Praesent nec lectus eu neque fringilla volutpat. Nulla tristique ullamcorper quam, ut dapibus massa hendrerit ac. Etiam a tempus lectus. Proin nibh nisi, vehicula non enim sed, venenatis faucibus orci. Fusce ut elementum nunc, vel eleifend sem.
INTRODUCTION
When evaluating the performance of an EMI filter, there are several ways to approach the test methodology. For single-ended components such as capacitors and inductors (lumped elements), a two-port network analyzer can perform a very accurate broad frequency characterization of components. These characterizations can prove valuable when building accurate circuit models. Still, when evaluating a dual-line filter, the complexity of measuring and modeling a filter’s performance over a broad frequency can produce results that differ substantially from real world out comes.
For accurate determination of common mode and differential mode noise performance on a dual-line filter, the circuit designer needs to understand the parasitics associated with the circuit, as well as the parasitics associated with the EMI filter. Parasitic variances induce impedance imbalance between lines as well as to the circuit’s reference (ground). To compensate for parasitic imbalances, it generally requires a complex EMI filter design resulting in larger values of lumped elements (capacitance and inductance). A properly designed and implemented EMI filter will balance a circuit’s parasitic impedance mismatch to combat common mode and differential mode noise, as well as the conversion from one to the other. The scope of this article is to propose a test methodology that will determine the balance and parasitics of an EMI filter both in and out of a circuit.
WHAT IS BEING MEASURED?
The first step in evaluating an EMI filter is to determine what is being measured and what information is important. To simplify these questions, an engineer can classify measurements in one of three ways.
Figure 1. “System classification” measurement. Notice that the test points are located at the IC and the connector.
Figure 2. “Filter-in-system classification” measurement. Notice that the test points are located just before and after the EMI filter.
System classification – shows the net result of adding an EMI filter to a circuit’s loop that measures the total impedance of the filter plus physical geometries such as PCB layout, vias, and plane spacing. For example, Figure 1 has an IC driving a differential signal pair to a connector at the edge of a PCB. Test points pairs should be located at the pins of the IC and at the pins of the connector. The net impedance includes the total impedance from both the filter and the PCB.
Figure 3. “Filter-only classification” measurement. Notice that the test coupon is just big enough for the EMI filter and connectors.
Filter-in-system classification – this measurement would be the same as “system classification” except that it measures only the impedance of the filter and the filter’s loop on the PCB. Figure 2 shows that test point pairs would be located just before and after the filter. The net impedance includes parasitics from the filter and its mounted loop.
- Filter-only classification – a specifically designed test coupon is used to measure an EMI filter outside of the intended circuit where all fixture and circuit parasitics can be removed. The net impedance includes only the filter’s parasitics (Figure 3).
It is important to note that classifications can overlap; however, this overlapping does not present a significant problem as long as parameters such as those listed below are fully defined and agreed upon.
- Test fixture setup/layout
- Location of test points
- Number of components that comprise the filter
TEST COUPON DESIGN FOR FILTER-ONLY MEASUREMENTS
A test coupon has many factors that can significantly affect measurements and these will, of necessity, affect correlation to other measurements. Most component/filter manufacturers supply information such as insertion loss curves or impedance measurements for their products. Unfortunately, no standard exists to control how this data is obtained. Evaluation of products from various manufacturers, or even different product lines from the same manufacturer, can pose an impossible challenge if further information is not available. The designer must determine how the test fixture was constructed, obtain a description of the test setup, find the type of calibration employed, and learn the size of the sample base. If additional information about manufacturer’s data is not available, then the engineers must determine if they will accept the data, or will need to perform a validation test of their own.
The good news is that test coupons, with a little thought, can be constructed very inexpensively and can provide good performance insight for most filter evaluations. Considerations for any good test coupon should be the frequency range of interest and the type of measurement classification desired. Defining these things will help determine test coupon’s parameters including dielectric material; its shape and size (length, width, and height); and its layout, construction, and test points.
Figure 4. Top coupon is an “open” and bottom coupon is a “short”. Both are used to apply a fixture correction to a network analyzer’s calibration plane.
NETWORK ANALYZER CALIBRATION, SETUP AND FIXTURE EXTENSION
Prior to performing a calibration on the network analyzer, the setup, channel, and state should be loaded. The setup should include start and stop frequency, measurement bandwidth, number of points, channel and trace setup, and type of measurement (e.g., balanced pair measurement). “Balance pair measurement” is also known as mix-mode or infinite balun measurement. (In the following discussion, the term “mix mode measurement” will be used.)The mixed mode measurement determines the common mode (CM) and differential mode (DM) responses, as well as the conversion responses—i.e., common-to-differential (CD) and differential-to-common (DC). (Mix-mode measurements are similar to using physical baluns without the calibration difficulty or frequency limitations.[1])
Another item to be aware of when using four-port network analyzers is the manner in which mix mode measurements are taken. In a “virtual” mix-mode, a signal is sent out one-port at a time. The CM, DM, DC, and CD responses are then calculated using software. In a “true” mix-mode, a true differential signal is sent out on two-ports simultaneously. Actual hardware is used to measure CM, DM, DC, and CD. (The cases for “virtual” and “true” mix-mode are explored in the example below.)
The calibration of a network analyzer usually includes the test cables as a minimum. This process can be carried out with either a manual calibration kit (through-open-short or through-open-short-matched) or with an electronic calibration device. Either method will result in a highly repeatable calibration of the network analyzers and cables. Still, depending on the classification of measurement being taken (system, system-in-filter, or filter-only), some thought should be given to include calibration of the test coupon and/or any probes attached to the test cables.
Extending the calibration plane to the test fixture (coupon) can be done by simply applying an “open” and “short” correction. The procedure for applying this correction can be found in the application notes supplied by manufacturers of network analyzers.[2] Examples of “open” and “short” coupons are shown in Figure 4.
To demonstrate the difference between a mix-mode measurement with and without a test fixture (coupon) correction, see Figure 5 and Figure 6. Figure 5 shows the test layout and equipment setup, and Figure 6 shows the results of a common-mode-choke with and without a test fixture correction. As shown in the data, applying the coupon correction has a very noticeable effect on the two conversion measurements CD and DC.