Radio Architectures, Pt 2: Receivers, LOs, and Mixers
By Christopher Bowick
The following is excerpted from Chapter 8 from a new edition of the book, RF Circuit Design, 2e by Christopher Bowick. You Can buy the book HERE
Moving up the scale in complexity, we come to the next evolutionary RF architecture: the tuned-radio-frequency (TRF) receiver (see Fig. 8-6). This early design was one of the first to use amplification techniques to enhance the quality of the signal reception. A TRF receiver consisted of several RF stages, all simultaneously tuned to the received frequency before detection and subsequent amplification of the audio signal. Each tuned stage consisted of a bandpass filter –which need not be an LC tank filter but could also be a Surface Acoustic Wave (SAW) filter or a dielectric cavity filter– with an amplifier to boost the desired signal while reducing unwanted signals such as interference.
The final stage of the design is a combination of a diode rectifier and audio amplifier, collectively known as a grid-leak detector. In contrast to other radio architectures, there is no translation in frequency of the input signals, and no mixing of these input signals with those from a tunable LO. The original input signal is demodulated at the detector stage. On the positive side, this simple architecture does not generate the image signals that are common to other receiver formats using frequency mixers, such as superheterodynes.
The addition of each LC filter-amplifier stage in a TRF receiver increases the overall selectivity. On the downside, each such stage must be individually tuned to the desired frequency since each stage has to track the previous stage. Not only is this difficult to do physically, it also means that the received bandwidth increases with frequency. For example, if the circuit Q was 50 at the lower end of the AM band, say 550 kHz, then the receiver bandwidth would be 500/50 or 11 kHz–a reasonable value. However at the upper end of the AM spectrum, say 1650 kHz, the received bandwidth increases to 1650/50 or 33 kHz.
As a result, the selectivity in a TRF receiver is not constant, since the receiver is more selective at lower frequencies and less selective at higher frequencies. Such variations in selectivity can cause unwanted oscillations and modes in the tuned stages. In addition, amplification is not constant over the tuning range. Such shortcomings in the TRF receiver architecture have led to more widespread adoption of other receiver architectures, including direct-conversion and superheterodyne receivers, for many modern wireless applications.
A way to overcome the need for several individually tuned RF filters in the TRF receiver is by directly converting the original signal to a much lower baseband frequency. In the direct conversion receiver (DCR) architecture, frequency translation is used to change the high input frequency carrying the modulated information into a lower frequency that still carries the modulation but which is easier to detect and demodulate. This frequency translation is achieved by mixing the input RF signal with a reference signal of identical or near-identical frequency (see Fig. 8-7). The nonlinear mixing of the two signals results in a baseband signal prior to the detection or demodulating stage of the front-end receiver.
The reference signal is generated by a local oscillator (LO). When an input RF signal is combined in a nonlinear device, such as a diode or field-effect-transistor (FET) mixer, with an LO signal, the result is an intermediate-frequency (IF) signal that is the sum or difference of the RF and LO signals.
When the LO signal is chosen to be the same as the RF input signal, the receiver is said to have a homodyne (or “same frequency”) architecture and is also known as a zero-IF receiver. Conversely, if the reference signal is different from the frequency to be detected, then it’s called a heterodyne (or “different frequency”) receiver. The terms superheterodyne and heterodyne are synonyms (“super” means “higher” or “above” not “better”).
In either homodyne or heterodyne approaches, new frequencies are generated by mixing two or more signals in a nonlinear device, such as a transistor or diode mixer. The mixing of two carefully chosen frequencies results in the creation of two new frequencies, one being the sum of the two mixed frequencies and the other being the difference between the two mixed signals.
The lower frequency is called the beat frequency, in reference to the audio “beat” that can be produced by two signals close in frequency when the mixing product is an actual audio-frequency (AF) tone. For example, if a frequency of 2000 Hz and another of 2100 Hz were beat together, then an audible beat frequency of 100 Hz would be produced. The end result is a frequency shifting from a higher frequency to lower—and in the case of RF receivers—baseband frequency.
