• Founded nBand with Krishna Yarlagadda, Sanjay Vishin, and Srinivas Lingham in early ’98. I was VP Systems Engineering, but then assumed the CEO role in March ’99
• The team developed a Software Defined Radio (SDR) solution to broadband wireless communications market needs via the development of a communications processor called nFlex. nFlex was based on a standard RISC processor that was augmented with vector extensions optimized for supporting modern OFDM baseband modulation techniques, and included hardware special function accelerators optimized for communications tasks.
• Engaged RedHat to build the compiler and tools chain for nFlex.
• Raised a Series A round in August ’99 of ~$5M from Angels and Axis Communications. Strategic partnership agreement executed with Axis.
• Raised a Series B round of $14M in Sept 2000 led by Vantage Point Venture Partners with participation by Axis Communications and Sunrise Capital fund.
• Built team of 50
• In June 2001 secured strategic partnership with Nokia Rooftop Broadband Wireless business unit and $2M equity investment.
• Signed MOUs with Spike Broadband and Piping Hot Networks. Engaged with a variety of other potential customers.
• I presented at DSP World 2000, Embedded Systems Conference, Commverge, Network Outlook 2001, DesignCon 2001, BWIF 2001, Robertson Stephens Semiconductor Conference 2001.
• Developed high quality Middle UNII band 5.25Ghz – 5.35Ghz) radio in partnership with a third party (Noel Marshall & Assoc)
• Developed baseband for 802.11a. BWIF (Broadband Wireless Internet Forum) baseband port was in progress.
• Developed MAC for 802.11a. BWIF MAC port was in progress.
• Partnered with CISCO to obtain source code to CISCO BWIF chip (Fireball) to port to nBand’s nFlex communications processor
• Acquired DOCSIS MAC from Pacific Broadband in Q2 2001 to provide turnkey product for BWIF and residential gateway products. The MAC and Phy were ported to nFlex.

nBand was a fabless semiconductor company that developed a communications processor (called nFlex) capable of software implementations of a variety of communications standards including 802.11x, xDSL, WiMAX, and WCDMA.  The company’s mission was to be the dominant supplier of programmable baseband and Media Access Control (MAC) technology to broadband communications equipment manufacturers.  Later in the company’s history, the go-to-market focus narrowed to broadband wireless (specifically Wireless LAN and last mile Fixed Broadband Wireless Access), for which nFlex could be characterized as a Software Defined Radio (SDR) processor.   Architectural emphasis was placed on support for OFDM (Orthogonal Frequency Domain Multiplexing), a modulation scheme being increasingly deployed in new standards such as 802.11a, ADSL, WIMAX, DAB, and DVB (digital TV).  The product would be the processor along with software that implemented the required communications protocols and baseband modulation.

nFlex offered the following values to the customer:

1. Faster time-to-market than ASIC approaches (1/3 the time)
2. Graceful support for changing standards or standards updates with “over the air” updates
3. Increased product longevity (via software updates/changes) aka “Future Proofing” the product
4. Reduction of product risk (using software patches)
5. Better management of complex standards and better support for multi-function, multi-protocol requirements (such as the trend then to move from standalone modems to multi-function gateways)
6. Allowing customers to differentiate and innovate
7. Reducing development and systems cost

The executive summary state that:  “nBand delivers a low-cost, low-power, single-chip solution which can run multiple broadband protocols (MAC and baseband) at full speed, simultaneously — and is fully programmable in C or MATLAB”.

nFlex implemented an embedded communications supercomputer – with the objective of bridging any analog front end, be it a radio or a twisted pair DAA, to the network.  nFlex was a fusion of vector DSP, RISC processor, communications focused dedicated streaming Special Function Accelerators, and on-chip DRAM that was conceptually portrayed via the diagram below:

nBand was founded in 1998 by Krishna Yarlagadda and was originally named Avaj.  Krishna assembled myself, Sanjay Vishin, and Srinivas Lingam to round out the founding team, and he funded much of early operations “out of pocket”.  However,  the team had a falling out with Krishna, and in March ’99 he left the company and I assumed the role of CEO and VP of Systems Engineering.  Shortly thereafter I renamed the company nBand Communications.

nFlex was implemented as a vector extension to a standard RISC processor that was integrated with a cache-coherent streaming Special Function Accelerator architecture to accelerate those low level communication or error correction functions that could be much more efficiently implemented in dedicated hardware.  The company made a wise decision to outsource the compiler and tool chain development to RedHat so that Wave could concentrate on building the processor and delivering baseband modulation firmware.  That relationship worked out well.  A high level block diagram of the chip is shown below.

