Bandwidth Relief To Be Phased In
ELECTRONIC ENGINEERING TIMES, Page 73, March 31, 1997
In the Internet era, where residential and small-business users alike
remain bandwidth-challenged by their analog phone-service dial-up speeds,
wideband and broadband relief is on the way from any number of
semiconductor vendors and OEMs-at least in theory. The problem is that
central offices, whether owned by the traditional regional Bells, emerging
competitive local exchange carriers, or standalone Internet Service
Providers, aren't being upgraded in a coherent way, to take advantage of
new analog and digital services.
Phone companies are tied up in a wealth of options. Analog modem suppliers,
to pre-vent erosion of an installed base, have turned to unique
pulse-code-modulation (PCM) schemes that require digital termination in
the central office in order to achieve downstream speeds of 56
kbits/second. But 56-kbit speeds don't seem that great to those who have
heard of digital subscriber lines (DSL), which offer rates as high as 6
Mbits/s downstream over standard copper twisted pair, or the tens of
megabits offered by cable-TV quadrature-amplitude-modulation (QAM) modems
over coaxial cable. And the two mutually non-interoperable 56K camps,
represented on one side by U.S. Robotics Inc. (to be acquired by 3Com
Corp.) and on the other by the Lucent/Rockwell/Motorola coalition, appear
at this writing to be nowhere close to reaching common interoperability
specs.
DSLs seem to be the answer to a phone company's prayers, since they can
take advantage of the installed base of copper twisted pair, and offer
downstream speeds as high as 8 Mbits/second in the asymmetric DSL (ADSL)
format. Unlike basic-rate ISDN, which requires circuit provisioning and
does little to alleviate the voice-circuit overload now seen in Internet
access, DSLs can be treated as raw bit-pump services, provisioned as
independent data-overlay services.
The problem is similar to that of PCM modems: ADSL relies on two different
line codes from two different vendor bases. While most vendors are working
on the discrete multitone (DMT) code for ADSL approved by American
National Standards Institute, ADSL pioneers Globespan Technology and
Paradyne Corp. are sticking with their carrierless
amplitude/phase-modulation line code.
Even outside the bitter ADSL dispute, phone companies are being offered
such a confusing alphabet-soup array of DSLs, based on either single or
dual pairs of copper wire, it is no surprise many carriers are throwing up
their hands and losing hope of offering a coherent mix of services. At
last count there were HDSL, HDSL-2, RADSL, MDSL, IDSL, SDSL, DSL/384 and
VDSL, with new options being proposed all the time.
Vendors adhere strongly to diametrically opposed views of what users will
want in DSL services. Chip and system vendors using single pairs of copper
wire and ISDN's familiar 2B1Q coding to achieve speeds of between 128
kbits/s and 1.5 Mbits/s argue that most users will not pay the anticipated
$500 end-user price for an ADSL modem, and will gain the maximum
Web-access advantage from speeds in the hundreds-of-kilobits range. Those
in the ADSL camp argue just as loudly that the advent of animation and
VRML on the Web will drive a need for multimegabit speeds, and that
mid-speed DSLs are too much like ISDN to become popular.
Studies released in the last month from a wealth of analysts, including
Forward Concepts Inc., International Data Corp. and Tele-Choice Inc.,
suggest that, for most carriers, not to decide is to decide by fiat. Those
that delay in wideband-access deployments risk being challenged by the new
Competitive Local Exchange Carriers (CLECs) and independent Internet
Service Providers (ISPs). Will Strauss, principal analyst at Forward
Concepts, said that the only certainties are that every carrier will
become an ISP, and that each carrier and ISP will look for a unique mix of
analog and DSL services to provide. And cable-TV multisystem operators
(MSOs) dare not push out cable-modem offerings if they wish to be players
at all in this market.
Is it any wonder most carriers are at a loss to decide what to provision?
