DSL Prime is always looking for news. Email Dave Burstein

The trade paper of a Net community
September 30, 2003

ECI Taiwan $60 DSLAM + modem
Public bid, confirming rumors in France and China
Bargains under a million lines
Power mad modems
The Bells want them fixed
Third generation DSLAMs
Non-blocking 8X throughput, 25%+ more coverage, IP control, testability
Alcatel's remarkable opportunity
Are new DSLAMs as good as they claim?
Untested, unproven, uncertain
Equipment isn't cheap if it breaks down
SBC to $26.95
Is low price the best sales tool?
Moore's law applies: Comcast to 3 mbps, Roberts talks 50 mbps
Competitive intelligence failed
Forum interop thank you
Remarkable diplomacy keeps the industry together

       Briefs: Qwest, Cablevision, DSL Prime bias, Belgacom IPO $1.5B oddity, Mario Mella of Fastweb, GoDigital, Mark Kersay, B2, Tiscali, Harry Edwards/ Michael Powell, Quad spectrum ADSL2, VDSL line code, again, Kevin Sheehan, Faraj Aalei, Taher Behbehani, Lyn Gulbransen, David Molony, Verizon earnings warning

Taiwan's Chunghwa is paying $35-40/port for state of the art ECI DSLAMs, and a total of $60 including modem. News of that open, public bid for 800,000 lines is spreading at internet speed, and will influence every contract in the world. France Telecom is getting similar offers, while LDCOM's price from Huawei is extraordinary. Price isn't everything, as my outages described below make clear. But these lower prices should change crucial strategies.

With today's prices, any claim it's "unaffordable" to install DSL wherever you have fiber is simply nonsense. That makes the hundreds of millions in "incentives" Chris Timm is offering a waste of the British taxpayer's money, while the $13.4M budgeted by Illinois is more than the total cost SBC will incur extending their DSL network in the state.

The improved reach of new DSLAMs will allow a telco to cut the number of "unservable" customers 10-25% - good business and good politics. The lower price makes plausible Italian and Netherlands prices around 15 euros for low end service, and Verizon's Telus offer under $20 for the first six months. With gear ready and affordable, France Telecom is planning a massive video deployment. They'll need it, with Telecom Italia planning to invest 200M euro, and Free.fr and LDCom doing VOIP at a cent or less a minute.

Say hello to the round fellow with a beard at the FTTH conference in New Orleans or ITU in Geneva. Kevin Schneider of Adtran has put together some strong technical presentations for USTA in Las Vegas I'm sorry to miss that week. But I learn more about the future by studying the countries like Japan and now France that are pulling ahead of the U.S., and care about how the net will be delivered to the poorer parts of the world.

ECI Taiwan $60 DSLAM + modem
Public bid, confirming rumors in France and China
"Watch Taiwan" ECI's Chaim Rainer advised. "They're opening the bids publicly, and look for surprises." Five separate people had told me DSLAM bids to France Telecom were thirtysomething, but no one would go on the record. Someone I trust tells me DSLAM plus modem were in the forties in Chinese bids, where they are moving millions of lines per quarter.

Baycom integrates the system, charging Chunghwa NT$2,082 (about US$61.23) for the DSLAM from ECI and a modem from Tecom, Digitimes reports. Globespan is supplying the chips, and presumably offered a generous discount to win the contract. Samsung, Comtrend, and ADI won last year, while Alcatel, Lucent, Nokia, and Ericsson were among this year's high bidders.

Bargains under a million lines
Buyers like Chunghwa and FT, looking to spend a hundred million, of course have more clout with vendors and expect great pricing no matter whom they deal with. But smaller purchasers who buy wisely can also get great pricing. LDCOM went to Chinese giant Huawei, who gave their first European DSLAM customer a great price, possibly lower than FT is getting. Paradyne, whose ad is above, is happy to sell you a couple of 24 port DSLAMs for $65/port, and even throws in some modems. Small buyers can get great prices as well.

Power mad modems
The Bells want them fixed
"Who's going to pay for fixing performance robbing systems?" is a big question for the telcos, who of course would prefer their suppliers, mostly Alcatel, pick up the burden. Design errors made several years ago mean most DSL gear in COs blasts 100's to 1000's of times the necessary power to many lines. That made for great results in lab tests initially, with few problems (except long reach customers) when take rates were under 10% and speeds capped under a meg. But with more active lines to interfere with each other, and marketing people insisting on 3 mbps and more, reliably delivered, the problem has become severe. The standards and specifications were not entirely clear, and currently everyone's pointing the finger at someone else to blame.

Third generation DSLAMs
Non-blocking 8X throughput, 25%+ more coverage, IP control, testability
"We changed DSLAM providers, and that's one reason" BT CEO Dr. Paul Reynolds pointed out in London. We were discussing throughput for video, and he guided me to look at new DSLAMs from Marconi and Fujitsu. Thousands of lines of video rated DSLAM now fit in a single rack.

DSLAMs are following Moore's Law, with new units being both faster and cheaper. They deliver ADSL2 (and soon ADSL2+), as well as fiber and VDSL, but what defines the ECI as third generation is that it's non-blocking to high levels of traffic. The new DSLAMs can serve multi-megabit video simultaneously to nearly every port, including many unicast streams for video on demand and time shifting. Second generation units can choke if more than perhaps 10% of users want on-demand video, or HDTV demand grows as I think it will. The ECI is designed for 70% take rates, as you might get at Olympics highlights, major football events, major crises (think 9/11), and other peaks.

ECI, Fujitsu and others, are moving the market forward impressively. The new generation is based around network processors and backplanes with 5-10 times the throughput of prior models. One board designer spoke highly of the Agere and TI controller chips, but chose the Wintegra that is getting lots of buzz. The current analog and digital DSL chips drive enormous numbers of ports in very little space, eliminating nearly all of the "glue" chips previously required. The splitters, also dropping in size, now take more room than the DSL components. Nokia, Marconi, AFC/Telliant, Calix and now Siemens claim similar performance levels, but I haven't looked closely at them.