During most of mid ’99 we focused on building the team and raising an A round.    Greg Blanck joined as VP Engineering, and in early 2000 Dmitri Rubin joined as Director Wireless LAN to manage the baseband firmware development under me as acting VP Systems Engineering.  Kent Kernahan joined as VP Marketing.   In general staffing during the ’98, ’99, and 2000 was a huge challenge as those were boom years and talent was both scarce and expensive.

During ’98 and early ’99 we pursued business opportunities with 3G cellular infrastructure suppliers such as Nortel and Ericsson as a cell site modem.  We flew to Sweden to meet with Ericsson – but ultimately we did not pursue that market as it proved glacial and difficult to penetrate.   Qualifying onto Ericsson’s AVL (Approved Vendor List) at that time was a difficult and opaque process.  The cellular infrastructure business is standards driven and involves massive deployment of new standards.  It did not matter how good your technology was – if you did not have a seat on the buildout bus, when the bus left the station if you were not on board it was just too bad.

In August of ’99 nBand closed an A round of roughly $4.75M from about a dozen Angels and from Axis Communications.  Many of the Angels were sourced from a “high net worth” individuals list provided by our lawyers, Tom Chaffin and Jorge Del Calvo of Pillsbury Madison & Sutro.

In Sept 2000, the company announced a $14M Series B led by a $9M investment by Vantage Point Venture Partners (VPVP) with participation by Axis Communications and Sunrise Capital Fund (Jack Balletto).  Matt Ocko was the lead from VPVP who was a passionate high energy board member who “got” hardware and systems.  I remember many of our cellphone conversations went well past midnight.   The guy never seemed to sleep, but he did fall asleep during one or two our board meetings 🙂   Matt is currently a founder and managing partner at Data Collective VC.

In June 2001, Nokia made an equity investment of $2M.  Nokia had recently (Sept ’99) acquired Rooftop Communications, and was developing a rooftop based wireless broadband distribution network product/business for which they wished nFlex to be the baseband processor.

In late 2000 we were running out of space at our location on Oakmead in Sunnyvale, so we scouted a new location with about 3x the square footage at 161 E Evelyn Ave in Mountain View, to which we moved in Feb 2001.  The new building need significant Tenant Improvements, and when I voiced my concerns about the additional cost (roughly $400K), I was told by one of the board members – “Pete, if you are going to hit the wall, we want you to hit the wall going 90 miles an hour”.   I mention this not as an indictment of the board, but rather to give the reader a indication of the general ethos of the times.  This profligate burn rate approach was a very different set of operational priorities from what I was to experience 10 years later as CEO of Wave Computing.

nBand actively presented at conferences during 2000 and 2001, including presentations at:

1. DSP World 2000
2. Hot Chips 2001
3. Embedded System Conference
4. Commverge
5. Comdex
6. Network Outlook 2001
7. DesignCon 2001
8. BWIF (Broadband Wireless Internet Forum) 2001
9. Robertson Stevens Semiconductor Conference 2001

With the exception of the Hot Chips presentation, which was delivered by our Chief Architect, Sanjay Vishin, I gave the majority, if not all, of the other presentations.  The last nBand presentation I gave was to have been at the Lehman Brothers Semiconductor Conference on 9/11/2001.  I was driving up to San Francisco to present at that conference when I received a call from the conference organizers notifying me that the conference had been cancelled – this is how I learned of the terrorist acts at the World Trade Center.

The company saw good business development progress during 2000 and early 2001.  By mid 2001 Wave had signed development agreements with both Axis Communications and Nokia (as well as equity investments from both).  We had signed MOUs with Spike Broadband and Piping Hot Networks.  Hughes Network Systems was allocating budget for two engineers to start coding for nFlex.  Netro was interested in using nFlex for a BWIF (Broadband Wireless Internet Forum) market product.  We had a signed JMA (Joint Marketing Agreement) with rTrust (a developer of elliptical encryption software).  An MOU was under negotiation with Radia for an RF partnership with similar discussions underway with RF-Magic.  An MOU was expected with Cal-Amp.  We had also been working fairly closely with Cisco for some time on BWIF related products.   Cisco was a major driver of the BWIF initiative which was created to facilitate cost-effective, broadband wireless access solutions based on Vector Orthogonal Frequency Division Multiplexing modulation (VOFDM), which was a technology CISCO obtained through their acquisition of Clarity Wireless.  Cisco gave us the source code to a broadband wireless subscriber ASIC, code named Fireball, that they had been developing for the BWIF market -with the goal of developing a port to nFlex.  For  support of the BWIF  opportunity with CISCO as well as other opportunities (such as residential gateway products) Wave needed access to a DOCSIS MAC, and after a thorough search we acquired a DOCSIS MAC and PHY from Pacific Broadband in Q2 2001.

