A Versatile Test Bench for Wireless RF/Microwave Component Characterization
A low cost RF/microwave test bench comprising a network analyzer, spectrum analyzer and signal source fro measurements between 0.3 and 3 GHz
A Versatile Test Bench for Wireless RF/Microwave Component Characterization
Recent advances in instrumentation, motivated by the volume and cost demands of the commercial wireless marketplace, allow a powerful RF/microwave test bench to be assembled at relatively low cost. A versatile test bench comprising a vector network analyzer (VNA), spectrum analyzer (SA) and signal source is described, along with test techniques for a wide range of RF/microwave measurements in the 0.3 to 3 GHz frequency range.
Lawrence P. Dunleavy, Thomas M. Weller, Paul G. Flikkema, Horace C. Gordon, Jr. and Rudolf E. Henning
University of South Florida, Department of Electrical Engineering
Wireless product development requires many types of RF/microwave characterization measurements. These measurements cover the gamut from on-wafer testing to final system performance verification of a completed wireless product, such as a pager, cellular telephone or two-way radio. A well-equipped RF/microwave characterization laboratory would include several test stations with a cost of over $100K each. However, if measurements below 3 GHz are the focus, these needs can be satisfied using the few basic microwave instruments configured in the low cost test bench described in this article.
Here, the primary interest in a versatile low cost test bench is for undergraduate and graduate wireless-related education. However, the described techniques and equipment are equally applicable to design and manufacturing applications. Assisted by grant funding from the National Science Foundation and Hewlett-Packard Co., the University of South Florida (USF) has recently established an innovative new Wireless and Microwave Instructional (WAMI) laboratory.1 The described equipment is used on six identical benches in this new learning environment for a required undergraduate course entitled "Wireless Circuits and Systems Design" as well as senior-/graduate-level electives, including a course entitled "RF and Microwave Measurements."
The goal of these courses, and the purpose of this article, is to provide exposure to a wide range of measurement techniques of interest to the wireless RF/microwave circuit and system designer. Specific test configurations are described for several example measurements of interest, including basic S-parameter measurements, swept amplifier compression, mixer port match, isolation and conversion loss, and third-order intercept (TOI).
Test Bench Description
The basic USF WAMI lab test bench equipment is shown in Figure 1 . The three main microwave instruments are a 0.3 to 3000 MHz model HP8714 (B or C) RF VNA, a 9 kHz to 2.9 GHz model HP8594E SA and a 100 kHz to 3200 MHz model HP8648C synthesized signal source. Significant cost savings can be realized if measurements to 1 GHz (rather than 3 GHz) are determined to be sufficient, as listed in Table 1 . Further cost reductions are possible by replacing the VNA with a scalar network analyzer.
3 GHz Branch
1 GHz Branch
Total List Price
Total List Price
Because the VNA plays such a major role in the described measurements, the chosen analyzer will be examined closely. Figure 2 shows a block diagram of the HP8714 VNA, which is a one-path/two-port unit with one box containing a source, signal separation devices, receiver components and a display. Compared to higher cost full two-path/two-port network analyzers, this VNA requires the device under test (DUT) to be physically reversed to measure the reverse S parameters S12 and S22 , and does not inherently provide a full 12-term error model correction2 to the S parameters.
A particularly useful feature of the VNA is the broadband detection mode whereby the receiver path is switched to broadband detector diodes in place of the tuned receiver components. This feature allows broadband absolute power measurements to be made at the RF In test connector and conversion loss measurements to be made of components such as mixers (or complete receivers) for which the input (RF port) and output (IF port) signals are located at different frequencies.
In addition to the basic microwave instrumentation described previously, some other equipment is needed. Biasing of active devices requires DC power supplies and multimeters. Other test accessories include various test cables and adapters, fixed and/or variable attenuators, isolators, a coaxial calibration kit and a lowpass filter.
