ABSTRACT

State-of-the art frequency multipliers, particularly those that operate in the submillimeter wave band, suffer from several limiting factors that make them impractical for modern array applications. In this paper we review the unique requirements placed upon frequency multipliers by array systems, report on the progress of our multiplier development effort, and outline our future development goals toward a suite of multipliers for the NRAO Millimeter Array.

Keywords: frequency multipliers, local oscillators, LO sources, array receivers, Millimeter Array

1. OVERVIEW

The conventional approach to generating local oscillator (LO) power for millimeter and submillimeter wave heterodyne mixers is to generate power at a lower frequency using a suitable phased-locked source, and to convert this power to the desired commensurate frequency using a nonlinear diode such as a varactor in a frequency multiplier circuit. Although useful for single-dish telescope receiver systems, the conventional approach, using current state-of-the-art components, is highly impractical for large array-type radio telescopes for which manageable cost and high reliability are important factors. In this paper, we will examine new approaches to conventional LO generation in the context of the proposed NRAO Millimeter Array (MMA), underscoring the limitations of the current state of the art and highlighting the challenges of contemporary LO designs, with special emphases placed on the next generation of millimeter and submillimeter-wave frequency multipliers.

In order to improve upon reliability and decrease cost, the limitations of the two basic components in the LO system, namely the oscillator and the frequency multiplier, must be examined. For the power source, the mainstay is the Gunn-effect oscillator which has been used successfully for many years because of its adequate output power, inherently low amplitude noise characteristics, and electronic fine tuning which makes it well suited for phased-locked circuitry. However, for large array applications, its usefulness is somewhat compromised since the coarse tuning is accomplished through mechanical adjustment of a high-Q resonant cavity. It is an expensive task to make this mechanical adjustment automated, accurate, repeatable, and reliable. Also, such cavity systems can suffer from unwanted moding which results in narrow frequency bands in which the output power can drop to very low levels. The mechanical tuning is limited in range as well, and, hence, several Gunn oscillators are needed to cover a given waveguide band. The maximum operating frequency of second harmonic Gunn oscillators is about 150 GHz¹, and so to reach millimeter and submillimeter wavelengths, frequency multipliers are required. State-of-the-art multipliers are limited in performance because of several factors, including: 1) narrow instantaneous bandwidth requiring mechanically-adjustable tuning structures that may reduce reliability , 2) low conversion efficiency leading to difficulties in power distribution, 3) use of point-contacted varactors which are mechanically fragile structures, and 4) intricate mechanical details making component assembly rather difficult. Overcoming these limitations is essential if conventional LO systems are to be made practical for contemporary radio telescope arrays.

All-electronic LO tuning has the advantage of improving reliability for array systems at a modest cost. The most useful all-electronic power source up to 50 GHz is the Yttrium-Iron-Garnet (YIG) tuned FET oscillator (YTO). The tuning is over a very broad band and it can easily be phased-locked to a reference source. The YTO, followed by wideband, fixed-tuned frequency multipliers and low noise power amplifiers, form a viable alternative to the Gunn-effect oscillator. Wideband monolithic HFET power amplifiers are becoming increasingly more common up to 100 GHz primarily due to current military and commercial demands for systems operating in this frequency range. However, it is the wideband frequency multiplier technology that has lagged behind in development. Advances in this area will determine the success of future YTO-based millimeter and submillimeter wave LO systems.

Over the past few years, the single most important factor influencing future frequency multiplier development was the advent of versatile computer-aided design packages enabling the design engineer to analyze complex electromagnetic structures², create and simulate detailed equivalent circuit models (such as Hewlett-Packard's Microwave Design System), and predict semiconductor transport properties³, all to a high degree of accuracy. For the first time, the nonlinear dynamics of the varactor, the electrical properties of the semiconductor package, and the embedding circuitry of the multiplier can be analyzed together as a complete frequency multiplier circuit. Upon applying such tools, one begins to understand the reasons behind the limitations of existing multiplier designs, thus opening the door to exploring new approaches and techniques never before possible in order to meet the stringent demands placed on LO systems by next-generation telescopes such as the MMA.

