For more than eight years I have been battling manmade noise, or QRM. In this, I am not alone. The entire ham radio world is experiencing a rapid increase in manmade noise. The result of the many electrical devices used in the home. From phone chargers to LED lights and from e-bikes to solar panel systems. But the radio amateur would not be a radio amateur if it did not experiment with means to eliminate QRM from a radio signal.
Processing the signal before it enters the receiver
Many experiments have been done with processing the radio signal before it enters the receiver, such as with the QRM Eliminator X-phase and with special receive modes like Diversity. Both methods that rely on phase shifting. Manufacturers of transceiver use methods in the receiver itself with a variety of SDR based techniques such as Dynamic Noise Reduction and Noise Blanker. But, there is a different way to strip a radio signal of noise. Not in the high frequency part but directly at the audio output. With audio processing apps using Artificial Intelligence or AI.
AI learns which part of the signal is noise
AI is self-learning, as long as you provide enough data. So AI can actually learn how to strip a radio signal of noise. You only need to offer a data set of signals with noise and without noise. Then AI learns which part of the signal is noise and thus needs to be stripped. For each type of interference you would have to feed AI with a clean signal and a signal with noise. This is rather time-consuming, as there are quite a few different types of interference and many new types of interference appear daily.
Already available audio processing apps
Now there are apps available that have already learned this in way. Audiovisual editors use AI to strip the audio of background noise. For example noise from machinery, chatter, cars driving by, etcetera.
I used such online AI app, at MyEdit.online, to process a recording of a radio signal I picked up during a contest. It’s the contest station CR6K on the 15 meter shortwave band, with interference from the micro inverters of a solar panel installation on my neigbour’s roof. Receiver is the SDRplay RSPduo and antenna a Wellgood active loop v4.1.
Processed signal using AI:
The one disadvantage of this online app is that you have to upload audio as a file to edit it. This is not convenient in our case. We want to do this in real time with limited delay.
Realtime noise cancelling through your computer
Those apps do exist, often as plug-ins for well-known online meeting apps such as Zoom, Microsoft Teams and Google Meet. These apps, meant to suppress background noise, developed at lightning speed when Covid-19 made people start meeting online much more. One such app with some free trial time is that of Krisp.ai.
You install this app on your computer; Windows 11 in my case. It then installs a virtual microphone on which the app performs the operation in real time. In Windows, you can then set it up to listen to this microphone.
A few demos on medium wave and shortwave
Below is a demo of a medium wave signal in AM, disturbed by QRM from the inverter of my neighbor’s solar panel installation. Receiver is an SDRPlay RSPduo and the antenna a Wellgood active loop. Apologies for the bad quality…
It is certainly not perfect yet. As interference increases, the audio quality distorts. Also, with a narrow signal such as in SSB, it works much less well than a broadband signal like in AM on medium wave. The app has trouble distinguishing speech from QRM at a low SNR (signal-to-noise ratio).
Below is a demo with a narrow band SSB signal on the 20 meter band. Receiver is an SDRPlay RSPduo and antenna a Cobweb. Apologies for the bad quality…
The noise cancelling with the Krisp.ai app still leaves something to be desired on this type of signal. This app is clearly developed for ‘hifi’ speech.
Better realtime processing apps
I believe there are better realtime processing apps out there, like the online one at MyEdit.online, that can do a better job on a narrow band SSB signal. If you read this and found one, drop an email via my email address on my QRZ.com page. Or send me a message through Twitter.
Disclaimer: I am not an IT expert. If you need support setting up the audio on your computer, please contact the support of that app.