Under normal circumstances, if an adapter is added to a VNA port, one could measure its two-port response and then use de-embedding to remove its effect on the measurements. De-Embedding is a standard feature on Copper Mountain Technologies’ VNAs and is accessed under the “Analysis” menu. A balun is not a two-port device, however, so in this case, the standard method is not possible; but an approximate method does exist.

SDS21 is like S21, SSD12 is like S12 and SDD22 is somewhat like S22. If a two-port matrix is formed from these values and then used for de-embedding, the results are reasonable. To be clear, a normal two-port S-parameter matrix measurement of an adapter would be used but the differential equivalent may be substituted as an approximation. A normal S-parameter matrix for de-embedding:

Differential Approximation:

The reason why this method is only approximate is that the balun is not perfect and SCS21 of the first balun in a pair used for measuring a balanced DUT is not zero. A small common-mode signal is created on the output of this balun along with the balanced signal. This signal is passed along and creates a small, unbalanced signal on the output balun due to its SSC12. This signal adds to that created by SSD12. These balun imperfections result in small errors in the de-embedding. A more thorough treatment using all mixed mode parameters is certainly possible, but this must done with post-processing in Python or some other programming language while the de-embedding matrix (15) can be used with the built-in de-embedding feature of the VNA.

It is possible that the DUT between the two balun adapters might have a particularly large SCS21 and the small erroneous unbalanced signal created by the balun might be greatly amplified, causing an even larger error in the measurement.


A pair of baluns is evaluated, and de-embedding files are created. The ADC-WB-BB baluns from Texas Instruments are made from Mini-Circuits TC-1-13MA+ baluns mounted on connectorized circuit boards (see Figure 7). The frequency range is 4.5 MHz to 3 GHz.

Figure 7

Figure 7 Balun board.

Three-port S-parameters are measured for the two balun boards over the 4.5 MHz to 3 GHz range and Touchstone de-embedding files are shown in Figure 8 using entries from the de-embedding matrix (15).

Figure 8

Figure 8 Calculated Touchstone file.

The two balun boards are connected back-to-back with a pair of male-to-male SMA adapters (see Figure 9) and measured with and without de-embedding. Figure 10 shows the measured results without de-embedding. Return loss is poor and there is greater than 4.5 dB of loss at 3 GHz.

With de-embedding using the files created for the two baluns, the results are more reasonable (see Figure 11). The return loss is considerably better, and insertion loss is flatter. There are places where the insertion loss is slightly positive, which is clearly an error, but only within a few tenths of a dB. Evaluating the de-embedded baluns in this manner is a good way to estimate the measurement errors.

Figure 9

Figure 9 Balun measurement setup.

Figure 10

Figure 10 Back-to-back baluns without de-embedding.

Figure 11

Figure 11 Back-to-back baluns with de-embedding.


The mixed mode S-parameters for a balun are introduced and the method for calculating them explained. A method for creating two-port de-embedding files suitable for use with a 2-port VNA and a pair of baluns is shown and results demonstrated. Errors in the method are examined and for some applications, the convenience of standard two-port de-embedding techniques built into the VNA may outweigh the small errors caused by balun imperfections.


  1. K. Suto and A. Matsui, “Two-Port S-Parameter Measurement of Wide-Band Balun,” Proceedings of the International Symposium on Antennas and Propagation, October 2016.
  2. R. W. Lewallen (W7EL), “Baluns: What They Do and How They Do It,” ARRL Antenna Compendium, Vol. 1, 1995, pp. 157-164, Web:
  3. “Measurement Techniques for Baluns,” Anaren, May 2005, Web: