The Development Of Mems For Rf Applications

Robert Rhodes Final Cmpe 185 Introduction: With the increased demand for smaller portable cellular and other rf devices, the need for correspondingly smaller components has increased. A possible solution to the problem of finding such components lies in the field of rf mems design. Various rf components including phase shifters, switches, and oscillators may be implemented with mems and made to work at the microscopic scale. A device of particular interest for the purposes of this paper is the mems version of the crystal oscillator. Recent work has led to the development of implementing rf oscillators (and electronic oscillators in general) with mems based devices and manufacturing methods. These mems oscillators provide better thermal hysteresis and aging characteristics than discrete quartz -based parts at comparable prices [1]. Now we will explore the amazing research, technologies, and applications resulting from rf mems oscillators. As rf applications of this generation of oscillators will be the primary focus of this paper, it would be prudent to begin with an overview of rf systems and then turn attention to the workings of some crucial components such as the oscillator. Then sufficient background for appreciating mems applications in rf will be realizable. The radio will serve as an example since all rf systems are in some form or another a radio (transmitter, receiver, or transceiver). For simplicity, AM de/modulation will be considered. To begin, one assumes that an rf system is modeled as a single module like the diagram in Fig. 1

Single module system In

Out

Fig. 1 single module system block diagram Furthermore, assume that two in signals enter into this system and that two out signals exit it. Let one of the input signals be denoted as x(t) and the other as w(t) while the output signals are z(t) and y(t) where t represents the time dependence of these signals. The desired operation of the above module is to take the input signals and produce two new signals although in general there will be more than two output signals most of which can be attenuated by an appropriate filter network. The process that produces these signals will be a non linear process. Thus for signals described by: ∞

∑f i =0

i

= x j (t )

The output is: x j (t ) ⊗x j ±1 (t ) ≠ F ( x j + x j +1 )

Where x is a generic signal and fi is the frequency of the ith term or component of the signal. F( ) represents the Fourier transform(frequency spectrum) of the signal and ⊗denotes the operation of mixing which is equivalent to taking the product of the

individual transforms of the two signals(i.e. this should not be confused with audio mixing where the two signals are added instead of multiplied) The system response to these inputs will ensure that the mixed output will contain frequencies (fi ) not present in the frequency spectrum of either input signal which a plot of the frequency spectra of the input will show when contrasted against the mixed output. Applying Fourier analysis to the system again and assuming ( as is the case) the signals

may be represented as the sum (or integral) of singular frequency signals, the following result can be derived. i.e. if the signals may be represented as: ∞

∑cos( ωi t ) + j sin( ωi t ) and i =0

∞

∫e

∞

∞

i =0

0

∑cos( ωk t ) + j sin( ωk t ) or

∫e

ωi jt

and

jt ωk

0

(Where e is the sum of cos and sines) Then multiplication of any two sines or cosines of frequencies ωi , ωk yields: sin( ωi t ) • sin( ωk t ) =

1 • (cos{ ωi + ωk }t + cos{ ωi − ωk }t ) 2

This follows from: cos( ωi + ωk )t = cos( ωi ) cos( ωk ) − sin( ωi ) sin( ωk )

Since sines differ from cosines only be a phase delay, the above results state that for any two incoming sinusoidal or cosinusoidal components of two different or similar frequencies, the mixed output will consist of at least two signals one having frequency being the sum of the two signal frequencies and one having frequency being the difference of the two. Other higher order products as well as the original signals will also server as output but are usually not of interest and thus are dealt with in kind. This is the premise of AM modulation which may be done by means other than mixing such as the regenerative method; however regeneration still produces an envelope and so essentially accomplishes the same task with simpler implementation and corresponding

disadvantages. Despite the method, the two sums and difference signals form a modulation envelope which moves as a time varying signal with group velocity vg. The information of the original signal is to be found in these two envelopes as when they are unmixed or demodulated, another sum and difference pair is produced namely: ωi + ωk + (ωi − ωk ) = 2ωi , ωi + ωk − (ωi − ωk ) = 2ωk

Which are frequencies multiples of the transmitted signals. Referring to Fig. 2

it becomes clear that the bandwidth of demodulated signal will range between these two frequencies and thus the spectrum of the signal may be suitably modified with hardware to achieve the desired effect. Turning to the subject of hardware, it is now time to consider the source of the input signals that enter the two port single module system. One source will be from a device producing the signal and information to be transmitted and the other will be from a fixed frequency periodic device that will server as a carrier for that data. In many cases, this fixed frequency periodic signal is generated by an oscillator.

Oscillators may be of a single unchangeable frequency or they may be tuned to oscillate at a particular frequency. Regardless of which oscillator is considered, it must oscillate at a particular frequency for some length of time. There are two main criteria that must be satisfied for electronic oscillation in circuits to occur. A signal must experience some net gain greater than and later equal to 1 and the signal must be fed back for more amplification. The second condition is that the feed back signal must encounter a 360◦ phase shift so as to return in phase and provide positive feedback. Non linearities that develop as the amplifier output approaches its limit normally cause the oscillator to reach a stable amplitude oscillation. Starting of the oscillation may occur in various ways but mainly by means of two. Transients at start up are one method since they cause some initial small oscillations which are amplified and allowed to receive positive feedback. Noise start up is another method and the one most concerning this paper. This method relies on filtering the thermal noise generated in the amplifier so that these fluctuations will be filtered down to a signal of a singular frequency component. The initial noise output is filtered then fed back into the amplifier input to be amplified again and re-filtered. This continues until the above mentioned non linearities take effect.

The tuned filter network may consist of discrete passive reactive elements such as ceramic, electrolytic, or mica capacitors in conjunction with an inductor and/or resistor of some sort. Another option is to use a piezo electric quartz resonator. These devices consist of a crystal that has been shaped so as to resonate at a particular frequency. When the crystal receives a signal, it deforms due to electro restriction. The crystal later regains its original form and in doing so generates piezo electricity in response to the applied physical stress. This deformation and relaxation occurs at a particular frequency causing the device to oscillate. A signal of more precise frequency is output and can serve as a stable fixed frequency which is less affected by temperature. Thermal hysterisis is hysterisis( a requirement that the path to the current configuration be known to attain the next configuration) that corresponds to configurations generated by the thermal noise. This oscillator is less succeptible to temperature and thus has a more tractable hysterisis characteristic. This type of oscillator is an electrical to mechanical to electrical transducer.

Transducers are devices that couple energy from one form to another. They are a core component to any mems system especially an rf mems system and their mems oscillators. The principles of feedback and amplification that hold for oscillation of electronic signals outside the domain of mems hold for mems as well; the major difference is that the implementation of feedback and initial input are now different. A few notable implementations and applications of these oscillators will now be discussed The first example comes from the work of Zalalutdinov et al [2] in regards to an rf mems oscillator for use in an rf mems phase and frequency modulator. The oscillator is thermally actuated via the heat generated within a heating element from the current through it. The transducer used is a circular membrane than is capable of deforming like a drum head due to heat exposure ( not quite like a bi metallic strip but still a property of thermal deformation); thus deflection out of the plane of the circle results. The heat induces this thermal property in the circular membrane. The resulting movement is detected by a HeNe(Helium-Neon) laser which when used in conjunction with the membrane and other structures in the device as an interferometer which will detect the interference patterns produced by the interaction of the laser and the changing gap width between the aforementioned structures. The detected pattern is associated with an output signal from the oscillator. This output is amplified and sent through a feedback network. A dc bias( which may be shifted so that the device supplying it will operate at different bias/Q points) will be superimposed on the feedback signal. Steady oscillation results after the non linearities associated with the mechanical motion such as not being capable of infinite amplitude, take effect. Applying the correct bias will cause the heating element

to generate heat such that the membrane’s mechanical properties change and its resonant frequency along with it and thus the oscillator can be detuned and reset.

Fig. 4 One such application as realized by myself and actually stated by Zalalutdinov et al is the use of such a tunable oscillator in a more compact phased array like those used to electronically steer antenna patterns so as to achieve the same result as if some one were actually adjusting the antenna array by hand. The use of a tunable mems based oscillator would allow the production a frequency that varies with the bias. This would be analogous to using a delay line array as seen in Fig 4. this will provide a delay of the signals arriving at each of its inputs. The phase difference between the signals causes the transmission or reception pattern of the array to shift or rotate once the signals from each array output are added together. This method is suited towards larger systems like radar systems and could be implement for smaller devices by a mems based oscillator which with proper additions, could not only scale down the size of such an array but also provide for (automatic) continuous adjusts of the antenna direction based on signal strength in any particular direction.

Fig 4 courtesy of wikimedia.org Also, as a high quality factor (sharp resonance) tunable frequency source that can be included into ic manufacturing methods, this oscillator serves a good candidate for developing radio on chip communication systems. The next implementation that illustrates the usefulness of mems based oscillators to rf comes from the work of Kubena et al [3]. Although mems based technology can allow the production of much smaller systems at higher frequencies, quartz still has some desirable properties. Among these properties are its status as a low loss high Q piezoelectric material with zero temperature coefficient for certain crystal cuts( cuts of the crystal along a plane with a given orientation in the lattice). Also, the chemical inertness of its surface prevents any undesired reactions from inadvertently changing the behavior of the oscillator. With this in mind, Kubena et al sought to develop a mems based quartz oscillator. More specifically, since current methods of producing quartz resonators do not facilitate scaling them down, they wanted to provide a more straight forward method of producing scaled down quartz oscillators that would work well with rf

electronics. Taking advantage of recent advances in mems micro fabrication processes such as precision wafer bonding and plasma etching, the group achieved benefits they wanted with the size required. In contrast to its discrete counterpart, this mems based device provides the desirable properties of a quartz oscillator with the UHF-VHF performance required for modern communications such as programmable radios and GPS.

Fig. 5 courtesy of Kubena et al. On the left, Fig. 5 displays the completed resonator part of the mems resonator which can be interfaced with rf electronics, such as amplifiers, to complete the mems based oscillator. On the right is a completed view of a filtering network for the oscillator. Both photographs were taken using a scanning electron microscope.

Although rf mems in general is still somewhat exotic, rf mems are receiving notice from the military. NASA is qualifying rf mems oscillators for its space exploration missions.[4] Of particular interest are the applications of such oscillators and other rf components for use in communications during space walks and for deep space probes since the mems longevity matches those required for such probes venturing to the beyond. Another drawing factor to rf mems being used in such probes, according to Aaron Partridge the chief technology officer at SiTime, is their “…insensitivity to the radiation and extreme temperatures of space."[5] The private sector has also shown interest. Companies like Discera Silicon Clocks have gone into production of on chip resonators. SiTime Inc already makes mems oscillators that operate at 125 MHz . Rf mems components will develop systems that will allow the integrating of the DCS, PCS, GSM, EGSM,CDMA, WCDMA, GPS, and Wi-Fi bands or standards while progressively reducing the size of handsets and maintaining battery life.[4] Conclusions: Rf mems oscillators are providing great solutions for the problem of scaling down rf systems. They will allow the development of systems small enough to place on a single chip and will be integrated into systems of mind boggling complexity and compactness. Hopefully, with developments in the field of Nems( nano electromechanical systems) rf Nems will become a prominent successor to rf mems and provide us with even more

fine tuned devices. Until then, mems can only continue to develop and prosper.

Works cited:

[1]Courtney Dimpel,"Mems oscillators mark the beginning of the end for quartz", Hearst Electronic Products, 2007

[2] Reichenbach,R.B; Aubin, D.L; Zalalutdinov, M: Parpia, J.M; Craighead, H.G; "A Microsystems.2005. Digest of Techical Papers. TRANSDUCERS '05 The 13th international Conference on Vol 1, pp 1059-1062

[3]R.L. Kubena, F.P. Strtton, D.T. Chang, R.J.Joyce,T.Y. Hsu, M.K. Lim,and R.T.M Closkey "Mems-Based Quartz oscillators and filters for on chip intergration" HRL Laboratories(Kubenal et al) UCLA (Kim and Closkey)2005

[4]Kamaljeet Singh and K. Nagachenchiah"RF MEMS: MaturingTechnology is GettingReady for Prime Time" High Frequency Electronics, August 2008

[5] R. Colin Johnson "Mems finds niche in space exploration"EETimes Asia,2007

commons.wikimedia.org for phased array image.