VNA Reflections

Cold Source Method Brings More Confidence to Multi-Port Device Testing

February 15, 2019

The complexity of today’s high-speed designs has created many a challenge for signal integrity engineers developing communications devices and systems. When time-to-market and tight cost parameters are added to the equation, finding the right test solution to meet emerging applications can be the difference between a product performing according to specifications and one that doesn’t make it into mass production.   

When it comes to vector network analyzers (VNAs), engineers need an instrument that can effectively conduct noise figure measurements. This has become particularly more daunting due to the proliferation of multi-port devices in emerging designs.

In recent years, numerous improvements have been made in noise figure measurements through better algorithmic understanding of the measurements, more sensitive receivers, and less error-prone methods of processing noise power measurements. One advance is the utilization of the cold source noise figure measurement method, which has benefits compared to the more traditional Y-factor technique.

Cold Source Method

The cold source noise figure measurement method was developed to eliminate the requirement for a multi-state noise source. A simpler, better controlled noise source – nominally a termination at room temperature – can be used in the cold source approach. In this case, the noise figure is found from a more easily populated equation, shown below.

To calculate the noise figure, a few key steps must be taken:

1. Because an absolute noise power is required (numerator N) with the cold source method, a means of a receiver power calibration is necessary. Anritsu has developed highly accurate power calibrations and very linear broadband receivers to accommodate this requirement.
2. An effective measurement bandwidth (B) is needed. Since measurement bandwidth is largely determined by the digital IF system of the VNA, B can be pre-determined.
   
   
3. To isolate the noise figure of the device under test (DUT), the receiver’s noise contributions must be taken into account. Similar to the conventional Y-factor method, a measurement of receiver noise is required, however, this approach only requires the cold source to be attached to the receiver input. Taking the receiver noise into account, the equation above can be re-expressed in the form below:

One thing is immediately obvious when looking at these equations. Errors in gain or the noise power measurement will propagate to noise figure on roughly a dB-for-dB basis, if the composite receiver gain is sufficiently high.

Advantages Over Y-factor Method

The cold source approach, which is used in the noise figure measurement capability of the Anritsu VectorStar^® VNAs (figure 1), addresses some of the limitations associated with Y-factor. Y-factor uses a noise source that can produce a low noise output power (cold = Nc) and an elevated one (hot = Nh). This noise source becomes the input noise signal to the DUT.

Figure 1

The ratio of measured noise powers in these two states was termed the Y-factor (Y=Nh/Nc) and could lead to a quick calculation of noise figure. By making a noise figure measurement of the receiver itself and the system (DUT + receiver), it is possible to deconvolve the DUT noise figure.

One advantage the Y-factor noise figure measurement method is that no absolute power calibrations are needed; they are all based on ratios. The measurement also is not dependent on DUT gain. 

On the other hand, calibration and mismatch errors are a concern when using  the Y-factor method. Also, the noise source used in the Y-factor method needs to be periodically calibrated to confirm measurement accuracy. Mismatch errors become a consideration due to the match changes of the source when switching between the hot and cold states. This can result in measurement errors, particularly as the DUT input match worsens and effort is not made to correct.

Cold Source Method

The cold source method simplifies the measurement by only requiring a load at the input of the DUT.  A power calibration is used to determine the noise power of the receiver and thus the measurement becomes an absolute noise power difference. Anritsu uses the cold source method as part of its Single Ended and Differential Noise Figure option for the VectorStar that allows the VNA to measure 2-, 3- and 4-port devices in single-ended, differential, and common mode operation with a variety of processing options. The new differential noise figure option adds the ability to perform levels of vector correction in both 2-port and multi-port devices for even greater accuracy, particularly when mismatch is significant.

To learn more, download a new application note from Anritsu entitled Performing Differential Noise Figure Measurements. The educational document explains the two methods, as well as outlines the steps to conducting the measurements using the VectorStar VNA.

How High-frequency Designs are Changing De-embedding Approaches

November 1, 2018

Signal integrity engineers have increasingly been reliant on de-embedding methods, particularly in PC board and cable assembly spaces. Constant challenges are always repeatability issues and standards availability problems at the device-under-test (DUT) plane.

Some of the same issues occur in millimeter (mm-wave) fixtures, where repeatability can be even more of a challenge. For certain fixtures, the repeatability and standards sensitivity associated with newer de-embedded methods can be orders of magnitude better than classical methods, while showing similar sensitivities to first tier calibration issues. The absolute errors can, however, be substantial for certain distributions of mismatch within the fixture.

