Optical switching, like the United Nations, is arguably a somewhat premature and perhaps troublesome construct and yet one to which a lot of thought is given by the great and the good. What is clearly needed, in both cases, is a means of bridging a series of seemingly incompatible areas. Though it remains something of a cult item in the networking fraternity (it is currently more likely to be employed in such areas as large-area projection display applications), Texas Instruments' Digital Micromirror Device (DMD) has many of the characteristics of 'the next big thing' in an industry desperately eager for fresh blood.

And while all agree that the product is reliable (TI has even gone to the trouble of producing a document entitled "Why is the Texas Instruments Digital Micromirror Device so Reliable?"), the likely push into next-generation optical switching applications creates doubt and optimism in almost equal measure.

However, it is also arguably the case that the device offers more complications than many are willing to countenance at this time. At a technical level, the individual mirrors in the DMD operate in a binary fashion with two stable and well controlled tilt angles as well as an intermediate state that is - according to some accounts at least - not well controlled.

One can use a set of mirrors in one of the stable states to efficiently couple light from an input fibre to an output fibre, switching some or all of those mirrors to the other stable state to provide attenuation of the light coupled to the output fibre. Collection of the light from the mirrors in the second state and the subsequent efficient coupling to another output fibre, is somewhat more challenging.

Because TI's Micromirror product is a 2D device it is, accordingly, not easy to convert into a cross-connect. For an NxN cross-connect it is desirable to have a basic 1xN switching functionality with 2N of those giving an NxN switch. Other criticisms leveled at the product are that MEMs-based 1xN switches need complicated controls and that the TI mirrors are smaller than desirable for a switching application.

And while it is clear that TI has the expertise to develop larger-size multi-position tilting mirrors, this would involve many design challenges involving significant new costs, which in the present state of the market might not be easy to recoup particularly as such a radical redesign would almost certainly mean sacrificing a proportion of the reliability and cost advantages that make TI's current product - optimised for its current binary projection application over 15-20 years of development - so attractive. TI would certainly need some major incentive and quite a bit of time to develop a cost-effective multiposition mirror design.

Finally in terms of predicting the possible incremental impact of the TI DMD chip on the optical switching market, there is the issue of timescale and the likely two to three years required for the qualification process of any future large switch design based on the TI component.

In conversations with Wes Stalcup, TI's CMO for Dynamic Optical Networking, a number of interesting points were made regarding the state of development in commercializing the DMD in optical networking applications.

To the company's credit, TI is candid about the prospects and suitability of the product in optical networking applications, stating, "At a per fibre level we don't think this is necessarily the best technology for doing large cross-connects although we still have some customers who are interested in looking at it." However, according to Stalcup, "It's out there, it's available and we've built a variety of reference designs".

Also, TI is reluctant to overplay the level of its efforts in finding optical networking applications for its Micromirrors, stating, "We're trying to be fairly serious about it but we're not making what I call huge amounts of direct investment at this time". TI has also made the pragmatic decision to take a laid back approach and leave much of the utilization and commercialization of the product to the market: "We thought that if you can put this technology in the hands of people who are interested in doing things, such as in universities or start-ups, and make it easy for them to buy, then they will be able to come up with a lot more than we would by ourselves."

To this end, TI strives to provide chip sets that allow the greatest flexibility, including mixing and matching DMDs that cover various wavelengths, in order that companies which have experience in detailed small form factor optical design and manufacturing channels can then communicate their requirements to TI. TI's second-generation chipset, tentatively called OSP-2000, is likely to use a different DSP due to cost reductions and improvements in the overall processing performance with a six four platform. TI plans to explore these developments with its partner customers in more detail rather than coming up with the whole idea by itself.

Although Siemens looked at the DMD some time ago, the most serious player to have signed on so far is CIDRA Corp of Connecticut. According to TI, CIDRA's aggressive sales mode has resulted in a fair amount of work for TI. In addition, a couple of unspecified U.S.-based optical networking companies are said to be in fairly heavy evaluation currently.

A key TI development has been the abandonment of the use of one mirror per one lambda in favour of allowing one lambda to cover hundreds or thousands of mirrors at a time. According to TI, this offers the unique ability to fine tune, manipulate and reshape that light, to attenuate it and also to impact its phase. TI states, "You can destructively interfere and achieve a higher contrast ratio by turning certain mirrors back on in a spot. If you turn selected mirrors back on based on the size of that spot and the area that's illuminated then you can achieve 3-5db more of extinction ratio or contrast in the display. It's clearly destructively interfering with phase and we know it preserves the wave front so we think there's some 'undiscovered country' there in terms of grooming a signal". The company states that it has not yet fully explored the possibilities but that it has merely "played with the algorithms and mapped the behaviour".

In the end, the key ideas for networking, as TI sees it, are "scalable number of lambdas per fibre; real time dynamic control; proven productions parts; and a business model which finally makes this stuff available".

TI believes that rather than large OXC type architectures, more scalable, modular and 'meshy' approaches based on smaller switches are more practical going forward. According to the company, a pair of four-port ROADMs set into EAST-WEST configurations fits pretty well with today's existing approaches and allows the incremental functional changes network operators prefer.

Scalable optical real time DGEs with built in OPMs adjacent to amps combined with such ROADMs make intelligent links, which adapt to amp aging, load changes, and provisioning needs on software command. The point for TI is that each of the aforementioned functions is on the same electrical platform, possibly the same optical platform. This, according to TI, represents an unprecedented economy of scale.

TI believes that the real possibility for totally new price points in networking exists if it can land in the right place and architecture that will drive units. Accordingly, TI believes that price points and volumes could align to make this technology and architecture viable for regional and metro applications.