High-tech

Due Diligence: Phyworks

By The Chilli analysts

Introduction

Phyworks was founded in 2001 by CTO Nick Weiner (formerly of Microcosm, Motorola) and CEO Stephen King (formerly VP of sales at Microcosm). Phyworks is a post Chilli R2 fabless semiconductor startup, based in Bristol, UK. The company is specifically targeting 10Gbps optical modules for networking and storage communications applications. It has fully functional products available now.

Vital statistics

Value proposition

The Chilli perspective

Vital statistics

Additional management team members includes Brian Williams as VP of manufacturing, Paul Denny as director of design and Chris Bryson as director of marketing; all the management team had worked together at their previous gig at Microcosm, another startup that was sold to Conexant for £128 million. [Editor's note: Out of interest, at the time Microcosm was sold, 40% of the equity was held by the founders and employees while the VCs held 40%. The team had an 18 month earn out period which lasted until May 2001.Some of the VCs which received great returns for the Microcosm deal included Pond, Vertex and 3i).

Stan Boland (CEO of Icera Semiconductor, formerly CEO of Element-14, sold to Broadcom) is an adviser to the company.

The two principal founders and seed investor Gary Steele (founder of Microcosm) invested their own seed capital of $1.2m in July 2001. The company went through a roller coaster time trying to raise its R1 round as they began fund raising just before September 11. R1 funding of approximately $9.8m came from Add Partners, Atlas Venture, Prelude Ventures and friends & family in February 2002. The R1 funding allowed the company to complete the second iteration of their electronic dispersion compensation (EDC) chip and tape out their second forward error correction (FEC) chip. [Editor's note: in the semiconductor industry, it is quite common for the chip to go through at least two iterations before fully meeting all the latest functional specification].

The company began raising R2 in January 2003. Eventually it managed to raise $13.3m and closed in September 2003. Participating in this round was Advent Ventures in addition to R1 investors Add Partners, Atlas Venture and Prelude Ventures. King expects that this R2 round will allow Phyworks to reach breakeven position.

Phyworks currently has a headcount of 35, and is expected to grow this to 55-60 by the end of 2005. The company is looking to strengthen its analogue design team with additional experienced designers and will later invest in sales and marketing infrastructure.

King divides his time between the Bristol HQ and Oregon sales office, with one business development specialist staffing an office in Japan. Repping (sales agents) houses will be used where language/business culture is a challenge. The executive team has previous experience (from Microcosm) of doing business in Taiwan and China, but according to King, 'Japan is going for 10Gbps now, but Taiwan will be a little further down the road at this point.'

Prospective customers for the EDC chip include the top five optical module makers like Agilent, Infineon, E2O, Intel and Finisar. King claims to have a firm order for its forthcoming laser driver chip.

Value proposition

Optical communication involves the transmission of data as light pulses along a glass fibre. It has a number of advantages, including high bandwidth (data rate). The high bandwidth of optical communications makes it suitable for a number of applications, including:

• Storage area networking (SAN): a high-speed network interconnecting storage devices with servers on behalf of a larger network of users
• Local area networking (LAN): a network in a small geographic area, e.g. a building
• Metropolitan area networking (MAN): a network interconnecting multiple smaller networks in a geographic area, e.g. a city
• Wide area networking (WAN): a geographically dispersed telecommunications network, which may use telephone lines, fibre optic cable or satellite links

Of course, cost must be traded against performance, so optical communications are currently used as a high-speed 'ring road', that slower copper based networks can feed into and out of.

Electrical data is converted to and from optical light pulses by an optical module. A typical optical module includes the following components and integrated circuits (ICs):

• Transmission portion
  • Laser component: to transmit light. Laser type depends on the transmission distance required
  • Laser driver IC
• Receiver portion
  • Light detector component: to receive light
  • Transimpedance amplifier (TIA) IC: converts output from light detector into a voltage for the limiting amplifier
  • Limiting amplifier IC: converts the voltage output of the TIA into digital logic levels for the outside electronic world

Optical modules come in two basic variants:

• Serial electrical interface: data moves in and out of the module over one data link, giving advantages in a smaller module size.
• Parallel electrical interface: data moves in and out of the module over multiple data links, allowing data to move in/out much faster, at the expense of greater module size, affecting packaging density of the overall system of line cards (which the module resides on) and cost

