RF front ends

Overview

The RF front end is everything between the antenna and the data converters. It amplifies a received signal that may be weaker than −100 dBm without adding much noise, moves signals between the carrier and baseband, filters out everything else in the band, and amplifies the transmit signal to watts without distorting it. Its imperfections set the receiver’s noise figure and dynamic range, and the transmitter’s spectral purity and power efficiency. The digital baseband inherits all of them.

This hub is for baseband and software engineers who need to know what the analogue blocks do to their samples, and for RF engineers who want the system view.

Key ideas

Receive chain and architectures

The low-noise amplifier comes first, after a band filter. By Friis’s cascade formula the first stage’s noise figure dominates the receiver’s, provided its gain is high enough to swamp the stages after it. Any loss ahead of the LNA, from cables, switches or filters, adds directly to the noise figure. The LNA must also survive strong signals: its 1 dB compression point and third-order intercept point (IP3) set how large a blocker it can tolerate before distorting the wanted signal.

The mixer multiplies the signal by a local oscillator to shift it in frequency. A superheterodyne receiver mixes to a fixed intermediate frequency first, where selective filters are practical, then to baseband. It needs an image-reject filter, because a signal on the far side of the LO from the wanted one lands on the same IF. A direct-conversion (zero-IF) receiver mixes straight to baseband with a quadrature mixer. It needs no image filter and integrates well, which is why it is common in integrated transceivers and SDRs. Its own problems are DC offset from LO leakage and self-mixing, flicker noise near DC, and I/Q gain and phase imbalance that creates mirror images. The papers by Abidi (1995) and Razavi (1997) listed below work through these trade-offs. Low-IF receivers place the signal slightly off zero to dodge the DC problems.

Oscillators and synthesisers

The local oscillator is usually a phase-locked loop that multiplies a crystal reference up to the carrier. Its frequency accuracy is the reference’s, in parts per million, and its phase noise is the reference’s noise multiplied up inside the loop bandwidth and the voltage-controlled oscillator’s noise outside it. Phase noise is quoted in dBc/Hz at offsets from the carrier. It hurts twice. In the receiver, a strong adjacent signal mixes with the LO’s noise skirt and lands on top of the wanted channel, which is called reciprocal mixing. In both directions, it rotates and smears the constellation and, in OFDM, leaks energy between subcarriers.

Data converters

An ideal N-bit ADC sampling a full-scale sine wave has an SNR of 6.02N + 1.76 dB, about 74 dB for 12 bits. Real converters fall short, so datasheets give the effective number of bits (ENOB) and the spurious-free dynamic range. Sampling clock jitter sets a separate ceiling of −20 log10(2π f tj) dB for an input at frequency f and RMS jitter tj: 1 ps of jitter limits a 1 GHz input to about 44 dB. Walden’s 1999 survey is a classic reference on how real converters compare with these limits. The receiver’s automatic gain control keeps the signal high in the ADC’s range without clipping on peaks. On the transmit side, the DAC’s output images and its sinc-shaped roll-off must be filtered or pre-compensated.

Power amplifiers, linearity and DPD

The power amplifier is usually the largest consumer of power in a transmitter. It is most efficient near saturation, where it is least linear. OFDM and high-order QAM signals have a high peak-to-average power ratio, so a linear PA must back off from saturation and loses efficiency. Non-linearity shows up in two places. Spectral regrowth leaks power into neighbouring channels, measured as adjacent channel leakage ratio (ACLR). In-band distortion raises the error vector magnitude (EVM). 3GPP TS 38.104 (Release 19 version) sets the NR base station ACLR limit at 45 dB in most FR1 bands, and EVM limits from 17.5 % for QPSK down to 3.5 % for 256QAM and 2.5 % for 1024QAM at carriers up to 4.2 GHz. TS 38.101-1 sets the corresponding user equipment requirements.

Digital predistortion (DPD) lets the PA run closer to saturation. The DPD block applies the approximate inverse of the amplifier’s non-linearity to the baseband signal, so that the cascade is close to linear. Wideband PAs have memory, meaning their distortion depends on recent inputs as well as the current one, so practical DPD uses memory polynomial or generalised memory polynomial models. An observation receiver samples the PA output, and the coefficients are re-estimated as temperature, frequency and load change. DPD algorithms often run in FPGA logic or dedicated hardware. FPGA Guide’s DSP on FPGA hub covers how such filters and polynomial evaluators map onto FPGA resources.

Where to start

The beginner lessons cover the system side first. Read /learn/beginner/01-what-is-a-radio-link/, then /learn/beginner/04-noise-snr-and-capacity/ for noise figure and sensitivity, and /learn/beginner/02-complex-baseband-and-iq/ for what IQ imbalance and DC offset do to samples. /learn/beginner/09-first-gnu-radio-flowgraph/ gets you to a working receive chain on an SDR. Intermediate lessons on transceiver impairments and DPD are in preparation. Radio and transceiver IC specifications are on /reference/hardware/.

Tools and platforms

Common mistakes

• Putting a lossy filter or long cable before the LNA and wondering where the sensitivity went.
• Driving an SDR’s receive gain to maximum. A strong out-of-channel signal then compresses the front end or clips the ADC.
• Reading a PA’s saturated output power as its usable output power for OFDM.
• Ignoring the DC spike and I/Q image of a zero-IF receiver, or tuning the wanted signal exactly onto the LO.
• Running multi-channel or coherent experiments with free-running oscillators on each radio.

Going deeper

• H. T. Friis, Noise figures of radio receivers, Proceedings of the IRE 32(7), 1944.
• A. A. Abidi, Direct-conversion radio transceivers for digital communications, IEEE Journal of Solid-State Circuits 30(12), 1995, and B. Razavi, Design considerations for direct-conversion receivers, IEEE Transactions on Circuits and Systems II 44(6), 1997.
• D. B. Leeson, A simple model of feedback oscillator noise spectrum, Proceedings of the IEEE 54(2), 1966.
• R. H. Walden, Analog-to-digital converter survey and analysis, IEEE Journal on Selected Areas in Communications 17(4), 1999.
• F. H. Raab et al., Power amplifiers and transmitters for RF and microwave, IEEE Transactions on Microwave Theory and Techniques 50(3), 2002.
• J. Kim and K. Konstantinou, Digital predistortion of wideband signals based on power amplifier model with memory, Electronics Letters 37(23), 2001, and D. R. Morgan et al., A generalized memory polynomial model for digital predistortion of RF power amplifiers, IEEE Transactions on Signal Processing 54(10), 2006.
• 3GPP TS 38.104 (NR base station radio requirements) and TS 38.101-1 (NR UE radio requirements for FR1).