Distortion Effects Prevail In RF Design

With the wave of wireless devices proliferating in the world today, there is a corresponding need to perform analysis of RF circuitry.


It’s an exciting time for consumers of wireless devices, but it’s a challenging time for system designers who must design, analyze and verify that all of the components in those wireless devices interoperate.

In wireless designs, distortion effects play an important role in the performance of RF circuits, including mixers, low-noise amplifiers (LNAs) and power amplifiers (PAs) and managing these nonlinear effects is critical. These effects have an important impact on the performance of radio frequency (RF) circuits, which include power-amplifiers, low-noise amplifiers, mixers, local-oscillator/synthesizers, baseband analog processing, mixed signal sampling, and baseband time-domain processing.

“The distortions broadly fall under RF coupling/interference among the hardware modules, multiple classes of non-linearity’s including coexistence operation with other RF devices, dynamic range in stand-alone, channel fading, and coexistence operation, first and third order inter-modulations and intercept points in coexistence, peak to average power ratio (PAPR), carrier leakages, in-phase and quadrature phase (IQ) imbalances, frequency drifts, time alignments, power supply noise, and temperature effects,” said Eric Ruetz, director of engineering at Synapse Design. “The distortions manifest by causing error vector magnitude (EVM) degradations, FCC mask compliance failures and emissions compliance failure. These effects could also cause degradation of signal to noise ratio (SNR) in the receive chain by affecting demodulation of the data in the baseband.”

Nebabie Kebebew, senior product marketing manager for custom simulation products at Cadence, explained the way distortion manifests in day-to-day activity is when talking on a cell phone, for example, and all of a sudden it goes silent. “If your design has distortion what it ends up meaning is that you won’t hear the person you are talking to. You basically lose the actual signal being transmitted and you don’t hear the person on the other side or vice versa.”

There are many ways to analyze this distortion is also try to correct it, she said. “In the past, circuit designers would do a trial-and-error method, but we have come a long way from that – where now RF designers specifically when we talk about distortion for an RF type of design, some of the components we refer to are the low noise amplifier or the mixer. And in that scenario what they do now is run simulation to figure out where the potential source of distortion is. But more important, once they find out there is distortion, they find out what is causing it. If you don’t find out which part of your circuit is causing it you are still going to be going through that trial and error.”

Pascal Bolcato, manager of the analog/RF simulation group at Mentor Graphics, agreed and explained further that what is required to characterize distortion is a spectrum of the signal at the output of the circuit to be characterized. “Then this output spectrum can be obtained using different methods. The first one could be a transient analysis with the Fourier transform after that, but this is probably the worst because it’s not very efficient and not very accurate. After that you can use steady state analysis to directly compute the steady state of the signal, then the spectrum of the signal at the output of the circuit.”

There are two methods used to compute steady states. One is in the frequency domain called ‘harmonic balance,’ while the other is in the time domain called ‘shooting methods.’

“Harmonic balance directly computes the signal in the frequency domain and is probably more accurate and provides better resolution and dynamics,” Bolcato said. “That’s why I think harmonic balance should be preferred to characterize distortion. And now, using steady state analysis you can directly compute total harmonic distortion, or intercept points that are standard figures to characterize distortion when the input signal is a pure tone.”

This technique is adapted for simple modulation schemes or analog modulation schemes but for modern modulation schemes, and when dealing with the digital modulation, for instance, in modern wireless circuits, such as 3G or 4G circuits that use CDMA, HPSK or OFDM digital modulation, steady state is probably too simplistic to characterize distortion.

In this case, what can be done is to use techniques like modulated steady state or envelope transient simulation, Bolcato pointed out. “With these techniques, it’s a mixed time and frequency analysis…It can handle digitally modulated signals and we can characterize circuits and characterize distortion of circuits stimulated with a circuit corresponding to real life conditions and figures to characterize distortion are no more intercept points or total harmonic distortion. We use what we call NPR (noise power ratio) or ACPR (adjacent channel power ratio). These are the main measurements to characterize distortions when we deal with modern digital modulations.”

At the same time, it’s a very difficult thing to solve and also to model, pointed out Navraj Nandra, senior director of marketing, DesignWare Analog and MSIP Solutions Group at Synopsys. “The tools that are out there do a pretty good job, but it is very complicated because you are dealing with waves that are in the air, and that requires some very sophisticated modeling.”

As engineering groups are doing more RF-type integration into SoCs, they also are integrating things like IP blocks, he observed. Along with this, the question comes up about whether the IP will act as an aggressor to the sensitive RF blocks or will it be a victim?

Rather than looking at the issues related to design of low distortion analog circuits for RF, or tricks to measure distortion, it should be looked at from the complete system perspective, Nandra said. “Even if the input intercept point (IIP) of each analog block cannot be improved, it is possible to reduce its system-level effect if fewer of these blocks are used. A typical RF chain (e.g.: RX) includes low-noise amplifiers (LNAs), mixers, filters and other amplifiers in cascade. The distortion impairments are multiplied by the gain across the chain. If the RF chain is simplified by removing one or all mixers, amplifiers and/or filters, then the cascaded effect is reduced and the overall IIP is improved. System implementations like digital-IF demodulation and/or full band conversion are steps in this direction, by moving the digital/analog interface closer to the antenna. They are enabled by the availability of the ADC (DAC) that supports effectively every sampling rate and is able to convert high frequency signals accurately.”

Further, in the case of ADC/DAC, distortion is also an important factor, but the concepts of interception points are meaningless, because of the mechanism through which ADC/DAC generate distortion which has a less direct dependency of signal power, he said. Concepts such as SFDR, two-tone intermodulation and multi-tone power ration are used to describe these phenomena.

Mitigating distortions
The challenges in mitigating the distortions start with requirements of lower power dissipation in the system and reducing die size of the radio chips. At a system level, several techniques use feedback mechanisms, and measurements that combine information from RF and baseband algorithms. The techniques use RF circuits to work as digital signal processing based digital RF system to create inverse operation of non-linearity and distortions, which are popularly referred as digital pre-distortions. At system on chip, required level of isolation, signal integrity, stable operation for the temperature range, and power supply noise also play a role,” Ruetz noted.

Simply put, distortion is caused by circuit non-idealities and must be reduced because there is no way to remove them completely. “The trend is to correct the distortion not by changing the RF circuits of the analog part but treat the causes with software or with the digital part that will be a feedback loop between the RF part and the digital part – so that the digital part will correct the RF blocks in order to reduce these distortions and non-idealities. That’s why it’s important not to only simulate the RF part alone, to characterize distortion, but also to simulate the complete feedback loop containing the RF parts, analog parts and the digital parts, and the feedback loop between all of these. Then it’s important to be able to simulate RF together with analog and digital,” Bolcato concluded.