What's the difference between the signal generated by hNRReferen​ceWaveform​Generator.​m and wireless waveform generator?

Hi,

I want to use MATLAB to generate a 5G NR downlink TM signal, replay it on an R&S SMW, and then demodulate it using an FSW. Initially, I used the Wireless Waveform Generator from the 5G Toolbox to create a TM3.1a signal with a 100 MHz bandwidth, 30 kHz SCS, and FDD configuration. After generating the signal and importing it into the SMW, everything worked smoothly. However, I encountered an issue during demodulation. Using the same demod configuration for this TM signal, I measured an EVM of 4%, which is excessively high for a simple replay. Upon further investigation, I discovered that switching the "Demod → Calculation Method" option from the default "EVM 3GPP Definition" to "At Optimal Timing Position" reduced the EVM to 0.3%, a much more reasonable value.

Next, I tried generating the signal using the hNRReferenceWaveformGenerator.m script from the EVMMeasurementOfNRDownlinkWaveformsExample example. I applied the same configuration and process to import it into the SMW. This time, the demodulated EVM was only 0.3%, even with the "Calculation Method" option set to "EVM 3GPP Definition." This result puzzled me because the signal was generated and demodulated with identical configurations in both cases, yet the EVM outcomes differed. What could explain the difference between these two approaches? I find the Wireless Waveform Generator from the 5G Toolbox more convenient for quick signal generation. How can I configure the signal in this tool to ensure the EVM during demodulation is consistently low and accurate?

For hNRReferenceWaveformGenerator.m, below scripts are used:

addpath("EVMMeasurementOfNRDownlinkWaveformsExample\");

wavegen = hNRReferenceWaveformGenerator("NR-FR1-TM3.1a", "100MHz", "30kHz", "FDD");

For Wireless Waveform Generator from the 5G Toolbox, the config generator.m is attached.

Below two screenshots are the Calculation Method and results for two different settings.

Later, I also tied to use the Wireless Waveform Analyzer. It seems also have same issue to the 3GPP mode.

Thanks for your guys help!

Answer 1

[Sriram Tadavarty]

Hi Deng,

Thank you for sharing the MATLAB scripts used for waveform generation. They are helpful for investigation.

I noticed that the WindowingPercent property differs between the two configurations (hNRReferenceWaveformGenerator and generator.m), which caused the difference you observed. In the attached WWG-generated configuration (generator.m), the WindowingPercent value is [], indicating that the Windowing source is set to Auto in the App. In contrast, hNRReferenceWaveformGenerator sets WindowingPercent to 0. By default, the WWG App sets the Windowing source to Custom and the Windowing percent to 0, matching the configuration generated by hNRReferenceWaveformGenerator.

To generate identical waveforms, set the WindowingPercent property to 0 in the generator.m code. When using the WWG App, ensure that the Windowing source is set to Custom and Windowing (%) is 0. This will produce the same waveforms as those generated by hNRReferenceWaveformGenerator.

The impact of windowing on EVM is seen in 3GPP mode, as the measurement involves EVM computation at EVM window edges.

Hope this helps.

Regards,

Sriram