Although not used in the present treatment, the USF WAMI lab test benches also have a Tektronix model TDS430A 400 MHz digitizing oscilloscope for analysis of baseband signals and a Pentium PC for data acquisition and running CAD software. An additional special project bench in the WAMI lab includes a 6 GHz model HP8753 VNA, 6.5 GHz model HP8595E SA, LPKF circuit board milling machine and J microTechnology model JR2727 Personal Probe Station wafer prober. The SMA connectorized components (DUTs) used in the test examples in this article were selected from a bench-top 915 MHz wireless receiver.4
Basic S-parameter Measurements
The basic VNA calibration consists of connecting short-open-load (SOL) standards on port 1 to establish a reflection calibration and a thru connection between ports 1 and 2 (usually desired at the ends of a pair of test cables) to establish a transmission response calibration. This procedure provides a full three-term reflection model accounting for imperfect directivity source match and reflection tracking, and a one-term transmission error model accounting for transmission tracking errors.
With additional effort, source match errors can be corrected for in transmission calibrations but not load match. Although the VNA provides a more limited error model than used in full two-path/two-port VNAs (such as the HP8753 or Anritsu 37200A), it usually provides more than adequate measurement accuracy up to its 3 GHz upper frequency limit. The effects of (uncorrected) source/load mismatch in transmission measurements can be reduced by using attenuators (or pads) on either side of the DUT.
An example two-port S-parameter measurement made with the VNA for the case of a Piezo Technology lumped-element 915 MHz bandpass filter is shown in Figure 3 . The filter has a center frequency of 914.3 MHz, 3 dB bandwidth of 31.1 MHz and in-band insertion loss of 4 dB. These measurements are determined easily using a soft-key option.
S-parameter Measurement of a High Gain Amplifier
Moving on to a slightly more complicated measurement, Figure 4 shows a test configuration for a high gain amplifier measurement. For amplifier S-parameter characterization, the input signal must be low enough to ensure linear operation of the DUT. The standard configuration for the VNA does not include an internal attenuator, limiting the power range from –10 to +10 dBm. Hence, depending on VNA options, external attenuation may be required to achieve the proper DUT input level. The output signal level also must be controlled (with external attenuation, if necessary) such that it is within the linear range of the VNA’s receiver. The same basic configuration is used for swept power 1 dB gain compression point (P1dB) measurements with the VNA in a CW frequency mode.
Figure 5 shows the measured small-signal responses for a MITEQ 2 to 8 GHz amplifier. Reflection and transmission calibration was achieved as described previously using a 3.5 mm calibration kit’s SOL standards and a thru connection. The swept frequency test configuration also was used. Here, a VNA with the internal attenuation option was employed so that the amplifier input power could be set to approximately –30 dBm without the use of an external input pad. The presence of a high attenuation input pad precludes a good reflection measurement due to dynamic range considerations (as the reflected signal also is attenuated by the pad). The measurements show that this amplifier, although rated from 2 to 8 GHz, has flat gain performance (31 to 32 dB) down to 150 MHz. The input return loss is better than 12 dB from 150 MHz to 3 GHz.
Power Compression Measurement
The power for P1dB measurement is a commonly used figure of merit for amplifiers and mixers (in which case P1dB is the power for 1 dB conversion gain/loss compression). The swept frequency test configuration also can be used for compression testing. In this case, the VNA is set in CW mode and the power sweep feature is used to sweep input power (instead of frequency) as the x-axis variable in the display. Once the desired input power sweep range is set through a combination of internal (if available) and external attenuation, calibration is achieved with a thru connection in place of the DUT. The output pad must provide sufficient attenuation to maintain linear VNA receiver operation at the maximum expected amplifier output power. In the case of a high power amplifier measurement, additional precautions, such as the use of a directional coupler and high power load, may be necessary. For amplifier measurements, the tuned receiver (normal VNA mode) is used. For mixer measurements, the broadband (diode detector) mode is used along with some other test accessories and considerations.
Figure 6 shows an example compression measurement made on the same 2 to 8 GHz amplifier whose small-signal characteristics were shown previously. Gain vs. corrected input power is plotted on a 1 dB scale. This plot was obtained from two sets of measurements: gain vs. input power setting (the initial result) and input power vs. input power setting. Input power vs. input power setting can be obtained by connecting the cable with input pad directly to the RF In connector of the VNA and using its broadband power measurement feature. For the amplifier measurement shown, the frequency was 915 MHz, and the results show an input P1dB level of –17.3 dBm, or approximately +13.1 dBm output P1dB.
