# Updating the Bob Smith Termination Technique

For designers of Ethernet systems, the Bob Smith Termination technique is regarded as an important approach for migrating potential RF emission and susceptibility problems for Ethernet infrastructure. But, what if the technique has flaws?

Unfortunately, there are some flaws with the termination technique patented by Bob Smith IN the 90s. This article points out potential flaws and then shows how a proposed termination technique can solve some of the problems encountered using a Bob Smith termination approach. The article also provides a physical verification method that designers can use to analyze the impact of the proposed termination technique.

The problematic results of improper termination are difficult difficult to measure and verify. They show up as problems with electromagnetic compatibility (EMC certification and worse as intermittent and difficult to identify performance problems in service. Thus, this article will also provide a sound theoretical approach for evaluating the proposed termination scheme.

**Problems with Mr. Smith**

In a U.S. patent, Robert W. (Bob) Smith described a method to reduce the longitudinal or common mode current on multipair conductor systems where the pairs are interrelated in a uniform manner. He alludes to the fact that the pair to pair relationships of a CAT5 cable form transmission lines in themselves.

Smith asserts incorrectly that such transmission lines exhibit a characteristic impedance of approximately 145 ohms. His description is confusing and misleading because he does not define the configuration in which this is supposedly exhibited. He correctly goes on to say that the cable can support and radiate common-mode signals. Moreover, he states that there is a method to reduce the common mode radiation by matching the common mode impedance of the cable, and then presents a method of terminating the common mode transmission line, that does not in actuality accomplish that goal well. His termination technique does reduce the common mode radiation, but not optimally and not from the correct understanding.

For a CAT5 cable, there are two common modes of interest. The first is the pair-wise common mode and the second is the cable-wise common mode. In the article to follow, we'll notes will address the pair-wise common mode.

The pair-wise common mode has a common mode characteristic impedance that can be readily characterized, while the cable-wise characteristic impedance is grossly dependent on implementation. Actually, the characteristic impedance of the cable-wise common mode is not a meaningful concept except in some well-defined circumstances.

With that background in mind, let's look at the pair-wise common mode characteristic impedance relationships within a CAT5 cable. As **Figure 1** points out, the pair-wise mode calls for four pairs with an intrinsic characteristic impedance between each pair. Due to the approximate symmetry in the construction of the cable, each pair has the same relation with any other pair. It is prudent to define the relationships to be described (Smith doesn't do that). Since the inherent characteristic impedances cannot be measured directly, to verify the theory it is necessary to measure them in combinations and resolve the individual elements.

*Click here for Figure 1*

The first relationship is between any two pairs (Z_{1-1}). This parameter is somewhat meaningless in application because all four pairs are interrelated, but it is useful and can be characterized none-the-less. The second and more meaningful relationship is the characteristic impedance of one pair relative to the other three pairs (Z_{1-3}). The third is the relationship of two pairs to the other two pairs (Z_{2-2}). For the case of one pair versus any other pair with the two remaining pairs floating, the measured characteristic impedance is approximately 100 ohms.

For the case of one pair versus the other three pairs tied together the measured characteristic impedance is approximately 70 ohms. For the case of two pairs tied together versus the other two pars tied together the measured characteristic impedance is approximately 50 ohms. Nowhere does there exist a relationship that the characteristic impedance is anywhere near Smith's stated 145 ohms.

**Figure 2** shows the definitions of some of the possible measurable characteristic impedances.

*Click here for Figure 2*

**Table 1** shows some measured data samples. These were taken using the Agilent 4396B Network Analyzer.

**Table 1: Measured Data Samples.**

**Analysis**

Analysis of the measured values to determine the inherent characteristic impedance values is complex. Some techniques include:

Z_{1-3(1,234)} = Z_{14}// Z_{12}// Z_{13}

Z_{2-2({12,34)} = Z_{14}// Z_{23}// Z_{13}// Z_{24}

Z_{1-3(1,234)} = Z_{12}//(Complicated combination of Z_{14}, Z_{34}, Z_{23}, Z_{13}, Z_{24})

This does not appear to be readily manageable. Therefore, a simulation was developed selecting values for Z_{nn} to give values something near the measured data. From this, the intrinsic characteristic impedance Z_{nn} appears to be in the vicinity of 200 ohms. From this simulation, it can be suggested that the value of Z_{nn} varies along the length of the cable as well as from pair to pair. Thus, to properly terminate the common mode transmission lines, the technique in **Figure 3** must be used.

*Click here for Figure 3*

One way to visualize a proposed termination method is to terminate the Z_{1-3} characteristic impedance. First, that transmission line must be terminated with a value equal to its characteristic impedance. A single resistor termination will accomplish this, however the other three pairs are not tied together in normal use.

A more general approach, in theory, is to form a resistor network comprised of four resistors of equal value, one in series with the other three in parallel. Once the network is set up, designers can then compute the resistance value such that the network matches the Z_{1-3} characteristic impedance of the cable **(Figure 4)**.

Smith claims that all four resistors should be 75 ohms **(Figure 5)**. This means that a 75-ohm resistor in all four terminations would match a common mode impedance of 100 ohms for Z_{1-3}.

