An examination of the small tech solutions and applications expected during the next five years. Panelists: Dr. Barry Alexia, John Deere Worldwide; Mr. Erik Puik, TNO Industries, the Netherlands; and Jim Walker, Tellium Inc.
 |
| Walker Puik |
Alexia: The focus of this response will be on MEMS applications, thus allowing the developers of MEMS to respond to the current processes:
Based on a strategic view in both application and technology, there is a need to develop the technology to meet the application (not the application to meet the technology). Reason being: if MEMS is going to be a “household word,” it must be introduced as being non intrusive to both the customer and current application(s). Those applications could include:
- wireless communication (cell phones, RFID products, and wireless computer networks);
- law enforcement products (speed radar and microwave non lethality devices); and
- lawn and garden products (bio-nutrient detection).
By replacing the current discrete electronic components with MEMS type devices (i.e., RF switches, tunable slot antennas, wave guides, and filters) the end product should be smaller and cheaper with an improvement to product durability.
The consumer is not concerned with the discrete components/MEMS devices, they are concerned with the functionality of the end product. And, if MEMS devices provide an improvement to quality and usability, then MEMS could be herald as a defining feature of the next technology revolution. If not, then MEMS may very well suffer the fate of a failed technology that was introduced to the market place before its time.
Puik: The best selling feature for any product nowadays is probably a cost advantage. This won’t be different for microsystems. So considering the big picture for small tech applications, this will be a major element in the coming five years. There will be other breakthroughs, such as better functionality, more compact design (that enables portable use) and increased reliability. In spite of all these other advantages, the success of microsystems is going to depend heavily on the reduction of their cost.
Since microsystems need state-of-the-art production techniques — that causes high investments for production equipment — they need to be produced in large quantities to meet cost demands. This will be the challenge for the next five years.
If we succeed in doing this, the most surprising element for the near future will be the fact that all these high-tech microsystems become available to the masses.
Walker: Even during the current slowdown in the optical telecom arena, service providers are continuing their push toward higher capacity networks. This push is happening along two main fronts;
- higher channel density and channel count OC-48 (2.5 Gbit/sec) and OC-192 (10 Gbit/sec) systems, and
- higher data rate systems, specifically OC-768 (40 Gbit/sec). Each path offers significant transmission challenges to providers that at present do not have proven solutions. In concert with this drive to higher capacity is the desire to add switching functionality through optical and all-optical cross-connects and programmable wavelength add/drops. These are desired to provide dynamic wavelength allocation, rapid network provisioning, and immediate service restoration in event of failure. These applications in total provide a window of opportunity for MEMS technology to develop inroads onto the telecom component marketplace, as there are no clear solutions using more traditional telecom technologies at present.
At 40 Gb/s, chromatic dispersion and polarization mode dispersion (PMD) are quite difficult to address. PMD is particularly challenging due to its random nature caused by non-uniform cross-section of optical fiber, particularly in older optical fiber (produced prior to 1995 [1]). Power fluctuations and non-uniformity across all wavelengths create additional difficulties such as channel mixing and crosstalk for both OC-192 and OC-768 systems as channel spacings and bit-periods shrink [2]. Static compensation for these impairments is difficult to achieve, therefore, the dynamic capabilities of MEMS components makes them attractive solution candidates, particularly for OC-768 systems. MEMS devices have been combined with free-space optics to demonstrate dynamic gain equalizers, chromatic dispersion compensators, gain tilting elements, tunable WDM filters and sources, variable optical attenuators, and the beginnings of a PMD compensator.
Since MEMS technology is still in its relative infancy as compared to other technologies that could be considered for these applications, it is essential that product development happen quickly and that a reliability story be established as rapidly as possible. Due to the dearth of solutions to their perceived problems, systems developers are at this moment more willing to consider unproven technologies such as MEMS. That window of opportunity could be closing however, as more mature technologies are brought to bear on these problems. In the very recent past, MEMS approaches in telecom were met with a genuine excitement as evidenced by the incredible rate at which MEMS device and technology companies were being acquired by larger systems companies. Now however, MEMS technology is at risk of becoming the next “bubble memory”. It is therefore imperative that the real promise of this technology is realized soon and the “over-the-top” hype surrounding MEMS is avoided.
2. Describe a specific small tech application that offers a unique solution:
Alexia: This response is centered on the current market place applications, instead of an invention where MEMS would be introduced.
