Testing multicarrier power amplifiers (MCPAs) for wideband-code-division-multiple-access (WCDMA)/third-generation partnership project (3GPP) wireless systems places great demands on the test-signal source. To obtain the optimum test setup for MCPAs, specific spectral and statistical characteristics of WCDMA multicarrier signals must be taken into account. Fortunately, modern vector signal generators such as the SMIQ03HD from Rohde & Schwarz (Munich, Germany) can produce signals with adequate dynamic range for testing these amplifiers.

The complementary-cumulative-distribution function (CCDF) defines the statistics of a test signal for WCDMA/3GPP, specifying the signal's probability of exceeding a specific power threshold. To minimize costs, the dynamic range of the signal should be limited. This can be achieved in different ways. Tests of the US CDMA System IS-95 (cdmaOne) have shown that the CCDF of a CDMA carrier can be influenced by selecting specific Walsh code combinations.^1,2 The approach can also be applied analogously to 3GPP although because of its different spreading factors (in contrast to IS-95), the approach is somewhat more complex for 3GPP. Adjacent orthogonal codes in the code domain usually yield high crest factors (Fig. 1). Combinations with the code channels spread evenly across the code domain are better. Figure 1 shows the CCDF of a 3GPP base-station signal with the five obligatory control channels and eight traffic channels. The crest factor can be decreased by about 3 dB through code selection.

Moreover, 3GPP allows a timing offset of carrier traffic channels. Since the pilot data of the DPCH is not channel-coded, all DPCH signals have the same pilot symbols. This inevitably causes constructive interference and high power levels (if the pilot data of all channels is transmitted simultaneously). Timing offset circumvents this effect, reducing the crest factor by another 2 dB.³

A similar method is used with 3GPP multicarrier signals to avoid high peak envelope power values. Although the 3GPP standard defines different scrambling codes for different carriers, this is not sufficient in practice where, for this reason, the timing of the overall signals is offset on the individual carriers. The 3GPP standard advises a delay of 1/5 slot (133 µs) for signals on adjacent carriers,⁴ thus helping to effectively minimize the crest factor of a multicarrier signal (Fig. 2).

Clipping is another method of minimizing the crest factor. The sum signal of a base station is mathematically reduced by means of a saturation function directly before pulses are shaped in the baseband. The advantage of this method is that it works for all signal configurations, independent of code combinations or timing offset. However, correct transmission of constellations with high instantaneous power is no longer ensured, and the bit-error rate (BER) increases. Clipping therefore requires improved forward-error correction. For this reason, producers of base stations combine all of these methods to maximize dynamic range.

Test signals for 3GPP PAs must emulate these possible variations. Moreover, nearly all parameters of a 3GPP test signal should be accurately determined to achieve comparable results. For this reason, 3GPP has defined test models. The 3GPP TS 25.141 specification⁴ also defines unambiguous configurations for multicarrier signals (Fig. 3).

Every amplifier affects a signal in two ways: it generates additional noise, and its transmission function is linear only over a limited domain. Nonlinear components cause intermodulation. For 3GPP signals, this results in unwanted spectral components in the adjacent channels (Fig. 3). A 3GPP carrier has a width of 3.84 MHz and channel spacing of 5 MHz. The third-order intermodulation products (IM3) are in the 1.92-to-5.76-MHz range (relative to the carrier center frequency). The IM3 and wideband noise therefore contribute to the adjacent-channel leakage power. These power components may interfere with the transmission in the adjacent channel and must be minimized by achieving maximum amplifier dynamic range. Good results are usually obtained if IM3 and wideband noise make the same contributions. Wideband noise and fifth-order intermodulation (IM5) occur in the alternate channel. Since IM5 is one order less than IM3, the IM5 contribution is negligible compared to the wideband noise. The measurand is the adjacent-channel leakage ratio (ACLR), i.e. the ratio of the power in the useful channel to the power in the adjacent channel.

The situation is slightly different for multicarrier signals. A signal with four adjacent 3GPP carriers has a width of 18.84 MHz (Fig. 3). The IM3 now occurs in the 9.42-to-28.26-MHz range, both in adjacent and in alternate channels. In this case, the amplifier has to be driven to a lower level to achieve optimum ACLR.

Measuring instruments also generate intermodulation and noise and may contribute to the measured ACLR. Figure 4 shows a quantitative recording for a spectrum analyzer. If the inherent noise of the analyzer is 5 dB less than the measured value (consisting of the input signal and inherent noise), it is still necessary to deduct just under 2 dB from the measured value to obtain the correct value of the input signal. To ensure that the measuring instruments do not significantly influence the overall result of the ACLR, they must exhibit ACLR values that are at least 10 dB better than those of the device under test (DUT). The measurement uncertainty of the ACLR value of the DUT is significantly higher if the contributions of the measuring instruments are of the same order as those of the DUT, or even if they are dominant (Fig. 4).

The 3GPP base-station standards specify a minimum value of 45-dB ACLR in the adjacent channel. Most producers aim for an ACLR of 50 dB for the entire base station. As a result, ACLR values of minimum 60 dB are obtained for the associated PAs. For these reasons, a signal generator should exhibit an ACLR of 70 dB or better in the adjacent channel, which is a great challenge when designing signal generators (Fig. 5).