Page images
PDF
EPUB

Semiconductor Measurement Technology:
Modulation Measurements for Microwave Mixers

James M. Kenney*

Center for Electronics and Electrical Engineering
National Bureau of Standards

Washington, D.C. 20234

The measurement of mixer conversion loss using periodic or
incremental modulation of the local oscillator, and the evalua-
tion and minimization of the associated systematic and random
uncertainties, are discussed in terms of an X-band mixer mea-
surement system constructed at NBS. It is shown that the sys-
tematic uncertainty in the incremental modulation method of
measuring conversion loss results largely from the uncertain-
ties in the calibration of microwave attenuation and power.
It is also shown that the "modulation" (periodic modulation)
and "incremental" (incremental modulation) methods of measuring
conversion loss are essentially identical, the only practical
distinction being in the somewhat different instrumentation re-
quired by the different modulation rates.

Several improvements in the periodic and incremental modulation
techniques are introduced. Novel circuits for measuring
intermediate-frequency output conductance and local-oscillator
return loss are described which may also be useful for other
immittance measurements.

Key words: Conversion loss; diode measurement; intermediatefrequency (i-f) output conductance; measurement uncertainties; microwave mixer diodes; modulation; point-contact diodes; reflectometer; return loss; Schottky-barrier diodes; semiconductor diodes; standing-wave ratio (SWR).

[blocks in formation]

The manufacturers of microwave mixer diodes currently lack easily traceable standards with which to calibrate their production line measurements. There are no standards laboratories to provide diode calibrations, as were once maintained by the Department of Defense.

There are no specified measurement tolerances, and no allowance is made for measurement uncertainty at the specification limits as required by MIL-STD-750B (4.2.3), which also requires an annual calibration which is not being performed at present. As a result there are large uncertainties in the absolute values of the measured parameters, even though manufacturers attempt to maintain a uniform product by comparison with groups of diodes measured in previous years. These uncertainties have, in turn, frequently resulted in appreciable differences in the average quality of diodes made by different manufacturers to the same specification limits.

* Now with Polymer Science and Standards Division, Center for Materials Science, National Measurement Laboratory, National Bureau of Standards.

As a consequence of these differences, competitive bidding based only on price cannot be conducted fairly, and inferior diodes are being purchased for replacement use.

Compounding these problems is a lack of detailed descriptions of modern measurement methods in the literature, and an influx of new diode producers unacquainted with traditional orally transmitted practices. In addition, very rarely is there any incoming inspection by the Department of Defense and most other users, who are even more handicapped by the lack of standards.

Microwave mixer measurements have traditionally been regarded as something of a "black art," presenting great difficulties in obtaining reproducible or even repeatable results, and with reputed systematic differences among competitive measurement techniques. Unfortunately, there has also been a tendency for the original diode manufacturers to regard their measurement techniques as proprietary, with no consideration made for the possibility of remeasurement by diode users, thus making it difficult to establish common standards. Manufacturers are understandably reluctant to accept new standards that may affect yield or force revisions of specification limits. The standard (MIL-spec) point-contact diodes, which still dominate the mixer field, have been reduced in price to the extent that there is little incentive for industry to improve measurements. Perhaps as a consequence of this experience, there has been little effort at standardizing Schottky-barrier types, although many are made with identical specifications.

The primary objectives of this program were to refine and evaluate existing and new microwave mixer measurement methods so that satisfactory accuracy and repeatability could be obtained. The program also sought to establish procedures for determining the accuracy and repeatability of these measurement methods.