Direct conversion or homodyne (zero-IF) receivers use an LO synchronized to the exact frequency of the carrier in order to directly translate the input signals to baseband frequencies. In theory, this simple approach eliminates the need for multiple frequency downconversion stages along with their associated filters, frequency mixers, and LOs. This means that a fixed RF filter can be used after the antenna, instead of multiple tuned RF filters as in the TRF receiver. The fixed RF filter can thus be designed to have a higher Q.
In direct-conversion design, the desired signal is obtained by tuning the local oscillator to the desired signal frequency. The remaining unwanted frequencies that appear after downconversion stay at the higher frequency bands and can be removed by a lowpass filter placed after the mixer stage.
If the incoming signal is digitally encoded, then the RF receiver uses digital filters within a DSP to perform the demodulation. Two mixers are needed to retain both the amplitude and phase of the original modulated signal: one for the in-phase (I) and another for a quadrature (Q) baseband output. Quadrature downconversion is needed since two sidebands generally form around any RF carrier frequency. As we have already seen, these sidebands are at different frequencies. Thus, using a single mixer, for a digitally encoded signal, would result in the loss of one of the sidebands. This is why an I/Q demodulator is typically used for demodulating the information contained in the I and Q signal components.
Unfortunately, many direct-conversion receivers are susceptible to spurious LO leakage, when LO energy is coupled to the I/Q demodulator by means of the system antenna or via another path. Any LO leakage can mix with the main LO signal to generate a DC offset, possibly imposing potentially large DC offset errors on the frequency-translated baseband signals. Through careful design, LO leakage in a direct-conversion receiver can be minimized by maintaining high isolation between the mixer’s LO and RF ports.
Perhaps the biggest limitation of direct-conversion receivers is their susceptibility to various noise sources at DC, which creates a DC offset. The sources of unwanted signals typically are the impedance mismatches between the amplifier and mixer. As noted earlier in this chapter, improvements in IC integration via better control of the semiconductor manufacturing process have mitigated many of the mismatch-related DC offset problems.
Still another way to solve DC offset problems is to downconvert to a center frequency near, but not at, zero. Near-zero IF receivers do just that, by downconverting to an intermediate frequency (IF) which preserves the modulation of the RF signal by keeping it above the noise floor and away from other unwanted signals. Unfortunately, this approach creates a new problem, namely that the image frequency and the baseband beat signals that arise from inherent signal distortion, can both fall within the intermediate band. The image frequencies, to be covered later, can be larger than the desired signal frequency, thus causing resolution challenges for the analog-to-digital converter.
In contrast to the simplicity of the direct-conversion receiver, the superheterodyne receiver architecture often incorporates multiple frequency translation stages along with their associated filters, amplifiers, mixers, and local oscillators (see Fig. 8-8).
But in doing so, this receiver architecture can achieve unmatched selectivity and sensitivity. Unlike the direct-conversion receiver in which the LO frequencies are synchronized to the input RF signals, a superheterodyne receiver uses an LO frequency that is offset by a fixed amount from the desired signal. This fixed amount results in an intermediate frequency (IF) generated by mixing the LO and RF signals in a nonlinear device such as a diode or FET mixer.
Generating local oscillators
The LO is often a phase-locked voltage-controlled oscillator (VCO) capable of covering the frequency range of interest for translating incoming RF signals to a desired IF range. In recent years, a number of other frequency-stabilization techniques, including analog fractional-N frequency synthesis and integer-N frequency synthesis as well as direct-digital-synthesis (DDS) approaches, have been used to generate the LO signals in wireless receiver architectures for frequency translation.
Any LO approach should provide signals over a frequency band of interest with the capability of tuning in frequency increments that support the system’s channel bandwidths. For example, a system with 25-kHz channels is not well supported by a synthesized LO capable of tuning in minimum steps of only 1 MHz. In addition, the LO should provide acceptable single-sideband (SSB) phase-noise performance, specified at an offset frequency that coincides with the system’s channel spacing. Referring to an LO’s SSB phase noise offset 1MHz from the carrier will not provide enough information about the phase noise that is closer to the carrier and that may affect communications systems performance in closely spaced channels. Phase noise closer to the carrier is typically specified at offset frequencies of 1 kHz or less.
The LO source should also provide adequate drive power for the front-end mixers. In some cases, an LO buffer amplifier may be added to increase the signal source’s output to the level required to achieve acceptable conversion loss in the mixer. And for portable applications, the power supply and power consumption of the LO become important considerations when planning for a power budget.