Adding to their woes are the different architectures proposed for the
central office for multiplexing DSL lines together. A group of regional
Bells known as the Joint Procurement Consortium argues for a complex
DSL-access multiplexer (DSLAM) with an ATM interface. OEMs such as Tut
Systems Inc. (Pleasant Hill, Calif.) argue that the easiest way for ISPs
and other newer carriers to get involved in DSLs is through a simplified
DSLAM, which Tut calls a "mini-SLAM." The latter argument may be augmented
by the fact that the CLECs and ISPs, with far less money to spend than
regional Bells, tend to be the ones most aggressive at DSL deployment.
Cable MSOs can laugh at this chaos in the central office, since two-way
cable modems being introduced by several vendors can support downstream
rates approaching 30 Mbits/s and upstream rates of 1.5 Mbits/s and beyond.
But the laughter may not last long.
Scientific-Atlanta Inc. has found interest in a rather mundane
telephone-return-path cable modem in recent months, precisely because many
MSOs are several quarters behind in building out the hybrid fiber/coax
(HFC) infrastructure necessary to support two-way cable modems. For a few
weeks in late 1996 and early 1997, it appeared that the severe cash-flow
problems at leading MSO Tele-Communications Inc. (Englewood, Colo.) was an
early indicator of wholesale abandonment of HFC upgrades. While the
capital equipment crisis at TCI now seems limited to that company alone,
many MSOs are watching their budgets much more carefully than they were in
1995-96. And many are dismissing the need for complex ATM switching
hierarchies within their HFC networks.
Indicators this spring are that the dark-horse entrants in Internet access
may try for a bigger slice of the user pie because of the carrier
infrastructure turmoil. Such options include space-based access offered by
Hughes Network Systems' DirectPC operation, and terrestrial wireless
Internet access through high-frequency networks known as multichannel
multipoint distribution services (MMDSes) and local multipoint
distribution services (LMDSes)
American Telecasting Inc. (Colorado Springs, Colo.) began limited trials
of MMDS Internet access in February, and Texas Instruments Inc. and
Hewlett-Packard Co. are lining up partners to offer LMDS systems this
summer, once the FCC finalizes its auction plans for the 28-GHz spectrum.
Gary LaBelle, marketing manager for wireless systems in HP's video
communications division, said that HP launched the "First Mile Project"
last year to spur LMDS for Internet use, since it was seen as the most
cost-effective broadband-access system that could accommodate low customer
take rates.
"We will be leveraging our low-cost MMICs to achieve systems with up to 52
Mbits/s downstream and 3 Mbits/s on the return path," LaBelle said. "LMDS
can offer better returns in situations where there are moderate customer
take rates, something that is harder to justify for DSL or cable modems.
We looked at MMDS, but true two-way interactivity would require a denser
footprint, and hence more infrastructure costs, than a properly configured
LMDS system."
The key loser among the wideband and broadband options is basic-rate ISDN.
When ISDN was proposed in the 1980s, overload on the circuit switches was
not a problem, and no one could foresee that long circuit-oriented calls
for file transfers and Internet access would cause overload on 5ESS
switches.
Ten years ago, two bonded "B" bearer channels of 64 kbits/s each seemed to
be plenty of bandwidth for home and small-office users, because desktop
users had not gotten used to LAN options like the 10-Mbit/s Ethernet and
its 100-Mbit and 1-Gbit successors. Now, even the bonded 128-kbit service
of BRI seems laughable, and phone companies hesitate to provision BRI
service because of the circuit-switch provisioning problem. BRI maintains
important markets in Europe and the Pacific Rim, and moderate markets in
some U.S. regions, but few analysts see a long-term place for this service
in the United States.
Analog lives on
The moderate Web-surfing needs of many residential or small-business
customers could be met by the new pulse-code-modulated modems, which can
offer downstream speeds up to 53 or 56 kbits/s-admittedly small potatoes
compared to DSLs, but within the general range of basic-rate ISDN, while
avoiding the provisioning and tariff issues of the latter service. If the
International Telecommunications Union approves a two-way PCM service
(compared to current U.S. trends of mandating a V.34-compatible return
path), PCM modems could offer symmetrical service at near-ISDN speeds.