Response: Other than the obvious application (or so it seems) of wireless communication, the next not too obvious applications could be
- ground speed radar used by the law enforcement agencies,
- railroad industry (time and distance measurements),
- agriculture manufacturers (ground speed indicators for tractors, harvesters, planters/seeders/sprayers),
- amusement parks (roller coasters),
- automobile companies (RFID tags and adaptive cruise control), and
- bio medical (monitoring patient breathing).
Each of which would benefit from a MEMS type component. Last add, in order for the MEMS application to flourish in the market place, the end product would need to represent a reduction in both size and cost.
Puik: Actually, I don’t like this question because I so often see big headlines with interesting new microsystems that are supposed to change the world but in fact are just niches that don’t bring microsystems further in commercial or technological point of view.
But to answer the question, first a little history. It all started with inkjet systems, an interesting market but just for a few large companies.
A few years ago we saw the hype with automotive sensors and accelerometers. For the production of airbag sensors there are only a handful of producing companies left.
About two years ago we were confronted with optical switching and RFMEMS. We’ve seen about sixty startups in optical systems, and next year probably half of them won’t exist anymore.
Now the money has been put on Biomedical Systems. No one knows if history is going to repeat itself but misery will take place here as well.
I don’t want so sound all too negative here. Every hype has had its personal winners and there will be future opportunities as well.
To choose a promising application, it must be one of these biomedical systems because they are in the latest hype. Personally, I’m very impressed with DNA chips.
Walker: Clearly, the small tech application that has received the most attention and offers the most unique solution is the large port count all-optical cross-connect (OXC). Unfortunately, this attention has come due to an extreme example of the overhype that threatens the credibility of MEMS technology in general. The goal in this answer is to provide a backdrop of reality against which the incredible promise of these systems should be viewed.
The all-optical cross-connect has been viewed as a pivotal network element for a number of years now. MEMS based systems offer the potential of protocol and bit-rate independency especially at the switch fabric level. In addition, since they do not require an optical to electrical to optical (OEO) conversion and associated high-speed electronics they are potentially lower cost than electronic counterparts. Typically, the MEMS-centric view of an OXC is dominated by the chip design and fabrication. In reality however, the MEMS device development is but a small piece of the effort to make a cross-connect. OXC systems with port counts of more than one hundred normally comprise at least one bay (roughly 2’x2’x7′) of optical and electrical equipment with a high degree of inter-related complexity as will be described in more detail below.
The justification for the attention paid to this application lies in a survey of all available technologies that can be exploited to develop large port-count all-optical cross-connects. One is led to the conclusion that for port counts of greater than 16 or possibly 32, MEMS technology is the leading and perhaps only candidate. For port counts of 32 or less, MEMS technology has significant competition from technologies such as thermo-optic waveguide switches, bubble switches, LCD arrays, and electro-optic devices. All of these technologies face significant challenges to scaling to higher port counts however, leaving MEMS virtually without competition.
MEMS OXC fabrics come in two flavors, 2D and 3D based on the manner in which switching takes place. In a 2D fabric, the switching takes place in a 2D-plane via an array of N2 MEMS pop-up mirrors (for an NxN switch fabric) [3]. The optical signals enter and exit the switching plane via 1D arrays of collimated fibers on adjacent edges of the switching plane. The desired connection between an input and output port is made by inserting a mirror at 45 degrees to the chosen incoming beam path that reflects the signal toward the desired output port. 2D fabrics are limited to small port counts of about 16 due to practical manufacturing and optical loss considerations.
There are several companies marketing such systems and lab trials using 2D OXCs are ongoing. For port counts larger than 16 or so, the MEMS solution is necessarily based on 3D fabrics. In typical 3D fabrics, the switching occurs in the 3D-volume located between two 2D arrays of two-axis gimbal mounted mirrors. The optical signals enter and exit the switching volume via 2D arrays of collimated fibers. These collimated fibers typically consist of separate 2D arrays of fibers and lenslets held in careful alignment with each other. There are five essential parts of a 3D MEMS OXC: fiber bundle, lenslet array, MEMS mirror array, packaging, and control electronics, each with related challenges.
Fiber bundle manufacture status is at present best described as in advanced development. Arrays numbering in the hundreds can be obtained in small quantities with the required micron-scale fiber positioning tolerance. Long-term reliability of these arrays (particularly fiber position stability over 20 years) has not been addressed as yet, but is likely to play some role in overall OXC reliability story. The state of fiber arrays numbering of order 1000 is best described as advanced research today, although the techniques used to make smaller bundles are expected to be applicable to larger arrays as well.
