RF Receiver Front-End Topologies for Software Radios
A number of different RF front-end topologies are appropriate for software radios, each with its own advantages and disadvantages. This article explores the tradeoffs involved with each approach.
The most common types of RF front-ends for software radios are dual conversion, single conversion, and tuned radio frequency receivers. The suitability of a particular receiver topology depends on a number of parameters that may include the following:
- Sensitivity defines the weakest signal level that a receiver can detect and is usually determined by the various noise sources in the receiver.
- Selectivity- represents the ability of the receiver to detect the desired signal and reject all others.
- Stability indicates the lack of change in the receiver gain and operating frequency with temperature, time, voltage, etc.
- Dynamic range is the difference in power between the weakest signal that the receiver can detect and the strongest signal that can be supported (either in-band or out-of- band) on the receiver without detrimental effects.
- Spurious response is a receiver's freedom from interference due to internally generated spurious signals or to their interaction with external signals.
The tuned radio frequency (TRF) receiver, shown in Figure 1, consists of an antenna connected to an RF bandpass filter (BPF). The BPF selects the signal and the LNA with the automatic gain control (AGC) raises the signal level for compatibility with the ADC. This BPF bandwidth relative to the carrier frequency can be quite narrow, while in absolute bandwidth, it may be quite broad. For example, a second-order inductor and capacitor filter would require a filter quality factor of 107 to extract a 30 kHz signal at 900 MHz with 60 dB of attenuation for a channel 60 kHz away, which is highly impractical.
Figure 1: TRF digital signal processing receiver
The primary difficulty in creating a practical TRF receiver is the limitation of the ADC, which must handle high-frequency signals. In addition, given the bandwidth and roll-off limitations of the RF filter, the sampling rate of the ADC must be very high to avoid significant aliasing. High power consumption is inevitable with high sampling rate conversion. The ADC must accommodate multiple signals over the wide bandwidth of the RF filter (potentially tens of megahertz or more) with high dynamic range of approximately 100 dB. Achieving this sampling characteristic is difficult, expensive, and power-intensive, and extreme demands are made of the tunable RF filter to remove interference signals that consume the dynamic range of the ADC. Non-idealities of the ADC, such as jitter and finite aperture size, lead to distortion of the signal.
In practice, the RF filter can select only a general band of interest: subsequent filtering within the DSP is required to extract the desired channel. The AGC adjusts its gain to accommodate varying signal levels to utilize the full range of the ADC without overloading it. However, the especially high gain required for a single-stage AGC in this application may be difficult to control. Nevertheless, the advantage of this approach is the minimal number of analog parts required.
A very popular topology for low-power application is the single conversion receiver (also known ashomodyne, direct conversion, or zero-IF receiver), which uses a single mixing stage to convert the signal to baseband or near baseband. This receiver architecture, shown in Figure 2, has one stage of downconversion. In the case of a phase or frequency-modulated signal, I&Q downconversion is required since the upper and lower sidebands of these signals contain different information and the sidebands would overlap for a real downconversion. Mixers tend to have high power consumption, and since only one mixer stage (possibly I&Q) is used in the single conversion receiver, the receiver potentially offers good power consumption characteristics. Typically, improved power consumption at the mixer can be traded for dynamic range.
Figure 2: (a) Single-conversion receiver for binary phase-shift keying (BFSK)
and amplitude modulation (AM); (b) single-conversion for frequency-
and phase-modulated signals
LO leakage has the potential of creating leakage across input ports, causing the mixer to downconvert a received version of itself (self-mixing), which may result in a large DC bias at the mixer output. Isolation between the LO and input to the mixer or other components is very desirable but difficult to achieve. An alternating current (AC) coupling capacitor helps but may remove important DC information in the signal. A more effective though more costly approach is to track the DC error after digitization and feed back a correcting bias signal using a DAC and subtracter.
A non-ideal I&Q downconversion may result if the phase and amplitude of the branches are not matched and cause a warping of the received signal constellation diagram as shown in Figure 3 for the case of a quadrature phase-shift keying (QPSK) signal. Furthermore, the phase stability on the local oscillator is extreme given the high and precise frequency needed to convert the signal to baseband since phase noise falls within the baseband. Good circuit design with digital signal processing based compensation can help mitigate these problems. Note that these problems are absent when using the TRF receiver.
Figure 3: Impact on constellation due to imperfect mixing process
(Note that with noise and distortion, a symbol is more likely to
lie in the wrong quadrant, leading to a bit error.)
In some cases, rather than directly downconverting the signal to baseband, it may be more convenient to downconvert to some low intermediate frequency at which the signal may be digitized and downconverted by subsequent digital signal processing operations. A more complex LPF with better roll-off characteristics can help reduce out-of-band interference and thus lessen the dynamic range requirement of the ADC, but it could also allow more noise to enter the system (less sensitivity.) resulting in non-linear distortion products from the filter.