Second generation DSLAMs have some staying power - Fastweb just bought an upgraded Alcatel 7300 (the 7301) for a 122 channel video system, and there probably isn't a DSL deployment in the world today that needs more power than that. But third generation cost no more.

Alcatel's remarkable opportunity
Alcatel won the North American market by getting in early, offering by far the best price, and doing a generally fine job of supporting their customers. Every few months, a Bell puts out an RFP, considers a few bids lower than Alcatel's, and decides the savings aren't enough to breach a relationship that's working well.

Now, world prices are typically half what the Bells pay. The Bells also want to solve the power mad modem problem. Simultaneously, they are considering spending money to implement QOS. It's far better for them to simply replace the obsolete old DSLAMs with 3G DSLAMs with QOS built in. Even a 5% improvement in reachable homes (an underestimate) adds several hundred thousand customers, saving marketing dollars. Fewer problems mean lower support costs. Necessary market responses (speed, video) are no longer crippled. Direct savings of replacing about 4M ports of DSLAMs can easily be documented at $100-200M per year, and considering the competitive struggle the number is much larger.

Alcatel can make that changeover pay in 12-24 months, with big dividends from better services for many years. That simply requires offering North Americans the same price similar units sell in China. It's time for them to go to the Bells with a pro-active plan, one that includes a "price too good to refuse" on a full upgrade of the DSL network, not just a lower price for incremental units.

The Bell's project to throw money into BRAS QOS was already a brewing disaster, because they need far more bandwidth and equipment performance to compete with cable. Alcatel has enormous incentive to do right by the Bells on this one. ECI has invested in OSMINE certification to sell to the Bells, with Samsung and others willing to price dramatically to win U.S. contracts.

Folks like Krish Prabhu won the Bell business by being a partner who guided them to the right choice. There's enormous inertia at the Bells. Without some new thinking, they'll stick to plans that have failed, and throw more money into QOS bandaids. Paul Reynolds at BT has seen further, and Alcatel's best move is to bring that vision into their product recommendations.

Alcatel has another trump: researchers advanced in DSM. They know how to use individual line parameters to dramatically improve performance, and could easily be the first to market. That's the kind of real product differentiation that justifies a premium. They'll need to press Broadcom (their indicated chip supplier) to include all the single port control into their CO chip, which will probably require a quick rev of Broadcom's chip, but both Broadcom and Alcatel have the engineering talent needed.

Are new DSLAMs as good as they claim?
Untested, unproven, uncertain
The new generation of DSLAMs is so much better, you'll want to move over quickly but not blindly. "Have you ever tested your DSLAMs using the features your pr is promising?" I've asked at least eight major manufacturers. No one has. One shot back "What do you want us to do, hook up 350 television sets simultaneously? None of the test and simulation vendors can do anything like that." Actually, I think that's exactly what they need to do (real or well simulated), if their marketing VP is swearing to customers such a service will work. Everybody in technology knows how often products don't work as promised in the real world. DSL's key example is the early DSLAMs and modems. They promised to serve customers reliably to 18,000 feet, but most telcos found far too high a problem ratio beyond 12,000-15,000. Similarly, Verizon's "DSL Hell" problems was partly due to control routers that failed unpredictably when loaded.

A division CTO pretended not to understand my questions when I asked if his DSLAM could deliver video to 70% of the users. He kept pointing to the spec sheet, and said I should just compare the "rated" bandwidth with the data rates the video required. But loads of other factors - jitter, latency, resonance effects, channel changing - all could affect whether a real provider, like BT, could deliver commercial quality service to their customers. There's every reason to expect some problems won't be clear without extensive testing and real world experience. The fiasco of overdriven DSLAMs is just now becoming obvious, three and four years in, because realworld testing was inadequate.

Numbers on data sheets often don't reflect real performance, even if not direct mis-statements. My friend Lee Goldberg of Analog Zone, working with Georgia Tech, is currently testing the actual throughput of serdes transceivers, a key backplane component. Georgia Tech has already tested Accelerant and Vittesse, with National, Quellan, and hopefully others to follow. Using three different physical backplanes, they test ultimate capacity before the bit error rate gets too high, jitter, and other properties. One test that seemed particularly relevant to performance under load is noise immunity measured with an active adjacent channel. That's looking for the kind of problem often ignored until reliability varies in the field.

Lee writes "While a DSLAM may have 128 10-Mbit/sec ports, can it really support the 1+ Gbits/s worth of bandwidth required to feed each customer a a video stream? There are many things that affect whether a DSLAM can handle anything close to its rated capacity, including the switch fabric and traffic management hardware, but one of the most significant choke points is the backplane - the high-speed bus network that connects the line cards to the other elements of the DSLAM. Unless the backplane and the silicon that drives it has sufficient capacity to support the total aggregate traffic, it does not matter how much fiber you feed the box with. "

Because I believe in the value of independent test results, I'll promise to publish at least a brief about any DSL vendor who arranges testing at Georgia Tech, Telcordia, or similar. I hear from DSLAM designers favorable comments about Broadcom, Agere, Wintegra and TI, but I'd like to have some objective data. If you care about DSLAM design, check Analog Zone http://www.analogzone.com/io_fram.htm If your DSLAM deserves kudos because its components are well-rated there, please send me details to report.

Equipment isn't cheap if it breaks down
Buying on price is clearly a mistake if the reliability of your system is affected. My network and DSL connection didn't work for three days, unless I held my finger on an ethernet plug and wiggled. My phone and DSL went down last month, reportedly because of a splitter failure. "The equipment is less than 10% of your total cost," a friend from Alcatel reminds. "The wrong equipment really costs far more. He's right - just the support calls, much less the replacement modem, far out way any possible saving in initial price.

