A cellular to WLAN roaming specification (dubbed UMA - Unlicensed Mobile Access) was published last week by a group of 14 noteworthy vendors and wireless operators (Alcatel, AT&T, BT, Cingular, Ericsson, Kineto Wireless, Motorola, Nokia, Nortel Networks, O2, Rogers Wireless, Siemens, Sony Ericsson, and T-Mobile USA).  This represents a first important milestone in establishing seamless roaming between the 2.5G/3G network and an 802.11 WLAN.  The key goal is for subscribers to be able to rely on the ABC  (Always Best Connected) model, which should enable their handsets to always attempt first to seek out an 802.11 connection (be it from an enterprise/home WLAN or a public WiFi hotspots) before attempting to connect to a carrier's wireless network.  Of course, from a consumer viewpoint, the goal is savings: access via a WLAN should typically be cheaper (or sometimes even "free") than via the mobile network.  But from the vendors and carriers' perspective, this new standard should pave the way for the inclusion of WiFi in most cellular packages and support for seamless roaming in future handsets.

As far as pricing models go, I expect that the wireless operators will initially bundle the WiFi minutes, albeit at a cheaper rate.  Therefore, a typical monthly package will have X minutes worth of GPRS/CDMA access and Y minutes worth of WiFi access (where Y>X).  But eventually, competitive pressures will probably drive some service providers to offer unlimited WiFi access for an extra monthly flat fee (which could cost anywhere from $5-7 to $15-20).

   more »

The existence of bandwidth-intensive applications is one of the key drivers for fiber to the premise, be it for consumer (FTTH - Fiber To The Home) or business (FTTD - Fiber To The Desktop) applications. Bandwidth demand varies by application and customer type. At the low end, a couple of G.729 calls entail only roughly 48 kbps (assuming 8 kbps of data and 16 kbps of overhead, or a total of 24 kbps for each). At the high end, a consumer application such as HDTV can chew up 19.2 Mbps worth of bandwidth.

One of the key determinants for both the business and consumer market is the cost of the “endpoints” (i.e. connectors and the electronics). Currently, there is a significant pricing differential between fiber and copper for connectivity items such as NICs.  As the price gap becomes narrower, fiber gets to be more attractive.  This NWFusion article, for instance, shows the trend: Foundry was the first to ship a pre-standard 10 Gig port on its BigIron switch in 2001 at a price tag of roughly $80,000 for the module and the optics.  By contrast, nowadays, vendors such as Cisco, Extreme, Force10 and Foundry have products at the $4,000 to $7,000 price range for a fiber-based Gig Ethernet port.    

The future development of new bandwidth intensive applications can also favor the fiber, particularly once the bandwidth demand reaches a point where copper can no longer meet it.  That said where is the "killer app"?  Probably not out there yet.  But still, there are enterprises that are deploying fiber as a means of future proofing their networks.  Similarly, fiber is being deployed in new high-end residential projects (with the cost typically being passed on to the owner) and the retrofit market (fiber is being rolled out in homes being remodeled and also passed on to the owner).    

But a true disruption can happen with the advent of new copper specifications. The emergence of the 10 Gigabit copper standard (please be sure to check the IEEE P802.3 10GBASE-T Study Group Public Area Index) can pose a threat to the fiber optic cable market, because it will push the bandwidth threshold further up, thereby taking away the “future proofing” case for fiber deployment. 

One company to watch for in this space is SolarFlare, an Irvine (CA) startup that is backed by VCs such as Sequoia Capital, Foundation Capital and Intel Capital (Anthem Partners and Intel came on board on the second $17.5 million round secured by the company in early 2003).  SolarFlare is trying to replicate the success that Level One had in the 10BASE-T segment or Broadcom and Marvell had in the 100BASE-T space.  The company has a strong IP (Intellectual Property) portfolio, consisting of over 9 patents in process (and counting). 

SolarFlare believes it can solve the limits of the "conventional wisdom" Shannon model and achieve a 10GBASE-T copper solution via its unique approach.  The algorithms used in the chipset include sophisticated media-optimized signal processing and customized DSP engines.  These algols effectively reduce the noise levels, and therefore increase the channel capacity. 

---

Note: The noise level reductions in are in both near and far-end crosstalk (crosstalk is the noise or interference caused by electromagnetic coupling from one signal path to another). The near-end crosstalk is measured at the end from which the disturbing signal is transmitted, whereas the far-end crosstalk is measured at the opposite end from which the disturbing signal is transmitted.   more »