Direct Conversion Receiver

Direct Conversion Receiver
The superheterodyne radio has clearly been the dominant architecture in cellular radio receivers. Therefore, it is discussed as an introduction at the beginning of this chapter. Next, the direct-conversion architecture and its properties, both benefits and drawbacks, are explained and recent implementations reported in the literature summarized. The other radio receiver architectures are not discussed since they are beyond the scope of this thesis. The purpose of a radio receiver is to detect the potentially weak desired signal in the presence of noise and unwanted signals. The power of the other signals might be many orders of magnitude larger than the power of the desired signal channel. Because of the harsh environment, a high selectivity is required. The channel selection at radio frequency (RF) would require filters with very high quality factors and selectivity. In UTRA/FDD, the channel bandwidth and carrier frequency are approximately 4MHz and 2GHz, respectively. The channel-select filter quality factor would be 500 and the adjacent channel attenuation at a 5MHz frequency offset should be at least 33dB in a cellular phone . In GSM, the signal bandwidth is only 200kHz, which increases the required quality factor to 4500, when the carrier frequency is 900MHz. The filter order should be at least five. Since such filters are not available, the problem has to be circumvented . The solution to the problem is heterodyning, in which the RF signal is downconverted to an intermediate frequency (IF) using a local oscillator (LO) signal at a different frequency from the carrier. At a lower IF, the requirements for the channelselect filter become easier to achieve . After selecting the desired signal channel, the transmitted information must be detected. In modern cellular systems, which use digital modulation and coding, the detection is performed digitally. Most of the signal processing is implemented in the digital domain where the limitations of the analog domain can be avoided. However, the direct digitization at RF is not technically possible at the moment, nor will it be in the near future because of the lack of appropriate ADCs. In the future, the analog front-end of the radio receiver will therefore remain necessary in order to reduce the dynamic range and maximum signal frequency before the analog-to-digital conversion. However, this interface is moving towards the antenna as a result of developments in analog-to-digital conversion techniques.


COMMENT Uncategorized

  1. Guru

    After the down-conversion mixers, the signal is at baseband where the channel selection can be performed with integrated lowpass filters. In addition, the baseband signal is amplified to a suitable level before the analog-to-digital conversion. In cellular systems, the power of the desired channel at the antenna connector can vary by many orders of magnitude. For example, in UTRA/FDD, the power in the desired 3.84MHz frequency band can vary by approximately 80dB. The desired baseband signal can be amplified with a variable gain to reduce the required dynamic range in the following ADCs. The channel-select filtering and amplification with a variable gain are usually chained or merged to optimize the performance, i.e. to achieve a sufficiently low input-referred noise without degrading the out-of-band linearity. Variable gain can also be realized in the RF front-end either to decrease the RF gain to achieve higher linearity with high wanted signal levels or to reduce the required variable gain range at baseband. If the dynamic range of the ADCs is sufficient for the entire input signal range of the receiver, no VGAs are needed. A variable RF gain can be used to decrease the dynamic range requirement of the ADCs even if no variable baseband gain is implemented. The gain can be changed using analog or digital control. In the former case, a continuous- or discrete-time continuous-value gain is achieved. In practice, the continuous-time approach is used. In the latter case, the gain value is selected from a pre-defined set in a discrete-time manner. Later in this thesis, the former approach is called variable gain and the latter programmable gain. Since the LNA load is typically an on-chip resonator, the resonator quality factor is limited to such a low value that only signals far from the desired system band are attenuated. Therefore, only the pre-select filter effectively limits the spectrum before the baseband circuits. An additional offchip filter can be used between the LNA and down-conversion mixers, but it is an additional band-select filter and therefore does not perform any channel selection. Since there is no preceding channel-select filtering at RF, the required dynamic range at baseband is high. Since RF voltage gain is typically between 20dB and 35dB, a very low input-referred noise is required at baseband.

    At least three off-chip filters are needed in the superheterodyne architecture. These are the pre-select, image-reject, and IF channel-select filter. In a direct-conversion receiver, the preselect filter is the only unavoidable filter in the signal path that cannot be integrated. Therefore, the direct-conversion receiver achieves a considerably higher integration level. The directconversion receiver is suitable for multi-mode receivers since the bandwidth of the integrated lowpass filters can be made programmable.

  2. Guru

    In a direct-conversion receiver, the desired channel is downconverted to DC in the first mixing stage. The direct-conversion architecture is also called zero-IF. A coherent LO is not typically used . In direct-conversion receivers, which use quadrature modulation, two downconversion mixers are required to avoid unrecoverable loss of information. The LO signals of the two mixers have a phase shift of 90°. The down-conversion mixers are part of the demodulator. The gain and phase errors between the I and Q branches corrupt the signal. The down-conversion mixers and baseband chain produce gain error. The phase shift in the quadrature downconversion differs from 90°, while the error depends on the generation of the LO signals. At baseband, the pole and zero locations in the s-domain are slightly different in the two channels due to mismatches. The results are frequency dependent gain and phase mismatches. Wide-band baseband amplifiers cause only gain error. Since the desired channel is downconverted to DC, the image is the channel itself and the power of the image is equivalent to that of the desired channel. Therefore, the required image rejection is moderate and can be achieved in IC implementations at RF frequencies. Since phase and gain mismatches are relatively constant as a function of time, their effect can be mitigated using calibration if necessary . In UTRA/FDD, the effect of the phase and gain mismatches can be compensated in the digital back-end because of pilot symbol-assisted channel estimation scheme.