SBC to $26.95
Is low price the best sales tool?
I wrote a year ago "$30 is the natural price", and got hooted at by telcos thinking they must charge more. The Bells had been at $49.95, rather than the $30 their own marketing surveys called for in 2000. DSL equipment and bandwidth are now so inexpensive, the natural price is $5-10 more than dialup.

We've just had several "natural experiments" about what drives demand. In Q2, Verizon's price drop had barely begun, and they attributed the poor quarter to it "not kicking in till the end of the period." SBC's net adds were three times as high, as their price drops were earlier. It will be interesting to see the "SBC-Verizon gap" in Q3, which I'm sure will be narrower. They still are are $35 without the bundle, however. Similarly, we saw a major uptick in Bell Canada's net adds when they eliminated the hated bandwidth cap, a hidden price rise they had imposed.

Moore's law applies: Comcast to 3 mbps, Roberts talks 50 mbps
The official announcement by Comcast of 3 mbps, and similar from all the other cablecos, shouldn't be news in the industry. DSL Prime and DSL Reports have been covering it for months, and Cablevision and Cox have long been there. Amazing how many people "forget" that @Home was serving 6-10 mbps to all the cable customers three years ago.

The cost of bandwidth to the carrier is primarily the fiber and the switches/routers. In most of the developed world, the fiber is already in place. Switch/routers thus are the driver, halving in cost every 18 months or so as Moore's Law drives down their chip cost. Carriers in many areas have to fight the backbone oligopoly, but SBC for example is already peering most traffic and others can learn from them.

The difference in cost per user between offering 1 mbps and 3 mbps to all users is less than $2, probably less than $1, based on actual marginal backbone and data carrying cost of a Bell or an ISP large enough to use dark fiber. 10 mbps adds perhaps as much again, with few web sites effectively going over a meg or two. Peer to peer redirection (ask Sandvine) keeps that down, among other techniques far smarter than customer losing bandwidth caps.

DSL Prime loves low consumer prices, but SBC's bottom line doesn't want them dropping quite so rapidly. That requires a network that more than matches cable, however, so they can offer comparable speeds. They are ready, with some struggles, to go to 3 mbps if management gives the ok, but will tap out soon after that.

Competitive intelligence failed
In the U.S., the Bells suddenly are discovering they aren't ready to match the 3 mbps service all the cablecos are moving to before the end of the year, much less the 30 and 50 mbps they are ready to upgrade to. NTT, surprised by Yahoo's success, has reallocated $10B to fighting competition. These trends were visible 12 months ago in most cases; I reported in 2002, for example, that Mark Coblitz of Comcast was designing a network to deliver 15 mbps to his typical home. Anton Wahlman has reported Japanese cable companies are offering 30 mbps service, with Harmonic expecting U.S. customers shortly for similar.

Few telcos listened. "Nobody will want more than 1.5 mbps" was an amazing email I got from a very smart telco engineer, whose company was obviously out of touch with the outside world. Reminds me of 1982, when nobody at IBM thought you'd ever need more than 640K (not M) in a microcomputer. The results were years of terrible kludges for MS-DOS, with more kinds of memory extenders than I can recall, and far too many operating problems. Top management knows better now, but the technical departments haven't necessarily translated that into their plans. One telco is buying units for remote terminals that top out at 1.5 meg or so, half today's cable speeds and even further behind in a year or two. A second is considering a deployment of a shared network splitting 6 meg among far too many users to get performance. Folks like Ed Whitacre and Larry Babbio are trained engineers; they need to make sure their technology choices fit their strategic goals, because there's a clear disconnect between their engineering team and what the rest of the company knows will be happening in the market. An easy test in the U.S.: if you and your teenage daughter can't watch different HDTV programs on your planned networks, you will starting bleeding your best customers in two or three years. That speed already is inadequate in the Japan and Korea, probably further in the future in most of Europe.

Forum interop thank you
Remarkable diplomacy keeps the industry together
Shin-Jou Fang, president of TrendChip, writes "Our engineers just came back from the ADSL2/2+ Interoperability Test at UNH. It was a great event for us. We want to thank all DSLAM participants for their kind support." This is the kind of response the quiet work of the Forum often produces, providing common ground where every part of the industry can work together. More co-operation means greater success for all. It's a remarkable job of leadership that has kept everyone working so well together. Tom Starr, Chair of the Forum, has a tough job, uniting behind the Forum an extremely competitive industry. Like predecessors Bill Rodey and Hans-Erhard Reiter, he's maintained the Forum as a safe haven that supports crucial industry work. Quietly, standards are emerging for network intelligence, modem and gateway interoperability, and home networking. This is remarkable in an industry fighting so hard for a billion dollar market.

email:

       "Someone should be paying you a fortune in marketing for one of the DSL guys -- I marvelled at the double-speed/half-opex catch phrase you constructed." Haven't made a fortune, although I'd like to and should spend more time selling advertising and the like to make it happen. But it wasn't just a slogan: the advances in engineering, especially reach and DSM, making doubling speed likely in the next 24 months. The better performance solves many customer problems. Add improved testing, call center automation, better modem software and the like, and cutting Opex in half is a realistic goal as well.

       A Bell planner wrote "Dave, I agree that VoIP and Wi-Fi should be added to the DSL modems (just to lower the overall CPE costs, reduce complexity to the end users, AND to keep up with the cablecos). Many current generation cable modems already have 2 built-in RJ-11 ports to support analog phones and voice/fax applications as part of CableLabs' PacketCable architecture. Now, Wi-Fi is being added to chipsets in cable modems to faciliate transfer of digital media (such as movies) within the home network. BTW, TV manufacturers are starting to add Wi-Fi interface to high-end digital TVs (SONY is using 802.11a, while one of the European manufacturers is using 802.11g). The view is that Wi-Fi will enable one to play movies downloaded from the Internet (stored on a PC hard-drive such as MSFT's MediaCentre PC) directly onto the digital TV. " That's the kind of response TI was hoping for when they wrote and sponsored the VOIP chip article. Even sponsored articles get edited at DSL Prime, but this one seemed so directly to the point I didn't ask for any changes. (P.S. DSL Prime welcomes articles, which get summarized in the news letter and posted on the web. TI's sponsorship meant it was also directly mailed to subscribers, which I allow a few times a year.)