The current state-of-the-art in lenslet array manufacture is reasonably good. Large lens arrays can be obtained from multiple sources that exhibit high yield to tight tolerances, although availability in quantity is still difficult to find. For both lenslet arrays and fiber bundles, a close working relationship with suppliers is necessary. These lenslets are required to provide collimation over a broad wavelength range (1200-1650 nm) which creates a challenge to both the lenslet formation as well as the AR coatings required to suppress back-reflections with high transmissivity.
The number of MEMS chip producers working in this space far exceeds the likely market potential, although this suggests that the likelihood of a reasonable supply of mirror arrays is high. Exact mirror specifications are dependent upon the optical system details and unique to each OXC developer. While development of suitable MEMS mirrors provides a significant challenge, it is not believed to be insurmountable and existence proofs can be found. Numerous chip suppliers, some captive to systems houses and others not, have demonstrated good switching capabilities for arrays of limited numbers of mirrors.
Perhaps the greatest challenge for the MEMS chip producers lies in the number of electrical I/O required and the overall size of the mirror array chips. Each mirror requires approximately four electrical signals to actuate in two axes, making the total electrical I/O from a 256-mirror chip roughly 1000. This assumes of course 100 percent yield of the mirrors on the die. This number of electrical I/O strains the boundary of chip packaging capabilities today. In addition, these mirrors typically require voltages in the 200 V range. Achieving high reliability of the electrical connections and associated mirror drive electronics becomes part of the MEMS device development challenge. As mirror array sizes grow, the size of the MEMS chip itself can reach several centimeters on a side. Obtaining reliable die attachment over the range of temperatures the chip is likely to see is also quite challenging. Finally, it is widely held that the MEMS die must be hermetically packaged for reliability reasons, which means providing a reliable glass seal to the package and the same AR coating performance as for the lenslets.
Packaging of the optical train requires attaining and maintaining precise positioning of the fiber bundles, lens arrays, and MEMS arrays with accuracy on the order of one micron and a few milliradians over a 20-year lifetime of the system. For optical loss reasons as well as equipment bay constraints, the volume within which the package resides is compact. This creates an additional challenge for routing the thousands of electrical signals as well as the hundreds or thousands of optical fibers into the package. An additional little-discussed hurdle for packaging is dealing with the thermal range over which these systems must operate as well as the heat dissipation from the fabric itself.
The final piece of the OXC to discuss is the control electronics and algorithms required to affect low-loss optical connections. The cross-office environment that these systems must operate within have a limited loss budget for the fabric itself due to transponder limits. The loss budget for intermediate reach and high-power very short reach transponders is roughly 12 dB. The portion of this budget allocated specifically to the cross-connect fabric is a fraction of this. Therefore, obtaining highly accurate positioning of the MEMS mirror is critical to achieving low loss. The exact accuracy required is dependent upon system specific parameters, however it is generally expected to be a fraction of a milliradian. This angle is roughly equivalent to that subtended by a golf ball at a distance of one mile. Once this accuracy is obtained and held over time, it must also be able to provide stochastic immunity.
Because shock and vibration have the potential to drive the mirrors from their low loss connection positions, it must be addressed through either packaging or control methods. An inordinate amount of hype has been focussed on 3D MEMS OXCs, but practical implementation of such systems may be delayed for some time due to economic and technical reasons [4].
Cost competition from OEO (electronic core switches) systems at 10 Gbit/sec has reduced the pressure to deploy OOO (MEMS optical core) systems until significant OC-768 fiber has already been deployed. Recent analysis suggests that significant deployment of OC-768 systems will not take place until sometime in the 2003 timeframe [5].
Interestingly, this may ease the task of OOO switch developers as the desired port count of these systems is also under some reconsideration. Over the past year, the view that port counts in the hundreds were merely stepping-stones to 1K+ systems has shifted to where these “smaller” systems may suffice in many cases. There is clearly still a desire to achieve port counts of up to several thousand, but the deployment of these larger systems has perhaps been put off for several years. At present, several companies have small OOO switches in customer lab trials. True widespread deployment of such systems however will be gated by two factors; 1) development of a compelling MEMS reliability story and 2) reduction of the optical insertion loss exhibited by current OOO OXC systems. As the bit-rate increases, the penalty from a channel failure increases as well; therefore, reliability of the MEMS device becomes paramount. As of today, the only MEMS telecom system to have successfully passed Telcordia qualification testing is the 2D cross-connect developed by OMM.
