Higher Data Rates Require New De-embedding Techniques

By Jon Martens and Bob Buxton

Higher data rates can create challenges for traditional de-embedding techniques. As frequencies approach 70 GHz or even 110 GHz, errors related to fixturing can be greater than those of the “device under test” or DUT. Fixtures and connectors to DUTs come in many forms. Poor de-embedding can lead to both passivity and causality errors. In addition, high fixture loss may affect the accuracy and repeatability of de-embedding. Achieving accurate measurements at these higher frequencies offers the advantage of improved ability to locate discontinuities, impedance changes, and crosstalk issues.

Newer more flexible and repeatability-tolerant methods of calibration and de-embedding techniques help resolve complex 28 Gbit/s problems. This white paper discusses the available de-embedding techniques and the tradeoffs between them.

Fixture De-Embedding

There are many situations where it may not be possible to connect directly to the DUT. In this case it is necessary to de-embed the DUT from the surrounding test fixtures. The opposite is sometimes required: it may be useful to take a device and assess its performance when it is surrounded by other networks. Figure 1 illustrates this.

Figure 1. De-embedding can be used to remove test fixture contributions, modeled networks and other networks described by S-parameters (S2P files) from the measurements. Embedding is the reverse process.

Poor de-embedding can lead to both passivity and causality errors. Poor causality, where outputs can appear to happen in negative time, can be caused when there is insufficiently high frequency content. Passivity errors occur when it appears that a passive device has gain or is otherwise converting energy. The passivity effect can be subtle and due to small de-embedding problems, but it can have large effects on follow-on modeling or simulation. For example a minor calibrations standards problem can lead to small changes in the measured fixture match with apparent gain.

Using Flexible De-Embedding Techniques

The solution is to have a wide range of techniques available that can handle different situations. By matching the calibration and de-embedding method to your specific DUT and fixture structures you will improve the accuracy and repeatability of your measurements. Table 1 lists several of the possible extraction methods for de-embedding.

Table 1. De-embedding Methods

As can be seen there are many extraction methods available and the choice is somewhat context dependent. For signal integrity applications, the most common methods are B/D/F/G/H. Types A/C/E are more commonly used for active device testing such as transistors or power amplifiers.

The Bauer-Penfield method, or Type B, offers the best accuracy and is the most commonly used method when the fixture can be treated as uncoupled. It requires that you have a good set of standards – the equivalent of open/short/load or offset shorts. Figure 2 highlights the sensitivity of this method to the quality of the standards being used. This method also requires good repeatability.

Figure 2. Type B is the most accurate method, but requires quality standards. This example shows the impact of a load standard return loss of 10 dB instead of the intended 30 dB.

When one has an interface with poor repeatability in terms of placement accuracy sensitivity), then types B and E may respond badly, whereas types D, F and G are more tolerant (Figure 3).

Figure 3. Contact repeatability is a determining factor in selecting your de-embedding method. This example shows the result of the different techniques on a fixture with large pads and not very accurate positioners.

The methods of type D/F/G only require transmission line interconnects and while they offer a reduced accuracy from type B, they perform well in situations where a complete set of standards is not possible. These methods are designed to de-emphasize inner plane match (on the fixture side) and mainly go after insertion loss. As such, if the loss is high and/or the inner plane match is good, they do well. If the fixture is low loss and poorly matched, they will not do well. An example is shown in Figure 4.

Figure 4. Types D/F/G are designed to de-emphasize inner plane match and work well with high insertion loss and is less sensitive to matching (figures show results for type F).

The main difference between the D/F/G methods relate to the number of ports needed and whether the transmission line coupling is an issue or not. For some low-end products, the insertion loss is the critical specification and coupling issues such as cross talk are not required and so type D or F may be used depending on whether the application is 2 or 4 port. For products where energy transfer between lines is a concern, such as backplanes, the type G method is required. Figure 5 highlights the difference between the coupled and uncoupled methods.

Different from the other methods, type H is based on time domain processing and uses spatial separability to remove fixture effects. As such, it works best when the fixture arms are long relative to the wavelengths involved (else inaccuracies can increase). Only very simple standards are needed (often just a short at the DUT plane). It also tends to be relatively insensitive to repeatability problems.

Conclusion

The de-embedding challenges created by higher data rates are meet with the wider variety of methods now available. Higher frequencies require that signal integrity designers use careful considerations to which method best meets their fixtures need. Considerations such as fixture loss and probe/contact repeatability must be considered at these higher frequencies. Achieving accurate measurements are even more critical in reducing passivity and causality errors so that there is an improved ability to locate discontinuities, impedance changes, and crosstalk issues. Table 2 shows a summary of recommended methodologies for de-embedding in signal integrity applications.