When U.S. Robotics and Rockwell Semiconductor made the first announcements
of PCM technologies last September, analysts wondered how widely the
technology would be accepted. Unlike the Trellis-code upgrades of standard
QAM modems represented in the V series of ITU standards, PCM modems use
tricks in local-loop synchronization between end node and central office,
requiring a digitally terminated central office to achieve speeds. Since
last fall, however, vendors and carriers alike have signed on to the "56K"
upgrade plans with fervor.
"Most ISPs realize they have to have digital backbones anyway," observed
Martin Rauch-werk, marketing manager in Lucent Microelectronics' modem
group. "The issue of digital termination is going away. The problem of
incompatible central offices remains, however. Until a standard arrives,
end users have to know whose equipment is installed in their local central
office."
The mix of digital and analog services favors the type of hybrid
auto-sensing modem-pool equipment for central offices offered by vendors
such as Motorola Inc.'s Information Systems Group (Mansfield, Mass.). The
digital lines used in most such architectures, however, are ISDN B
channels. The key for finding the right mix of service will be to mix the
DSLAM features from the DSL world with the hybrid access devices offered
by ISDN routers and first-generation modem pools.
That's why developers of equipment for the central office, such as Ascend
Communications Inc. and U.S. Robotics, will play key roles in defining the
mix of services offered by carriers. Some of the early plans, such as
Ascend's MAX xDSL strategy and USR's V.Everything, rely on midspeed
single-pair DSL options that preserve the 2B1Q coding of ISDN.
These systems are based on low-cost DSP chips that achieve speeds between
200 kbits and 1 Mbit/s, without requiring the special line drivers and
high-power budgets of ADSL or HDSL.
Ascend has even tried to make the best use of ISDN B channels by creating
what it calls "IDSL," in which 64-kbit services are offered in DSL
channels, taking the ISDN traffic off the main voice-circuit switches. All
these OEMs are predicting that speeds significantly under 1 Mbit/s will
prove important in the DSL market.
These bottlenecks may be real, but this doesn't keep Level One or Rockwell
from looking at higher-speed options. Dan Cordingley, director of
marketing at Level One's communications division, said that the ANSI
efforts to define a single-pair HDSL-2 may represent the ideal end game
for lower-cost, high-performance DSL options.
In the meantime, however, ADSL system vendors like Amati Communications
Corp., OrcKit Communications Ltd., Westell Corp., Aware Inc. and Alcatel
Data Networks Inc. are telling carriers in specific regions that they can
distinguish themselves by jumping to ADSL today, while competitors are
stuck in midspeed DSL ranges. The proof in cost reduction will come in
April, when Motorola Semiconductor Product Sector's wireline IC division
samples its long-awaited CopperGold single-chip transceiver, which
marketing director Jacques Issa said will make a turn to ADSL painless for
most carriers.
Amati did not wait for CopperGold, instead turning to both NEC Electronics
Inc. for DSP ASICs, and Texas Instruments Inc. for the 320C6x DSP
processor, in order to reduce costs for ADSL as quickly as possible. U.S.
Robotics, now backed with 3Com money, plans to take an entirely
proprietary route to developing DSP chips for all relevant DSL standards.
Carriers that try to leap too quickly into DSLs, however, may find that the
access muxing equipment is a gating factor. The full-featured DSLAM is
most likely a preferred end state for most carriers, since the 51-Mbit/s
ATM interface will probably be the transport interface of choice. But ISPs
and small carriers that want to slam a DSLAM into service may want to look
at cheaper options.
This is why Tut Systems Inc. developed its mini-SLAM, and why Ascend
advocates a graded, step-by-step upgrade to full-featured DSLAMs.
Tut president Sal D'Auria said that ATM is still an uncertain transport
technology, even in the WAN. But assuming ATM becomes a preferred
interface, there will be an important transition time in which central
offices will want to combine DSLs in T1 or T3 channels in the simplest way
possible.
Copyright c 1997 CMP Media Inc. |