The anticipated advantage gained from a lack of electronics for OOO systems is actually tempered considerably when one considers the loss of intelligence and switch capability that results. The intelligence associated with the high-speed electronics used in OEO switches provides the ability to perform many valued network level functions including automatic topology discovery, core grooming, multi-vendor interoperability, automatic restoration, and wavelength conversion. The cost saving normally associated with deploying OOO hardware is likely lost when one compares the savings associated with overall network management using OEO. The actual comparisons are dependent upon specific details associated with particular networks, but these management capabilities must be considered as part of the OOO vs. OEO debate.
3. What natural extensions are there for this technology for future use (next five years):
Alexia: Our Agriculture vision of MEMS would be an upswing in bio chemical/agriculture technologies. Reason being: the need to develop a robust crop or soil is last piece of high yield crop puzzle. The seed producers are on a plateau in developing a higher yield and disease resistant seeds.
That said, now is the time for the MEMS industry to introduce the next generation bio/ag technology which could monitor soil nutrients (potassium, nitrogen, and phosphorus) and early warning of soil health (i.e., micro irrigation release sensors). The detection process can be applied either on the go during the crop harvesting, or release by overhead assets.
Puik: DNA chips are going to significantly increase the number of DNA tests that will be performed in the next five years. Because of the miniaturization it will be possible to perform about 1,000 to 10,000 different tests with only one drop of blood and have the results in minutes. This is really going to have impact on our lives in the next five years.
Walker: Before suggesting natural extensions for MEMS technology in general, and MEMS OXCs in particular, it might make more sense to suggest that the MEMS field would be well served to make real what it has promised already. That being said, a natural extension of MEMS technology in optical telecom lies primarily in addressing the issues faced by network architects when deploying agile OC-768 systems going forward into the future.
Another fertile direction for MEMS in telecom lies in the integration of electronics, MEMS, and photonic devices. This could be accomplished through a combination of monolithically integrated intelligent MEMS and hybrid integrated active photonic elements or material. By combining the attributes of these three essential technologies, one could imagine eliminating the barrier between the fast, dumb all-optical cross-connect and the slower, smart OEO type of cross-connect for example. Certainly other intelligent, high-speed photonic MEMS devices would also be enabled and likely find warm reception by system developers.
4. What applications are likely to be the most “disruptive” to industry:
Alexia: Short answer is the bio/Ag sensors (see number 3).
Puik: I don’t know too many applications that were disruptive to technology in the past. Of course, we’ve seen personal computers change the world, and the same has happened with the Internet. But I don’t think that there will be microsystems causing comparable impact to this, at least not in the next five years. Looking at a further horizon I think that “robots” are going to be causing disruptive changes to the world and there will be a large population of micro robots as well. I can hardly wait.
Walker: It could be argued that the all-optical cross-connect has already been the most disruptive to industry without having yet achieved successful introduction. This argument can be made on pure economic means based on the many billions of dollars that have been spent pursuing this one network element. Additional justification for this position lies in the number of companies that either started up to chase this application (without much in the way of experience in telecom) or diverted themselves from other nearer-term MEMS applications in telecom in order to work on OXCs. Although this use of the term “disruptive technology” is different than is typical, in this case it seems to fit.
Again using perhaps my own definition of “disruptive” and believing it to accurately describe the deployment of OC-768 systems, this will be gated by the availability of the devices described earlier, dynamic gain equalizers, dispersion compensators, PMD compensators, etc. As such, advances in these components are critical to the successful deployment OC-768 systems could therefore be considered disruptive to the telecom industry. Even with the slowdown in deploying these systems, the enabling technology development for 40Gbit/sec must continue and is still being driven by the service providers, if only to cover themselves. As Ciena CEO Gary Smith was quoted as saying, “No one wants to be the Rich McGinn of OC-768” [5].
null
References:
[1] M. Fuller, “PMD compensation for 40 Gbits/sec remains a question mark,” Lightwave, cover article, August 2001.
[2] D. McCarthy, “Faster vs. Denser: Networks Reach Another Crossroad,” Photonics Spectra, September 2001, pp. 110-118.
[3] L. Lin, “Micromachined free-space matrix switches with submillisecond switching time for large-scale optical crossconnect,” OFC ’98 Technical Digest, OFC ’98, San Jose, CA, Feb. 25, 1998, pp. 147-148.
[4] J. McGarvey, “All-Optical Snag?” The Net Economy, August 6, 2001.
[5] O. Malik, “Networking’s next big thing may be slow to shine,” Red Herring, July 15, 2001, pp. 28-29.