Consumers continue to demand cell phones with increased ...



1. Technology Development

1.1 Introduction and Executive Summary

Consumers continue to demand cell phones with increased functionality and bandwidth, requiring larger and more expensive batteries. To provide a cost-saving and weight-saving solution, Resonant Microsystems, University of Wisconsin-Eau Claire, NanoRite and Purdue University are teaming to commercialize an adaptive Radio Frequency (RF) chip-set based on Micro-Electro-Mechanical Systems (MEMS) technology to improve cell phone transmitter efficiency by up to 2(.[1] Since cell phone batteries cost about $4-$10 each, the proposed adaptive RF chip-set can have a significant impact on the cell phone bill of materials at $2 to $5 per phone.[2] This represents a multi-billion dollar market with more than a billion handsets produced per year.

While handset features are multiplying, battery research has not kept pace (Figure 1). Thus, with the proliferation of power-hungry features such as multi-band operation, MP3, GPS and broadband data, improving efficiency is critical. While the initial focus is the cell phone handset market, the same technology will improve the performance and the efficiency of 3G/4G, WLAN, and WIMAX data terminals.

1.2 Proposed Product and the Current State-of-the-Art

Cell phones today utilize components that lose their efficiency when the environment surrounding the antenna changes.1 To address this problem, the proposed MEMS-based adaptive RF chip-set (Figure 2) constantly tunes the cell phone transmitter. As a result, optimal efficiency is maintained under all operating conditions, distances from base station, and antenna impedance conditions. The proposed RF chip-set also improves reception and reduces dropped calls.

The increase in efficiency is dramatic, equating to 40%-50% longer talk time or smaller, cheaper and lighter batteries. With increased talk time rated as a critical commodity in most surveys, our technology addresses a market demand that has not been successfully met. As discussed, battery research has not kept pace. In particular, battery energy density roughly doubled while CPU speed increased by >100× over the last ten years.2

[pic] [pic]

|Figure 1. Slow pace of battery advancement.2 |Figure 2. Enabler of the Adaptive RF chip-set: MEMS varactor. |

1.2.1 Tasks and Milestones

The proposed 1-year project is divided into 4 tasks, each with one critical milestone. These are also the necessary tasks and milestones required to sampling parts to handset developers.

• Task 1: Develop the ability in Eau Claire, Wisconsin to manufacture the key MEMS varactor device (Figure 2, Section 1.2.3) that enables the adaptive RF chip-set.

o This will be accomplished by transitioning the relevant technology from Professor Dimitri Peroulis (Purdue University) to the production tool-set at Resonant Microsystems / NanoRite (Section 2.2).

o Milestone: Qualification of equipment set.

• Task 2: Develop a first-generation control algorithm for the RF chip-set that can be subsequently implemented with a CMOS chip.

o The control algorithm will be developed in C++ to control the feedback loop of the amplifier based on the sensed output power level.

o Milestone: Successful laboratory demonstration of the algorithm.

• Task 3: Fabricate MEMS varactors with Resonant Microsystems / NanoRite tool-set.

o Perform the 5-mask varactor process consisting primarily of a number of lithography, sputtering and plating steps.

o Milestone: Functional MEMS varactors.

• Task 4: Fabricate an adaptive RF chip-set for cell phone frequencies.

o A group of MEMS varactors devices will be heterogeneously integrated with an RF amplifier (Figure 5, Figure 6, Section 1.2.3). The feedback loop of the amplifier is controlled by setting the capacitance level of the MEMS varactor.

o Milestone: Successful laboratory demonstration of the adaptive RF chip-set.

[pic]Figure 3. Gantt chart showing scheduled tasks and key milestones.

1.2.2 Methodology / Technical Details

The aforementioned 2× improvement in efficiency is achieved by addressing two problems:

Problem #1: Power Efficiency as a function of the cell phone output power

A cell phone power amplifier constantly changes its output power level depending on the path to the nearest cell tower. Unfortunately, the efficiency of handset power amplifier drops sharply as output is reduced. Several studies have been conducted to determine this distribution. The average power added efficiency (PAE) exhibited by power amplifiers is actually much lower than the maximum efficiency possible. As Figure 4 shows, under typical operating conditions, the probability that the cell phone power amplifier will operate under its optimal conditions is very low (3% in this example). This situation can be corrected if the output impedance of the amplifier is automatically adjusted to be optimal depending on the specific output level.

