How We See CE

When I first talked in this blog about my views on the future of the networked Consumer Electronics (CE) industry, I tried to share my optimism about how networked CE products were rapidly becoming "must have" products even for mainstream users. The demand for IP-enabled devices like gaming consoles, networked DVRs, network attached storage, VoIP adapters and digital media adapters is rapidly increasing, and according to Mark Kirstein from Multimedia Intelligence, by the time my Tivo-addicted son becomes 8-year old, the market for network interface semiconductors (including ethernet, wireless, powerline and coaxial networks) will grow to nearly $2.5 billion.

Given that most of the necessary technologies are already available, this growth trend only requires two key elements to continue: consumer demand and interoperability standards. 2008 will probably not be a great year regarding consumer demand, as most people are more worried about paying their mortgages than getting new TVs, but I'm confident it will shortly be back back to normal levels. The other side of the equation - interoperability standards - is a different story, at least for the powerline networking industry.

A comment to a previous entry in this blog asked about the opinion of my employer regarding the IEEE P1901 effort. Although I responded there briefly, I think a longer answer here would be appropriate. I'll spare the complex details of the IEEE P1901 working group structure (the actual work is divided into something called "clusters", one for creating a proposal for "access applications", another for "in-home applications", and another "coexistence mechanisms" between access and in-home systems) and I'll talk specifically about what EDN readers concerned about networked CE are probably most interested in: the proposal for "in-home" applications (the same applies to the "access proposal").

The "in-home cluster" proposal selected by the IEEE P1901 WG on October 2007 is very peculiar and will likely look strange to any engineer familiar with IEEE standards. The document is the perfect example of how not to write a specification: it includes two completely incompatible specifications for the specification“Physical Layer” (PHY) and for the “Data Link Layer”. One of the specifications is based on OFDM modulation while the other is based on Wavelet modulation. Vendors are not required to implement both options (and I predict that nobody will). The modulation schemes are mutually incompatible, so IEEE P1901 products based on OFDM modulation will not interoperate with products based on Wavelet modulation.

Let me repeat that again, just in case you have not fully understood the implications: "IEEE P1901 products based on OFDM modulation will not interoperate with products based on Wavelet modulation". So much for standards! The current specification not only includes two incompatible modulation schemes, it also includes two completely different Forward Error Correction (FEC) schemes - one based on “Turbo Codes” and another based on “Low-Density Parity-Check Codes” (LDPC). Other unnecessarily duplicated elements of the specification include functionality traditionally found at or above the Data Link Layer, such as mechanisms for Quality of Service (QoS) management, Security, MAC Frame Format, TDMA/CSMA time-slot allocation or Multiple Network Operation.

What does this mean for a consumer? When a consumer buys a TV with embedded powerline technology based on Wavelet P1901 and gets home to find that it does not interoperate with her powerline-enabled router based on OFDM P1901, he or she is going to feel frustrated and disappointed. Either the TV, the router, or both will be returned, which will be a problem for both retailers and service providers. This kind of dual-PHY standard also creates problems for device manufacturers and semiconductor companies. If a semiconductor vendor decides to implement a P1901-compliant chip, it will have three options:

• Implement the OFDM version (and thus be incompatible with the other half of the market that is based on Wavelet)
• Implement the Wavelet version (and thus be incompatible with the other half of the market that is based on OFDM)
• Implement both versions of the standard (and thus have an expensive and non-competitive chip that requires twice the silicon area and twice the effort.)

The problems associated with this approach are very well described in this excellent interview with Intellon's Mark Hazen, where he explains why Intellon's HomePlug AV chips are not interoperable with the company's own HomePlug 1.0 chips:

Though it is possible to build a backward-interoperable device, the complexity would be prohibitively expensive to our customers. Such a device would have to include the ability to switch between FEC techniques and many other features, which places cost at an unworkable level.

Here, he was talking about implementing interoperability of two versions of the same specification. Imagine if you had to include 2 different modulation schemes, 2 different FEC schemes, etc. "Prohibitively expensive" is a good description of the problems faced here.

None of the 3 options mentioned above is very attractive and all of them represent too much risk. Most likely, any semiconductor vendor that considers this will decide not to implement a "two-PHY/two-MAC" spec, and will instead focus its resources in less risky markets. Unlike other well-established IEEE standards such as 802.11 that have progressed faster, the current “two-PHY/two-MAC” proposal at P1901, has been unsuccessful at getting a confirmation vote at any of the recent P1901 meetings in October 2007 (Boston, MA), December 2007 (San Diego, CA) or March 2008 (Fukuoka, Japan). Confirmation votes require 75% of approval by the Working Group and are designed to ensure that successful proposals enjoy significant consensus by the WG members.

So, getting back to the question from the reader: what is my employer's position on this issue? We still think the problems of P1901 can be fixed. We are convinced that a standard should have a single PHY and a single MAC. We think the problems in the current proposal can still be removed if common sense and basic engineering principles are used to address the problem. We will continue to push for full interoperability at IEEE P1901 for the benefit of all industry stakeholders, including consumers, service providers, device manufacturers and silicon vendors. We are committed to work with other members of IEEE P1901 WG to address the shortcomings of the current proposal and to ensure that a standard with a single PHY and a single MAC is developed at IEEE.

Otherwise, the powerline industry will not get a significant part of the $2.5 billion pie that we have been promised for 2012.

—Chano Gomez, DS2