Measurements are vital to research as well as production. Over the last 50 years, ever-improving test equipment led to greater understandings of component, device, and system electrical behaviors, paving the ways for dramatic advances in those areas. But early RF/microwave test instruments were primitive by today's standards, lacking automation and limited in bandwidth. Still, they were a starting point for many of the high-frequency test instruments in use today.
Among the most essential of RF/microwave measurements are those related to signal power and frequency, and several companies in the Long Island area of New York State offered power meters during the early 1960s with General Microwave Corp. (Farmingdale, NY) providing their model 454A Thermoelectric Power Meter (Fig. 1). The power meter operating over 23 selectable power-measurement ranges, with total measurement range of 0.3 W to 300 mW with 1% accuracy from 2 MHz to 10 GHz when using coaxial heads. Waveguide power heads were also available for X-band and Ku-band frequencies.
Model 454A employed power heads with a thin-film metallic load which promised the accuracy of a calorimeter and the speed and sensitivity of a bolometer. The thin-film load acted as a well-matched termination, absorbing incident RF power. The load was formed by means of bismuth and antimony films vacuum-deposited on a thin Mylar substrate in a geometric configuration that produced a number of thermoelectric junctions. By sinking some of the junctions to the transmission line and the rest to the air space within the power head, these latter junctions would serve as the calorimetric mass and provide high sensitivity and fast response times. By controlling the area of the thermoelectric load, a considerable range in transducer sensitivity and power-handling capability can be achieved. The output of the thermocouple fed a DC amplifier with low noise and low drift (about 3-dB noise figure). Each power head, which contained the thin-film detector, could be disassembled with a screwdriver. The model 454A sold for a mere $475, with each power head costing only $175.
Another low-cost power-measurement instrument of that time was the model 440 power meter (Fig. 2) from Narda Microwave Corp. (Plainview, NY) which sold for $250. Powered by a nickel-cadmium (NiCd) rechargeable battery, it was ideal for on-site power measurements and was designed to be used for any frequency range for which bolometers or thermistor mounts were available. It provided as much as 18 mA bias current for the power sensor, and could measure continuous-wave (CW) or pulsed power from 0.1 to 10 mW with 3% accuracy.
In terms of noise measurements, Kay Electric Co. (Pine Brook, NJ) provided noise-figure measurements on their model 792-AFE automatic noise-figure meter, released in 1965. Working with a noise source, the basic model provided noise measurements from 10 to 900 MHz but could be extended to 26.5 GHz with an external source. Noise measurements were shown on a 3-in. analog meter.
A major development in 1967 was the introduction of a "digital" power meter from Pacific Measurements (Palo Alto, CA) for use from 10 MHz to 12.4 GHz. The mainframe worked with crystal diode detectors to measure power levels from -40 to +10 dBm with 0.1-dB resolution. Although it was not truly a digital power meter, it did use a digital readout to show measured power levels with accuracy of 5%. As an option, it could provide binary-coded-decimal (BCD) digital output signals, and the meter could be triggered by an external computer to perform as many as 1000 readings per second.
One of the more important measurement tools to the RF/microwave industry, and one used predominantly in this rather than in other industries, is the vector network analyzer (VNA). Because RF/microwave components and systems must handle such physically small signal wavelengths, understanding the forward and reverse transmission characteristics of a device under test (DUT) can not only provide great insight into the electrical behavior of the device, it can also offer measured starting points for building a model of that device.
In January 1967, the industry received a glimpse of what was to come in terms of VNA measurement capabilities, when Hewlett-Packard Co. (Palo Alto, CA) introduced a basic vector network system (Fig. 3) consisting of the model 8411A frequency converter, along with the model 8410A mainframe with 8413A indicator and 8414A polar display plug-in modules. The VNA measurement system offered a coaxial measurement range from 110 MHz to 12.4 GHz and could be extended to higher frequencies through the use of waveguide adapters. The separate modules worked together very much like a receiver, with the local oscillator (LO) in the 8411A automatically tuned and locked to a submultiple of a tuned signal frequency by means of phase-lock loop (PLL) in the 8410A mainframe. Described by its own literature as a form of microwave vector voltmeter, the measurement system could indicate phase, amplitude, impedance, admittance, and reflection coefficient. However, it would still be some years before the embedded computing technology was available for practical VNA error correction. In spite of great care in manufacturing, variations in amplitude and phase in internal passive components, such as directional couplers, limited the accuracy of these early systems.