But Smith claims the characteristic impedance of "some" common-mode configuration that he is attempting to match is 145 ohms. Since it's difficult to understand in the patent the configuration Smith is describing and configurations typically are not anywhere near 145 ohms, it can only be concluded that Smith's numbers and approach to deriving them are wrong. Additionally, his matching technique, given a common mode characteristic impedance, does not produce a match.

Thus, Smith hasn't presented a really workable solution to the problem he has recognized. For example, for CAT5 and CAT5E cable an improvement in return loss achieved by replacing the 75 ohm resistors by 52.3 ohm resistors, would be from 15 dB to more than 28 dB, given a variation in common mode impedance of +/-5 ohms. The variation of common mode characteristic impedance is taken into account in the calculation of return loss **(Table 2)**.

**Physical Verification**

For the above conclusions to be accurate, it's important to have a physical verification method. Fortunately, there is a relatively simple method for doing verifying the above claims.

Under the proposed verification method, designers configure a length of cable with the termination resistors, then insert a common mode pulse while monitoring that insertion with an oscilloscope (a poor man's TDR). A pulse generator with an output impedance of 50 ohms and a rise time of approximately 1 ns was used **(Figure 6)**.

*Click here for Figure 6*

In the verification method, if the cable is properly matched, no reflection should be seen. If not, a reflection will be observed.

In the example described in this article, terminations of three values were applied at each end of a cable seven feet long. The resistor values employed were sets of 52.3, 75 and 100 ohms. The cable employed was 7 feet long.

The series of scope presentations in **Figure 7** show implementations of the various terminations for comparison. The fact that a termination in the vicinity of 52.3 ohms is proper for the match can be clearly seen.

*Click here for Figure 7*

For calibration purposes, several conditions are shown are shown in Figure 7. The first scope presentation shows the output of the pulse generator with no load. The second shows the cable connected to the pulse generator as in Figure 6 and the far end open circuited. The third is the same conditions with the far end short-circuited. The remaining three scope presentations show the results for the three values of the termination resistance.

It is clear that for the 100 ohm and 75 ohm resistances there is a reflection, whereas, with 52.3 ohms there is not. The conclusion is that in the vicinity of 52.3 ohms is correct and that Bob Smith's 75-ohm value is not correct.

As a note, CAT6 cable has a higher common-mode impedance. This is expected due to the significant difference in its construction. It is anticipated that also due to its method of construction that the characteristic impedances will be much more uniform both versus each other and along the length of the cable.

**Theoretical Analysis**

**Figure 8** shows the model for computation of the termination resistor given Z_{nn} and R, the termination resistor. The theory is that when the input impedance is equal to the value of R/3 + Z_{1-3} then the cable is matched. The value used for Z_{nn} was 200 ohms, which is close to correct. With this value for Z_{nn}, R turns out to be 52.3 ohms. This agrees with the pulse reflection test.

*Click here for Figure 8*

Once viewed, the solution is simple and obvious. In Figure 8, short pair 1 to pair 2 to pair 3 **(Figure 9)**. In this case we have the Z_{1-3} characteristic impedance and the three resistors, R, are tied in parallel, yielding R/3 for the resistors. Then looking into the terminals indicated, the impedance will clearly be Z_{nn}/3 plus R/3. Thus, R+ R/3 equal to Z_{nn}/3 would be the proper match resistance.

*Click here for Figure 9*

Now, if the short is removed, due to symmetry, nothing is changed since no current flows through Z_{13}, Z_{12}, and Z_{23}. This verifies that the approach reported above is correct.

As the values of Z_{nn} depart from symmetry, the analysis is no longer strictly correct. However, for practical purposes, the departure for real CAT5 cable is not really significant.

As stated above, the common-mode characteristic impedance of CAT6 cable will be higher than for CAT5 and CAT5e cable. The proper common-mode termination for CAT6 cable will be about 66 ohms versus the 52.3 ohms for CAT5 cable. If it is envisioned that there will be a mixture of CAT5 and CAT6 in a predominance of installations, then a compromise value between the two would probably be appropriate.

**Further Work**

Others may further this work by formal characterization of more cable from a number of manufacturers. A stastical and documented mean for the characteristic impedances of interest would be useful to designers. It is necessary to characterize more cable of different types and from different manufacturers. this work is currently under development.

**Author's Note:** This study was performed while the author was employed at Cicada Semiconductor, which has since been purchased by Vitesse Semiconductor.

*
About the Author
James Satterwhite is the president of Teltest Electronics Laboratories, Inc. James has a BEE from the University of Florida and an MSEE from Purdue University. He is also a senior member of the IEEE and can be reached at jims@teli.us
*

Author

satter 2/4/2014 7:04:21 PM

Thanks for reading the article,

Jim

Author

dougp01 8/17/2013 6:40:55 PM

An assumption seems to be made that the coupling bewtween the conductors is a constant impedance over the entire legnth of the cable. In fact, when Bell Labs first developed the concept of paired bundles for telcom wire many years ago, the problem of coupling between pairs was addressed. The solution they adopted on pair-wises cables was to introduce a different twist rate on each pair of wires. The effect they were seeking was to ensure that field coupling between pairs was cancelled every few centimeters. With 8-pair cable this can be easily verified. With 25-pair or 50-pair is it a little more difficult but this is where twist of pair-bundles comes into play.

-doug