The most common RF front-end for radios is the heterodyne receiver. This receiver, shown in Figure 4a, is commonly used in analog radios. A heterodyne receiver works by frequency translating the incoming signal to an IF that is fixed and independent of the desired signal's center frequency. When this IF frequency is lower than the center frequency of the received signal's carrier frequency and higher than the bandwidth of the desired signal, the receiver is called a superheterodyne receiver. The desired signal that is now frequency translated to a fixed IF can be more easily filtered, amplified, and demodulated. Plenty of good quality RF components are available for standard IF frequencies.
Often a superheterodyne receiver involves using two stages of downconversion. Such a dual conversion receiver has the advantage of relaxed filtering requirements. Because the filtering occurs in stages, the filtering requirements at each stage can be more relaxed than in a single conversion receiver. That is, by lowering the center frequency of the signal using the first stage of downconversion, the filter quality factor can also be relaxed because the ratio of center frequency to filter bandwidth is reduced.
Figure 4: (a) The heterodyne receiver (Note the similarity to the single-
conversion receiver, except for the use of a BPF after the first mixer.);
(b) Dual-conversion superheterodyne receiver
Gain can also be achieved in stages, reducing the LO power on the mixers and relaxing the isolation needed between the LOs and the mixer inputs. The distribution of this gain throughout the front-end impacts the overall dynamic range. DC offset is of no concern in this architecture since the LO frequency ωlo is not equal to the center frequency of the desired signal at the input of the mixer. The additional mixer and LO result in higher power consumption and a larger circuit than that of a single conversion receiver, and often the second filter can be expensive and may exist off-chip. The characteristics of the I&Q mixers need to be matched to prevent distortion like that shown in Figure 3. Phase noise also impacts overall performance by causing unintended modulation on the desired signal because of the time-varying frequency of the non-ideal LO.
At each mixer stage, not only is the signal downconverted, but also a portion of the band at ωI, the image frequency, is upconverted, which places it on top of the frequency translated desired signal. For instance, a 68 MHz LO (ωlo) will downconvert the desired signal by 68 MHz. but the adjacent band, located 136 MHz below the desired signal, will be upconverted to the same frequency range (ωIF, the intermediate frequency) in which the desired signal now lies. This problem is illustrated in Figure 5.
Figure 5: The image frequency problem of heterodyne receivers (Note the
LO frequency ωLO is 68 MHz.)
To mitigate this self-induced interference, an image filter precedes the mixer to suppress the low-frequency band that might interfere with the desired signal after the mixing operation. A consideration in choosing the LO frequency is ensuring standard filters can be utilized after the mixer. Designing the image filter becomes especially challenging if the band of potential interference is heavily occupied with high-power signals. In general, trade-offs exist in the selection of the IF frequency, the image filter, and the post-mixer filter. In some situations, it makes sense to use an LO frequency that is higher than that of the desired signal to up-convert the negative frequency image of the desired signal to a positive IF frequency if this reduces interference from the image. Other approaches to mitigating the image problem are using I&Q downconversion, such as the Weaver or Hartley mixing process, and using mixer structures that use I&Q conversion and phasing to reject the image.
Which Is Best?
The designer needs to weigh the multitude of combinations before arriving at the optimal design and, generally, a trade-off occurs between sensitivity and selectivity. Higher-order conversion receivers (with multiple downconversion stages) may be the best solution in some situations, but more mixers may mean more spurious signals, particularly when high-power signals are present.
The TRF receiver is better suited for a software radio that supports multiple air-interface modes and multiple bands than the single conversion receiver and particularly than the heterodyne receiver because the filter requirements for the IF stages make it difficult to support the multiple bandwidths that might be required of a multimode receiver. Retuning a receiver can result in a complex interaction of multiple components comprising the RF chain. The simpler the RF chain, the more predictable its response will be after retuning. The choice of a single or double conversion receiver depends on a number of factors including channel spacing, frequency plan, spurious response, and total gain. In general, the smaller the channel spacing, the more attractive the double conversion receiver becomes because of its ability to narrowly filter the desired signal.
This article is excerpted from Reed, Jeffrey H., ed. Software Radio: A Modern Engineering Approach. New Jersey: Prentice Hall, 2002. Used with permission.
About the Author
Dr. Jeffrey H. Reed is a professor in the Bradley Department of Electrical and Computer Engineering at Virginia Polytechnic Institute and State University (Virginia Tech) in Blacksburg, Virginia. He currently serves as Director of the Mobile and Portable Radio Research Group (MPRG). Dr. Reed received his B.S., M.S. and Ph.D. from the University of California, Davis (UC Davis), in 1979-1987. Dr. Reed was employed by Signal Science, Inc. from 1980 to 1986 and worked as a private consultant and part-time faculty member at UC Davis before coming to Virginia Tech in 1992.
Dr. Reed's areas of expertise are DSP implementation and software radios. He has co-authored or co-edited fourteen books and over ninety-eight journal and conference papers. Dr. Reed is a past recipient of Virginia Tech's College of Engineering Award for Excellence in Research. He has served as principal investigator or co-principal investigator on over forty-four projects while at Virginia Tech. Dr. Reed continues to work as a consultant and has provided short courses to many companies.