Design and performance of millimeter-wave thermocouple sensors - the HP Q8486A and HP R8486A Power Sensors - technical

Hewlett-Packard Journal, April, 1988 by Lee H. Colby

Design and Performance of Millimeter-Wave Thermocouple Sensors

WITH THE INCREASE in design activity in the 33-to-50-GHz frequency range, Hewlett-Packard wanted to extend HP's power measurements into this millimeter-wave frequency range. The HP Q8486A and HP R8486A Power Senors shown in Fig. 1 incorporate a special method that inserts a 50-MHz calibrating signal into the waveguide. The HP Q/R8486A sensors have the same 50-ohn thermocouple used by HP's lower-frequency coaxial power mounts and share the same one-microwatt-to-100-milliwatt power range, low drift, low SWR, 50-MHz calibration, and accuracy that these coaxial thermocouple power sensors have. The HP Q8486A operates in the 33-to-50-GHz waveguide band and the HP R8486A operates in the 26.5-to-40-GHz band.

Since the thermocouple was designed for 50-ohm systems, it was necessary to use a TEM structure for the 50-ohm thermocouple and connect it to a lower-reflection coax-to-waveguide adapter. The adapter's center conductor, which is normally screwed into the waveguide wall opposite the coaxial entry point, is isolated from ground by a high-frequency choke. The choke allows the 50-MHz energy to be fed across the waveguide through the coax to a coplanar transmission line and into the thermocouple. However, for millimeter-wave power, the choke reflects a short at the center conductor where the center conductor is normally screwed into the waveguide. The millimeter-wave signal coming down the waveguide is transformed from the waveguide impedance to 50 ohms because of the transformer action of the multistepped waveguide.

The ability to diplex the 50-MHz and millimeter-wave signals allows the power meter's gain to be set for the sensor head in use and removes the requirement that the thermocouple and its amplifier have a constant gain versus time, environment, and reasonable overloads.

The thermocouple's low output voltage, approximately 60 nanovolts at one microwatt, is chopped at 220 Hz to eliminate dc drift, amplified, and then fed to the power meter.

Calibration information is typed on each sensor's label and includes the calibration factor, which is the efficiency at which the sensor converts the absorbed power into a power reading, and the reflection coefficient. The calibration data is supplied at 1-GHz intervals across the waveguide band and the calibration factors are traceable to the U.S. National bureau of Standards (NBS), or to a dry calorimeter for frequencies where NBS does not have standards. The power mount's microwave performance is tested and calibrated using an automated network analyzer.

Waveguide-to-Coax Transition

The R-band transition shown in Fig. 2 was designed originally as an outgrowth of the lower-frequency X, P, and K281C waveguide-to-coax adapter designs. The Q-band transition was scaled from the p281C design since P band has a similar 2-to-1 waveguide width-to-height ratio.

A combination of empirical and computer-aided modeling was applied in such a manner that the design of the transition and four-step impedance transformer was not done in one step. The adapter design was broken into two parts. First, the coax-to-low-impedance (reduced height) ridge waveguide.sup.1 transition was designed. The waveguide is broadly ridged to decrease impedance variations and increase the bandwidth. Second, sufficient transformer steps were then added to get the low reflection coefficient desired (approximately 0.01). The impedance of the waveguide was modeled using the relationship: Z.sub.o = Z.sub.o*./[square root of]1+([lambda]/[lambda].sub.ca.).sup.2 ohms

where Z.sub.o* = 377([pi]/2)(b/a) ohms (voltage-current definition) for rectangular waveguide at infinite frequency, a is the wide dimension of the waveguide, b is the narrow dimension, [lambda] is the wavelength in free space at the frequency of interest, and [lambda].sup.ca is the cutoff wavelength of the waveguide.

One goal of the coax-to-waveguide junction design was to arrive at a complex reflection coefficient that would look like the input impedance of the last step before the 50-ohm load of a four-step Chebyshev impedance transformer. After substantial empirical adjustment of the short position in the waveguide, ridge length, and other mechanical dimensions, the desired impedance was obtained. Finally the remaining three transformer steps were added. The combination was measured and the measurements were deembedded using a computer program called Opnode to obtain the step discontinuity capacitances. The steps were then optimized for length and impedance with Opnode to obtain the desired 0.01 reflection coefficient, machined, and measured again.

50-MHz and Millimeter-Wave Diplexer

The need to calibrate the sensors at a low frequency requires a mechanical layout of the senor front end that is different from most other waveguide senors. The coax calibration port (see Fig. 3), beginning at the 50-MHz coaxial input at the bottom and moving u pward, consists of a short length of 6-ohm line and a shorted radial transmission line choke.sup.2 whose dimensions are chosen for a high series impedance at the center frequency of the waveguide band on the coax line. This is followed by a quarter-wavelength 6-ohm coax line that transforms the radial choke's high series impedance to a short at the entrance to the waveguide.