Power measurements on RF transistors and RF integrated circuits (RF ICs) have grown in complexity as a function of the modulation formats used with these devices. One of the most important measures of high-power device performance is saturated power. Because the parameter is difficult to evaluate with CW techniques, it is often tested under pulsed conditions. The approach presented here eliminates some of the key drawbacks of the typical method employed to make these measurements. It does not require an external personal computer, but takes advantage of some lesser-known capabilities of the SMIQ signal generator from Rohde & Schwarz as well as the company's FSP signal analyzer functioning as a high-dynamic-range peak power meter.

By using trace math and marker features, direct readings can be obtained of gain and power at any compression level up to the saturated power level of a device. Measurements on a high-gain LDMOS power RF IC from Freescale Semiconductor designed for the UMTS band (the model MW4IC2230MB) show the advantages of this method.

Saturated power is an important device or amplifier characteristic because of digital predistortion systems used to linearize multicarrier cellular base-station power amplifiers. Saturated power is usually considered the greatest amount of output power possible from a predistorted power amplifier. Even though LDMOS devices are more rugged than bipolar transistors, it is still difficult to measure high CW power levels. In fact, the self-heating of the device makes it almost impossible to produce accurate and repeatable measurements. As a result, saturated power measurements are usually performed with pulsed signals. A signal generator with a pulse input and a peak power meter with two sensors are typically employed. The input power at the device can then be increased and a plot of output power versus input power can be created with the help of a PC.

However, this method suffers from limitations in accuracy. Two-way peak RF power meters require that both sensors be operated in a given dynamic range for better accuracy. This condition is easy to fulfill if the test bench is appropriately designed. But if the device under test (DUT) has high gain, such as a multistage RF IC, another source of inaccuracy appears: The sensors can't operate in the same dynamic range both during calibration (when the device under test is replaced by a thru standard) and during the measurement. Thus, there is an interdependence between the measurement result and the power level at which the bench was calibrated.

Test Bench
The test bench (Fig. 1) uses a pulse generator connected to the "Pulse" input of the SMIQ RF signal generator. To use the SMIQ in power sweep mode, the power sweep must be synchronized with the time sweep in the signal analyzer. Fortunately, this family of signal generators has features that allow them to be used as a scalar network analyzer (SNA) when associated with diode detectors and an analog display like an oscilloscope in XY mode. On the rear panel of the SMIQ, there are several BNC connectors with a power sweep ramp to drive the X axis of the oscilloscope, and markers to calibrate the X axis of the display. In this case, the "Marker" output is used as a trigger signal for the signal analyzer.

The "Marker" output of the SMIQ is connected with a BNC cable to the "External Trigger" input of the FSP. "Marker 1" is set to the "Sweep Start" value, and the RF output of the SMIQ is connected to a variable attenuator. This allows the power level at the input of the DUT to be adjusted without having to change the "Start" and "Stop" values of the sweep in the signal generator.

This method is preferred because if the "Start" and "Stop" values of the sweep are modified but the position of the "Marker 1" is not, the synchronization of the spectrum analyzer can be erratic or even lost if the marker is outside the power sweep range. A high-power amplifier is used to drive the DUT to make sure that the driver amplifier will not saturate before the DUT. Input and output couplers allow sampling part of the signal to be sent to the spectrum analyzer. A calibrated attenuator is used as a load in order to have an accurate power reference that can be measured with a standard power meter after entering the load attenuation as an offset.

Before making a measurement, the input and output paths to the spectrum analyzer must be calibrated. As usual, the DUT is replaced by a thru standard and the signal generator is operated in CW mode. The power meter reads the power level going through the thru standard, and the spectrum analyzer in "Zero Span" mode reads the absolute power at the coupled path of either the input or output coupler. This makes it possible to determine the attenuation of the input and output paths to the spectrum analyzer. These values are called "IN_OFFSET" and "OUT_OFFSET" for future reference.

All the parameters are chosen so that the current consumption by the DUT does not deviate from the quiescent current, ensuring stable thermal behavior. The signal generator is switched to pulse mode by selecting the pulse option in the analog modulation menu. In the sweep menu, the power sweep mode is selected. The start level is set to –20 dBm and the stop level is set to 0 dBm. A step size of 0.2 dB allows 101 measurement points. The dwell time must be carefully selected. If too small a value is selected, some transients occurring during the power sweep can cause the current drained by the DUT to deviate from the quiescent current. A dwell time of 200 ms keeps this effect negligible, maintaining a reasonably short sweep time of 20 s. In the same menu, the marker 1 is set to the start value of the sweep, which is –20 dBm, and is activated by selecting the "ON" state. Figure 2 shows the detailed configuration sequence.

