Architectural and Specification Considerations

As in the Weaver architecture, a low-IF architecture inherits many of the attributes of a homodyne receiver, but it has lower sensitivity to DC offsets and higher ͳ

݂ noise. The

trade-off, however, reappears as the image rejection issue. If the goal for the receiver design is to avoid the use of expensive filters, then the burden of image rejection can only be solved by using suitable architecture. The Weaver architecture, which has the ability to resolve the difference between negative and positive frequencies, also endows it with the ability to resolve the difference between a signal and its image.

Although the direct-conversion receiver architecture has many drawbacks, it has still gained popularity in RFID system due to the improvement in IC technologies. Well controlled and suppressed IC technologies enable its possibility in discrete implementation. Since direct-conversion receiver architecture directly converts RF to baseband, the DCR employs only LPF for filtering out unwanted interference and no image rejection filter is required. Moreover, LPF and one stage of LO has also reduced the architecture complexity (BPF design in superheterodyne is much more complex compare to LPF) and power consumption of the system.

Likewise, in order to approach low power consumption and simple circuit structure, the active mixer is employed as passive ring mixer and post-mixer amplifier (PMA). Even though an active mixer has worse linearity when compared to a passive mixer, it performs better in isolation and has better conversion gain. In addition, the extra PMA stage will introduce extra noise and signal processing delay.

A down-converted band in direct-conversion receiver architecture also faces some serious problems that require further discussion. Since the signals are down-converted to baseband, any DC offset voltages can corrupt the signal. Large DC voltage can affect the bias voltage of the transistors and further degrade the performance of other stages. In RF front-end blocks, LO leakage often occurs when there is an imperfect isolation among the LO port, IF port and the RF port. The feed-through from capacitive and substrate coupling will heavily degrade the performance between LNA and mixer. This offset problem can be more severe and unpredictable if the LO is provided externally, which will introduce additional bond-wire coupling. The LO leakage signal will feed- through the mixer and subsequently mixed with the desired LO signal, causing a “self-

mixing” issue. The dc term generated by self-mixing can saturate the front-end stages in the receiver or even leak to the antenna since LNA and mixer have limited reverse isolation capability. This leakage problem could further contaminate itself or nearby

receivers to become “self-jammer” [38]. A similar situation could also occur the other way around from mixer input to LO port, known as interferer leakage. Therefore, maintaining high isolation between ports and high linearity for LNA, mixer and VCO are the major concerns in further design.

Down-converted signal after the mixer is usually feeble and very sensitive to any noise. Especially, the signal of interest at zero-IF frequency is susceptible to the flicker noise (also known asͳȀ݂noise). A closer investigation regarding noise behaviour around baseband frequency should be carried out. A relatively high gain in the RF range is desirable in the later design considerations. Therefore, an active mixer is preferred to passive mixer in this design to achieve higher gain.

In baseband, the even-order distortion in homodyne structure becomes critical enough to be taken into consideration. The second-order intercept point (IP2) should remain at a high value to overcome the distortion problem. Applying capacitive degeneration and ac coupling techniques in the mixer design can improve even-order linearity. Differential mixer topologies are less susceptible to the even-order distortion and will be employed in later designs. Unfortunately, this problem may not be alleviated in the LNA stage since single-ended LNA is required to work with antenna and duplex filter in a typical single-ended system. Converting the single output to a differential signal may require extra blocks such as baluns and lead to additional power consumption. Unfortunately, a transformer is avoided at high frequency since it generates high losses as a consequence of a higher noise figure [136]. Therefore, optimizing the trade-offs between power consumption, system simplification, performance in system design and evaluation for different structures should be carried out in detail.

In this work, each block operates in UHF range at 866 MHz frequency. The LNA and mixers design should provide a high voltage or conversion gain (>12dB) and low noise figure (<2.5dB for LNA and <6dB for mixer) at this operating frequency. Since the direct-conversion receiver architecture is applied here, the VCO is required to generate the same oscillating frequency (866MHz) and low phase noise (<120 @ 3MHz). The supply voltage and power dissipation are also critical for the design consideration and aim to remain as low as possible.

In the following sections, each block of the RFID front-end is investigated in detail with respect to both design methodology and circuit topology to overcome design problems. To evaluate overall performance, more than one possible solution is described. At the same time, derivation of accurate models for each component was done in order to optimise overall performance. Figure 3-3 demonstrates the direct-conversion receiver

architecture with image-reject mixers that are employed in this thesis for the UHF RFID receiver front-end. Differential LNA LPF Rx,OUT-I Rx,OUT-Q Rx Front-end Quadrature Down- Converting Mixer t LO Z sin t LO Z cos Quadrature Down- Converting Mixer 0/90 Quadrature VCO Rx,OUT-I LPF Rx,OUT-Q

Figure 3-3: Proposed Direct-conversion receiver architecture in this thesis for RFID receiver front-end.