Problem #2: Power Efficiency as a function of the cell phone antenna impedance

This dramatic increase in efficiency is obtained from controlling the antenna impedance. Figure 5 shows that a 60% improvement is obtained with optimal impedance. In comparison, transmission conditions without adaptive tuning can translate into an 80% loss in efficiency.

|[pic] |[pic] |[pic] |

| |(a) |(b) |

|Figure 4: . Statistically adjusted PAE of a typical |Figure 5. Adaptive RF chip-set provides >60% improvement: (a) Simulated and (b) measured |

|cell phone power amplifier. The maximum PAE is 41%. |power added (PA) output power and efficiency as a function of the antenna impedance. The |

|However, the probability that the cell phone will |PA is optimized for loads (impedances) of 2-4 (. Measurements after Kaajakari et al[3]. |

|work at this PAE is only 3%. | |

1.2.3 Adaptive RF Chip-Set Based on Micromachined Tuning Elements

Figure 6a shows a schematic of the chip-set architecture. Figure 6b shows a X-band demonstration (non-cell phone frequency) without automatic feedback5. The output port of the power amplifier (PA) is connected to a MEMS-based adaptive RF chip-set that transforms the output impedance of the antenna to the optimal impedance for maintaining the maximum power added efficiency (PAE). The amount of radiated power will range from 0.1W to 2W, and will depend on the distance/path between the handset and the closest cell tower. Due to non-linearities of the power amplifier, each power level will require a different impedance level. As already mentioned, this scheme is expected to improve the cell phone efficiency by about 2(.

The adaptive RF chip-set includes (a) an adaptive matching network (AMN) that covers the required impedance range; (b) a RF coupler; (c) a RF power detector; and (d) CMOS circuitry that controls the AMN based on the output power level.[4] Conventional matching network approaches have limited power handling due to the introduction of strong resonances (resulting in high voltages and currents) in order to reach high impedance points on the Smith Chart. In contrast, the team has developed a new technique[5] to design tunable matching networks simultaneously optimized for impedance coverage and power handling.

[pic] [pic]

(a) (b)

Figure 6: Adaptive RF chip-set demonstration (a) The chip-set absorbs antenna variations by an adaptive matching network (AMN) and maintains optimal output power and power added efficiency. (b) MEMS-based RF chip-set that utilizes MEMS components with high power handing (10W). There is no feedback loop in this power amplifier –to be implemented in the proposed project.

1.2.4 Adaptive Matching Network Design Improves Power Handling

The adaptive matching network (AMN) design is critical for handling the necessary power levels. Conventional matching network approaches typically consider only the impedance coverage viewpoint to cover as large of an area on the Smith Chart as possible. However, this usually leads to the introduction of strong resonances in order to reach high impedance points on the Smith Chart. Consequently, power handling is limited because of the strong currents or voltages associated with these resonances. To address this, the team has developed a new technique[6] for designing tunable matching networks simultaneously optimized for impedance coverage and power handling. The fundamental idea is based on avoiding strong currents and voltages in the circuit by optimizing the location of the tuning elements in the circuit. Figure 7 shows an example of such a design that was recently developed for K-band applications.

A) [pic] B) [pic]

Figure 7A) Example of K-Band impedance tuner simultaneously optimized for power handling and impedance coverage. The placement of MEMS components on the transmission line is optimized to ensure a nearly equal voltage swing across all the MEMS parts. This increases the power handling by 4× compared to a conventional design6. The impedance coverage, shown on the right, is very good and remains nearly unaffected. The design will be scaled for cell-phone frequencies. B) Measured and simulated impedance coverage of the matching network shown in the left.

The design can be optimized for impedance coverage. As shown, the design has excellent coverage (simulated and measured) at 30 GHz. It is important to note that due to analog tuning, the entire area shown in Figure 7A) is covered. The figure shows some representative points since it is impossible to measure the entire area. This type of matching network also has reasonable bandwidth as shown in Figure 7B). In the proposed project, a similar approach will be taken to optimize the tunable matching network at 800-2000 MHz with 2-4W power handling.

1.3 Industry and Academic Partnerships

This project is enabled by the Center for Adaptive Microsystems (CAM). CAM is a partnership involving UW-Eau Claire, UW-Stout, NanoRite, Chippewa Valley Technical College, the Adaptive Radio Electronics and Sensors Group (ARES) at Purdue University and four companies working in the area of microsystems, including Resonant Microsystems. The mission of CAM is to provide a link between industry and academic institutions for advancing research and development in the field of reconfigurable and adaptive microsystems, furthering the goal of commercialization. Key interests include materials research and analysis, quality control, manufacturing processes, validation and verification, applications and the development of higher levels of integration. UW-Eau Claire and Resonant Microsystems have been working together for two years to pursue funding for equipment that will benefit both parties. This will be the first project intended for commercialization.