As already mentioned, the spectrum analyzer is used in the "Zero Span" mode. Both the resolution bandwidth and the video bandwidth are set to 10 MHz because the spectrum analyzer is used to measure peak power. For the same reason, the detector must be configured in the "Max Peak" mode. A sweep time of 25 s is chosen in order to have a whole sweep on screen. The external trigger option is selected. It is judicious also to use the "Trigger Offset" feature in order to center the trace on the screen. A value of –2 s is appropriate. Figure 3 shows the detailed configuration sequence.

The pulse length and duty cycle must be chosen not to perturb the thermal state of the device under test, but they must also be in line with the response time of the spectrum analyzer. A pulse duration of 1 µs and a repetition period of 1 ms produce good results.

The results presented in this section are based on measurement of an LDMOS power RF IC from Freescale Semiconductor designed for the UMTS band: the MW4IC2230MB. It has small-signal gain of about 30 dB and saturated power greater than +47 dBm. Because of its high gain, it is a perfect example of the advantages of this method.

The input variable attenuator is first set to its maximum value. The DUT is connected and the spectrum analyzer is connected at the coupled path of the output coupler. While being in "Clear/Write" mode, the output power ramp is plotted on the analyzer's screen as an asymmetrical saw tooth. The variable input attenuator is then released and the effect of the DUT saturation should begin to appear (the top of the ramp starts to bend). Attenuation is further reduced until the top of the saw tooth is truncated, ensuring that saturation is reached.

The sequence of functions to use on the spectrum analyzer is shown in Fig. 4. The spectrum analyzer is now connected at the input and the input power is acquired in the second trace ("Trace 2") and kept in "View" mode. The black trace on Fig. 5 is obtained. After reconnecting the analyzer at the output, the output power is acquired in the third trace ("Trace 3") and also kept in "View" mode. By setting the reference level offset to the value of "OUT_OFFSET," which was determined to be 43.7 dB during the calibration phase, "Trace 3" now provides a direct reading of the absolute output power in dBm, and the green trace in Fig. 5 is obtained. By positioning a marker in the flat area at the end of the saw tooth, the saturated power is read as shown on the green trace in Fig. 5. The MW4IC2230MB shows saturated power of +47.23 dBm.

Sequence Of Functions
The sequence of functions to use on the spectrum analyzer is shown in Fig. 6. Trace 1 is configured to compute "Trace 1" minus "Trace 2," the latter containing the input power. A new acquisition of the output power will thus provide a plot of the gain in dB. The "Trace Position" feature is used as an offset to obtain a direct reading of the absolute gain in dB. The trace offset is set to "OUT_OFFSET" minus "IN_OFFSET." In this case, "IN_OFFSET" was determined to be 30.7 dB during the calibration phase, resulting in a value of 13 dB as offset. This value can't be entered directly as an offset in dBm since the "Trace Position" menu only accepts inputs as a percentage of the Y scale. The rotating knob is used to obtain the right percentage corresponding to an offset of 13 dB. The blue curve on Fig. 7 is obtained.

Both the gain (blue curve) and the output power (red curve) are plotted simultaneously and both plots are "calibrated," i.e., the marker readings provide absolute values. Marker 1 is set at the small-signal gain on the blue curve and a value of 29.8 dB is measured. This marker is used as the reference for the "Delta Marker." Still on the gain curve, a marker is used in "Delta Marker" mode to determine the 1-dB compression level (marker 2) and another one to determine the 3-dB compression level (marker 3). After selecting the power trace, a fourth marker (marker 4 on the green trace) is set at the same abscissa as either marker 2 or marker 3 in order to have a direct reading of the output power at either 1 or 3 dB compression level. In this example, a 1-dB compression level of +45.74 dBm and a 3-dB compression level of +46.69 dBm are obtained. This method of performing power measurements under pulse condition allows characterization of high-power RF transistors and RF ICs to be performed quickly and easily, eliminating some of the limitations of previous methods.