Industry-favored solutions for GSM-EDGE have converged to primarily one of two choices for the receive architecture: a) direct-conversion or b) low-IF. Aside from complexity and low-cost features that these architectures provide, several technical issues have warranted these choices.

GSM-EDGE requires IP2 performance on the order of 50 dBm or more when referred to the antenna input. This requirement amplifies the already challenging issues pertaining to dc offsets in the receiver. Direct-conversion receivers struggle with this problem more than low-IF receivers since the dc component falls directly within the receive bandwidth. The dc offset is also time varying because it is driven by dynamic adjacent-channel interferers. It also is affected by local oscillator (LO) leakage, low-noise amplifier (LNA) gain, and temperature.

CMOS designs must also contend with fairly severe 1/f noise in the sensitive IQ gain stages that immediately follow the down-conversion mixer. Detailed 1/f noise parameters depend signifi-cantly on oxide thickness and channel length. RF CMOS technologies in the 130 nm to 180 nm realm generally exhibit 1/f corner frequencies on the order of several hundred kHz, thereby making the low-IF architecture attractive for this reason. Issues of dc offset are not eliminated entirely with the low-IF architecture, but the severity is reduced.

Most WCDMA receivers have adopted the zero-IF architecture. Owing to WCDMA's much wider modulation bandwidth, dc offset issues are more easily addressed than for GSM-EDGE. The bandwidth argument also reduces the 1/f noise issue because (i) its impact on the overall receive signal-to-noise ratio (SNR) is considerably less, and (ii) this noise is spread across multiple chips of the WCDMA waveform where it can be effectively tracked out by the baseband signal processing if desired.

One of the major problems facing WCDMA receiver design pertains to transmitter leakage that falls through the duplexer filtering into the LNA input. This leakage adversely impacts attaining receiver IP2 and IP3 requirements, and normally requires band-specific SAW filters to be used between the LNA outputs and the mixer input. In WCDMA low-band, the transmit signal is offset from the receive signal by a scant 45 MHz whereas the offset is increased to 190 MHz for the IMT band near 2 GHz. These offsets combined with the required filter attenuation that is needed make the on-chip filtering option quite challenging. Since the filter follows immediately after the LNA, its insertion loss must be reasonable or else additional constraints are imposed on the LNA gain. As shown in Figure 1, the ratio of inductor-Q to filter-Q must be at least a factor of four in order to have a reasonably small insertion loss.

Previous 1 2 3 4 5 Next