially on fiber optic cable to the home, but some thought it might be feasable to transmit 1 video stream in 1 direction over a 1.5M bit/sec twisted pair, using the higher frequencies above 100K Hz. In 1992 the T1E1.4 standards committee took on the Asymmetric Digital Subscriber Loop (ADSL) project. There were several entries in the official T1E1.4 standards competition:
- Stanford/Amati DMT led by John Cioffi and his grad students
- Bellcore/UCLA/Broadcom QAM
- ATT’s CAP (Carrier Amplitude Modulation is actually a DSP based version of QAM)
The adaptive multicarrier known as Discrete Multi- Tone or DMT (with bit swapping between bins to track noise changes) won the competition by having much better noise margins. The closest test was a 11 dB advantage for DMT. Some of the tests showed 30 dB improvements. T1E1.4 picked DMT on March 10, 1993. That standard did not become popular. Instead, ATT spin-offs (Globespan, Lucent and Paradyne) started making CAP based DSL chips and equipment, which ATT and other telcos started to deploy. In 1996, T1E1.4 took on an ADSL version 2 standard (T1.413v2). Due to a lot of controversy, there was a primary standard based on DMT (which I contributed to and wrote several sections) and an Appendix on CAP. DMT quickly prevailed as Alcatel started designing and deploying DSLAMs based on it. THe ADSL Forum only recognized DMT and not CAP. Carriers all over the world (like Pac Bell, Singapore Tel and many others) signed exclusive deals with Alcatel and DMT based DSL became dominant. CAP was then dead.
Aware lead a consortia of groups who did not want to license Stanford/Amati patents by introducing “G.lite” that tried to remove the very essential bit-swapping of DMT and reduced the number of tones from 256 to 128 to “reduce cost.” While G.lite became an ITU-T standard (the editor was from Intel, which exited the xDSL business), it failed in the marketplace. Instead, G.dmt (and then ADSL2+) went the other direction of higher speeds and actually increasing the number of tones. DMT emerged as the worldwide transmission system for ADSL.
VDSL was moving on a parallel track, but was a perceived to be a smaller market due to distance limitations. There was an equally fierce CAP vs DMT standards war such that T1E1.4 could not select a clear winner. VDSL was envisioned to use ATM exclusively for layer 2 transport and not Ethernet. But the IEEE 802.3ah Ethernet First Mile (EFM) committee selected DMT-VDSL as the short range copper interface and defined a convergence sublayer to make Ethernet MAC ride over VDSL (and SHDSL-2). Despite much fanfare, I don’t believe that standard was ever implemented. But now, ATT has deployed VDSL-2 as part of its U-Verse triple play transmission system from the Fiber Cabinet in the node to the customer premises. Unlike the earlier versions of the VDSL and EFM standards, there is a POTS band reserved in VDSL-2 for analog telephony.
Note: I exited the DSL space over 10 years ago, so have not kept up with its progress. However, I got the following information from a very credible source that must remain anonymous:
“There was a second VDSL Olympics in 2003, also hosted by Bellcore and also BT. DMT VDSL systems were submitted by Alcatel and by Ikanos, while QAM/CAP systems were submitted by Lucent and by Metalink/Infineon. The results were similar to the first VDSL Olympics in that the advantage was roughly 10 dB. That did it, CAP/QAM died at a June 2003 T1E1.4 meeting where DMT was selected for VDSL2.”
Yesterday we learned that ATT is using VDSL2 for its U Verse FTTN transmission system -between the Optical Network Unit (ONU) and the subscribers Network Termination unit (NT) over a copper twisted pair. They are also using it as a U Verse distribution system within Multi- Dwelling Units. I was astonished ATT claimed they got very good performance at 5K feet line length and often deployed longer VDSL loops. They are now testing 4 HD video streams + high speed Internet + POTs over a single VDSL2 loop! We would expect test results to be announced later this year.
We will get filled in on all the gaps by our ASSIA speaker on Sept 8th. I invited Prof Cioffi to the meeting, but he wrote that he might have to travel to Asia. Cioffi and I taught DSL classes at IEEE Infocom and we co-authored several T1E1.4 standards contributions. Later, I taught many ADSL architecture classes at private companies that were implementing or deploying ADSL/SDSL. My partner was Amati’s chief engineer John Bingham (who I had worked with at Fairchild in the Spring and Summer of 1970). Bingham and I got so good at teaching the ADSL class that we joked that we would swap sections (he did the modulation and transmission while I did the architecture and OAM section). He asked me to write the introduction and the second chapter of his book on ADSL Network Architecture. You can read it on line (isn’t this a copyright violation?):
You can buy the book here:
Book Review from IEEE Communications Mag, Sept 2001:
http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=948376&userType=inst (IEEE Explore account required to access full text)
End of Story……………………………………………………………………..