2. Economic Impact

2.1 Value Proposition to Customer and Competitive Advantage

The value proposition is longer-life battery and higher performance at reduced cost and weight. Efficiency is critical, as battery technology advances slowly while new features proliferate.

The adaptive RF chipset products have a number of unique advantages:

( Reduced power consumption: For cell phone handsets, the chipset will increase power efficiency by 2×. This equates to 40%-50% longer talk time or reducing battery cost by $2 to $5 per handset. Increased talk time rates as a critical commodity in surveys.

( Future opportunities: WLAN and WiMAX/4G/VOIP will benefit in a similar fashion. Wireless device manufacturers face similar power challenges as handset manufacturers.

There were significant barriers to entry that the team overcame to develop our chip-set products:

( Extremely-reliable micromachined products.

( RF circuit designs that enable tuning of the transmitter to maintain high efficiency for all range, power and impedance condition and provide phase shifting for beam steering.

• Low-cost production process that can scale to millions of units with minimal equipment investment. For the varactor, estimated device cost is $0.13 at 50% yield and $0.075 at 90% yield, more than 10× cheaper than the cheapest MEMS components today.

2.2 Manufacturing

Chip-set manufacturing leverages Resonant Microsystem’s wafer-scale micropackaging capabilities. Wafer scale packaging is critical as hermetic packaging constitutes 80%-90% of MEMS manufacturing cost. Resonant Microsystems’ process has been utilized for over 10 MEMS devices, and typically reaches over 90% yield by the 3rd fabrication cycle of a design.

Complete wafer-scale batch processing is critical because it decreases the number of processing steps from tens of thousands of steps per wafer to only hundreds of steps per wafer (Figure 8, Figure 9). This also significantly reduces the ramp-up equipment and setup costs.

These cost advantages are critical for regional economic impact. These components can be economically manufactured in Wisconsin, with the potential creation of 50-100 jobs.

[pic] [pic] [pic]

Typical Today – tens of thousands of steps True Wafer Scale Processing – hundreds of steps

Figure 8. Wafer-scale processes reduce the number of steps from tens of thousands to hundreds per wafer.

[pic] [pic]

Figure 9. Wafer scale processing is key to low-cost: More than 10 device types produced using this process.

3. Intellectual Property

The team has been awarded fundamental patents covering MEMS devices (micro-machined component design) based on the low-cost wafer-scale packaging process.

Furthermore, Resonant Microsystems and University of Wisconsin-Eau Claire have agreed to a fee payment in which University of Wisconsin-Eau Claire will receive a 2% royalty off revenues from every adaptive RF chip-set sold by Resonant Microsystems until a total of $100,000 has been paid to University of Wisconsin-Eau Claire by Resonant Microsystems. After that, subsequent royalty arrangements will be negotiated.

4. Grant History

The proposed project is not part of a current or past grant application. My colleagues and I are not receiving other grant funds for the project.

5. Budget Narrative

Funding from the grant will be used to provide partial summer support ($6000 for approximately 3 weeks of full time work) for the PI to oversee the project, supervise undergraduate student researchers and work with both Purdue and Resonant Microsystems to keep the project on schedule. The PI has additional summer support (one month’s salary) as Director of the Materials Science Center. Part of those director duties include interacting with companies to support research and development. Funding will also help support summer appointments for students (1-2) to work on the project (140 hours at $10/hour). Students may be involved in computer programming aspects of the project and in fabrication/characterization of the chipsets.

For success the project needs someone familiar with the commercial processes and techniques. Funding is requested as a contractual agreement with Resonant MicroSystems to have a process engineer work on the project in collaboration with UW-Eau Claire and Purdue University.

Funding for travel for the PI and UWEC students between Eau Claire and Purdue will be funded through the Materials Science Program/Materials Science Center at UW-Eau Claire. This funding will allow the PI and student(s) to meet with the research team at Purdue. The budget includes two trips by fleet vehicle with one trip including a student. (fleet vehicle with mileage $165 (hybrid), two nights lodging in West Lafayette at state rate $220, per diem for meals $120 for 3 days, total of $505 each trip. Cost for student lodging $220 for two nights, meals $120 for 3 days. Total $1350)

6. Bibliography and Resume

1. D. Medendorp, “Impact of reduced insertion loss on performance of cellular/PCS handsets”, O2 Wireless Private Study;

2. I. Kang, J. Kim “The design of an improved gain dynamic envelope tracking amplifier in cellular radio handset” Microwave Journal, August 2001.