ToI Point Measurement
Another example of the versatility of the test bench is shown in Figure 7 . With the addition of a few test accessories, including two isolators, a combiner and a variable attenuator, a two-tone TOI measurement can be made. This measurement relies on the use of the VNA in CW mode as one signal source (at f1 ) and the frequency synthesizer as the second signal source (at f2 ). The spectrum analyzer completes the measurement with its ability to separately measure the frequency components of interest, which usually are very close together.
The TOI is the theoretical point at which the power from the third-order distortion products (at 2f2 – f1 ) is equal to the power from the fundamental signal tones for increasing power levels, as shown in Figure 8 . This condition never occurs in practice (due to compression of the output powers) and the measurement is made at low enough power levels such that the slopes of the fundamental and third-order power levels (with increasing input power) are 1:1 and 3:1, respectively. The simple equation output TOI = Io = So + D1 /2 is derived from the assumption of these slopes and provides a means of calculating TOI from a set of measurements made at a single input power level. Alternatively, data from several low power measurements are fit and TOI is determined by linear regression or graphical methods.
Frequency-converting Component Measurements
The measurement of frequency-converting components presents a more complicated and challenging set of measurements. The test bench has good capabilities in this area as described previously3 although some important accuracy considerations and limitations exist as well.
In this treatment, attention is restricted to the measurement of mixer conversion loss. A general test configuration for performing conversion loss/gain measurements with the WAMI test bench is shown in Figure 9 . As illustrated, conversion loss is defined as the difference in decibels between the power presented to the mixer RF port (at the RF) and the power exiting the IF port (at the IF). Conversion gain is the algebraic negative of conversion loss and is actually what is displayed on the VNA during the described measurement. Conversion loss/gain measurements can be made with either the VNA (using its broadband detection mode) or the SA. A more detailed discussion of these techniques has been presented previously.3
Note that for either measurement the mixer has attenuators connected to each port to improve SWR conditions during measurement and to minimize the effects of multiple reflections of mixer products. A lowpass or bandpass filter must be included in the IF cable path during mixer-under-test (MUT) measurement so that only the IF component is incident on the VNA’s broadband detector diode. Corrections must be made as described previously3 for IF cable (and filter where it is used) insertion loss differences between the RF used for calibration and the corresponding IF present during MUT testing.
Figure 10 shows example measurements made using both of these techniques for a Mini-Circuits model ZEM4300 mixer. The RF signal was swept (VNA method) or stepped (SA method) from 785 to 1485 MHz with the LO frequency held at 985 MHz, producing IFs ranging from 0 to 500 MHz. The LO signal was +4 dBm and the RF power level was –20 dBm. Accordingly, a 500 MHz lowpass filter was used in the IF cable during VNA mixer measurement.
The measurements show nominally the same (5.5 to 6.5 dB) conversion loss with some discrepancies attributed to inaccuracies and nonlinearities in the amplitude measurements made by the SA and VNA (in broadband detector mode). Neither of these instruments is, by design, a highly accurate power meter. The addition of a conventional power meter to the test bench would allow more accurate conversion loss measurements to be made where required, albeit at the loss of both speed and convenience.3
The basic USF WAMI lab test bench has been shown to be a versatile tool for measuring RF and microwave components of the type used in modern wireless systems. In particular, the broadband detection mode and absolute power measurement features of the model HP8714 VNA are valuable supplements to the conventional VNA tuned receiver mode. Although this relatively low cost VNA does not provide full two-path/two-port error-corrected measurements, it provides more than adequate accuracy for most component measurement applications when proper procedures (for example, the use of pads) are followed.
For measurement of conversion gain/loss, while an uncertainty analysis is outside the scope of this article, experience suggests that the measurements described previously probably have an uncertainty on the order of ±0.5 dB. The bench could be expanded to accommodate noise figure measurements using spectrum analyzer methods with the addition of a solid-state noise source.
