Improving the EVM Results of Modulated Wideband Signals

Visão geral

When signal power is low, bandwidth is extensive, and the modulation scheme is large, it becomes very difficult to measure the error vector magnitude (EVM) because of noise. In this video, we’ll look at how to overcome this problem using cross-correlation EVM measurements.

Improving the EVM Results of Modulated Wideband Signals

When signal power is low, bandwidth is extensive, and the modulation scheme is large, it becomes very difficult to measure the error vector magnitude (EVM) because of noise. In this video, we’ll look at how to overcome this problem using cross-correlation EVM measurements.

Video transcript 

The Wi-Fi standard continues to evolve to include more challenging measurements. New Wi-Fi 7 channels span 320 MHz. New modulation schemes include 4096-QAM.

 

Here’s a big problem when measuring the modulation accuracy of these tightly packed I/Q constellations in such large channels: When the signal power is low, it becomes very difficult to measure Error Vector Magnitude accurately due to noise. In fact, noise can cause the measurement system to report false EVM results that look better than they are, because noise pushes I/Q symbols into neighboring demodulation decision areas. 

 

In this demonstration, we’re going to look at how to overcome this problem using cross-correlation EVM measurements. Here in my setup, I have a signal generator applying a Wi-Fi 7 signal with 320 MHz of bandwidth and 4096-QAM to a power amplifier under test, operating in the new frequency band from 6 to 7 GHz. After the DUT, this splitter routes the signal to two independent vector signal analyzers. 

 

In terms of software, we have the cross-correlation EVM measurement application that is part of the NI RFIC Test Software. This application controls each of the analyzers to acquire their corresponding signal samples and apply an iterative cross-correlation algorithm between the two sets of samples.  

 

Let’s start the cross-correlation measurement, sweeping from a low power all the way up to amplifier saturation. This way we will get a cross-correlation EVM bathtub plot, or EVM versus Output power. 

 

Notice that the algorithm runs for multiple iterations at each power set point. It acquires the repeating Wi-Fi packets with the two analyzers and demodulates each signal. The cross-correlation math preserves the common Wi-Fi signal information while progressively gets rid of uncorrelated instrument noise from each of the analyzers.  

 

The outcome? Better and better EVM results with every iteration, giving us a TRUE representation of the performance of this power amplifier. 

 

By suppressing uncorrelated noise, this measurement system can produce EVM results that you would never be able to achieve with a single receiver dominated by noise.  

Was this information helpful?

Yes

No