Still, VNA systems were rare in the late 1960s, and most measurement laboratories relied on fundamental performance testing by means of power meters, frequency counters, and signal generators. Frequency measurements were important during an era when the frequency of even the best test signal sources tended to drift over time and temperature. In August 1969, one of the more innovative frequency-counter suppliers of that time, EIP Microwave (Santa Clara, CA), introduced a pair of flagship products in their models 350A and 350B (Fig. 4) automatic frequency counters with 11-digit NIXIE tube readouts. The "A" model operated from 20 Hz to 12.4 GHz while the "B" model provided measurements from 20 Hz to 18 GHz. Each instrument relied on a YIG-tuned comb generator locked to a 1-MHz crystal-controlled reference oscillator, which also served as the driving signal for the comb generator. Signals from the comb generator were used for heterodyne frequency conversion of input signals under test in the frequency counter. Two heterodyne frequency converter circuits were used in each counter, covering 20 Hz to 1.2 GHz and either 1 to 12.4 GHz or 1 to 18 GHz. The B model provided three input ports to cover the entire frequency range from 20 Hz to 18 GHz, with 1-Hz measurement resolution. It sold for $5450.
Continue on Page 2
Another key measurement for the early RF/microwave industry was spectrum analysis, and in November 1970, Hewlett-Packard Co. introduced their model 8555A microwave spectrum analyzer with impressive performance numbers (Fig. 5). It could measure absolute power levels from -130 to +33 dBm (2 W) at frequencies from 10 MHz to 18 GHz. While many spectrum analyzers of that era relied on a broadband backward-wave oscillator (BWO) as the signal source for heterodyne downconversion, the 8555A broke ground by using a solid-state YIG-tuned transistor oscillator. The analyzer's RF section featured a hot-carrier diode mixer with wide dynamic range and a variety of thin-film microwave integrated circuits (MICs), enabling the instrument to achieve an on-screen dynamic range of 70 dB. The spectrum analyzer provided better than 100-Hz frequency resolution and was available with a variable-persistence 5-in. screen or a fixed-persistence 12-in. screen.
Hewlett-Packard would improve upon the 8555A some years later with the development of the 8566A, which became of the best-selling laboratory microwave spectrum analyzers of all time. With a basic frequency range of 100 Hz to 22 GHz, it could display signal amplitude levels from -134 to +30 dBm. But competitors were manyamong them, Polarad Electronic Instruments (Long Island City, NY) and their model 2400 spectrum analyzer, with a 70-dB dynamic range from 10 MHz to 12.4 GHz and -105 dBm sensitivity and Tektronix, with their model 7L18 spectrum analyzer capable of covering 1.5 to 60.0 GHz, and later their popular 492 and 494 series of portable microwave spectrum analyzers.
Before VNAs became a fixture on many microwave test benches, many RF/microwave component manufacturers relied on microwave scalar network analyzers (SNAs) for component characterization. While a VNA can measure phase and amplitude with frequency, an SNA can only measure amplitude as a function of frequency, typically using a mainframe unit and a diode detector or directional bridge. Many of the classic SNAs, such as the 8757 series from Hewlett-Packard/Agilent, have been discontinued, and some of the one-time leading suppliers, such as Pacific Measurements and Marconi Instruments, are no more, but commercial SNAs are still available through microwave and millimeter-wave frequencies from numerous sources, including the model 8003 from Giga-tronics is an SNA with 10 MHz to 40 GHz.
One of the leading SNAs of the 1980s was the model 560 SNA introduced by Wiltron Company (now Anritsu Co.) of Mountain View, CA in 1979 (Fig. 6). It allowed simultaneous transmission and reflection measurements to be displayed from 10 MHz to 18 GHz, and was designed to be used with an additional microwave signal generator or sweeper. In an age when most microwave measurements were performed manually, the 560 was the first microwave SNA available with built-in GPIB interface for making automated measurements under the control of a GPIB-compatible external microcomputer. The 560 sold for $5900 without the GPIB interface and $7250 with the interface. It featured calibrated offset controls that were continuously variable from -80 to +80 dB with 0.1-dB resolution for precision amplitude measurements.