2. A PROPOSED LO SYSTEM FOR THE MMA

The NRAO MMA is a good example of a new generation array radio telescope whose requirements will drive millimeter and submillimeter-wave local oscillator development for years to come. Reliability is an important issue not only because of the number of components required but also due to the remoteness of the observing site⁴. Reliability can be greatly enhanced by using all-electronic tuning and by replacing the fragile point contact varactor with the more rugged planar varactor. Due to the relatively large current densities in varactors, anode temperatures can reach well over 100 degrees C above ambient, thus compromising the long-term lifetime. The lifetime can be increased indefinitely through the use of cryogenic cooling which is typical in modern receivers and therefore should not increase the cost of the LO. However, the cost of building frequency multipliers is rather large due to the current complexity of the micro fabrication required. This cost can be reduced substantially at the circuit design stage by using monolithic (MMIC) technology, minimizing the machining operations required, and reducing the need for close tolerances during machining steps so that efficient duplication can be achieved. Finally, the higher frequency multipliers should be designed as cascaded components of doublers and triplers for interchangeability. Further requirements on frequency bands, output power, amplitude noise, and phase noise will now be examined.

2.1 Frequency requirements

A proposed band plan for the MMA is shown in Table 1⁵. The first three columns indicate the type of receiver, either HFET for the lower frequencies or SIS for the higher frequencies, the RF band, and the RF band-delimiting frequency ratio defined as f_max/f_min for each band. The highest frequency band is a future possibility and is not planned to be implemented as part of the initial construction. Assuming an IF band from 4-12 GHz, columns four and five show the LO tuning range required and the LO band-delimiting frequency ratio defined as f_max/f_min for each LO range. A prime feature of this plan is that all LO frequencies above 60 GHz can be derived from two phase-locked sources: #1 covering the range 60-90 GHz, and #2 covering 80-120 GHz.

2.2 Power requirements

A specification for the LO power level is derived from the pump power required by the SIS mixers which is approximately 1 W. In the worst-case scenario where only single-ended SIS mixers are used, a waveguide or quasi-optical LO coupler, having a coupling factor of -20 dB, will be required to combine the LO with the RF signal. As a result, the amount of LO power required at the input of the receiver will be approximately 100 W. An estimate of frequency conversion efficiencies that form realistic yet challenging goals for new broadband, fixed-tuned, planar frequency multiplier designs is given in Table 2. The first three columns give the LO tuning range from Table 1, the driving source along with its tuning range, and the multipliers needed. Columns four and five give the multiplier efficiency and output power for a driving power of 50 mW.

The output from a single phase-locked source will be split between the two channels of a dual-polarization receiver. This 3 dB power division at the drive source output implies the following: 1) given the current state of the art in power amplifiers, the maximum output power of the all-electronic source will be about 100 mW, yielding the required 50 mW of drive power per receiver polarization, 2) a separate frequency multiplier chain will be required for each receiver polarization, 3) each multiplier chain could be tied to the cryogenic refrigeration system to improve conversion efficiency and increase varactor lifetime, and 4) an LO leveling circuit, perhaps using the SIS mixer current in a servo loop while adjusting the bias current on the frequency multipliers, will have the dynamic range shown in column six. This dynamic range can be increased substantially if balanced mixers are used⁶, since the RF and LO ports will then be separated, eliminating the need for the -20 dB LO coupler. With this configuration, the dynamic range will not be compromised at the higher operating frequencies.

2.3 Amplitude noise

The specification for amplitude noise is to meet an acceptable value for the gaussian noise that will be added to the front-end noise of the receiver by the SIS mixer LO. The contribution of LO noise to the HFET front-ends will be negligible. The mixer LO noise manifests itself as noise sidebands associated with the CW source, but far enough away from it that the noise will ultimately appear in the RF passband of the receiver. It has been suggested⁶ that the LO amplitude noise contribution to the noise temperature of a single-ended SIS mixer be limited to one degree Kelvin, and with a typical LO pump power of 1 W per mixer, a specification of 1 Kelvin per microwatt is therefore defined. If the CW LO source power is assumed to be confined to a 1 Hz bandwidth, this specification translates to an LO signal-to-noise ratio (SNR) of approximately 168 dB at the input port of the mixer. A relatively low-Q bandpass filter, centered about the signal frequency, can be used to reduce this noise if needed. If balanced mixers are used, the specifications for the filter can be relaxed in proportion to the LO isolation that is provided, typically on the order of 10 to 20 dB, or perhaps eliminated entirely.