       Matt Byrd of Net to Net wrote because I hadn't included their field proven unit in an article about bonding, and Lesley Hansen demo'd it for me in London.

       "I've heard no complaints. On the contrary, there's been nothing but praise for the job we've done in the last couple of years." John Badal, Qwest New Mexico President (NMBW)

       Cablevision charges $5 more for cable modem service to customers who only take basic cable service. This is similar to Verizon, which also charges $5 extra to basic phone customers in the most common bundling choice.

       "I really like DSL Prime because you have no bias," a VC surprised me saying in Boston, especially because I had recently been unkind to one of his key portfolio companies. I had to reply he was wrong, however; I have a bias for the consumer who deserves a better, faster internet at an affordable price.

       Mario Mella of Fastweb is now a prophet with honor, as his visionary 11,000 kilometer fiber build across Italy has attracted 250,000 customers and begun multicasting 120 channels. Even the most skeptical were impressed with EBITDA positve results.

       Repeaters and extenders don't make sense, you say? Tell that to the 100 customers GoDigital announced for their DSL products. Most telcos have already picked the "low-hanging fruit", and it time to move to "DSL Anywhere".

       Mark Kersay of Current Analysis points out how far ahead SBC has been, compared to the other U.S. telcos "Of the five major DSL providers (the four RBOCs and Covad), SBC has established itself as the DSL market leader, having added more new subscribers in Q2 than the other four combined." Verizon will come back some this quarter, but is unlikely to come close to SBC.

       B2 in Sweden, one of the world's key innovators, launched VDSL service and video. Josephine Kenny of CA opines "Telia Sweden has no excuse for lagging behind B2 or Bostream in the VDSL race." Suddenly, France, Netherlands, Sweden, and parts of Germany (Hansenet) are seeing competition that matters.

       Tiscali is signing up 17,000 DSL subscribers per week, Christopher Emsden reports, well on their way to 750,000 by the end of the year. That's probably more than BellSouth, and more than Verizon in many quarters. (Verizon is overdue for a strong quarter after the price drop, and I believe is getting one.) That kind of achievement is allowing CEO and controlling stockholder Renato Soru to run for political office, hopefully with more diplomacy than Berlusconi.

       All the papers I checked missed a juicy story Tuesday, the extraordinary conflict of interest of Judge Edwards of the D.C. Circuit. Googling "Harry Edwards Michael Powell" would seem an obvious reporting step, after Edwards took over the FCC Triennial case from the assigned Saint Louis court, on which Powell has staked his career. Powell remarkably has called for the court to overrule his own agency, because he was outvoted. The second reference finds Powell saying "special thanks to my dear friend and mentor Judge Harry Edwards," as well as "You are, and always will be, my friend." Once upon a time, Washington reporters covered up the sexual escapades of Jack Kennedy and other politicians. Perhaps today they believe this kind of conflict of interest by a judge is so common it's not worth reporting, like the generally overlooked fundraising ties of Congressmen. The only ethical thing for Edwards to do is recuse himself from the case.

       The Belgacom IPO deal contains an inexplicable $1.5B payoff to SingTel and SBC, in the form of the government taking over $5B in pension obligations. Most reports missed that part, not noticing most of the "profit" is a government handout to the telcos.

       Quad spectrum ADSL2 can deliver speeds of 50 mbps for about a kilometer, and may soon be deployed in Japan. Centillium Communications, eAccess, NEC, Aware, Panasonic, Sumitomo, & Sharp are proposing a standard. "Quad spectrum ADSL2 offers the potential to significantly increase the service area coverage for fixed data rate services at 20+ Mbit/s, allowing service providers to offer 'triple play' services (video, data and voice) to a large enough number of customers to garner critical mass and ensure service viability" they offer. Whether the right choice for speeds of 20-50 mbps is ADSL2+, quad ADSL2, VDSL or VDSL2 is a key industry question. I wish my providers in New York were part of the solution.

       VDSL line code, again. A surprising coalition of ADC, Cisco Systems, Hatteras Networks, Lucent Technologies, Motorola, Next Level Communications, Paradyne Corporation write. "As system providers we have customers around the world requesting systems with QAM as well as DMT VDSL line codes. " Transwitch and Photonic Bridges are new names to me that have signed on for the DMT only side. Meanwhile, work plans are moving forward to align ADSL2 with VDSL for compatibility.

       Super salesman Kevin Sheehan moves to CEO at Hatteras (per LR), whose primary product bonds multiple lines for high speed symmetric service. That market, typically 10 mbps, seems appealing, with speeds higher than T1 and prices lower than T3. Unfortunately, few have been buying, either from Hatteras or competitors like Spediant and Actelis. Sheehan writes "There's a ton of activity in our space." One trump Sheehan will have is a board that includes U.S. West veteran Joe Zell and Verizon ex-exec Fred D'Allessio, both now investors.

       Faraj Aalei of Centillium joined the Silicon Iran community foran MIT sponsored event, the Iranian-American Technology Forum (ITF2003). Aalei writes "There is a huge Iranian American high tech community, including Pierre Omidyar of eBay, Farzad Nazem CTO of Yahoo, Hossein Eslambolchi CTO of ATT, Shygan Khradpir CIO of Verizon, Yousef Javadi President of Sprint International, Amad Bahaei CTO of National Semi, and Afshin Mohebbi former president of Qwest."

       Taher Behbehani of UTStarcom, at the same event, reports some of his company's success comes from using both an "international management team and ex-pats, which enables us to combine local and US styles of management for best practice in each account."

       Lyn Gulbransen, a good friend to many of us since she worked with Jim Southworth building the Concentric Network, is now at UK ISP Brightview.