Additional microwave SNAs included the SWOB 5 from Rohde & Schwarz (Munich, Germany), an SNA with 76-dB dynamic range from 100 kHz to 1000 MHz and the model 1038-N10 from Pacific Measurements, with a 76-dB dynamic range from -60 to +16 dBm at microwave frequencies. It showed plots of amplitude versus frequency with a unique light-emitting-diode (LED) display and could be automated via external GPIB computer.
By the 1980s, microwave test instruments would be leveraging the new-found computing power in more powerful microprocessors and more practical computer memory. By January 1984, Tektronix would add to its popular 492 and 492P (for programmable via GPIB) portable microwave spectrum analyzers with its models 494 and 494P (Fig. 7) for measurements from 10 kHz to 21 GHz with coaxial connectors and to 325 GHz with waveguide adapters. What had once been a dedicated test instrumentthe frequency counterwas now built in as an automatic function in these analyzers.
As with many other microwave test instruments, the 494/494P spectrum analyzers would make use of the microprocessor to improve accuracy: these instruments featured an internal calibration routine that helped offset the effects of frequency drift in the intermediate-frequency (IF) filters. The center frequency accuracy of these portable spectrum analyzers was cited as 37 Hz at 100 MHz, 41 Hz at 4 GHz, and 115 Hz at 60 GHz. The average noise level for measurements in a 1-kHz resolution bandwidth was -110 dBm at 7.1 GHz, -85 dBm at 50 GHz, and -65 dBm at 220 GHz. The instruments featured an 80-dB on-screen dynamic range, 1 Hz resolution, and 30-Hz resolution bandwidth through 60 GHz.
The first half of the 1980s saw more affordable test signal sources, as instrument manufacturers sought to provide support for an emerging commercial communications market. In October 1984, for example, John Fluke Manufacturing Co., Inc. (Everett, WA) introduced their model 6060A synthesized signal generator, one of the first test signal sources aimed at the growing cellular communications market. Priced at less than $5000 ($4800 with GPIB and $4500 without), it tuned from 400 kHz to 1050 MHz with 20-Hz resolution and signal levels from -137 to +13 dBm. The signal generator was characterized by -30 dBc harmonic levels and -60 dBc spurious levels. Fluke, no longer involved in RF/microwave instrumentation, is now a leading supplier of network analysis, calibration, and biomedical analysis equipment. Around the same time, Wavetek San Diego (San Diego, CA) would introduce their fast-switching model 5155A frequency synthesizer for under $15,000. With 0.1-Hz frequency resolution from 100 kHz to 1 GHz, it featured 1-s switching speed with phase noise of -100 dBc/Hz offset 10 Hz from the carrier and -90 dBc/Hz offset just 1 Hz from the carrier.
Continue on Page 3
Many important microwave test instruments entered the market in 1984, but perhaps none as significant as the HP 8510A from Hewlett-Packard Co. which graced the January 1984 cover of Microwaves & RF (Fig. 8). This test system marked the next stage of evolution in microwave vector network analysis, with a dedicated internal computer that enabled automatic error correction for precise measurements of signal amplitude and phase from 45 MHz to 26.5 GHz.
The HP 8510A offered a wide range of displays, including logarithmic and linear magnitude versus frequency, deviation from linear phase, and group delay versus frequency, and could show test results simultaneously in the frequency and time domains. The analyzer was essentially a sensitive four-port receiver designed to work with one of the firm's microwave signal sources, such as the HP 8350B sweep oscillator or the HP 8340B synthesized signal generator. The performance levels of the HP 8510A are impressive even by today's standards, with magnitude resolution of 0.001 dB across a 100-dB dynamic range, phase resolution of 0.01, group-delay resolution of 10 ps, and magnitude and phase accuracies of 0.05 dB and 0.03, respectively.
