Many different sources can degrade 802.11a/g transmitter performance — non-linearities, I/Q imbalances, additive noise, transient effects — but pulsed phase noise poses particular difficulties in recognizing and identifying its presence. However, using advanced WLAN test instruments, the designer can assess the impact of pulsed phase noise.

The multicarrier OFDM signals employed by 802.11a/g are of the most interest — such signals support the highest data rates under the 802.11a/b/g standards and are the most demanding in terms of modulation accuracy. The reference crystal and VCO of an 802.11a/g transmitter are very sensitive to interference, and coupling with the signals generated by dc-dc converters in the power supply or other baseband circuitry can result in pulsed phase noise. Such coupling can be a consequence of poor signal routing on the printed circuit board, inadequate shielding or insufficient grounding.

Figure 1 illustrates the measured phase variations in an OFDM signal corrupted by coupling with a 20 kHz source. As can be seen, the phase error varies regularly in a pulsed manner. Figure 2 shows the phase noise power spectral density of the same signal, where peaks at the 20 kHz fundamental frequency and its harmonics are evident.

Although direct measurement of the phase noise in the transmit signal as shown in Figures 1 and 2 reveals degradation, typical analysis of transmitter performance by using the error vector magnitude (EVM) as a figure of merit can mislead the designer into thinking that the transmit signal quality objectives are fully met.

In the presence of pulsed phase noise, the difficulty in using EVM as a measure of transmitter performance stems from the way that the phase is tracked in the measurement process. For example, the IEEE 802.11a/g standards stipulate that EVM should be calculated after correcting for the carrier phase error, which is measured and corrected for symbol-by-symbol from the OFDM pilot carriers. Using this symbol-by-symbol estimation and correction method with the signal of Figures 1 and 2 results in a calculated EVM of -31 dB and the signal constellation of Figure 3. Because the phase variations are relatively slow with respect to the symbol rate, they are tracked and compensated symbol-by-symbol with no apparent significant impact.

Commercial 802.11a/g receivers, however, do not necessarily estimate the phase on a symbol-by-symbol basis. To compensate for poor channels or signal dynamics in which one or more of the pilot tones may be adversely affected, some WLAN receiver designs may estimate the phase by applying a moving average over several OFDM symbols. In the presence of pulsed phase noise, however, this phase estimation method does not track the phase variations fast enough, and, as a result, the pulsed phase noise will degrade the link performance. The impact will vary depending on the characteristics of the pulsed phase noise and the specific receiver algorithms.

To reveal such effects, a test instrument must be capable of using different phase estimation algorithms. For example, the IQview^™ 802.11a/b/g WLAN test solution lets the designer assess EVM using various different receiver algorithms for phase tracking, channel estimation, and frequency synchronization, thereby providing the means to expose impacts that might be otherwise hidden. If, for the signal of Figures 1 and 2, the phase is estimated by applying a moving average over the last 10 OFDM signals, then the calculated EVM is -23 dB, and the signal constellation is as shown in Figure 4.

The figures are showing the 16 QAM mode for clarity. If the same transmitter would be configured in the 54 Mbps, 64 QAM mode, the phase noise would cause a similar EVM and the transmitter would fail the 54 Mbps, 64 QAM mode EVM requirement of -25 dB.

By testing with a moving average phase estimation method as opposed to a symbol-by-symbol approach, both the EVM and the signal constellation reveal the effects of the pulsed phase noise. If such testing had not been conducted, this 802.11a/g transmitter design might have been assumed to be within specifications but would evidence poor performance with different receivers, resulting in product returns, reduced sales and costly redesign.

ABOUT THE AUTHORS

Carsten Andersen and Dick Walvis are members of the technical staff at LitePoint Corporation, San Jose, Calif. Andersen has more than 18 years of experience in designing systems for wireless communication. Walvis has more than 20 years of experience in designing DSP algorithms for wireless communications.