Simplifying frequency converter testing

Frequency-translating components - devices like mixers or frequency converters that shift an input signal from one frequency band to another - are key elements in most RF and microwave systems. They are found at the core of satellite receivers and transponders, radar systems and telecommunication networks.

While in the past it was often sufficient to only characterise the magnitude response of these components, today’s engineers must also know the phase and group delay response as well.

Having both magnitude and phase information for all of the components of a system means designers can simulate and optimise the overall system performance.

Traditionally, vector network analysers (VNA) have been the tool of choice to characterise the magnitude and phase performance of a mixer or frequency converter. They offer two or four test ports, one or two internal RF sources, fast sweeps (compared with using stand-alone signal sources and a spectrum analyser or power meter) and provide complex data of both reflection (eg, input and output match) and transmission responses.

VNAs can offset their source and receiver frequencies, a necessary capability to test frequency-translating devices.

Typically when using a VNA, two additional mixers are needed to characterise the device under test (DUT). The first mixer provides a signal to the VNA’s reference receiver at the same frequency as the output of the DUT.

This signal provides a phase reference for ratioed measurements of phase and group delay versus frequency (Figure 1).

In older VNAs, the reference mixer was also used to phase-lock the internal source to the correct offset frequency. In contrast, modern VNAs use fully synthesised RF sources that do not require a reference mixer to generate the proper stimulus signal.

Figure 1. The traditional approach to testing mixers and frequency converters requires a reference mixer to provide a signal at the same frequency as that present at the DUT output, for ratioed phase measurements between the R1 and B receivers.

Calibrating a VNA for these measurements requires both a power reference and phase reference, as well as standard S-parameter calibration standards like an ECal module or a mechanical calibration kit. Power calibration is done using a power sensor.

Instruments like the PNA family of network analysers calibrate all the receivers for the test ports based on a single source-power calibration at port one and the standard S-parameter error terms between the ports. Calibrating the phase response of the receivers is most commonly done with a second mixer.

This calibration mixer is used as a characterised through standard, meaning its vector reflection and transmission responses are known. One way to do this is to perform three reflection measurements at the input of a reciprocal calibration mixer with an appropriate output intermediate-frequency (IF) filter, while presenting three different impedances to the output of the mixer/filter pair.

These impedances can come from open, short and load standards or from an ECal module. From this, the input match, output match and two-way transmission response can be calculated. Using a reciprocal mixer, the one-way transmission response can be determined by dividing the two-way response in half.

While this method is relatively straightforward and has been commonly employed, in practice it is often difficult to find the proper reference and calibration mixers and the appropriate IF filter, especially when a broad range of DUTs must be measured and especially above 26.5 GHz.

Providing a local-oscillator signal to the reference and calibration mixers can also be cumbersome, particularly when using older VNAs with no built-in second source.

A simpler alternative would speed up measurement set-up and calibration, making design and test engineers more productive.

One solution to this dilemma is the new SMC+phase method, which was developed for the Agilent PNA and PNA-X series of microwave network analysers. This method eliminates the need for both a reference and calibration mixer when testing the phase or group delay of frequency-translating devices.

Instead of relying on ratios of test and reference signals at the same frequency (which requires a reference mixer), it ratios single-receiver phase measurements done at different frequencies (those corresponding to the DUT’s input and output).

This technique relies on the inherent phase coherency of the fractional-N-based synthesis architecture used in the PNA’s sources. Relative phase coherency is maintained across a frequency sweep by digitally incrementing the phase accumulators embedded in the fractional-N hardware, and by employing synchronous IF detection and digital-signal processing.

At band-crossings, where changing synthesiser-divide numbers cause discrete phase jumps, the phase is mathematically stitched together to maintain phase coherency across the sweep.

Sweep-to-sweep starting-phase variation is removed by normalising one point in the phase sweep to zero, which allows sweep averaging to be employed as an effective noise-reduction technique.

In conjunction with single-receiver phase measurements that eliminate the reference mixer, a new phase calibration method has been developed that eliminates the need for a characterised through mixer. Instead of using a mixer to determine the phase-versus-frequency response of the PNA’s receivers, a harmonic comb generator is used to create a broadband set of signals with 10 MHz spacing, extending all the way to 67 GHz.

The comb generator, which is calibrated by electro-optical methods developed at the National Institute of Standards and Technology, is identical to the phase references used with Agilent’s nonlinear vector network analyser (NVNA).

Once phase calibration is done, accurate measurement of phase deviation and absolute-group-delay can be performed. As with previous SMC measurements, a power sensor is used to accurately calibrate the magnitude-versus-frequency response of the receivers (Figure 2).

Figure 2. The SMC+phase set-up is simple, since reference and calibration mixers are not required. Test system calibration is achieved using a power sensor (left), a broadband comb generator (middle) and an ECal module (right) or a mechanical calibration kit.

SMC+phase measurements can easily be done on devices with embedded LOs that are not locked to the system 10 MHz reference. The measurements use previously developed search algorithms that tune the PNA receivers to the actual output of the DUT, instead of relying on nominal LO frequencies (Figure 3).

The tracking is done sweep by sweep to keep up with LO drift. Unlike the modulated-carrier envelope approach to testing DUTs with embedded LOs, the PNA’s method is simple and does not require two RF sources and a signal combiner.

In addition, it enables phase-versus-drive measurements to characterise phase change versus RF input-power levels, something that cannot be done with the envelope method.

Figure 3. When measuring devices with embedded LOs, the PNA uses a coarse frequency sweep (top) to determine the nominal LO offset. A phase-versus-time sweep (bottom) is used to finetune the estimate of LO offset.

With the PNA and PNA-X family of microwave network analysers, mixer and converter test has never been easier. The new SMC+phase method uses single-receiver phase measurements to eliminate the need for reference mixers.

Connecting to the DUT is as simple as connecting input and output cables and, if needed, connecting a third cable to the instrument’s built-in second source for use as an LO signal. A new phase calibration method uses a broadband harmonic comb generator as a phase standard to eliminate the need for calibration mixers and their associated IF filters.

The method promises to revolutionise mixer and converter test by providing a simple set-up with fast, accurate results based on easy-to-use broadband calibration standards.

Simple and fast characterisation of mixers and frequency converters makes design and test engineers more productive, helping them decrease time to market and lower the cost of test.