Perhaps the most amazing part of the performance achieved by the HP 8510A was the complement of internal computing components which made it possible. The HP 8510A's mathematics processor ran at a clock rate of 16 MHz and was controlled by an algorithmic state machine (ASM) with 1 kB of microcode. The VNA sent measurement data that had been captured by detector circuitry to a front-end high-speed analog-to-digital converter (ADC) and then in digital form to the processor, using an internal 16-b bus to transfer the data. The analyzer's internal computer boasted total memory of 256 kB random-access memory (RAM), 256 kB bubble memory, and 26 kB read-only memory (ROM). About 16 kB of the ROM space was dedicated to built-in test software and the boot-up operation of the system. The main system memory was taken from 1 MB of nonvolatile bubble memory and placed into 128 kB of main RAM. And 60 kB of bubble memory were used to store the coefficients employed in making vector error corrections. The analyzer had a built-in tape drive to store and load graphics and data from the system cassette. Of course, a great deal of high-frequency analog engineering went into the HP 8510A as well, including a new broadband directional bridge which contributed to the test set's 50-dB directivity.
This one instrument's introduction would impact the industry profoundly, bringing new measurement capabilities, new levels of accuracy. The scattering parameters (S-parameters) that the HP 8510A would measure to depict the forward and reverse transmission characteristics of high-frequency components would eventually become the basis for behavioral models that could represent those components in software simulations. Other microwave VNAs would follow, among the most noteworthy the model 360 VNA system from Wiltron with versions available in frequency to 40 GHz (Fig. 9), but none would ever change the industry in the manner of the HP 8510A.
Ironically, in December of that same year (1984), Hewlett-Packard would introduce a product that was almost "lost in the noise" of the HP 8510A: the HP 70000 series modular measurement system (MMS). Although it didn't storm the industry in the manner of the HP 8510A, it proved that "serious" instruments could also be modular and was, in fact, a forerunner for other modular industry instrument formats in use today, such as PXI and LXI. When it was introduced, the HP 70000 MMS appeared as a spectrum analyzer (Fig. 10). It could operate from 50 kHz to 22 GHz, or to 26.5 GHz with an option, but could operate to as high as 325 GHz when configured for millimeter-wave use. Function modules slid into a 19-in.-rack-mount, mainframe unit, with initial modules that included a power supply and an LO/controller/digitizer, which was a microwave sweeper that operated from 3.3 to 6.6 GHz in a package only 4 x 4 x 19 in.
During the time that the industry pursued improved VNA performance, the then-National Bureau of Standards (NBS) had developed an alternative means of measuring microwave signal amplitude, phase, and S-parameters, using an approach they called a six-port automatic network analyzer (ANA). Through information dissemination at technical meetings such as the Automatic RF Techniques Group (ARFTG), NBS spokespeople for the six-port approach, such as Algie Lance, would explain the accuracy possible with this broadband system. The six-port ANA, developed by Glenn Engen of the NBS, now the National Institute for Standards and Technology (NIST), was a reflectometer capable of measuring the amplitude ratio or phase difference between two electromagnetic (EM) waves. In doing so, it could measure the complex reflection coefficient (amplitude and phase) of a DUT.
The six-port ANA was essentially a passive linear network with two input ports and four output ports, so that the four outputs provided four different linear combinations of the EM waves at its input ports. By using a pair of six-port reflectometers in a dual six-port configuration, it is possible to perform the same measurements as with a traditional network analyzer, that is, to measure not only the reflection of waves in order to determine the input impedance, but also the transmission of waves through a two-port DUT in order to determine the gain or attenuation of the DUT. Although this approach never surfaced as a commercial product, measurement services from 0.1 MHz through 18 GHz based on the approach are still offered by NIST.
Although these instruments offer a small sampling of the RF/microwave metrology advances over the last 50 years, it is important to note that many of the advances were due to "behind-the-scenes" work on oscillators, MICs, software, and even coaxial connectors. Gifted "machinists," including Julius Lange and the famed microstrip Lange coupler; Julius Botka of Hewlett Packard Co. and his considerable work on connectors and other low-loss structures that would enable the development of such landmark instruments as the HP 8510A VNA; Bill Oldfield of Wiltron Co. and his innovations on coaxial connectors, including the K connector of the 1980s which would break though the frequency limitations of the standard SMA connector; Mario Maury of Maury Microwave whose work on probes, adapters, and impedance tuners would make possible many previously impractical measurements; and even Jim Kubota of Gilbert Engineering (Glendale, AZ) who, with John Morelli of Automatic Connector, would contribute much to the improvement in performance of push-on type coaxial connectors.