Given the need for conducting de-embedded measurements at higher frequencies, especially for emerging 5G and advanced wireless designs, an improved process is necessary. In this post, we will outline limitations associated with conventional methods, how to address them, and which techniques are best suited for today’s high-speed emerging designs.  

mm-wave De-embedding Techniques

Dozens of de-embedding routines are commonly used at mm-wave frequencies. Most are based on limited assumptions, often reciprocity and perhaps symmetry, about the fixture parameters. Engineers have found at lower frequencies that if the DUT interface has repeatability or standards issues, these classical approaches may be sub-optimal for various reasons:

1. Some classical methods (e.g., Thru-reflect-line [TRL]) are very sensitive to changes in launch impedances/admittances
2. Some classical methods (e.g., defined-standards approaches) rely on geometrically well-controlled structures at the DUT plane (not always possible)
3. Classical methods deterministically solve for inner and outer plane match and, as those match levels degrade, there is a nonlinear coupling to insertion loss extraction

All these potential problems are compounded when designing for mm-wave frequencies.

A second approach - partial information de-embedding – has also been used at lower frequencies. It may be more useful in addressing the mm-wave fixture problem. While there are many permutations on these methods, one variety that uses spatial separability of dominant mismatch centers from the main path appears to have more advantages.

Some Different Approaches

While there are many different de-embedding methods, they can be categorized by two types: classical (or full information; i.e., the fixture network is solved with minimal assumptions) and partial information (where more assumptions and structural information about the fixture are used). Examples of the classical methods are TRL and Bauer-Penfield. TRL uses a two-tier calibration based on transmission line characteristics to allow computations of fixture parameters while Bauer-Penfield is based on one-port calibrations for each fixture arm at both inner and outer planes. The latter technique can be viewed as a generalization of open-short de-embedding that is popular in on-wafer measurements.

In partial information techniques, fewer standards are used but more assumptions are made, including the distribution of mismatch within the fixture. The reflection data is correlated with a series of propagation kernels to separate the response portions due to the reflect standard and those caused by the internal mismatch centers. The process is shown schematically in figure 1. Because the contribution separation is based on phase resolution, there is a limit when a mismatch center near the DUT interface cannot be separated from the reflect standard at the DUT interface. There are also variations on this approach that use time domain processing. 

 

Figure 1

Let the Experiment BeginWe conducted a somewhat controlled measurement experiment using the two method classes on a WR-10 fixture to show the results when repeatability is poor. The DUT plane did not have an anti-cocking support and had about 200-μm concavity of the flange surface. Two additional holes were drilled in the flange surface, each 2 mm from the aperture narrow walls, to mount the DUT.

For each connection, the waveguide screw torque was randomized with a uniformly distributed range of 0.5-0.7 cN-m on each screw. The same first tier calibration was used for all runs and 20 second tier runs were done.  Figure 2 shows that the scatter was limited to largely one portion of the frequency range using the partial information method.

 

Figure 2

The scatter on Bauer-Penfield, which uses a short-short-load (SSL) set of standards, was larger (figure 3). Not surprisingly, the distribution is not particularly well-behaved, as there are multiple non-linear geometrical mechanisms involved with both positional alignment and skew gap formation.

Figure 3

The absolute errors of the partial information methods are a function of the spatial structure of the fixture because of the phase correlation techniques used. Another significant factor is the processing details, making it difficult to come to a 100% certain conclusion but the relative repeatability immunity trends seem to hold.

Sensitivity to Standards Errors

Another point of interest was the sensitivities to standards errors. For the Bauer-Penfield technique, SSL standards were the basis when we conducted the experiment but conclusions for an SSS set are similar. As one of the shorts is common to both methods, a basis for comparison was a simulation based on an error in that short offset length. Plots are shown in figure 4 that yielded roughly equal peak excursions but the distributions of the length errors were wildly different for Bauer-Penfield (+/- 10 μm) and for the partial information method (+/- 2 mm).

Figure 4

The reduced sensitivity of partial information method is not surprising since the phase interval used in the correlation was on the scale of cm. The result is that an offset length error resulted in relatively minor magnitude impact for the new method. S21 phase is transparently affected for the partial information method if the offset length was used explicitly, rather than an auto-rotation scheme. Two aspects of the Bauer-Penfield behavior need to be noted:

1. The two offset short lengths were chosen for a 180° reflection phase difference at 90 GHz, so the sensitivity to length error is minimized at that frequency
2. Altering the length entry directly effects the S22 as well as the S21 extraction and there is feedback between these terms

Conclusion

As the experiment indicates, classical de-embedding techniques can cause noticeable measurement problems in some repeatability-challenged mm-wave fixtures. To address this challenge, Anritsu has developed a measurement technique for its VectorStar^®  vector network analyzers (VNAs) that can help engineers more accurately measure design performance. You can learn more by downloading this white paper entitled MM-wave Partial Information De-embedding: Errors and Sensitivities.