Millimeter-wave frequency bands hold promise for both military and commercial applications, including broadband communications links. But development of technologies for the miniaturization of millimeter-wave modules has lagged behind advances in transmission-line and materials technologies at lower frequencies. Circuit losses and parasitic elements at frequencies above 10 GHz pose challenges in terms of achieving miniaturization and compact three-dimensional (3D) circuit structures with the performance levels needed for millimeter-wave frequency designs. But a novel transmission-line technology developed by Nuvotronics, known as PolyStrata, appears promising in terms of offering high circuit density with low crosstalk and signal loss.

PolyStrata is an air-dielectric shielded transverse-electromagnetic (TEM) transmission-line technology with low loss and crosstalk compared to conventional microwave stripline and microstrip transmission-line technologies. It was first developed with the help of funding from DARPA as part of the three-dimensional (3D) micro-electromagnetic radio-frequency systems (MERFS) program. The program included such goals as reducing the costs of phased-array data links and fabricating an antenna on a wafer.

PolyStrata circuits allow as much as 3 m of transmission line to be packed into a cubic centimeter of circuit board volume for extremely compact constructions. Attenuation of a PolyStrata 50-Ω line can be less than 0.1 dB/cm at 35 GHz . Because of this feature, as well as its miniaturization and size, weight, and power (SwaP) benefits in military electronics systems, PolyStrata offers a means of achieving higher power/efficiency transceiver modules. PolyStrata coaxial structures can be useful in constructing passive components, such as filters and power combiners/dividers, feed networks, and hybrid components.

PolyStrata features a 99.8% air dielectric with small dielectric straps that suspend the center conductor. Because of the unique suspended construction, it would be helpful to know how the transmission-line technology behaves under conditions of high vibration, as might be experienced in some harsh environments (including avionics systems). Until performing tests in the current study, little was known whether the center conductor would micro-phonically add phase noise during vibration. After performing phase-noise measurements under vibration, the PolyStrata circuits were found to exhibit little effects of vibration, with the phase-noise results actually dominated by the noise contributions of the test cables in the measurement setup.

PolyStrata components are fabricated by sequentially depositing thick photoresist layers, plating copper, and performing chemical mechanical planarization (CMP). The simplest coaxial transmission lines can be fabricated in five layers, while more complex structures, such as RF crossovers, can be constructed with as many as 15 layers. Figure 1 offers an example of how a four-port coupler might be realized. The layers can be between 30 and 200 m thick. The smallest coaxial cross section that can be fabricated is typically 250 m, and the largest (with the lowest loss) is about 1 mm tall. Wider rectangular coaxial lines can also be produced. While the power-handling capability of an air-dielectric coaxial line is thermally limited by the resistive heating of the center conductor, over 75 W CW microwave powerhandling capability has been reported for 300-m coaxial-line cross sections formed with PolyStrata.

In order to evaluate the performance of this technology under various operating conditions, several samples of 50-Ω air-dielectric coaxial lines were obtained from Nuvotronics for testing. The coaxial samples were characterized by outer cross-section dimension of 0.5 mm, with a 0.1-mm center conductor. One sample was tested in a brass "stop-sign" test fixture (Fig. 2), which can accommodate four-port circuits and was fitted with 2.92-mm connectors for operation to 40 GHz. For comparison, a control sample with 50-Ω microstrip on alumina substrate was prepared and mounted in an identical fixture for testing.

PolyStrata samples were processed on a high-resistivity silicon (HRS) wafer. In typical use, these air-dielectric coaxial structures would be released from the HRS substrate. But to ease handling and mounting, they were mounted with the HRS substrate directly in the test fixture. The sample for testing included 10 separate 50-Ω lines of approximately 1-cm length. The RF grounds required at both ends of the sample were accomplished by means of a pair of coplanar-waveguide-tomicrostrip transitions from J-Micro Technology. Wire-bond interconnects to the PolyStrata sample were staked down with conductive epoxy, rather than thermosonically attached. Since the center conductor of PolyStrata coaxial lines is suspended by means of miniature dielectric straps, and an in-house wire-bonding process was not available, this approach helped minimize damage to the test sample.

A recess was cut into the fixture so that the top surface of the PolyStrata sample would line up with the fixture's alumna interconnects. Even with this modification, the sample is in a less-than-ideal situation, with long wire bonds to the center conductor and ground connections. During attachments at one end of the sample, connections were inadvertently misaligned, most likely explaining limited small-signal response at frequencies beyond X-band (although this misalignment did not factor in the vibration experiment). Figure 3 shows the S-parameters for the test sample prior to vibration. At 10 GHz, the input return loss is about 9 dB (approximately a 2.0:1 VSWR). Phase-noise measurements under vibration were performed at 4 and 10 GHz, although there was little reason to believe that microphonic effects investigated in this experiment would be any different at millimeter-wave frequencies.