       David Molony of Total Telecom gave me a copy of his September issue, with a great cover of a huge Samurai warrior striking at a brave little Yahoo BB. Molony, Dix, and Donegan do some first rate reporting. You should subscribe.

       Verizon gave an earnings warning, and the U.S. telcos dropped billions in market cap. Sounds logical, but most pros on the street already assume that telco's claimed earnings are overstated by half or more. Temporary tricks to prop earnings up can only go so far (legally), so shortfalls are almost inevitable. I've been reporting this over two years, and several top analysts have said similar, not always in such clear language. Standard and Poor's in particular, has been devastating.

September 20, 2003

• Reynolds of BT - 90% reach on way
• Help him solve the next 7%
• ADSL3/DSM- "It will work"
• Chips hitting market will adjust each port
• TI is about to deliver
• eAccess public offering
• Japan August back to 340K
• ADSL2 - it matters 
• Reach, testing, power and noise savings
• ADSL2+ isn't working yet; ADSL2 has problems to solve
• 24-26 meg Japanese service all proprietary
• Briefs: ZTE, PacketFront, Dasan, SBC/Adtran, Infineon small single chip VDSL, KT QAM and DMT, Centillium/UTStarcom, NTT/VDSL, Vince Oddo, Vik Grover,  Globespan, a telco CFO

• Speeds delivered to customers can easy double. DSM is working in early testing, as is ADSL2 and improved equipment. The majority of customers can get 10-20 mbps; those at a distance, usually a meg or two.
• Operating costs can be cut in half. Improved technical performance, automatic testing and service configuration makes a difference. Bandwidth costs have a dozen obvious cuts. Hard work and best practices, of course, crucial.
• Growth will be explosive. British Telecom intends to go from 1M to 5M in three years, China from 4M to a probable 30M, SBC is geared for 500,000 a quarter. A virtuous circle: lower prices bring volume, and volume dramatically lowers costs. Equipment for each new customer costs less than $100.
• Deployment is 90% solved. 
• The most efficient early adaptors will achieve these results in two or three years. Even elephants can dance. 

• Move customers with short loops to the higher frequency bins, clearing the lower bins for the long loops. With less noise (and a possible power boost), long lines do dramatically better, while the short lines still reliably deliver the 1.5 or 3 mbps promised the customer.
• Check each line for impulse noise, such as dimmer switches. 90% of lines do not have an impulse noise issue, allowing changes to the maximum margin. This can yield a remarkable 15 db improvement on most lines.
• Change the switch point, moving some bands from upstream to downstream. Because these are lower frequencies, the result is possibly more than proportional. Result: 200-400K more downstream, still delivering 256K upstream.
• Measure and police power use, reducing the overall noise in the binder. ADSL2 powerdown, below, is part of the solution. Current ADSL chips are creating a problem in the field, with most of them actually running over the specified power limits already. This made for better results on rate/reach testing, but has at least one telco incensed at the noise as they add more active lines. Result, a frantic rush by chip vendors to retune current chips for new telco testing. They are also trying to find power solutions for existing modems.

• Reach on long loops improves 2,000-3,000 feet, demonstrated on a line simulator by Analog Devices. Real world results may not be as good, but any reduction in long loop problems saves money. If you extend your servable customers 1,000 feet, that's significant growth in sales.
• ADSL2 test modes give you a much clearer picture of what customer problems can be, providing much better data on the loop near the customer. Spirent, using Aware's Dr. DSL test software, showed how you can quickly  pinpoint problems like bad in-home wiring. These are now among your costliest to troubleshoot and solve. This will only take a software upgrade on the gear most U.S. telcos already own.
• Powerdown does more than just save on electricity costs, although that's significant as well. Unnecessary power use create interference on other lines, hurting performance in the field. (Telcos are telling me this is a major problem.) Reduced power/heat is especially important in remotes. Infineon is ready to test powerdown with telcos.
• Dozens of small improvements are incorporated in the new chip sets. Better filters and hybrids, improved design algorithms, and much more were all possible under ADSL1, but have not been incorporated until the newest designs. I'm hearing from many sources the modems are significantly better.

• "A colleague of mine said you're now writing for the New York Times. Is that true?" I don't know how that rumor got started, but it's not so. This is a good chance to thank the folks at the Times, WSJ, and many others who treated this online reporter as a colleague, and often have been helpful with ideas and sources.
• Covad's investors are vocal "I'm still rather stunned that you wrote that Covad will need to raise 100 million extra in funds.  I suppose others are too, as it is so clearly incorrect. Exactly how many times does Mr Hoffman have to publicly state that Covad will reach positive cash flow with a 100 million cash cushion for it to be understood?" No disrespect for Hoffman, but I've rarely heard from a CEO who didn't have a scenario for a profitable future.  They are projecting a "net loss to be in the range of $29 million to $33 million" for the third quarter, and I refuse to consider EBIDTA and cashflow an adequate substitute for real earnings. Competition is getting tougher, with Comcast at 3 mbps and SBC and Verizon having many offers at $30/month. SBC has completed the technical planning to raise speeds to 3 mbps this year, although the final decision isn't set. Covad will have to respond to all this, and much more next year. Equipment depreciation is real, likely to result in a less competitive offering. I'm assuming tougher competition than Hoffman is using in his projections, and we'll see who's right. I still wish them the best of luck - with anemic regulation in the U.S., we need stronger competition. Article next issue about the co-founders claims of "looting" in the Bluestar deal and Covad's outlook in the future.

