PRACTICAL DESIGN CONSIDERATIONS TO INCREASE MEASUREMENT ACCURACY IN A LOW COST VECTOR NETWORK ANALYZER WEI LI, JAN VANDEWEGE Department of Information Technology (INTEC) University of GENT St.Pietersnieuwstraat 41, B-9000 Gent BELGIUM Abstract: Precision design work and tight manufacturing tolerances demand highly accurate measurements. A Vector Network Analyzer (VNA) is one of the most important aids for linear RF analog designs. It can give accurate information about the passive network performance in both the frequency domain and the time domain. Designers heavily rely on these measurement results to locate the design problems and to evaluate the network performance. But the measurement errors always exist. In any high frequency measurement, there are measurement errors associated with the system that contribute uncertainty to the results. Therefore, measurement accuracy of a VNA is an important concern to the instrument designers and suppliers. This paper describes the main issues of determining measurement accuracy associated with VNA measurement. Based on the low cost Vector Network Analyzer developed by INTEC, University of Gent, Belgium, the practical design considerations have been discussed and employed in this instrument. The measurement results are present to illustrate the improvement of the measurement accuracy after taking into account these measurement issues in the design procedure. Key-Words: Vector Network Analyzer, Measurement, Analog circuit design. 1 Introduction 1.1 Importance of measurement accuracy of Vector Network Analyzer Precision design work and tight manufacturing tolerances demand highly accurate measurements. A Vector Network Analyzer (VNA) is one of the most important aids for linear RF analog designs. It can give accurate information about the passive network performance in both the frequency domain and the time domain, such as insertion loss, return loss, phase response, group delay, impulse response and so on. Designers heavily rely on these measurement results to locate the design problems and to evaluate the network performance. But the measurement errors always exist. In any high frequency measurement , there are measurement errors associated with the system that contribute uncertainty to the results. Therefore, measurement accuracy of a Vector Network Analyzer is an important concern to the instrument designers and suppliers. As well as the VNA measurements themselves, most errors in VNA measurements are complex quantities that vary as a function of frequency. By characterizing and virtually removing the systematic errors, measurement accuracy can be improved by several orders of magnitude. A plenty of commercial network analyzers are now offered with built-in, high-speed computational hardware that can perform the complex mathematics required for sophisticated error correction. 1.2 Main factors measurement accuracy of determining Commercial network analyzers, such as the series of network analyzers produced by Hewlett-Packard, Rohde & Schwarz, etc., are instruments that measure transfer and/or impedance functions of linear networks through sine-wave testing. A network analyzer system accomplishes these measurements by configuring its various components around the device-under-test (DUT). The first part of a VNA instrument is the VNA transmitter, a broadly tunable sine-wave signal source to stimulate the DUT. Since transfer and impedance functions are ratios of various voltages and currents, a means of separating the appropriate signals from the measurement ports of the DUT is required. Finally, the network analyzer receiver must detect the separated signals, form the desired signal ratios, and display the results in the format requested by the user. A perfect measurement system would have infinite dynamic range, isolation, and directivity characteristics, no impedance mismatches in any part of the test set up, and a flat frequency response. It is obviously impossible to make perfect hardware which apparently would not need any form of error correction. Even making the hardware good enough not to require any error correction for most devices would be extremely expensive. The best balance is to make the hardware as good as practically and economically possible, balancing performance and cost. Error correction is then very useful to improve measurement accuracy. Generally, there are two basic types of test sets that are used with network analyzers. It is illustrated in Fig.1. Source Source Transfer Switch R R A A B Port 1 Port 2 Fwd DUT Transmitter/Receiver-based analyzers (T/R test set), however, it offers only one-path, two-port calibration. Some errors still exist after this procedure. 2 The low cost Vector Analyzer (VNA99) Network A low cost VNA(VNA99) has been developed by INTEC-design team, University of Gent, Belgium. It comprises a T/R test set with one directional coupler, two measurement channels and one reference channel for measuring the magnitude and phase responses of the forward S-parameters S11 and S21 of a linear two port. The analyzer ‘s built-in synthesized source produces a swept RF signal in the range of 50kHz to 200MHz with a resolution of 100Hz. It can deliver the output power from –25dBm to +6dBm with a flatness of less than 1dB. VNA99 works in 50-ohm environment. The system block diagram is illustrated in Figure 2. B Port 1 Port 2 Fwd DUT Rev Coupler “Transmission/Reflection Test Set” “S-Parameter Test Set” Fig. 1 The two basic types of test sets that are used with network analyzer A real life VNA measurement is the vector sum of the DUT response plus all error terms. The precise effect of each error term depends upon its magnitude and phase relationship to the actual test device response. Network analysis measurement errors can be separated into systematic, random, and drift errors. In a well conceived high frequency measurement, the systematic errors are the most significant source of measurement uncertainty. Since each of the systematic errors can be characterized, and the DUT is operated in its linear range, their effects can be effectively removed to obtain a corrected value for the DUT response. These errors are qualified as directivity, source match, load match, cross-talk, and frequency tracking errors. Directivity and cross-talk errors are related to signal leakage, while source and load match errors are related to signal reflections. The final class of errors is related to imperfect frequency response of the receivers. Meanwhile random and drift errors can not be precisely quantified, so they must be treated as producing a cumulative uncertainty in the measurement results. In general, calibration can be applied to remove the system errors before measurement. For the RF Transmitter Transmitter LO DUT DUT source load Receiver Receiver Data Data Processing Processing & & Display Display Fig.2 The system diagram of the VNA99 3 Practical design considerations to increase measurement accuracy of VNA99 Practically, the system characteristics is affected by magnitude at the device input, frequency response accuracy, directivity, crosstalk, source match and load match. A more detail discussions are presented in the following sections about how to take into account the practical design issues to increase the measurement accuracy. 3.1 Directivity of the Directional coupler An ideal directional coupler has low loss and high reverse isolation. However, due to the difficulty of making a truly broadband multi-decade coupler, bridges are often used. The bridge configuration of the VNA99 directional coupler is illustrated in Fig.3. 50 ohm Source 50 ohm Detector 50 ohm Test Port Fig. 3 The configuration of the directional bridge Bridges are very useful for measuring reflection because they can be made to work over a very wide range of frequencies. The main trade-off is that they exhibit substantial loss to the transmitted signal, resulting in less power delivered to the DUT for a given source power. A bridge ‘s equivalent directivity is the ratio of the bridge detector signal level measured at maximum balance (measuring the perfect Zo load) and at minimum balance (measuring a short or open) respectively. It is known that the reflection measurements are limited by directivity only. The following factors should be taken into account. 3.1.1 Coupled arm power loss Directional couplers separate signals based on the direction of signal propagation. If a directional coupler is used to sample power in the measurement configuration, the coupled arm power loss should be taken into account, and compensation may be required for the directional coupler response. 3.1.2 HF components It should be noted that control of the behavior of the passive bridge components at higher frequencies is essential to make the directional bridge coupler truly balanced. Parasitic components from the board material as well as the 50-ohm resistor HF behavior have impact on the coupler performance, as the frequency for the network analyzer is sweeping from 50KHz to 200MHz. The passive component behavior is practically more critical here in the bridge than in other places of the VNA99. At least 3dB ~ 5dB improvement of directivity could be obtained in this Vector Network Analyzer by using HF resistors. Some flexibility should be designed in, to compensate for both the cable effect and the unbalance from the bridge arms. Very careful HF layout is required. 3.1.3 Interaction of Directivity with the DUT (without error correction) Directivity error is the main reason why in some cases, a ripple pattern occurs in the measurement of the return loss. At the peak of the ripple, the directivity error signal is adding in phase with the reflection from the DUT. In some cases, the directivity error signal may cancel the DUT’s reflection, resulting in a sharp dip in the response. So it is essential to obtain good directivity over the whole frequency range. The solution to this issue is the combination of the design considerations above and in the following section. 3.2 High dynamic range in the transmission channel Dynamic range is very important for measurement accuracy. To get low measurement uncertainty, more dynamic range is needed than the DUT exhibits. For example, to get less than 0.1dB magnitude error and less than 1 degree phase error, the noise floor should be more than 35dB below the measured power levels. System dynamic range in the transmission channel is limited by the noise level relative to a “ through “. It is calculated as the difference in dB between the maximum receiver input level and the receiver noise floor. The system dynamic range is applied to transmission measurements only. The dynamic range is affected by several factors, such as the test port input power, the test port noise floor and the receiver crosstalk. An efficient way to remove the effect of random noise is to apply weighted successive measurement traces, called “averaging”. But its consequence is that the speed of measurement will decrease when more averaging is applied, as more measurement samples are needed for each frequency point to be measured. Another way to reduce the noise floor is to reduce system bandwidth, i.e., the bandwidth of the heterodyne receiver intermediate frequency (IF). In the VNA99, the RX IF bandwidth is fixed to 3 kHz, as a compromise between measurement speed and RX dynamic range, and cannot be changed by the user. In the transmission channel, the dynamic range of the mixer has been measured to approach 90dB, defining a 0.1dB compression point, and in the case of a 30MHz working frequency. The A/D converter used is a high linearity sigma-delta converter with a 18-bit serial output. The effective number of bits (ENOB) of such a A/D converter is 15bit, which means it can provide maximum 92.06dB dynamic range. This ADC can perform direct IF sampling as the RX is downconverting towards a low IF frequency in the upper audio range. The analog input drive circuitry of the ADC must amplify the IF output signal of the RX mixer to make it swing the full ADC input scale, in order to get the maximum dynamic performance. It is noted that the input impedance of the integrated RX mixer has a frequency-dependent characteristic, i.e., the input impedance is not constant in the whole 200MHz frequency range. The characteristic of the amplitude of the input impedance of the mixer is plotted in Fig.4. The unit of frequency is MHz. broadband RX matching network, which has to take peak level into account in order to avoid RX input overload, and the RX dynamic range, which is reduced by an amount roughly equal to the peak to peak RF signal variation. Moreover, a small change of the reference signal, which is splitted by the injected RF signal, can result in a considerable change of the transmission and reflection signals, adding to the overall measurement uncertainty. In the VNA99, the RF signal variation is limited to less than 1 dB, to guarantee a low frequency tracking error contribution. Actually, an ideal directional coupler can never be implemented. A variation of the directivity over the 200MHz frequency range will also cause frequency tracking errors. In the design of the directional coupler, a capacitor is used in order to compensate the impact from the coaxial cable. This compensation includes both the parasitic capacitance and the parallel capacitance in the equivalent circuit of a cable. 4 Measurement results Fig.4 The characteristics of load impedance of matching network In order to get the maximum dynamic range for the receiver, while realizing a proper receiver input match to 50 ohms, broadband RX input matching networks have to be designed. To not exceed the 0.1dB compression point of the mixer with the loopedthrough TX output signal, a 6dB attenuator should be added in front of the RX mixer connected to the transmission channel, and a similar attenuator lowering the reflection power level has to be added in the RX mixer connected to the reflection channel. Meanwhile we have to keep a high return loss at Port 1 and Port 2. Actually, the broadband matching networks for both the transmission channel and reflection channel have been designed to meet the demands above. The practical design considerations discussed above have been employed in the design procedure. The measurements of the Transmission Characteristics, Reflection Characteristics, Time Domain Reflectometry (TDR), as well as the directivity of the direct coupler have been done to evaluate the system performance. Figure 5, 6 show the transmission characteristics of MCL-PBP-21.4 band-pass filter from MiniCircuits and the reflection characteristics of a Tlinear network respectively. The final measurement errors of S21(dB) and S11 (volt) are calculated as 0.58dB and 0.001volt respectively. And the directivity of VNA99 illustrated in Fig.7 is at least 62dB. 3.3 Reducing the frequency tracking error When estimating the frequency tracking errors, the following factors should be taken into account. First aim is to reduce the variation of the injected RF signal, and the reference signal from Transmitter, with frequency. This will significantly increase the measurement accuracy. RF signal variation will cause variation of directivity, and consequently cause directivity error. It will also penalize the design of the Fig. 5 The transmission characteristics of MCL-PBP21.4 measured by VNA99 . Fig. 6 S11 behavior of the T-linear network measured by VNA99 Fig. 9 The TDR measurement of the cables with different characteristics impedance 5 Conclusion Fig.7 Directivity measurement of VNA99 The measurement configuration of TDR is illustrated in Fig.8. The measured coaxial cable consists of two sections, one is 5 meter 50 ohm characteristic impedance cable and the other is 20 meter 75 ohm characteristic impedance cable. This cable is terminated by a 50 ohm load. The measurement result illustrated in Fig.9 shows the location of the mismatch of the different characteristic impedance of the cable. 5 m 50 ohm cable 20 m 75 ohm cable Port 1 50 ohm load VNA VNA Port 2 Fig. 8 The TDR measurement configuration The main issues of determining the measurement accuracy of Vector Network Analyzer are discussed. The practical design considerations to obtain the accurate measurement of VNA99 are applied in the design procedure, particularly in the optimization of the whole system. To illustrate the performance obtained after optimization, several measurement results are shown of transmission and reflection characteristics, and of TDR (Time Domain Reflectometry) measurements. It can be seen that the measurement errors of S21(dB) and S11 (volt) have been reduced to 0.58dB and 0.001volt respectively. Meanwhile the directivity of VNA99 can reach at least 62dB. It indicates that in the limited frequency domain up to 200 MHz, the measurement accuracy of the low budget VNA99 compares very well with results obtained by expensive laboratory equipment. References: [1] D. Ballo , Network Analyzer Basis , Application note from Hewlett-Packard, 1997. [2] Vizmuller, Peter, RF Design Guide – System, Circuit and Equations, ARTECH HOUSE,INC, 1995. [3] Herbert L.Krauss, Charles W.Bostian, Solid State Radio Engineering, New York, 1983.