For the PDF version of this article, click here.
Testing is one of the critical operations necessary to ensure that GSM/EDGE mobile phones meet their required specifications. Consequently, reducing this time is important while also maintaining high accuracy and repeatability. There are several tools available in modern RF signal generators, spectrum analyzers, and power meters that can speed the process. Features such as fast restore mode and gated list mode provide an advantage over traditional measuring methods in terms of speed and repeatability.
Typical measurement setups
In a typical test setup for mobile phone amplifiers (Figure 1), a signal generator with digital modulation capability supplies the input signal for the device under test via a low-pass filter and a directional coupler. The low-pass filter suppresses the harmonics of the signal generator. Two electronically switchable filters are needed for multiband power amplifiers that operate from 800 MHz to 1.9 GHz. Some of the amplifier input signal is fed to a power sensor via the coupling branch of the directional coupler.
An attenuator connects the output of the power amplifier to the RF input of the spectrum or signal analyzer. The trigger output of the signal generator supplies a trigger signal (edge frame trigger) that synchronizes all test equipment.
To attain minimum measurement uncertainty, the power sensor is used for the path calibration of the test setup, as well as for the absolute calibration of the spectrum analyzer. The power sensor rather than the spectrum analyzer is connected in a calibration cycle, and its display is subsequently corrected to match that of the power sensor. This is best done at nominal output power of the power amplifier by entering an appropriate reference level offset on the analyzer.
The following RF measurements are typically made in production testing:
gain at nominal output power;
adjacent-channel power or of the modulation spectrum in the case of GSM/EDGE power amplifiers;
second harmonics and third harmonics in the case of a CW signal (bursted CW with GSM/EDGE);
Gain slope (dependence of the output power on the dc voltage at the bias connection of the amplifier); and
modulation error (error vector modulation, EVM).
All measurements are performed in burst mode, with only two of the eight time slots active. Measurements of dc supply voltages and currents must also be performed in order to obtain the important efficiency calculation.
Fast restore mode
When EDGE power amplifiers are tested, the signal generator that supplies the input signal to the amplifier is switched back and forth between different RF frequencies and between the modulated and unmodulated output signals. Modern signal generators usually switch the frequency fast enough. However, depending on the standard, this can take several hundred milliseconds.
In order to avoid this, the required frequencies and modulated and unmodulated output signals can be stored in advance. They can then be recalled during the actual measurement sequence, which can considerably speed up the measurement. For example, the fast restore mode in the Rohde & Schwarz SMIQ vector signal generator reduces switching time between different settings to only a few milliseconds.
For fast level measurement in the time domain, spectrum analyzers provide a list mode, which is a predefined list of measurement points that can be loaded and processed very quickly. The list contains application-specific settings for frequency, reference level, values of electronic and mechanical attenuators, filter types, filter bandwidths, and measurement times. List mode is well suited for all power, channel power, adjacent channel power, and harmonics measurements.
For example, in the Rohde & Schwarz FSP spectrum analyzer, the commands define the constant settings for the list during multiple power measurements. Parameters for peak, RMS and average measurements define which measurements are to be performed at the same time at the frequency point. Correspondingly, one, two or three results per frequency point can be returned.
Another simple command sets up the application-specific settings for each list point (frequency, reference level, attenuation, electronic attenuation, filter type, resolution and video bandwidth, measurement time and trigger level) and queries the results. Up to 200 list points are possible. The result is an output list whose length is dependent on the number of points and the constant settings (one, two or three results per point).
To measure the modulation spectrum at ±200 kHz and ±400 kHz of slot 1 of an EDGE signal with active slots 0 and 1, trigger offset is set to 0.995 ms and sweep time to 0.2 ms so that the analyzer sweeps during the second slot of an EDGE frame from 50% to 90% of the slot length, excluding the mid-amble. Both slot 0 and slot 1 are on in this example (Figure 2).
A typical result after the query results would be RMS powers at center frequency, ±200 kHz, and ±400 kHz measured with 30 kHz bandwidth of 1.55…,-35.24…,-35.7…,-53.656…, and -54.92.
Gated list mode
If the level or channel power of EDGE signals is measured over only one timeslot, the repeatability of results is often insufficient. To compensate for the statistical level fluctuations that occur as a result of modulation, several timeslots must be averaged. The standard method is to repeat the measurements as often as necessary in the analyzer or in the controller and to average the results.
The gated list mode offers a quicker means of performing such measurements. This mode is a combination of the gated mode and the list mode. The analyzer then measures a user-defined number of timeslots. This overhead is produced as a result of the GPIB transmission of the commands and their processing.
Using a suitable gating setting, the analyzer measures only during the desired time. For example, during 50% to 90% of the desired time slot, as is required with GSM/EDGE for measuring the spectrum resulting from modulation, the measurement continues until the set measurement time has been reached. For this reason, the measurement time is set to integer multiples of the gate length.
To measure the modulation spectrum at ±200 kHz and ±400 kHz, the gate delay (trigger offset =0.995 ms) and gate length (0.2 ms) are set so that the analyzer sweeps during the second slot of an EDGE frame. The sweep time is set to eight times the gate length (1.6 ms) to average the result over eight consecutive frames through use of the RMS detector. At the sweep end, the analyzer supplies the RMS value (or if desired, the average value) of the trace. The repeatability ofresults is now much better. Standard deviation is about 0.3 dB vs. about 1.3 dB without averaging. The measurement time typically increases from 40 ms to 200 ms. For even better repeatability, only the sweep time must be increased to a higher factor, but the measurement time will increase too (Figures 3 and 4).
The programming sequence for the above example would, after query results, produce RMS powers at center frequency, ±200 kHz and ±400 kHz of 1.535…,-35.234…,-35.7…,-53.656…, and -54.929.
Trace readout for gain slope measurements
GSM/EDGE mobile power amplifiers usually have a power control connection pin to control the output power over a dynamic range greater than 30 dB. The dependency of this output power on the control voltage (called gain slope) is an important measurement for controlling the mobile phone's exact output power when the amplifier will be in service. The gain slope measurement is performed by applying a ramp function at the amplifier's control input and simultaneously measuring its output power.
Use of the trace readout function of the spectrum analyzer allows the gain slope to be measured in a short time. Typically, 8001 sweep points are transferred in less than 80 ms. The gating function is used again to obtain several ramps for improved measurement repeatability without any time overhead. In a programming example for the FSP analyzer, the gating time is set to 1 ms, the sweep time to 8 ms, and eight ramps are measured with one shot. The analyzer outputs 32,004 bytes (four bytes for every sweep point) when using the binary format. The whole sequence runs in less than 300 ms (Figure 5).
Repeatability, accuracy, and speed are critical elements in testing any RF component or subsystem, and are especially important in the high-volume production world of phones for GSM/EDGE (and other standards as well). But effectively employing features found on modern RF test equipment, all three requirements can be satisfied without trade-offs.
ABOUT THE AUTHOR
Roland Minihold is a senior applications engineer at the Test and Measurement Division of Rohde & Schwarz based in Munich, Germany. He joined the company in 1976.