Use of Adapters in Network Analysis

ANA Techniques where Calibration Kits are Not Available

Measurements made with network analyzers are made with respect to reference conditions established using a calibration kit in a procedure known as "error correction". But what can be done when presented with a device for measurement and yet a calibration kit isn't available for that connector type? The usual course of action is to use adapters and the measure at the adapter interface. In many cases this method is adequate, although the measured values include the effect of the adapters which must removed in some way if less uncertainty is required. Many network analyzers have an "adapter removal" routine built into the operating firmware but, for this to work effectively, additional calibrations to characterize the adapters are required. Consideration must be given to the trade-off in accuracy against the time required to complete the measurements.

The first consideration is the acceptable quality of the measurement. In terms of measurement uncertainty, the aim should be for potential errors to contribute to the overall uncertainty budget in a reasonably "insignificant" way. A case can be made (see article Single Measurement Uncertainty) that "insignificant" can be defined as up to, say, 30% of the total uncertainty. Performing measurements using the "adapter removal" routines will usually produce the best results, but the measurement and set-up time will be much longer than the other methods that will be described later.

• Single Measurement Uncertainty

Effect of Adapters

The accuracy of measurements performed without using the adapter removal routine will depend on the quality of the adapters employed. Corrections will normally be required whose precision will also be influenced by the quality of the adapters.

The adapters chosen must enable the item to be measured to be connected to the test port cables directly. Ideally, no more than one adapter should be used on each lead to minimize the impact on measurement quality although, in some cases, two "high quality" adapters will produce smaller uncertainties than one "lower" quality adapter. Follow the procedure shown above but note that in most cases it is possible, if only correcting for the insertion loss (and phase) of the adapters it is usually possible to save the adapter measurements in the network analyzer's memory and then subtract them from the measurement of the adapters plus the item. If high-grade adapters are employed, it is often unnecessary to make corrections for reflection magnitude -- usually just increasing the measurement uncertainty (due to the uncorrected adapter) is adequate. Many high-grade adapters have a reflection coefficient of better than 0.02 across their operating frequency range.

Figure 1 -- Reflection coefficient measurement of an APC7 0.15 rho air line following a full, 2-port correction of the ANA using an APC7 cal kit. Adapters are added and the measurement remade using the same calibration constants. The difference between the red and yellow traces represents the effect of the adapters. Removal of the air line and connecting the adapters together provides the third trace.

Assumptions Ease the Problem

Adapters are easiest to measure in insertable pairs and it is best to use adapters that are of similar physical length. This allows an assumption that each contributes equally to the measured loss and phase shift. The reflection coefficient of the adapter-pair should be measured but can only be applied in corrections to the data for the device-under-test (DUT) and adapter combination if phase and magnitude measurements are made for all four S-parameters. This can be time-consuming and a short-cut is just to use the adapters' reflection coefficient magnitude as an uncorrected error contribution in the uncertainty budget. A common assumption is that adapter reflection coefficient magnitudes add in a quadratic fashion.

Figure 2 --Reflection phase measurements on the same port of the 0.15 rho air line. Again, the measured reflection phase of the adapters has been subtracted from the measurements made on the air line with adapters. Essentially, the two traces overlay each other so the third trace shows the phase difference of less than 0.5 degrees. Such small phase differences are generally difficult to discern as the stability of cables and system repeatability often exceeds this value.

Non-insertable devices pose some of the most difficult measurement problems, even when using the adapter removal capability of modern network analyzers Again, it is important to understand customer requirements as it might be acceptable to perform the ANA error correction in a convenient connector-type and then measure the DUT with an adapter. The measured values include the adapters but most precision-grade adapters have a fairly negligible effect at low to medium microwave frequencies. In cases where better accuracy is desired, the adapter characteristics must be determined. For non-insertable items with similar connector-type "X", it may be possible to use adapters that are sold in "phase and loss" matched pairs. These work on the principle that the X-to-male adapter has the same physical loss and phase shift as the X-to-female adapter so that the loss of two X-to-male adapters can approximated by measuring the insertable pair.