• ZTE, the lesser known Chinese giant, won a $19M contract to rebuild the network in Rumania, including DSL. This is one of their first actual contract wins in Europe. They've won mindshare already; Dr Miguel Horta e Costa of Portugal Telecom told me how impressed he was after visiting their Chinese plant, and Paul Reynolds of BT mentioned them to me in his analysis of DSLAM vendors, and why he chose a new one.
• PacketFront in Stockholm is equipping several municipalities and power companies with systems for broadband over fiber and VDSL. Management is from B2 and Cisco, and is working with Öresundskraft/Helsingborg in Sweden and NESA in Denmark. The challange to Tele Denmark may be especially important, possibly accelerating SBC's moves to cash out of all foreign holdings. The Belgacom stake is involved in political manuvering that includes a $1B payoff to SBC/TDC in a projected IPO.
• Dasan has the best test results for both QAM and DMT in the testing at KT, which is in the market for 300,000 lines. Previous reports from Digitimes that Lucent won the order are probably uninformed.

• SBC has been buying 24 port pizza box DSLAMs from Adtran, rather than the small Alcatels they were using to upgrade remotes. There's at least a $100M market the next few years as the telcos around the world now deploy more widely.
• Special thanks to M., a pr specialist who responded to my factcheck of an article with a clear refutation of another reporter's mistake. I have errors too, and strongly appreciate any reader who helps correct mine.

• Infineon gave me a sample of a remarkably small single chip VDSL part. Containing both digital and analog parts, it easily fits on my little finger. They promise very attractive prices for the new QAM chip, fabbed at .13 micron. The Israeli designers promised me a picture of their (very simple) reference design.
• KT's next tender looks to split between QAM and DMT, after installing a million lines of QAM. They point to a major saving from competition. The DMT chips still cost about 50% more than the QAM ($15 vs. $10), and would probably be higher if Ikanos and ST were the only ones shipping. A leading DSLAM manufacturer has not yet received samples from Globespan or Broadcom. DMT advocates are confident the difference will become nil over time. It was enormously flattering to hear from a KT executive they read my work and are influenced by it. Their achievements are so extraordinary the world should now be learning from them. I'm looking forward to exchanging ideas with many more Asian practitioners, and hope several volunteer to present at Fast Net Futures the end of March.
• Centillium's proprietary 50 mbps extension of ADSL finds a customer at UTStarcom, Anton Wahlman reports. Yahoo BB will trial it, while NTT has a tender out for VDSL instead. Ah, to be young in Tokyo.

• Vince Oddo moves from COO at network Telephone to CEO at Access Intregrated Networks in Georgia, a private company reporting 100,000 profitable lines. BellSouth has reported marked dropoff in the smaller business market as CLECs have built relationships with customers who feel they were ignored. AIN uses independent agents rather than direct sales, and attributes their success to good customer relations. 
• Vik Grover moves from KBRO (after their very successful investor conference) to Needham, bringing his strong advocacy for Covad with him. Grover recently issued a "sell" on Verizon, with a detailed report of their problems. That's gutsy for an analyst, who typically say "hold" when they mean "sell". As always, don't read these as market predictions from me - I don't have certainty on either of these calls.

• VC's James Coulter and Dr. Hermann Hauser were replaced on the board at Globespan by industry executives, strengthening the hand of management. In many companies, the moneymen are inclined to cash out and management prefers to build the company.
• There's a telco CFO trying to decide how to handle my note, "Apparent discrepancy in accounting for costs of DSL deployment." That was a drastic step for me to take, which I've never done before. There are apparent inconsistencies in the data they've presented, and several weeks of questions to their appropriate people haven't resolved them. The only logical way I've found to explain the discrepancy is to infer that other expenses are being loaded into capital spending, directly or as excessive overhead allocations. MCI's first public issue was misallocation of opex to capex. This isn't a parallel case, but is obviously explosive, as well as far beyond my usual subject. All the bells clearly have in capex a large figure that common sense would not put there, inflating profits significantly.

September 6, 2003

Q2 U.S. and Canada 740,000
Comcast's "DSL Killer" - 3 megabit service
2X, 8X, and higher speeds - T1E1.4 bonding DSL
MIMO, DSM will be faster yet
Huawei 1.4M port Q2 Giant sleeping no longer
China Netcom passes million subs
Alcatel 3.1M quarter
Bruce Kushnick's a player in D.C.
Editorial: The Poor Should Not Pay More

Briefs: Steve Levy, churn, KPN, Verizon bandwidth

*** Increase DSL revenues and reduce operational costs. The mPhase iPOTS" enables comprehensive loop management remotely, through the POTS Splitter. This truly modular solution allows service providers to deploy traditional POTS Splitters and when ready, add the intelligent splitter bypass function without interrupting the line. By being able to determine the true bandwidth potential of the loop, service providers will be able to confidently offer premium tiered services line by line. The mPhase iPOTS, bypasses the splitter impasse.
For more information visit our Web site, www.mphasetech.com or contact Peter Gasparini at sales@mphasetech.com. (ad)

John Cioffi raised the speed of his DSL line over 500%, giving hope to all of us who want faster net connections. At 22kft 24 gauge, his speed went from 64K to 768K. Some changes were old tech (he improved his home wiring), some from the toolkit of dynamic management (changed forward error correction parameters), and some the result of a new generation modem (Westell/TI) soon available to all. He's not done, with some more "easy" changes to current systems planned and MIMO and other steps still in the future. Coming soon is a computer program that takes inputs from lines, does processing, and makes recommendations on how to set the parameters in the modems. The recommendations of course change with line, binder situation,
and time. This requires active assistance from the provider, of course, and bravo to SBC for their assistance to Cioffi's testing.

In London, I can easily prove the regulatory attack on VOIP is hogwash, and often technically impossible to implement. I can hook up my Vonage phone through the conference wireless link, or use my laptop with a mike to make calls on FWD. The carrier has no way to tell whether I'm in London, home in New York, or visiting Jennie's dad in Denver.

The problem with regulating VOIP is that there's no good definition that doesn't include AOL Instant Messenger, Windows XP (which has a built in SIP phone), any cheap video conferencing system, and even the voice enabled multiplayer games coming on PlayStations and XBox. All enable cheap voice calls between people with computers, or a connection via a voice gateway anywhere in the world to actual phones. It's working great. Does Minnesota really expect Microsoft, AOL, and Creative Labs in Singapore (videoconference) to register as a phone company?

