Tunable Bandpass RF Filters for CMOS Wireless Transmitters

Wireless local area network (WLAN) proliferation has driven low-cost CMOS system-on-a-chip (SoC) solutions that place both the digital baseband processing and analog radio frequency (RF) front-end on a single chip. Spurious signals, such as mixing harmonics, synthesizer spurs, wideband circuit noise, and digital clocking harmonics, as well as their coupling paths, are difficult to predict in an SoC implementation. These spurs can appear in the transmitter output spectrum and violate requirements set by the communications protocol or the Federal Communications Commission (FCC). For some of these spurs, the protocol and FCC requirements can be satisfied by reducing the peak transmitted output power. Alternatively, RF filtering that is highly selective in frequency can reduce the power of all the spurs that lie outside the signal band, and would typically be provided by a discrete surface acoustic wave (SAW) filter. Frequency selectivity at gigahertz frequencies similar to that attainable by a SAW filter remains one of the few functions that has not proven amenable to integration. Semi-passive filters that use active circuitry to compensate for the losses in LC-tanks can achieve a highly selective frequency response. These Q-enhanced resonators, as well as their use in wide bandwidth filters, have been previously demonstrated in literature. However, synthesizing wide bandwidth filters using a classic design of coupled resonators that have Q-enhancement leads to severe tilt across the passband. Additional compensation techniques or complicated Q-enhancement schemes are required in order to reduce the in-band ripple for a coupled resonator filter. This dissertation introduces a WLAN transmitter in which wide bandwidth RF filtering is accomplished by using a cascade of independent high- Q resonators. A wide bandwidth is achievable by stagger tuning the center frequencies of independent high-Q resonators. The WLAN transmitter uses a sliding-IF dual up-conversion architecture, with a stagger-tuned filter following the mixers. A variable-gain amplifier (VGA) follows the mixer to provide the requisite gain control called for by most applications. A programmable divider is used in conjunction with the signal path up-conversion chain to create high-frequency tones to directly measure the frequency response of the resonators. Such circuitry is shown to be useful in creating an automatic tuning control loop. An experimental prototype transmitter suitable for 802.11 a applications and employing a stagger-tuned filter consisting of two Q-enhanced resonators has been integrated in a 0.18-mum CMOS technology. The transmitter provides spectral mask and error vector magnitude (EVM) compliant output power of -8.26 dBm for a 64 QAM OFDM symbol coded for 54 Mb/s data rate. The integrated Q-enhanced filter is centered at 5.125 GHz with 200-MHz bandwidth and 0.8-dB ripple. The transmitter signal path dissipates 190 mW from a 1.8-V power supply, of which the filter dissipates 41 mW.