Figure 3 -- Transmission (loss) magnitude measurement of the 0.15 rho air line. The effect of the adapters has been removed by simple subtraction. This technique appears to work well where the match of the device approaches 50 ohms (the bottom of the peaks) but as the match worsens (peak top) the difference between using or not using adapters becomes more obvious. Also note that the effect of mismatch loss uncertainty does not always produce higher loss; some combinations have produced lower measured values than might be expected.

Further Difficulties

Unfortunately, some assumptions are made which are difficult to prove. Many manufacturers make male and female adapters that they advertise as being of similar loss and phase characteristics. This might be "proved" by assuming that the reflection parameters are similar. The approach relies on these similarities to arrive at an assumed loss and phase shift value for each adapter. The accuracy of this technique can be improved slightly by measuring each adapter with a similar one of the opposite sex (to form an insertable pair) and using the average measured values as corrections. At the expense of increased measurement and calculation time, even better confidence can result from making at least three measurements using different adapters and solving simultaneous equations to determine each adapter's characteristics.

Air lines have a predictable response but, since they have little transmission loss, imperfections in the measurement equipment, adapters or cables can be imposed on all measurements. For an attenuator pad, the effects of mismatch loss uncertainty and the uncertainties in the measurement of the mismatch tend to be isolated to the parameter being measured. For example, the mismatch measurement on port 1 will be influenced by the termination of port 2 to a much lesser degree. Therefore, the isolation provided by an attenuator more readily enables investigation of the individual parameters.

Figure 4 -- This plot shows reflection magnitude measurements made on a 10dB APC7 attenuator. The effect of the adapters is clearly seen and because the adapters' mismatch is similar to the attenuator, the additional error becomes quite significant. It is interesting to try to correlate the plots but it's not immediately obvious as only the magnitude term is shown. The errors are generated as the phase vectors rotate, causing the effects of mismatch to interact with each other. The worst deviation might therefore occur away from the peaks of mismatch, but where the phase components have also combined to greatest effect.

Figure 5 -- The transmission magnitude measurement using adapters looks much more predictable. Notice again that although we refer to "mismatch loss uncertainty", the term can appear to have "gain" (i.e. in the 2GHz to 6GHz region of the plot). The attenuator has a reasonably good match at its ports, as do the adapters, so the apparent error when using the adapters is, in this case, quite small.

Non-insertable devices with dissimilar connectors are the most difficult to measure. Devices often falling into this category are couplers, splitters, combiners and sometimes cables. The method is to measure each adapter with a similar one of the opposite sex. As before, common practice is to ignore the effects of the adapters on the DUT reflection measurements, instead making additional allowance in the uncertainty budget. An approximation for this additional contribution can be obtained during the measurement of the adapters' insertion loss by setting the ANA to also display S₁₁ or S_22; each adapter contributing in a quadratic manner.

Figure 6 -- Here, the transmission phase of the 10dB attenuator is measured. With some isolation between the ports and reasonably good matches, phase measurement works well even with adapters. The differences shown are well within the repeatability of the analyzer and its cables.

The Time Domain Solution?

Time domain gating is a process whereby the measurements made in the frequency domain are mathematically converted into the time domain. Any discontinuities in the transmission line (such as connections) then show as a spike on the analyzer's display and can be "gated out". The measurements can then be converted back into the frequency domain and the characteristics of just the DUT displayed.

But time domain measurements can be problematic. In order to obtain the best resolution, high frequencies must be used but, unfortunately, measurements in the upper and lower 10% regions of the frequency band cannot be considered reliable. The Time Domain function on any network analyzer requires a lot of skill to use; it is easy to make repeated measurements but get very different results each time.

Summary

• Reflection coefficient of the adapter(s) often dominates the measurement uncertainties.
• The adapter(s) insertion loss and phase shift can be readily corrected for and will tend to have little effect upon the measurement uncertainties.
• Lower grade adapters are more likely to have reflection coefficients that are sufficient to degrade the equivalent load and source match terms of the network analyzer that consequently impacts the accuracy of the adapter insertion loss and DUT attenuation measurements.