The biggest Euro telcos do the featured speeches in London, but the most interesting speaker there is Michael Boukobza of Free.fr, speaking with his switchmaker, Frederic Potter of Cirpack. They've rolled out free voice calls accross France to their 30 euro DSL customers, bringing Yahoo Japan style competition to Europe. I'm speaking on DSL Everywhere, also Tuesday at 3, and will be at the show all week. Please make sure to say hello to the round fellow with a beard, and give me a news story.

*** ZyXEL awarded PC Magazine Editor's Choice ZyAir Wireless Router "The ZyXEL ZyAIR B-2000 wireless broadband router is the most stylish, well-rounded, and potentially most secure product & and has earned our Editor's Choice for 802.11b category." PC Magazine, Aug. 2003. IEEE 802.1x wireless security with built-in authentication server / Dynamic WEP Key / Console Port http://www.usa.zyxel.com (ad)

Q2 U.S. and Canada 740,000
Wachovia points to interesting growth at Sprint and Alltel
Sprint added 38K to 223K, starting to catch up. Alltel added 17K to 105K, while Citizens is probably passing 100K as I write. SBC did 304K, BellSouth 103K, Verizon 101K, Bell Canada 81K. Qwest did a dismal 9K, less than Citizens and not much more than Century.
Q3 is usually slow, but Verizon's price cut and ad campaign should be kicking in, while SBC is keeping the $30 promotion. Cable had a surprisingly weak second quarter, so they are preparing their response. Rich Church provides two helpful charts with detailed numbers for all carriers, on the web at http://dslprime.com/News_Articles/news_articles.htm

Comcast's "DSL Killer" - 3 megabit service
Deadly promise
"DSL killer" was the chilling message on a cable slide at this week's KBRO Conference,as Comcast is actively upgrading cities to 3M and preparing to go faster. Brian Roberts of Comcast was clear they were ready to fight with price if he has to, but would prefer to fight against telcos by offering faster service and extraordinary features including 50% VOD coverage by the end of this year. Arris is supplying cablecos their even more frightening weapon - voice gateways already installed capable of supporting 17M customers.

Will customers go for the speed and better service? Jennie tells me "I'd switch to cable if they gave me three meg for a similar price," and she's the co-author of our book on DSL. She had just downloaded several multimegabyte patches for our wireless network, and was impatient even with our 1.5/768 service. Ivan Seidenberg and other bell CEOs know that they need a service that matches and beats cable; their staff is suffering a disconnect, considering choices that will make that impractical. In the short term, pricing 30% less may be a necessary defensive measure, as SBC's Stephenson is claiming. But that's prohibitively expensive over time. If telcos don't respond, only financial problems at cablecos will keep the game alive. That's plausible, because the cable guys have no profits and one day wall street may prefer EPS to EBITDA. More next issue.

2X, 8X, and higher speeds
T1E1.4 bonding DSL
"Bonding's the next big thing for DSL," Dr. Loop (George Hawley) tells me. Traditional linebonding methods, like T-1 IMA, are dependent upon all lines operating reliably at the same speeds. DSL lines, however, are generally rate-adaptive, dynamically pushing what Shannon will allow on every line, getting maximum speeds. The older standards all assume a static speed, and hence limit performance as line conditions vary. Operationally, that proves a disaster.

T1E1.4 is developing M2DSL (multi-megabit DSL), which specify bonding, physical layer, and power spectral density. One version, ATM+ is working already in Hawley's labs, and they've developed IMD which requires less overhead. The U.S. Bells and a slew of manufacturers are working to define the standard, looking to offer 5 & 10 meg bonded G.shdsl to business. M-squared DSL works for ADSL as well, potentially merging lines for video speeds. There are almost always extra pair available in a business setting, and surprisingly often for consumers as well.

Hawley's Valo is developing LoopStream, which bonds up to 8 pair, initially G.shdsl, for higher speeds. Hundreds of loops can be connected through a single system, with a centralized bonding engine, that works no matter what the access system is - including fiber.

MIMO, DSM will be faster yet
One home DSL line is now five times faster, because the user is participating in an early SBC test of DSM. That's on a single line, with transmission parameters adjusted based on tests of the entire bundle. Similar results aren't guaranteed in all situations, and in fact are impossible in some. But with DSM adding little to the cost of next generation systems, it's drawing lots of attention at several major telcos.

DSM (dynamic spectrum management) includes the application to DSL of the established engineering technique of MIMO (multiple input, multiple output) processing and other improvements. Similar work is promising to substantially increase the performance of 802.11 networking and mobile telephony. Noise is the limiting factor in most transmission systems, which until recently were built based on carefully estimated typical noise patterns. But in the realworld, noise is not random, and by determining the actual characteristics performance can be dramatically enhanced. "MIMO's the next breakthrough," Hawley predicts, "We will be able to compensate for the noise on the actual lines in use." John Cioffi's lab tests have inspired him to set a goal of 100 megabits per second, both directions.

Huawei 1.4M port Q2
Giant sleeping no longer
"The one competitor we think about most is Huawei," the lead at the largest DSLAM manufacturer told me almost two years ago. Everyone who has visited Shenzen raves about the extent and quality of their research, with over 10,000 R & D employees around the world. They have the sophistication to deliver OC-192 backbone switches, Class 5's, and full wireless networks. China Telecom demand has grown so rapidly the export effort in DSL has been sidetracked, but expect them to establish a major international presence in 2004. Meanwhile, China Telecom is literally ordering ports by the million. Digitimes reports modems will be supplied by Asustek and Acer/Ambit, the latter also a key Alcatel partner.

China Netcom passes million subs
Leng Rongquan, Vice President of China Netcom, expects over 1.4M DSL subscriptions by yearend, after beginning 2003 at 600,000, according to the press release by Huawei, who's just sold him a million lines of equipment. I have conflicting second hand estimates of DSL at China Telecom, which covers four times as many homes and probably passed 4M subscribers.