3. Battery cost typically range from $4 for small capacity battery to $10 for larger battery:

a)

b)

4.

5.

6. V. Kaajakari, A. Alastalo, K. Jaakkola, and H. Seppä, “Variable Antenna Load for Transmitter Efficiency Improvement,” IEEE Transactions On Microwave Theory And Techniques, vol. 55, no. 8, pp. 1666- 1672 Aug. 2007.

7. D. Qiao, R. Molfino, S.M. Lardizabal, B. Pillans, P.M. Asbeck, C. Jerinic, “An Intelligently Controlled RF Power Amplifier with a Reconfigurable MEMS-Varactor Tuner,” IEEE Trans. Microwave Theory Techniques, vol. 53, no. 3, pp. 1089-1095, March 2005.

8. Y. Lu, L.P.B. Katehi, and D. Peroulis, “High-Power MEMS Varactors and Impedance Tuners for Millimeter-Wave Applications,” IEEE Transactions on Microwave Theory and Techniques, vol. 53, pp. 3672-3678, Nov. 2005.

DOUGLAS J. DUNHAM

Professor of Physics and Astronomy; Director, Center for Adaptive Microsystems; Director, Materials Science Center; University of Wisconsin-Eau Claire

RESEARCH

• In the last 10 years, over 20 students have been engaged in my student faculty-research projects. 23 publications in peer-reviewed journals.

• PI/Co PI on 4 external grants totaling over $450,000 while at the University of Wisconsin-Eau

Business/Research Projects

UW-Eau Claire Materials Science Research Lab Industrial Collaborations Working with local industries to help them satisfy their materials characterization needs. Interfacial Solutions, Phillips Plastics, Northern Crossarm, SGI, HTI, Northern Engraving, etc. Developed a research project with Minnesota Wire and Cable and Chippewa Valley Technical College to analyze nanotechnology related materials.

Characterizing silicon wafer contamination. Supervised the experimental portion of feasibility studies for the use of spectromicroscopy to analyze silicon wafer contamination. The Semiconductor Industry Association identified improved methods for silicon wafer materials analysis as a major “industry wide” concern. Contamination of the wafers can result in the shutdown of production. Quickly identifying the chemical characteristics of the contamination (usually in the form of small particles) can lead to correcting the problem more efficiently. A two-year $746k Collaborative Research and Development Agreement (CRADA) was signed as a result of the feasibility study. Intel contributed $531k.

-----------------------

[1] a) D. Medendorp, “Impact of reduced insertion loss on performance of cellular/PCS handsets”, O2 Wireless Private Study; b) I. Kang, J. Kim “The design of an improved gain dynamic envelope tracking amplifier in cellular radio handset” Microwave Journal, August 2001.

[2] Battery costs typically range from $4 for small capacity battery to $10 for larger battery: a) b)

[3] V. Kaajakari, A. Alastalo, K. Jaakkola, and H. Seppä, “Variable Antenna Load for Transmitter Efficiency Improvement,” IEEE Transactions On Microwave Theory And Techniques, vol. 55, no. 8, pp. 1666- 1672 Aug. 2007.

[4] D. Qiao, R. Molfino, S.M. Lardizabal, B. Pillans, P.M. Asbeck, C. Jerinic, “An Intelligently Controlled RF Power Amplifier with a Reconfigurable MEMS-Varactor Tuner,” IEEE Trans. Microwave Theory Tech., vol. 53, no. 3, pp. 1089-1095, 2005.

[5] Y. Lu, L.P.B. Katehi, and D. Peroulis, “High-Power MEMS Varactors and Impedance Tuners for Millimeter-Wave Applications,” IEEE Transactions on Microwave Theory and Techniques, vol. 53, pp. 3672-3678, Nov. 2005.

[6] Y. Lu, L.P.B. Katehi, and D. Peroulis, “High-Power MEMS Varactors and Impedance Tuners for Millimeter-Wave Applications,” IEEE Transactions on Microwave Theory and Techniques, vol. 53, pp. 3672-3678, Nov. 2005.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download