There has been some concern about the amount of phase and amplitude noise generated by power amplifiers. However, recent measurements by J. Kooi⁷ on a 92-97 GHz MMIC power amp indicate this additional noise is quite small.

2.4 Phase noise

The dominant contributor to the phase fluctuations encountered by the MMA will be atmospheric fluctuations along the line of sight of the instrument. Two distinct methods⁸ are currently being considered for phase calibration: 1) Fast Phase Calibration (FPC) and 2) Radiometer Phase Correction (RPC). FPC involves observing an astronomical calibrator on time scales short enough that the atmosphere does not change significantly between calibrations and using calibrators close enough to the target source that the path through the lower troposphere responsible for the phase fluctuations is essentially the same. FPC corrects for fluctuations on time scales longer than the time period between calibrations. The 11.2 GHz satellite phase monitors currently used for site surveys have baselines of 100 m and 300 m and thus directly sample the atmospheric structure function appropriate for evaluating the utility of the FPC technique. The technique works because the dominant atmospheric fluctuations correspond to structure in the atmosphere on the scale size of the baseline and on time scales given by the wind crossing time. An added benefit is that calibrating on a 10 second time scale greatly decreases the effect of temperature and other slow fluctuations on the phase stability of the antenna and electronics. However, the most likely frequency for calibration will be the 90 GHz band or perhaps even lower due to the limited availability of suitable astronomical calibration sources at millimeter wavelengths, so it is important that the LO system be built such that phase calibration at one frequency will also correct for drifts in the LO chains for the other bands as much as possible.

RPC is based upon scientific evidence that the dominant source of path length variations at millimeter wavelengths is fluctuations in the water vapor along the line of sight and that this water vapor can be measured by its continuum or line emission. This method requires radiometer/receiver systems with sufficient stability. RPC works on time scales of approximately one second up to the time scale set by the radiometer stability. A special radiometer, dedicated to monitoring the 183 GHz water line, is currently being considered.

Based upon the requirements of the proposed phase calibration methods, as well as the need to use holography to measure the surface features to the desired accuracy of better than 8 m (wavelength/100), the resulting phase stability specification for the electronics, as defined by the M.C. Phase Calibration Working Group, is 3 degrees over time scales greater than 1 second. This specification will be used as a guideline for M.A. LO development. However, specifications for allowable phase fluctuations on time scales shorter than 1 second have not yet been defined. Integration-and-dump times, associated primarily with mosaicing operations, will be the defining factor for this specification.

3. ALL-ELECTRONIC TUNING FOR THE PHASE-LOCKED SOURCES

The two phased-locked LO systems will now be described. The RF oscillator is shown in Figure 1. The 100 mW, 60-90 GHz oscillator consists of a 10-15 GHz YTO followed by a doubler to 20-30 GHz, a 10 dB power amplifier, a tripler to 60-90 GHz, and a second 10 dB power amplifier. The 80-120 GHz oscillator consists of a YTO, doubler, and power amplifier to 20-30 GHz as in the previous case, but here it is followed by a doubler to 40-60 GHz, a 9 dB power amplifier, and a high power doubler. Both systems will require appropriate interstage filters.

Figure 2 shows a block diagram of the phase-locking circuitry and reference source that is common among the various systems. The output frequency of the unit is phased-locked to the sixth or eighth harmonic of a 10-15 GHz frequency synthesizer in the case of either 60-90 or 80-120 GHz operation, respectively. The synthesizer is, in turn, phased-locked to a reference source. The output of the harmonic mixer is the difference frequency of 100 MHz which is filtered and amplified before entering the In-Phase (I) & Quadrature-phase (Q) detector. The other input port of the phase detector receives a signal from a 100 MHz phase reference that contains the required frequency offsets and phase changes for fringe rotation, sideband separation, and phase switching. The I and Q output of the phase detector are both low-pass filtered. The signal in the I channel, upon integration, is used to adjust the RF oscillator fine tuning to complete the phase-locked loop. The signal in the Q channel drives a lock detection circuit, which in turn controls the ramp generator used as a search oscillator. The search oscillator sweeps the fine tuning control voltage until the RF source is within the lock-in range of the loop. Upon successful lock, the search oscillator is disconnected and lock is indicated.

