In-Process RF Test Measures Phase Error

The Need for Phase Error Tests
GSM phones communicate on channels spaced at 200 kHz, with information transmitted in digital format at a rate of 271 kbps per channel. The transmitted signal’s power spectrum must be limited to prevent significant crosstalk between channels. To limit the transmitted signal’s power spectrum, GSM handset circuitry limits the spectrum of the modulation signal.

A GSM handset employs Gaussian minimum shift keying (GMSK) as its modulation method. The GMSK process includes two steps to limit the spectrum of the modulation signal: It integrates the information signal, then filters the resulting waveform using a Gaussian response filter having a bandwidth equal to 0.3 times the data rate (called a “.3BT” filter).

The filtered waveform signal then modulates the RF carrier. The modulation level is scaled such that prior to filtering a +908 phase shift represents a logic high, and –908 represents a logical zero.

Measuring Phase Error
For a basic phase-error measurement, an in-process functional test system performs these steps:

• Applies quasi-random information to the handset under test (or UUT) to modulate the handset’s carrier.

• Receives the resulting transmitted signal from the handset, downconverts it to an intermediate frequency (IF), and then digitizes it.

• Demodulates the UUT’s downconverted transmitted signal (Fig. 1).

The test system’s demodulator multiplies the sampled IF signal by a sine wave having a frequency equal to the average IF frequency (which it can derive from the measured signal) and filters the result. (This method is called “coherent detection.” An alternative method is to compare the zero crossings—which can be found by interpolation—of the two signals.)

Note that GMSK modulation employs pure phase modulation; if signal amplitude affects the test system’s demodulation method, then the system must apply amplitude limiting to remove any amplitude variation.

• Derives the binary information (the data signal) from the demodulated waveform.

• Integrates (Fig. 2) and filters (Fig. 3) the data signal to produce a pure modulation waveform called the reference waveform.

Figure 2. Next, the system constructs a reference phase-modulation waveform (the lower trace shown here) from the digital data.

Figure 3. Filtering the reference phase-modulation waveform yields the lower trace here.

A test programmer can best develop the Gaussian filter used to obtain the reference waveform by shaping it in the frequency domain, then transforming it to the time domain with an inverse FFT. The test system can then employ convolution to provide filtering directly in the time domain.

• Compares the UUT’s original phase-modulation signal with the reference waveform to derive a difference waveform (Fig 4).

Figure 4. Finally, the system removes slope and offset effects, allowing direct comparison of demodulated and reference waveforms.

A difference between the UUT’s actual IF and the tester’s derived IF carrier results in a slope in the reference waveform relative to the original waveform. This slope, together with any DC offset caused by misalignment of the two modulation signals, must be removed before calculating the difference waveform. The test system can employ a linear-fit function to determine slope and offset.

• Calculates the RMS and peak levels of the difference waveform and compares them with pass/fail
values.

The European Telecommunications Standards Institute specifies 208 max phase error and a 58 RMS error for GSM handsets. T&MW

FOOTNOTE
1. Dewey, Mike, “In-process Functional Test Cuts Cost of RF Products,” Test & Measurement World, December 1998, p. 29.

Max Khazam is a staff scientist at GenRad.