Page Title

For evaluating the test samples under vibration, a commercial vibration (shake) table (Fig. 4) and controller were used. The controller was a model VWIN-II from Unholtz-Dickie. The table provides vibration in the z-axis. Its vibration levels are detected by means of accelerometers mounted on a device under test (DUT). A total of 16 input channels were available for programming vibration, although only four channels were used.

The DUT's RF test fixture was bolted to an adapter plate (approximately a 2-in. square of 0.25-in.-thick aluminum) with four 2-56 stainless-steel screws. The adapter plate was mounted to holes added to the vibration fixture, with four -20 bolts. The vibration fixture is tightly screwed to the vibration table with eight -20 bolts. For vibration testing, it is important that the equipment be rigid within the vibration spectrum (no acoustic resonances), and that nothing comes lose. Four accelerometers were configured, two on the adapter plate, and two on the vibration fixture (the red wires in Fig. 4).

A vibration profile was chosen typical of a smart munitions device on its way to a target. The profile consisted of a ramp in vibration frequency from 10 to 100 Hz, followed by a flat power spectral density (PSD) response to about 2100 Hz. Once programmed, the controller maintains the profile within tight tolerances. Figure 5 shows the measured vibration output from the accelerometer at a level of about 4 g's root mean square (RMS).

Phase-noise levels were measured with a model PN 9500 wideband noise and jitter analyzer from Aeroflex configured for residual phase-noise measurements. The test signal was supplied by a model 83711B synthesizer signal generator from Agilent Technologies. A model HP 11667 resistive power splitter (DC to 18 GHz) from Agilent was used to create two identical versions of the test signal, while a coaxial phase shifter from 3.5 to 12.4 GHz from Narda Microwave-East was used to align the reference and DUT signals in quadrature (as this is a residual phase noise measurement).

Coaxial test cables from Teledyne Storm Products were also used in the test system; efforts were not made to find special low-vibration, low-phase-noise test cables. As a result, the measured data will reveal that the phase noise of these test cables dominates any contributions from the PolyStrata air-dielectric coaxial lines.

With the test setup at rest, the residual phase noise of the system was measured at a respectable -150 dBc/Hz offset 1 kHz from the carrier (Fig. 6). Initially, the first vibration fixture (Fig. 4) had a severe resonance problem. A 1-kHz resonance was found in the residual phase-noise data for the control sample measured at 2 g's RMS (Fig. 7). This was attributed to the plate being oversized compared to the bolt pattern, in spite of the plate being nearly 1-in. thick. The resonance results in a deafening audible noise, comparable to the sound of a high-amplitude, somewhat detuned C note two octaves above middle C. This fixture did not yield repeatable data, so it was modified with a second, thicker plate.

Figure 8 compares the test results for the PolyStrata samples and the control circuits at 1 g RMS vibration, while Fig. 9 offers a comparison of test results for the two transmission-line technologies under 4 g's RMS vibration. The vibration profile clearly increases the noise across the vibration bandwidth of 10 to 2000 Hz; however, the effect can be attributed to the flexible cables that are used to route the RF signals to the DUT. Flexible coaxial cables are notorious for their microphonic conversion of vibration to phase noise, perhaps due to the outer jacket being stretched slightly when it is displaced. No systemic differences were found between the control sample and the PolyStrata sample at either 1 or 4 g's RMS.

Measurements were attempted at a higher vibration level of 10 g's RMS, but it was found that the epoxy used to hold the PolyStrata sample to the test fixture failed and the part came loose. The epoxy was chosen for its quick drying characteristic, although slower-drying options may have provided better adhesion and allowed a continuation of the experiment.

The measured results offer no evidence of degraded phase-noise performance for the PolyStrata sample during the 1 and 4 g's RMS vibration tests. It should be noted, however, that the samples may have suffered from microphonics that were not detected with the test setup. The phase noise from the test cables dominated the phase noise of the test system, including the DUTs. The vibration spectrum at 4 g's RMS is representative of the vibration intensity used to qualify many aerospace microelectronics, so these initial results for PolyStrata are encouraging. When computer-aided-engineering (CAE) models developed by Nuvotronics based on ANSYS software was used to analyze the PolyStrata structures, the models indicated that resonant modes should not appear until vibration frequencies of 70 kHz or more, independent of the length of the transmission lines.