In order to fully develop the firmware for Wave’s software defined radio we required a high quality radio in the middle UNII (5.25Ghz – 5.35Ghz) band.    802.11a and BWIF could both be supported by a radio in this band.  Commercially available 5 Ghz radios were not readily available at that time that met nBand’s engineering requirements, so in Nov 2000 the company retained Noel Marshall and Associates to develop a radio to our specifications.  That radio was delivered in Q2 2001 and a photo is shown below.  The radio was also part of a development kit that Wave was making available to third parties (such as Hughes Network Systems) wishing to to develop on nFlex.  Some of the radio specs can be seen in the development kit brief that is part of the slideshow below.

Progress on the nFlex design was proceeding steadily, but the July 1 2001 tapout date to TSMC (in a 0.18 micron process) was slipped one quarter to Oct 1.  The biggest issue facing the team was that the decision had been made to place and route the entire chip flat, rather than use a hierarchical approach.  This turned out to be a mistake as the tools and the compute infrastructure were not really up to the task.   Dmitri and the firmware team also needed another quarter to finish their work as well.  The architecture group and most of the design team had already moved on to the next generation part (nFlex2).

In June 2001 Neal Margulis also joined the board.  I had worked with Neal previously on a startup called Travertine Systems when I was an EIR (Entrepreneur in Residence) at Benchmark, and Neal had just finished a tour as President and CEO of iCompression.  In June the future looked bright,  nBand was going “90 miles an hour”, and no one thought the company would cease operations 6 months later.

The company headcount peaked in the low 50’s, plus a half-dozen local consultants.  Almost all heads were in engineering, split between the chip team (the largest group) and systems engineering (which included MAC level software, baseband firmware, board design, and radio design management).

By mid 2001 nBand had amassed a technology and product offering summarized below as part of a presentation to Nokia.

By mid 2001 the Telecom Crash was well underway (which culminated in the bankruptcy filing of WorldCom in mid 2002), and the Crash plus the terrorist attack on 9/11 and the chill that followed was the death knell for nBand, as the company needed to close a C round to fund ongoing operations, fab the chip, and take the product to market. Unfortunately, with astonishing speed all of our partners, customers and business opportunities vaporized. Axis’ stock price cratered to $2, the Rooftop Wireless Business Unit at Nokia went away, and Cisco abandoned BWIF and VOFDM.

Unable to close a C round, Wave ceased operations in Dec 2001. During early 2002 I worked to sell the assets, and in March 2002 Wave’s assets were sold to Proxim and the two other founders, Sanjay Vishin and Srinivas Lingham, went to Proxim to transfer the IP. Proxim, along with Intel were to champion WiMax (802.16), and nFex was a great match to implement the WiMax baseband, but WiMax itself was to lose out to LTE in the 4G standards battle.

In early Q3 2001 when it was becoming clear that fundraising was not going well, the board urged me to jump into what was considered one of the few lifeboats available, which was consumer 802.11a WiFi. However, as I pointed out to the board, Wave had, from a go-to-market standpoint, been targeting higher end broadband wireless systems with Nokia, Cisco, and Axis, and without integrated CMOS RF we would not be competitive in the consumer 802.11a market against players like Atheros. We had already begun negotiations for single chip CMOS 5Ghz radios with RF providers such as RF Solutions, Radia, RF Magic, and Raytheon with a goal of a demo with nFlex by mid 2002, but such a solution would be a separate chip costing nBand (initially) ~$20.

After some consulting (see: COMMUNICATIONS-VENTURES-AND-LEOPARD-LOGIC), I was to dive into a different industry, thanks I believe to a recommendation by Bill Unger of Mayfield Fund, by accepting a VP Engineering role at Predicant Biosciences, a cancer diagnostic biotechnology startup that used patterns of proteins in blood as a diagnostic biomarkers, see: PREDICANT-BIOSCIENCES

SOME TAKEAWAYS:

1. Although tempting, don’t expend precious mindshare stacking nickels with other startups. Although nBand’s primary focus was on its big customers and partners, Nokia, Axis Communications, and Cisco – we let ourselves be distracted with biz dev for startups such as Spike Broadband and Piping Hot Networks.
2. A CEO should have a deliberate ongoing process to have a “Plan B”. Use your board. Although you may feel you are over constrained by capital limitations or other factors, there are usually “out of box” ideas that can be unearthed with a deliberate process.

nBand Slide Show:

1. nFlex Die Plot
2. nBand Middle UNII Band (5.25-5.35 Ghz) Radio
3. nFlex Developer Kit Product Brief
4. RedHat nFlex Toolchain Guide
5. nFlex Evaluation PCI Board
6. nFlex Technology Convergence Venn Diagram
7. Why Software Defined Radio (SDR)
8. nFlex Software Development Diagram
9. FIR Filter Matlab to Vector C Conversion Example