SAN and WAN/telecommunications applications demand higher bandwidth, moving from 2.5Gbps to 10Gbps. However, there are economic and engineering challenges. Economically, the cost of upgrading fibre infrastructure would be prohibitive; the ideal solution would be to change the optical modules whilst maintaining the existing infrastructure. Technically, the challenges are to:

• Maintain a reasonable bill of material (BOM) cost for the 10Gbps module, ideally aided by using a serial electrical, but solving the problem of routing 10Gbps serial data from the system, through the module, across the circuit board and through connectors
• Run 10Gbps data over existing 1 or 2Gbps infrastructure at a reasonable distance whilst tackling the interference that reduces the transmission distance.

Phyworks has addressed these challenges by developing a range of ICs, including one for electronic dispersion compensation (EDC) and one for forward error correction (FEC).

The EDC chip allows operation at greater transmission distances by reducing an effect known as 'chromatic dispersion' that causes inter symbol interference - essentially, signal distortion caused by digital pulses overlapping with another, confusing the receiver.

The use of FEC improves the optical signal-to-noise ratio (OSNR), a measure of signal strength in relation to background noise. Phyworks claims its FEC solution provides a greater improvement in OSNR at a lower overhead than competing solutions. Weiner had spent some time researching FEC solutions with Dr. Peter Sweeney at Surrey University and the design team had spent twelve months researching customer requirements.

Phyworks has filed ten patents on its technology. The solutions allow greater transmission distances to be achieved, but can also allow cheaper components, e.g. laser, to be used for short distances. In order to help reduce the overall BOM cost, Phyworks made use of standard digital CMOS process and planned a number of integration strategies for its devices.

The Chilli perspective

The telecoms market has been hit very badly during the last four years but is starting to stage a modest recovery. Unlike many optical vendors who didn't have a plan B, Phyworks has focused its 10Gbps solution on the storage area networking (SAN) and metro networks market. According to a recent study by iSuppli, the SAN market has been growing by 30% annually and looks set to continue, especially as legislation for better record keeping in healthcare and commerce becomes law.

Network infrastructure is measured in terms of total cost of ownership (TCO), including deployment costs, maintenance. In order to proliferate 10Gbps swiftly, it is mandatory to keep the TCO manageable by reusing as much of the existing infrastructure as possible. In terms of the module itself, a 1Gbps optical module costs around $30; a 10Gbps optical module comes in at $2000. Optical module makers are demanding 10Gbps modules in the low hundreds of dollars.

Phyworks' integration possibilities for their technology, and use of lower cost standard CMOS processes pave the way for a reduction in overall chip count and power consumption, thereby lowering the BOM cost and reducing overall TCO at the 10Gbps level.

It is quite a remarkable achievement that the company has managed to develop two complex ICs (EDC and FEC) at under $25M. The company has also just taped-out (i.e. a design ready for fabrication) a laser driver IC, increasing the possibility of a lower BOM This is due to the team's previous experience of laser driver development. The company is courting the top optical module makers and claims firm orders for some of its devices, including the laser driver IC, putting the company in a strong position to generate revenue traction during 2004.

Phyworks' competitors include BigBear Networks, Broadcom, AMCC and Vitesse. BigBear Networks is a Chilli R3 startup based in California, founded in 2000 with fifty staff (down from seventy). The company has raised $78M, claiming an ESD solution. This is a different tack to Phyworks, which has methodically increased its share of the optical module BOM by developing several devices on a fraction of BigBear's funding. AMCC offers several parts for 10Gbps modules, including FEC and dispersion compensation. Broadcom offer a number of solutions, including FEC. Vitesse offers FEC solutions as well as clock and data recovery (CDR) ICs for 10Gbps. Phyworks has to demonstrate its competitiveness in terms of overall BOM reduction; a task made easier with the launch of its laser driver IC.

In terms of outlook, Phyworks' executive team is seasoned, with relevant domain expertise, and a track record of success in this market. If the products meet the claim of enabling the 10Gbps market, then the company will be an attractive prospect for an exit, via acquisition or IPO within the next three years. The challenge now is to make the products fly from inventory, requiring investment to be steered from development towards commercial infrastructure, in terms of marketing, sales and customer support in its target geographies.

Comments on this story? Send an e-mail to editor@thechilli.com

© Chilli Publishing Ltd 2004

13JAN2004