Network analyzers measure network parameters of electrical networks, such as two-port amplifiers and filters. The instruments characteristically operate at 5 Hz to 1.05 THz, mostly at the higher end, though some models are sensitive down to 1 Hz.
These instruments are available in two varieties: scalar network analyzers, which measure only the amplitude of the test signal, and vector network analyzers, which measure amplitude and phase. Vector network analyzers are far more common. In fact, it is hard to find instrument makers that still make scalar network analyzers. Those that once manufactured these devices now generally argue that the prices of VNAs have come down to a point where it makes sense to simply go with the more capable instrument.
The now-obsolete Agilent 8757D scalar network analyzer.
It’s possible to find less expensive VNAs, but they usually are USB-type instruments that only work below 1 GHz. Mini-Circuits also makes a DIY VNA board designed as an educational tool that goes for about $3,000 and works to 6 GHz.
Thus there may be financial incentives to know the differences between scalar analyzers and VNAs. Though scalar analyzers are tough to find new, it’s possible to get the same functions by using a spectrum analyzer combined with a tracking generator. A tracking generator is a signal generator having an output that tracks or follows the tuning of a spectrum analyzer. The usual test setup is to have the tracking generator output feed into a device under test (DUT), then feed the DUT output to the spectrum analyzer. The resulting signal level depicted on the spectrum analyzer display is then a function of the DUT qualities. The response can be displayed over the range of frequencies swept out by the tracking generator. Typical measurements made this way include gain and gain flatness of amplifiers, filter responses, return loss and/or VSWR, mixer conversion loss, reverse isolation, and absolute power measurements.
The usual procedure is to do a calibration before running analyzer tests. For transmission-type tests, the tracking generator output goes directly to the spectrum analyzer. Then DUT transmissions are divided by the results to compensate for the transmission frequency response of the measurement system. For reflection-type measurements, the calibration procedure averages the short and open response for the same purpose. A point to note is that this calibration doesn’t account for systematic errors from factors such as source and load mismatch and thus contribute to measurement uncertainty.
An additional factor is that scalar analyzers tend to have less dynamic range than VNAs. The old Agilent 8757D, for example, lists a dynamic range of +16 to –60 dBm. VNAs can be found with -110 dBm sensitivity, helpful for working with low signal levels.
Phase delay (top) and group delay through the same filter as measured with a Tektronix TTR500 VNA.
The usual explanation of group delay uses the example of an amplitude-modulated sine wave. Group delay specifies the degree to which the amplitude envelope will be delayed as it traverses the DUT. Another way to visualize group delay is to recall that amplitude modulation introduces sideband frequencies on either side of the carrier frequency. Group delay can be thought of as information about how the sidebands will be delayed relative to the carrier frequency. The delay will change the shape of the amplitude envelope.
Usually, both phase delay and group delay measurements on a VNA involve a single button push once the test setup has gone through a calibration and the DUT is connected.
That brings us to S parameters. As a quick review, S-parameters describe how an N-port network responds to signals applied at any or all of the ports. The first number in the subscript refers to the responding port, while the second number refers to the incident port. So S21 refers to the response at port two due to a signal at port one. In circuit work, the most common S parameters are for two-port networks. Also, the S parameters of most interest are generally the input and output reflection coefficients S11 and S22. These are typically plotted on the Smith chart. Transmission coefficients (S21 and S12) are usually not.
The VNA determines S11 and S21 by measuring the magnitude and phase of the incident, reflected, and transmitted voltage signals when the output is terminated in a load that equals the characteristic impedance of the DUT. This condition guarantees there are no reflections from the load. S11 is equivalent to the input complex reflection coefficient or impedance of the DUT, and S21 is the forward complex transmission coefficient. Likewise, by placing the source at port two and terminating port one in a perfect load makes sure there are no reflections from the input. Then S22 and S12 measurements can be made.
As with group and phase delay, VNAs typically make S-parameter data available with the push of a button.
Finally, it can be noted that some of the same reflected signal information available with VNA S-parameter measurements may be available from a sampling scope equipped to do time-domain reflectometry measurements. The TDR instrument sends a pulse through the DUT and compares the resulting reflections to those produced by a standard impedance. The typical TDR display is the voltage-vs-time waveform that forms from a fast step signal propagating through the DUT and reflections generated when the step encounters impedance variations. The time-domain display can be transformed into S-parameters (frequency response) by performing a Fourier transform.
VNAs tend to be faster and exhibit a higher dynamic range than scopes able to run TDR measurements over the same frequency range. Additionally, scopes able to work at the tens-of-gigahertz range and higher, where a lot of S-parameter work takes place these days, also tend to be almost as pricey as the VNAs able to handle these frequencies.
While I am a fan of Scalar Analysers, the author has not been accurate regarding the costs. You may be able to get a 8757D for $1000, but the full measurement set requires a splitter, a directional bridge or a dual directional coupler, three detectors (specific to the 8757D model) or four if you want to level the signal source output, as a minimum to be of any use. Used detectors / bridges can cost anywhere up to $5000 each depending on the frequency coverage. The 8757D is only a processing and display machine, which processes the DC or low frequency AC signals output by the detectors, and as such it is misleading to specify a 8757D up to 50GHz or even 110 GHz. On top of all this, you need a signal generator (synthesiser or sweeper) as an excitation source. You will end up spending up to $15,000 to $20,000 to set up a 8757D-based scalar analyser today and that too using secondhand bits of unknown history. VNAs have all of the above built into a single box. VNAs have become competitive at lower microwave frequencies, but at mmWave frequencies, VNAs can be pretty expensive, and that is where a scalar analyser is still competitive if you can build one up with hardware from a number of manufacturers and use a laptop to do all the processing and display.