March 22, 2018
As engineers working at high frequencies understand, ensuring signal integrity (SI) in their designs involves analyzing and mitigating signal impairments. This is particularly important at the IC and package levels, as well as on the printed circuit board (PCB) and backplane. With more designs operating at faster speeds and higher bandwidths, the importance of conducting accurate SI measurements is only going to continue.
SI engineers have come to rely on vector network analyzers (VNAs) to conduct measurements, as they offer advantages over TDRs and oscilloscopes at microwave and millimeter (mmWave) frequencies. When using a VNA, engineers need to consider three S-parameter quality metrics – reciprocity, passivity and causality – to make accurate measurements. In today’s post, we will discuss each in some detail.
Reciprocity
One way to estimate the uncertainty of S-parameter measurements is to compare reciprocal S-parameters. If the uncertainty is high, SI engineers should recalibrate. Reciprocity is a much greater concern for TDT/TDR-based S-parameters. In addition to their time base jitter and noise problems, the fast rise-time step voltage used by TDT/TDR equipment can cause synchronization problems in the forward and reverse directions.
Poor reciprocity is especially troublesome when measurement results are cascaded in simulation. The reciprocity errors cause mismatches to propagate incorrectly and the errors can build as the file count increases. Fortunately, S-parameters measured on VNAs rarely suffer these “reciprocity” problems, which is one reason to use a VNA in SI applications.
Passivity
Passivity problems occur when a passive device appears to have gain. Source and load match will contribute to the phenomena, as well as unsuccessful fixture de-embedding, especially if the fixtures have high insertion loss. Passivity problems can be eliminated simply by exercising good lab techniques, including:
- Assure that the VNA receivers are not compressed during calibration and measurement
- Check that all connectors are in good condition and properly torqued
- Verify that any cables used in the measurement are in good condition and are properly de-embedded
De-embedding errors are often the source of passivity problems. Small errors can cause channels to appear to have gain. This is most prevalent in fixtures with high insertion loss and low loss devices under test (DUTs). Anritsu offers a wide range of extraction methods for de-embedding a variety of fixtures and other devices. For example, the VectorStar® VNA family (figure 1) has seven standard methods and enhanced tools using its Universal Test Fixture Extraction (UFX) option. Included in the option is an Enforce Passivity capability when saving characterization files (SnP).
VNAs that create traced-based eye diagrams allow all-important parameters to be updated on a sweep-to-sweep basis, eliminating the need to transfer SnP files. Each new S-parameter sweep creates a new SnP file to the eye diagram trace and is updated automatically. As the DUT is tuned, the updated eye diagram provides a new indication of the data transmission characteristics. A screen display of this process is shown in figure 2.
Fixtures often are used to transition from single-ended VNA test ports to the input of differential devices. It is challenging to extract them directly from measurements, so VNAs need extensive de-embedding capabilities to reduce the complexity of fixture and transition removal. If the fixtures are not de-embedded properly, they will contribute to the overall error of a predictive model.
Passivity is becoming increasingly important, as non-connectorized devices are often used in integrated systems, such as automotive radar, 5G backhaul, and WiGig, that operate in the mmWave spectrum. At higher frequencies there is an increase in uncertainties and a need for higher performance DUTs.
Expanded de-embedding tools are needed in applications in which full calibration at the DUT/test fixture plane is not available. VectorStar provides an extensive array of network extraction tools for enhanced de-embedding capabilities, as well as calibration menus that generate SnP files for fixture and probe de-embedding.
The UFX software option provides increased flexibility for on-wafer and fixture calibrations where typical standards are not available. Engineers can add standards as they become available. There are also four additional network extraction features:
- Generalized B allows incomplete sets of standards to be used for single-port de-embedding
- Multi-standard partial information techniques allow increased accuracy when multiple standards are available for two- and four-port de-embedding
- Phase-localized partial information techniques improve estimates of inner plane behavior and crosstalk for more accurate extraction
- Sequential extraction provides the ability to generate an SnP file for a portion of a fixture to help during the test fixture design phase
Causality
The biggest potential weakness in transforming frequency-domain measurements to the time domain is causality. The Fourier transform from frequency to time requires integration over frequency from 0 to infinity, that is, from DC to very high frequency.
Anritsu VNAs reduce the causality and convergence risks by reaching down to the lowest VNA frequency in the industry. For example, the VectorStar VNA is a bridge-based VNA below 2.5 GHz to minimize DC extrapolation errors, which is a superior approach to coupler-based microwave VNAs that have DC extrapolation errors. That is because VNAs that utilize couplers only have a start frequency of 10 MHz with increased signal-to-noise at frequencies below 500 MHz with degradation in TDR impedance measurements. VectorStar combines couplers in the microwave region and bridges at RF frequencies for excellent dynamic range down to 70 kHz.
To learn more about how to address reciprocity, passivity, and causality, download a free white paper entitled A Guide to Making RF Measurements for Signal Integrity Applications.
Comments