This video is part of the TechXchange: TechXchange Talks.
What you’ll learn:
- Solving frequency-dispersion issues typically requires use of channel estimation.
- Wider bandwidths call into play techniques such as orthogonal frequency-division multiplexing.
- Out-of-band interferers can be filtered, but in-band problems require null steering to solve.
More often these days, wireless-related design projects are migrating to the rarefied reaches of the millimeter-wave (mmWave) bands, where systems deliver wider signal bandwidths. You get the advantages of higher throughput, but you also must bear new requirements for wideband linearity and frequency flatness.
With carrier aggregation comes bandwidths of hundreds of megahertz, and with them comes phase compensation, equalization, and active linearization algorithms that must be tightly integrated with RF transceivers.
All of the above implies a new world for system architects, who must explore and coordinate the design and implementation of multiple system elements. Antenna arrays, RF transceivers, and DSP algorithms operate across multiple standards and in many scenarios that involve interfering signals.
In this first of a series of three videos exploring the arena of millimeter-wave system design, Giorgia Zucchelli, product manager for RF and mixed-signal at MathWorks, will cover the topics of frequency dispersion and interfering signals and how they impact the performance of wideband receivers. She discusses the consequences of those signals, such as saturation and desensitization, and examines filtering techniques that are commonly used to mitigate their effects.
Dealing with Frequency Dispersion
With increasing bandwidth, frequency dispersion becomes more of a problem because either the transmitter or the receiver may not provide flat performance in terms of amplitude and phase over that bandwidth. As a result, one may see distortion of the signal constellation that looks like noise. The dispersion can come from any number of factors, such as distributed elements such as transmission lines, PCBs, and antennas, as well as RF amplifiers.
These dispersion sources, says Zucchelli, are all static, meaning that they change very little, if at all over time. Thus, they can be ferreted out using static analysis tools or electromagnetic analysis. Other sources of frequency dispersion are found in the over-the-air (OTA) channel, which must be tested and diagnosed through OTA testing using anechoic chambers. Consider cases where either the transmitter or receiver, or both, are moving: Then, the channel effects may not be constant.
That brings us to equalization algorithms, which are embedded in the receiver to reverse the effects of frequency dispersion. However, such algorithms are unable to distinguish between static and dynamic dispersion. So, equalization algorithms, in turn, bring us to channel estimation. You can go beyond estimation by prototyping the equalization algorithm using FPGAs.
The Role of AI in Testing Equalization
In evaluating equalization algorithms, one would use pilot signals to estimate the channel. Then, the channel must be inverted to be equalized. But with the move into mmWave frequencies, systems become more complicated, incorporating more and more antennas in arrays. Multiple-in, multiple-out (MIMO) arrays are now giving way to massive MIMO (mMIMO) arrays. This makes the inversion of the channel into a massive problem, as it’s a numerically complicated process.
In this context, AI can really help. It does so by providing an alternative for channel estimation, inversion, or equalization using autoencoders. Rather than treating the design and optimization of all the algorithmic elements of the receiver as separate subsystems, autoencoders can be used by applying AI to jointly optimize both transmitter and receiver as a whole system.
Autoencoders add redundancy to estimate the channel and to estimate the entire transmitter and receiver information. This is achieved by using AI in an unsupervised way. Thus, AI is helping greatly to reduce the complexity of the huge numerical complexity of mMIMO systems by going beyond the equalization algorithm.
Addressing Wider Bandwidths
The move into mmWave spectrum brings much greater channel bandwidths, which makes adaptive or cognitive radios necessary to manage operation over multiple bandwidths. A technology that comes into play here is orthogonal frequency-division multiplexing (OFDM), which enables scaling up or down the boundaries of the signal. This makes signals very flexible and able to be retargeted over various frequencies (both center frequencies and bandwidth).
The first step here is to assess the system’s requirements for targeting multiple frequencies. However, because transmitters and receivers don’t operate in a flat fashion across frequency, some degree of algorithmic adaptability is essential to lend versatility to the system. A link-budget analysis might indicate use of highly configurable digital filters and/or analog equalization and calibration algorithms to perform digital pre-distortion for linearization.
All the above must be tightly integrated. At larger bandwidths and higher mmWave frequencies, RF systems must be digitally assisted and controlled. Everything is tunable, which adds a great deal of complexity. To evaluate such designs, requires going beyond classic CW tone measurements and using metrics such as error-vector magnitude (EVM) and/or adjacent channel leakage ratios (ACLR) to examine the quality of the receiver’s signal constellation.
Measuring and Mitigating Interfering Signals
Interfering signals can, unfortunately, be much higher in power than the signal of interest. As a result, the signal of interest is distorted with a higher signal-to-noise ratio (SNR). That noise will be spectrally correlated—and equally uncorrelated—to the signal of interest. Interferers can be either in-band or out-of-band relative to the desired signal, which makes a difference in terms of both anticipating and mitigating interference.
The good news about mitigating interfering signals is that you can actively address them. Out-of-band interferers are much easier because they’re farther away in frequency. If your band is from, say, 24 to 27 GHz, and the interfering signal is at 19 GHz or 30 GHz, it can be removed with filtering.
However, if the interferer is of much higher power than the signal of interest, it may send your receiver into saturation, potentially desensitizing the front end or ADC. Then, any digital filter you apply is, in a way, too late. It won’t help recover the desired information.
Therefore, out-of-band interference can be removed with filtering, but it must be analog filters. These are often expensive, bulky, and have limited ability for tuning. In-band interferers are trickier because you can't filter them out.
However, with 5G and 6G, and the introduction of beamforming and MIMO arrays, you can steer a null in the direction of the interfering signal if the interferer and the signal of interest are coming from different directions. Also, if you’re building a receiver with a very large bandwidth and your array has many antenna elements, you can steer the beams to make your receiver essentially noise limited.
So, it becomes an interesting problem because you need analog-to-digital converters (ADCs) with much wider dynamic range to ensure that you can recover the signal of interest. Those are some interesting technologies that open new doors for interference mitigation.
The Effects of Interferers on Receivers
In terms of the effects of interfering signals on receivers, when an interferer hits the antenna, which is a passive element, the antenna lets that signal through. Usually, the first circuit after that is a low-noise amplifier (LNA), which is designed to expect a signal very close to the noise level. If the interfering signal is high in power, that LNA can go into saturation. At this point, there's nothing you can do to recover the information from the signal of interest.
That's why it's important to use filters and/or have LNAs with a high dynamic range, or to use null steering. But even with a very good LNA, you might hit saturation of the ADC.
This is where building models and using simulation tools becomes very useful in determining the limiting factor in your receiver, to find out the first thing that saturates or clips. Very often, having digital filters is too late in the signal chain because everything else happens before that.
Filtering Methods and Techniques
A good filtering strategy is to place an RF or analog filter as close as possible to the antenna. But if you want an agile receiver, these aren't great options. Beamsteering is a far more interesting technology, but that will only work in the mmWave range. You can’t have a mMIMO system at sub-6-GHz frequencies because the antenna would be too large.
But the good news is that in the sub-6-GHz bands, you can use a different type of architecture that we’ve seen a lot in recent years. RF ADCs basically bypass the RF content and go straight into the digital domain. It requires fewer RF components and fewer components overall. This is the holy grail of software-defined radios (SDRs).
Depending on the operating frequencies and bandwidths, you may have different choices that reduce the impact of interfering signals.
Related Links: