[For a copy of this article in PDF format, which displays figures and equations, click

here

. Requires Adobe Acrobat Reader,

free download.

]

In the past, individual RF or IF components tended to be based on differential topology. Widespread use is common within the wireless industry because this topology offers inherent noise immunity benefits. Although these balanced components provide good noise immunity, their structure gives RF designers headaches when characterizing and matching, because most of the time a vector network analyzer (VNA) was used for measuring the single-ended application.

Because of the difficulty in obtaining accurate results, a method to accurately measure and characterize such circuits is desirable.

Because IF surface acoustic wave (SAW) filters require very accurate input/output (I/O) characterization for determining the matching circuit, an incorrect matching structure or component value in the receiver chain will cause the degradation of sensitivity. This is due to the mismatch loss and deformation of the SAW filter frequency response.

The method described here provides a compact way to characterize the balanced components using a single-ended S-matrix in a CAD environment. These single-ended, S-matrix measurements were used for calculating the terminal impedances for the balanced components cascaded in the receiver chain.

It will also present an illustrated example based on this method. It exemplifies the ease in receiver design, especially for SAW filter matching circuits.

The step-by-step procedure for designing a model for the simulation is shown in figure 1. All four ports should be measured using traditional vector network analyzer measurements (be sure all probe lengths — from the end of component to the input of VNA — are the same to minimize measurement error). This measurement is important because the single-ended S-matrixes will be transformed into a balanced structure for the computer simulation.

As a result of transformation, six single-ended S-parameters will combine to represent two balanced S-parameters. To characterize the DUT as a balanced structure, a total of six single-ended measurements should be performed, according to the following steps.

First, connect VNA to the DUT's ports 1/2/3 and 4/4/4. Save the generated CITI file as DUT14/24/34.

Next, connect the VNA to the DUT's ports 2/2/3 and 4/3/4. Save this CITI file as DUT24/23/34.

The six single-ended S matrices are imported into one of RF CAD environments for the balanced simulation. Figure 2 shows the overall simulation setup for the balanced simulation, using the six data access components, terminated with 100Ω/1000Ω impedances, respectively. The impedances apply to the filter as a 100Ω of differential input impedance and 1000Ω of output impedance.

It should be noted that terminal impedances do not necessarily need to be resistive when it comes to the receiver system. The input or output impedances will depend on what type of components are cascaded for the receiver lineup. Therefore, reactive terminal impedance can be specified after calculating the balanced terminal impedance through Christoffel's formula or simulation.

Christoffel derived the formula by splitting the balanced single port into an equivalent two-port network, with single-ended ports for the purpose of using VNA. The equation is as follows:

It turns out, however, that the Christoffel calculation step is unnecessary in determining the balanced reflection coefficients. These will be provided by the balanced simulation, using S-parameters, automatically.

The most important thing to consider is the method of S-parameter extraction. This is wholly based on the measurement. If the wrong parameter is used, the matching job will not be helpful in determining the proper matching performance.

The following real-world example illustrates the technique used to determine the proper matching performance. The code-division multiple access (CDMA) part of the receiver circuit is used to demonstrated how to practically apply this method (see figure 3).

In the example, the SAW IF filter becomes the component of emphasis because of its inherently sensitive impedance. The procedure in the example will attempt to simplify the normally intensive matching procedure.

The input to the SAW filter is, electrically, the same point as the output of the mixer. The terminal impedance is 50Ω, single-ended. The output of SAW filter is connected to the input of the demodulator chipset using a balanced structure with approximately a 1000Ω impedance.

Therefore, three S-parameters of the SAW filter were extracted from the real prototype board after calibrating out the board parasitics. Next the two-port, single-ended S-parameter of the demodulator chipset was measured and converted to a differential impedance terminated as Term2 in the schematic (see table 1).

From this S-parameter result, the balanced reflection coefficient of the demodulator chipset is calculated, either from the formula or the balanced simulation. The value is 452.15-j189.13Ω, and is the termination impedance at the frequency of interest.

While this works in the example, there are a number of assumptions that must be considered when moving to the real world. This particular case is an example to illustrate how to setup the schematic on the simulation bench.

Furthermore, both the SAW filter's insertion loss and impedance can be simulated. The criteria is how close one wants to get to the maximum power transfer point. The closer the impedance gets, the closer one gets to the maximum power transfer point.

The simulation-oriented methodology can be easily expanded to include active circuit matching components such as the LNA, the mixer and other components.

Hence the receiver RF line-up, up to the input of demodulation chipset, can be imported into the CAD environment once the parameter needed for simulation have been extracted.

The simulated result of the inter-stage filter match is show in figure 4. It shows an insertion loss with approximately 1 dB difference from the vendor specification. The inductor and capacitor values in figure 4 were used for the receiver sensitivity test of the CDMA mobile phone.

When built, sensitivity of the prototype phone matched the simulated results on the first try. This supports the fact that such a method saves both development time and costs by reduces component count and redesigns.

The general procedure for accurately modeling and simulating components with a balanced structure was formulated from a simulation perspective.

The procedure accounts for parasitic effects and traces on an actual board. These effects are real and significant factors that cause discrepancies between real-world applications and the simulations. The real-world example of the differential SAW filter has demonstrated that the concept of differential simulation using single-ended S-parameters is useful and accurate.

SangJin Lee is a staff engineer with eAnywhere Wireless Inc., 2F Shinwon Bldg 823-14, Yeoksamdong, Kangnam-Gu, Seoul, Korea. He can be reached at Sam.lee@eawinc.com