4. FREQUENCY MULTIPLIERS

The next generation of frequency multipliers will continue to be based upon varactor nonlinearity but will use a higher degree of integration of the multiplier circuit together with the varactor. The abrupt junction varactor will be the mainstay for both doublers and triplers but the quantum-barrier varactor (QBV), with its symmetrical capacitance-voltage function thus eliminating the need for an idler circuit at the second harmonic, shows promise for use in triplers⁹. At millimeter wavelengths, we have recently been successful in developing a 40/80 GHz balanced doubler consisting of several planar abrupt-junction varactors fabricated together on a single semiconductor chip which is then mounted across the input waveguide¹⁰. The results are shown in Figure 3 for room temperature operation. The peak efficiency increased to more than 60 percent upon cooling the doubler block to 20 K.

Development of balanced doublers based on this circuit topology is in progress for designs for 55/110 GHz and 110/220 GHz. Figure 4 shows a sketch of the 110/220 GHz design that is currently being fabricated at NRAO.

At submillimeter wavelengths, discrete varactor packages become electrically large and waveguide circuits become increasingly more difficult to machine. The monolithic (MMIC) approach is the natural progression towards fully-integrated multiplier circuits¹¹. We are currently pursuing such a design for 220/660 GHz. However, controlling unwanted substrate moding and working around the current saturation problem make such high frequency designs challenging.

We wish to acknowledge John C. Webber and A. Richard Thompson for discussions leading to the present MMA conventional local oscillator plan. We also wish to thank Kamaljeet S. Saini and Donald G. Stone for their contributions to the conventional frequency multiplier development effort at NRAO.

1. M. F. Zybura, S. H. Jones, B. W. Lim, J. D. Crowley, and J. E. Carlstrom, "125-145 GHz stable depletion layer transferred electron oscillators," Solid State Electronics, 39, 1996.

2. M. S. Mirotznik and D. Prather, "How to choose EM software, IEEE Spectrum, 34, pp. 53-58, December 1997.

3. R. E. Lipsey, S. H. Jones, J. R. Jones, T. W. Crowe, L. F. Horvath, U. V. Bhapkar, and R. J. Mattauch, "Monte Carlo harmonic-balance and drift-diffusion harmonic-balance analysis of 100-600 GHz Schottky barrier varactor frequency multipliers," IEEE Transactions on Electronic Devices, 44, pp. 1843-1850, Nov. 1997.

4. P. Napier and J. West, "High Altitude Medical and Operations Problems and Solutions for the Millimeter Array," Proc. SPIE, 3349, 1998.

5. J. C. Webber, A. R. Kerr, S.-K. Pan, and M. Pospieszalski, "Receivers for the millimeter array," Proc. SPIE, 3357, 1998.

6. S.-K. Pan, private communication.

7. Private communication, January 1998.

8. D. Woody, M. Holdaway, O. Lay, C. Masson, F. Owen, R. Plambeck, S. Radford, and E. Sutton, "The MDC phase calibration working group report," NRAO MMA Memo 144 (http://www.cv.nrao.edu), Oct. 3, 1995.

9. J. R. Jones, W. L. Bishop, S. H. Jones, and G. B. Tait, "Planar multibarrier 80/240 GHz heterostructure barrier varactor triplers," IEEE Trans. on Microwave Theory and Techniques, 45, pp. 512-518, April 1997.

10. D. W. Porterfield, T. W. Crowe, R. F. Bradley, and N. R. Erickson, "A high-power, fixed-tuned, millimeter-wave balanced frequency doubler," submitted to IEEE Trans. on Microwave Theory and Techniques, Jan. 1998.

11. R. F. Bradley and R. J. Mattauch, "Planar monolithic Schottky varactor diode millimeter-wave frequency multipliers," Technical Report RL-TR-92-187, Rome Laboratory, Air Force Systems Command, Griffiss Air Force Base, NY, June 1992.

1. * Further author information -

^{R.F.B. (correspondence): Email: rbradley@nrao.edu; Telephone: 804-296-0291; FAX: 804-296-0324. D.W.P.: Email: dwp8j@virginia.edu; Telephone: 804-924-6575. The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc.}