4.2 CDMA RF and IF Designing Principles

4.2.1  CDMA RF and IF Architecture

 For the RF part, it is a traditional analog structure where valid signals are translated into IF signals. The downlink channel of RF part mainly consists of automatic gain control (RF AGC), receive filter (Rx filter) and down-converter. The uplink channel of the RF part mainly consists of automatic gain control (RF AGC), secondary up-converter, wideband linear power amplifier and RF transmit filter. The IF part consists of the de-aliasing filter, the down-converter and the ADC for downlink processing, and the IF, a smoothing filter, the up-converter and the DAC for uplink processing. Regarding WCDMA digital down-converter, its bandwidth of output base-band signal is larger than that of the IF signal by 10%, therefore, called wideband signal, it is different from the general GSM signal and the first generation signal.

4.2.2  CDMA RF Designing Performance and Considerations

As mentioned above, CDMA signal is wideband signal. Therefore, the RF part must be designed to be suitable for wideband low-power spectrum density signal. CDMA’s large dynamic range, high peak factor (due to linear modulating and multi-code transmission), and precise high-speed power control loop are great challenges to the linearity and efficiency of power amplifier.

CDMA makes very high requirements for the linearity and efficiency of the RF front end. Linearity is demanded for strict output spectrum mask and, at the same time, the great fluctuation of output signal envelope. To ensure the power amplifier is efficient enough, we should keep its operating level around 1 dB point.

To make the mobile station more compact and power-efficient, one-step direct conversion should be implemented from baseband to RF or from RF to baseband at both the transmit end and the receive end. Such technology is difficult in that the frequency mixer must be completely linear to avoid any possible intermodulation product between two adjacent channels. In addition, input isolation of the frequency mixer must be good enough to avoid DC due to self-mixing.

The performances of AGC and LNA in RF part are crucial as well. In WCDMA designing, the noise index of AGC should be around 80dB, while that of LNA should be lower than 4dB, because it decides the overall noise index of the receiver.

Analog RF components cause great RF index changes and individual diversity. We should emulate the total receiver performance loss caused by each RF component in the worst case, so that a group of stable RF designing parameters can be obtained. Moreover, according to the latest designing scheme, the number of analog components should be made as small as possible, which makes it necessary to move ADC and DAC closer to RF part. However, when considering the present signal processing capability of the component, the digital IF technology is a commonly used for designing.

4.2.3  Digital IF Technology

The sampling law shows that, if we perform equal-interval sampling at interval of 1/2f_H second for the continuous time signal m(t), with a frequency band limited at (0,f_H) Hz, m(t) can be definitely determined according to the sampling result.

In this case, 2f_H is called Nyquist frequency.

Typically, a modern receiver is structured such that analog-digit conversion and sampling are performed by the IF component. The specific process is: IF signal M(ω) with the bandwidth of B undergoes IF sampling with fs ≥ 2B(1 +α/n), to get the resulting signal M_S(ω), which further becomes the quantized and sampled low IF signal M'S(ω) after passing the low pass filter H(ω). The final signal has the same frequency spectrum as that of the original one.

It can be seen from the above process that IF sampling can be done with a frequency lower than the highest valve of signal frequency as long as the frequency meet the specified conditions. In the meantime, frequency conversion can be achieved through IF sampling, that is, converting the signal to a lower IF, and multiplying the common frequencies in the numeric field, and the base-band diversities I and Q can be deduced.