1.1 Background

An important phase in the history of semiconductor electronics was the revival of the point-contact semiconductor "crystal detector" for microwave radar receivers long after it had been discarded in favor of the vacuum tube for lower frequency reception. Two types of microwave diodes were developed for radar use: one for use in the heterodyne conversion of microwave signals to a lower intermediate frequency, commonly called "mixing" (not to be confused with the linear addition of signals at aud.o frequencies which is also called "mixing"), and the other for use in direct demodulation of the microwave signals, commonly called "video deThe physical construction of these two types is quite similar, the only major difference being in the different methods of testing for these applications. Mixer diodes have been by far the more commonly used and have therefore received the greatest attention in the development of improved types. Their high usage has also resulted in a relatively low unit cost. For these reasons it is common for mixer types to be used in detector applications even though they have not been evaluated for this use. Schottky-barrier diodes are now being made for use as micro

wave mixers and detectors, but the much lower prices of point-contact diodes and the large amount of existing equipment designed for their use have supported the continued large production of the latter. The measurement methods devised for point-contact diodes are equally applicable to Schottky diodes, since the measured parameters are those of the entire mixer: the diode in a holder (mount) with local oscillator drive. If the holder and local oscillator meet the appropriately chosen specifications, the mixer characteristics can be attributed to the diode.

1.2 Mixer Parameters

Although the performance of a mixer at high signal levels (e.g., in terms of its intermodulation distortion) is of some interest, the most important characteristics of a mixer are generally those concerned with its ability to heterodyne very low-level signals with a minimum degradation of signal-to-noise ratio. As with any low-gain multiport network, a mixer can degrade signal-to-noise ratio in three ways:

(1) By attenuating the signal (or merely by failing to amplify it) so that the noise added by following stages (the intermediate-frequency amplifier) is significantly large by comparison with the noise accompanying the signal. The parameter expressing this gain factor is known as "conversion loss." (Note: Precise definitions of mixer parameters are given in Appendix D.)

(2) By adding noise originated by the mixer. This added noise has two basic components: (a) the "essential" (thermal) noise that would be added by any completely passive network having the same loss while in thermal equilibrium at the same ambient temperature, and (b) the "excess" (shot) noise added by the active device (diode). The essential noise is determined by the conversion loss and the physical temperature. The excess noise is commonly characterized by "output noise ratio." Modern high-quality mixer diodes, particularly of the Schottky-barrier type, generally contribute very little excess noise.

(3) By adding noise which enters via unused "spurious" frequency conversions ("spurious ports"). The primary spurious conversion for a mixer is its image response. "Signal" and "image" are arbitrary designations for the two frequencies on opposite sides of the local-oscillator frequency which differ from it by the magnitude of the intermediate freDue to the difficulty in selectively and reproducibly blocking the image by reactive termination, the mixers used in diode testing are broadly tuned to the local-oscillator frequency and are terminated in a line-match at both signal and image frequencies. Thermal noise from the image termination (at a "standard reference temperature," see Appendix D for definitions of terms) is accepted as a part of the mixer (redefined as a two-port which includes the termination of the spurious ports). Noise into the image port which is in excess of this standard noise temperature must be accounted for in the measurement. (The signal and image responses are identical in the broad-band mixer.) Other spurious responses may occur at frequencies in the vicinity of local oscillator harmonics, but this complication has not been investigated. In general, the

termination at the radio-frequency (r-f) port of the mixer should provide a good line-match at the signal, image, and local oscillator frequencies, and at all higher frequencies up to several times the local-oscillator frequency. Such a match not only insures reproducibility of the termination thermal noise, but also of the conversion loss, which depends upon the distribution of signal power among the various terminations and upon the match to the local-oscillator power and its harmonics. (Return of those harmonics to the mixer generating them could affect the conversion loss to an unknown degree depending upon the amplitude and phase of the reflection. The phase dependence would make a reflecting termination difficult to reproduce. The importance of the harmonic termination has

not been investigated.)

An additional important way in which signal-to-noise ratio can be degraded is by the presence of noise at the signal and image frequencies contributed by the local oscillator. A single-ended (one-diode) mixer is used for diode testing, and the local-oscillator noise adds directly to the excess mixer noise. A narrow-band bandpass filter must therefore be used between the local oscillator and the mixer. Balanced mixers having two or more diodes are used in receivers; these provide local-oscillator noise rejection by an opposing-phase relationship between diode outputs at the intermediate frequency for noise accompanying the local oscillator, while maintaining an in-phase relationship for conversion from the signal port. Some mixers, of still greater complexity, also provide rejection of the image frequency, thereby making it possible to tune the mixer over a wide range by varying the local-oscillator frequency without also having to vary a preselector (image-rejection) filter.