Mixers are an integral component in any modern radio front end (see Fig. 8-9). Frequency mixers can be based on a number of different nonlinear semiconductor devices, including diodes and field-effect transistors (FETs). Because of their simplicity and capability of operation without DC bias, diode mixers have been prevalent in many wireless systems. Mixers based on diodes have been developed in several topologies, including single-ended, single-balanced, and double-balanced mixers. Additional variations on these configurations are also available, such as image-reject mixers and harmonic mixers which are typically employed at higher, often millimeter-wave, frequencies.
The simplest diode mixer is the single-ended mixer, which can be formed with an input balanced-unbalanced (balun) transformer, a single diode, an RF choke, and a lowpass filter. In a single diode mixer, insertion loss results from conversion loss, diode loss, transformer loss. The mixer sideband conversion is nominally 3 dB, while the transformer losses (balun losses) are about 0.75 dB on each side, and there are diode losses because of the series resistances of the diodes.
The equivalent circuit of a diode consists of a series resistor and a time-variable electronic resistor. Moving up slightly in complexity, a single-ended mixer consists of a single diode, input matching circuitry, balanced-unbalanced (balun) transformer or some other means for injecting a mixing signal with the RF input signal, and a lowpass or bandpass filter to pass desired mixer products and reject unwanted signal components.
Single-ended mixers are inexpensive and often used in low-cost detectors, such as motion detectors. The input balun must be highly selective to prevent radiation of the LO signal back into the RF port and out of the antenna. Although the behavior of the diode changes with LO level, it can be matched for impedance at a particular frequency, such as the LO frequency, to achieve fairly consistent conversion-loss performance and flatness.
The desired frequency converted signals are available at the IF port; the filter eliminates the unwanted high-frequency signal components generated by the mixing process. The LO drive level can be arbitrary, although different types of mixers and their diodes generally dictate an optimum LO drive level for mixer operation. The dimensions of the diode will dictate the frequency of operation, allowing use through millimeter wave frequencies if the diode is made sufficiently small.
Some single-ended mixers use an anti-parallel diode pair in place of the single diode to double the LO frequency and use the second harmonics of the LO’s fundamental frequency, somewhat simplifying the IF filtering requirements. The trade-off involves having to supply higher LO power in order to achieve sufficient mixing power by means of the LO’s second-harmonic signals.
A single-balanced mixer uses two diodes connected back to back. In the back-to-back configuration, noise components from the LO or RF that are fed into one diode are generated in the opposite sense in the other diode and tend to cancel at the IF port.
A double-balanced mixer is typically formed with four diodes in a quad configuration (see Fig. 8-10). The quad configuration provides excellent suppression of spurious mixing products and good isolation between all ports. Because of the symmetry, the LO voltage is sufficiently isolated from the RF input port and no RF voltage appears at the LO port. With a sufficiently large LO drive level, strong conduction occurs in alternate pairs of diodes, changing them from a low to high resistance state during each half of the LO’s frequency cycle.
Because the RF voltage is distributed across the four diodes, the 1-dB compression point is higher than that of a single-balanced mixer, although more LO power is needed for mixing. The conversion loss of a double-balanced mixer is similar to that of a single-balanced mixer, although the dynamic range of the double-balanced mixer is much greater due to the increase in the intercept point (recall IP discussion from earlier chapters).
By incorporating FET or bipolar transistors into monolithic IC mixer topologies, it is possible to produce active mixers with conversion gain rather than conversion loss. In general, this class of mixer can be operated with lower LO drive levels than passive FET or diode mixers, although active mixers will also distort when fed with excessive LO drive levels.
For RF front ends, wireless receivers, or even complete transceivers fabricated using monolithic IC semiconductor processes, the Gilbert cell mixer is a popular topology for its combination of low power consumption, high gain, and wide bandwidth. Originally designed as an analog four-quadrant multiplier for small-signal applications, the Gilbert-cell mixer can also be used in switching-mode operation for mixing purposes. Because it requires differential signals, the Gilbert-cell mixer is usually implemented with input and output transformers in the manner of double-balanced mixers.