The equipment utilized in preparing this article was acquired through combined grants from the National Science Foundation (grant number DUE-9650529) and Hewlett-Packard Co. with matching funds from USF. Various forms of sponsorship of USF’s WAMI lab from Honeywell, Tektronix, J microTechnology, HP EEsof, Ansoft, M/A-COM, Cushcraft, Mini-Circuits, Piezo Technology Inc. and MITEQ also are appreciated. Additional information on the WAMI laboratory is available on the Internet at www.eng.usf/WAMI. Material for this article was first presented at the Wireless and Microwave Technology ’97 Symposium held in Chantilly, VA October 6–10, 1997.
1. P.G. Flikkema, L.P. Dunleavy, H.C. Gordon, R.E. Henning and T.M. Weller, "Wireless Circuit and System Design: A New Undergraduate Laboratory," Proceedings of Frontiers in Education ’97, November 5–8, 1997, Pittsburgh, PA.
2. J. Fitzpatrick, "Error Models for Systems Measurement," Microwave Journal, Vol. 21, No. 5, May 1978, pp. 63–66.
3. L.P. Dunleavy, T.M. Weller, E.W. Grimes and J. Culver, "Use Network and Spectrum Analysis for Mixer Measurements," Microwaves & RF, May 1997 (Part I) and June 1997 (Part II).
4. L.P. Dunleavy, P.G. Flikkema and A. Kuppusamy, "Characterization and Simulation of a 915 MHz Wireless Receiver," Automatic RF Techniques Group Conference, December 1997, Portland, OR.
Lawrence P. Dunleavy received his BSEE from Michigan Technological University in 1982, and his MSEE and PhD from the University of Michigan in 1984 and 1988, respectively. From 1982 to 1983 he worked for E-Systems Inc., and from 1984 to 1990 he was employed by Hughes Aircraft Co. (including three years as a Howard Hughes Doctoral Fellow). In 1990, Dunleavy joined the electrical engineering department at the University of South Florida where he is now an associate professor. Currently, he is on a sabbatical research appointment with the Microwave Metrology Group of the National Institute of Standards and Technology (NIST) in Boulder, CO (August 1997 to August 1998). He is a senior member of the IEEE and serves on several MTT committees as well as on the ARFTG Executive Committee.
Thomas M. Weller received his BS, MS and PhD degrees in electrical engineering from the University of Michigan, Ann Arbor in 1988, 1991 and 1995, respectively. He is currently an assistant professor in the electrical engineering department at the University of South Florida, Tampa. Weller’s research involves micromachining applications for microwave and mm-wave circuits, hybrid and MMIC packaging, numerical modeling and mm-wave sensors.
Paul G. Flikkema received his BS in computer engineering from Iowa State University, and his MS and PhD in electrical engineering from the University of Maryland, College Park. Since 1994, he has been an assistant professor at the University of South Florida. Flikkema previously worked in industry, most recently with Techno Sciences Inc., Greenbelt, MD, where he focused on signal processing for low S/N and spread spectrum satellite communications. His current research interests include spread spectrum/CDMA communication and the physical-layer analysis and design of wireless networks.
Horace C. Gordon, Jr. received his BSEE from the University of Florida at Gainesville in 1964 and his MSE from the University of South Florida (USF) at Tampa in 1970. He has been a full-time faculty member at USF since 1977 and is currently a lecturer in the department of electrical engineering. Gordon’s research interest is in the area of RF/microwave circuit theory and passive component modeling. He is a registered professional engineer in Florida.
Rudolf E. Henning received his BSEE, MSEE and D Eng Sc degrees from Columbia University. The first half of his engineering career was spent in industry with the Speny Rand Corp. where he headed the engineering staff of the Sperry Microwave Electronics Co. in Clearwater, FL for 13 years. For the last 25 years, Henning has been with the University of South Florida in Tampa where he is a distinguished service professor. Henning is a Life Fellow of the IEEE and was 1968’s MTT Adcom chair. He chaired or co-chaired MTT-S’ 1965, 1979 and 1995 International Microwave Symposia and continues to serve on MTT-S committees and on the IEEE’s chapter and section levels. He is a registered professional engineer in Florida.