A detailed analysis of the way in which the signal-to-noise ratio is degraded by the various factors mentioned above is not given. It can be seen, however, that the most important single factor associated with the diode is conversion loss. It can be shown that conversion loss degrades the signal-to-noise ratio by a factor equal to itself. The signal-tonoise ratio degradation by a network for a standard input noise is its noise figure; for finite bandwidths, it is referred to as "average noise figure," F, which is the gain-weighted average of the (spot) noise figure [1]. The overall average noise figure, Fo, of the mixer and the intermediate-frequency amplifier, in terms of mixer conversion loss, L, mixer output noise ratio, N, and intermediate-frequency average noise figure, Fi is FL (N + Fi = L (N + Fi - 1), with all quantities expressed

as power ratios (not in decibels). Since L always appears as a multiplying factor, the overall average noise figure in decibels is always the sum of the loss in decibels and a noise term in decibels.

Mixer diodes are marketed with overall average noise figure limits in 0.5-dB steps. Suffix letters are added to the generic designation to denote the diode grade. The better diodes have a small excess noise contribution (N≈ 1) which is not greatly different for the various diode grades. The 0.5-dB steps in overall average noise figure therefore result largely from 0.5-dB steps in conversion loss. However, this situation has not been reflected in the specification limits for conversion loss itself, which is still specified at the 6-dB level established for

most earlier high-noise-figure types. This came about because of a decision by the diode industry, concurred in by its military customers, to require 100-percent inspection based only on overall average noise fig

This decision followed the development of gas-discharge r-f noise sources which permitted overall average noise figure to be determined by a single measurement (in contrast with a piece-wise determination using the equation given above). Unfortunately, the usefulness of these r-f noise sources as absolute standards has been limited by their apparent lack of stability (possibly resulting from cold-cathode starting or to lack of precision in setting and maintaining the discharge current) and the limited calibration services available. In practice, diode suppliers tend to use r-f noise sources for relative measurements, with sets of previously measured diodes as their reference.

1.3 History of Mixer Standards

In the 1940s and 50s, coordinated diode calibration services were maintained by the U.S. Army Signal Corps Laboratories at Fort Monmouth, N. J., and by the Material Laboratory of the N. Y. Naval Shipyard, Brooklyn, N. Y.* These laboratories also maintained the primary standard holders against which other diode holders could be tuned. Following the de facto termination of these services, fixed-tuned holders reproducible solely from their mechanical dimensions were developed under military contract. It was hoped that these fixed-tuned holders, together with the development of the gas-discharge r-f noise source, would compensate for the lack of the calibration services. As has been mentioned, the r-f noise source proved to have limited usefulness as an absolute standard. The fixed-tuned standard holders were found by diode suppliers to be insufficiently reproducible, although there is disagreement between these suppliers as to the relative merits of the several holder types (one in each band) developed to date. The result has been that the major diode suppliers rely almost entirely upon their historic internal "standard" diodes for calibration of their measurements, and strive to maintain a uniform product rather than conform to an absolute standard. One supplier expressed the belief that this practice has resulted in an absolute drift of as much as 1.5 dB in the overall noise figure limit for some diode types. The development of the Schottky-barrier diode for microwave mixer use has brought new suppliers into the market. This and tightened military specifications requiring control of measurement uncertainty have made the lack of absolute standards a significant problem, both to industry and to the government. The NBS mixer measurement program resulted from expressions of concern by industry, through the JEDEC High-Frequency Diode Committee JS-3, and from support by the Naval Electronics Systems Command. Unfortunately, the latter agency was forced to curtail its intended support in 1972 because of funding difficulties (which affected the standards area generally), and NBS was unable to obtain sufficient expressions of need for this work from the private sector to warrant the use of Department of Commerce funds. The mixer program

These calibrations were started at the MIT Radiation Laboratory during World War II (Military Standard Test Methods for Semiconductor Devices, revised May 1970, p. 265).

« PreviousContinue »