Alcatel 3.1M quarter
Worldwide strategies paying off
It's nearly a decade since Alcatel invested in designing their own DSL equipment for European TV deployment, failed to find a market, then won big with forward pricing in the U.S. Bells. Europe, especially France, Spain and Italy, are showing major growth, and Asia is remarkable. Alcatel's purchase of Shanghai Bell gave them the needed local presence and low cost manufacturing. Alcatel has always had a solid and well supported product, and they have generally read the market effectively when setting prices. Result: early leadership continues despite a dozen major competitors.

Jouni Forsman and Gauri Pavate of Gartner also report UTStarcom, Siemens, and NEC between 700,000 and 800,000 in Q1. The total of almost 10M ports is 20% more than the modem shipments Gartner identified. DSLAMs are often installed fully loaded with line cards, resulting in increased DSLAM capacity well before the subscriber count matches. Thomson, Siemens/Efficient, Acer/Ambit, Xyzel and NEC led modem sales, with none over 10% of the market.

Gartner reported the average supplier revenue per port was $84, which sounds right to me. I've recently confirmed several large bids at $50, but many sales are at older prices and reflect deals for lower quantities. When you add features (subscriber management, built in splitters and test, control software) you raise the unit price, as does the more limited market of OSMINE certified units bought by the U.S. bells.

Don't buy DSLAMs just on price, and recognize the figures circulated are never based on apples to apples comparisons. Dramatically low bids make the news, but often have special circumstances. No one talks on the record. For example, the Chunghwa bid a while back was the first I knew with modem + DSLAM totaling less than $100. Taiwan has four competitors, so Chunghwa was fighting for a low price. Nokia had targeted it as strategic, then Alcatel decided to block Nokia, whose sales worldwide were too low to sustain the Diamond Lane division. So they bid easily 30% less than they were charging for similar equipment in North America. Samsung, working with ADI, had enormous production capacity and a predictable slowing in the Korean market, so they decided to win the contract anyway. They calculated just how low Alcatel could go for the million ports, and set their bid beneath that and won. I remember a wall street analyst pointing to the large "win" as a reason to invest in ADI/Aware - but I'm not sure if there was any profit on the deal. I've enough sources confirming that story to be confident, but no one willing to put their name on the record.

Bruce Kushnick's a player in D.C.
10 pages of the Triennial
America's most vehement critic of the Bell's, Bruce Kushnick, had a major impact on the FCC decision, reflected in the intensity of analysis he provoked. Bruce, with support from SBA and others, had filed that the Triennial was not considering the effect on small business, especially ISPs and data CLECs. He's absolutely right, the FCC comments in rebuttal miss the point, and he's got a strong case on the law if he brings it.

The Triennial deliberately creates a duopoly in high speed data, because Verizon, Intel and TIA convinced Kevin Martin that was the only way they would build fiber. The argument was they would not have the volume if more competitors were in the market, so no one would build in a shared environment and the consumer would be worse off. Whether that's true or not (I disagree), it clearly sets dramatically high barriers to entry and impairs access to essential facilities for ISPs. The clear result is that data carriers, whether regional ISPs, aspiring DSL vendors, or even giants like Earthlink or AOL, are now in life threatening circumstances, as the telcos and cablecos claim the market.

Twice, I found myself on a panel defending telcos against Bruce, unlikely as that may sound to anyone who knows how often I criticize them. Kushnick, a friend, is even more of an iconoclast than I am, and is roundly hated at the Bells. But the FCC, Telephony, and others take very seriously because he is often right on his facts. Unlike most commentators, he has no conflict of interest except an a passion for the rights of consumers.

Editorial: The Poor Should Not Pay More
If you make enough money to afford a fancy phone package, DSL costs less. If you are poor or prudent and just buy basic service, SBC wants $50. To anyone who cares about broadband for all, this is a rip-off. Bundling to save consumers money makes sense, but the cost saving is about the same whether no matter which phone service is included. The savings should correspond, and be available to basic phone customers as well.

Mike Powell says he believes "The country needs the deployment of broadband to all Americans at affordable rates." Let's see if he does anything about this. At minimum, he can work with the FTC on truth in advertising, and prohibit a headline that offers a low price only when the customer buys an expensive phone package. If most buyers don't qualify, that's simply fraud, old fashioned bait and switch.

Groups who profess to care about broadband should speak up. Sylvia Rosenthal and Matt Bennett of APT, Matt Flanigan of TIA, Gary Shapiro of CEA - where are you now? Affordable prices for all will do more to advance deployment than any of the "incentives" and "subsidies" you've been campaigning for.

Briefs:

Steve Levy, key wall street analyst, is at Lehman Brothers, not GS. Apologies - especially because now I'm getting from Lehman some very helpful reports.

On churn: "They have raised my rate to 49.95 so I switched to cable"

KPN in Holland reached 513K subscribers, and raised the yearend goal to 675K. Ambitious, but they are cutting prices to 20 euro. Versatel and Tiscali are offering similar, according to Telecom.paper.

"We are seeing customers demanding more and more bandwidth." Greg Evans, Verizon

August 31 ,2003

• Here comes VDSL2
• Aiming for fast standard by 2004
• Yahoo BB adds 116 games, 24 mbps ADSL2+
• Taizo Son, younger brother, promises faster access
• Covad's chance for cash
• Quick offering can guarantee their future
• Definitely "line-splitting" when a competitor does it
• Small pizza box DSLAMs suddenly a hot market for remotes
• Video is here
• Some skeptical warnings
• Our Prismiq set top is working well
• Hanaro in trouble
• Shareholders squabble, financing in doubt, AIG in lead
• Plain speaking from the editor
• The ISPs suing SBC have a strong case.
• The bell attack on the Triennial UNE-P is dependent on a lucky choice of judge