The applications of AI (machine learning) to radio are fascinating me at the moment. This is a simple experiment where I recorded some SSB and fed it in to an AI noise remover at https://audo.ai/noise-removal This video switches between the input and the resulting output.
This was done just with the default settings. No training on just the noise was done. I think it does a very good job. My guess is that with modern CPUs and ML chips it should be possible to do this sort of processing in near to real time.
Check out this trailer for a program covering the history of computers, and microcomputers, in Australia. So many memories for me.
"An in-depth look at the history of computers in Australia from 1949 to 2024. The series of episodes are currently in production and will be released as soon as possible."
I built many of the boards shown including the MiniSCAMP, EA 2650, Talking Electronics TEC-1, I also owned a Dick Smith System80 and a Kaypro II.
FreeDV, which bundles up the very low bit rate voice codec, Codec2 and an Orthogonal Frequency Division Multiplexed (OFDM) modem that can send the data over an imperfect HF channel using an SSB radio.
The project gives us the ability to have a pretty good sounding digital voice contact which can be free of the background noise we are used to with analog HF radio. Amazing.
The new mode, which applies machine learning to both the voice codec and the modem, is Radio Autoencoder (RADE). "It combines Machine Learning (ML) with classical DSP to send high quality speech over HF radio at SNRs as low as -2dB. The speech signal has an audio bandwidth of 8kHz, but the RADE V1 signal requires just 1500Hz of RF bandwidth."
I'm a bit sad to see codec2 being superseded by the FARGAN speech synthesiser which is based on work by Jean-Marc Valin. Codec2 by David Rowe VK5DR, is a world leading low bit-rate codec with the added utility of being open source. I'm sure it will continue to live on for constrained computing applications.
I like to use Linux in the ham shack and don't mind building open source software. To get RADE you need the version 2 fork of the FreeDV GUI application. It's early days and I ran into some build problems. I posted to the digital voice Google group and was quickly assisted by the author of the GUI application Mooneer Salem K6AQ and Barry Jackson G4MKT who has published a very nice build script that worked flawlessly for me on Linux Mint 22.
The machine learning parts are made with python, not uncommon for machine learning code, but it does make the build rather complex due to not only c/c++ dependencies but also python modules being required. The plan by the developers is to port the ML parts over to C++ in the future which will simplify the build and reduce the size.
The flow charts on the left show the "traditional" architecture of FreeDV on the left compared to the simpler ML based architecture on the right. The AI architecture is "simpler" in one sense but on the other hand is much bigger and more resource hungry.
There are ready to run builds for both macOS and Windows available here. (Note that this link will age so you should check for later builds). On macOS I ran into the issue mentioned at the top of the page where I opened FreeDV and got a message saying that the application was damaged. The solution is on the page, in terminal run:
xattr -c /Applications/FreeDV.app
The screenshot above shows it running on an M2 MacBook Air, receiving FreeDV's most enthusiastic user Jose Donnari from Argentina.
I'm having a bit of trouble finding people to talk to with this mode at present so to test it I transmitted on 40m and captured the signal on the excellent Ironstone Range SDR in South Australia. A path of about 700km. The recorded audio is played back in FreeDV which shows how it would be decoded by a station at the other end.
You can see there is deep selective fading on the signal but compared to past FreeDV modes there are no "beeps and burps". Another noticeable improvement is the muting when there's no actual signal seems vastly superior.
Given that this signal fits in about 2kHz of HF bandwidth, I think it's much easier to listen to than SSB once you get used to it.
There's starting to be quite a bit of activity. Here are two stations in South Australia talking on 40m, about 700km away. Even when one went to low power reception was excellent.
It's early days but this new application of machine learning to radio seems very promising. The pace of development of machine learning is breathtaking and my congratulations to David Rowe and the team for being at the cutting edge. The other big application of machine learning I can see is noise removal in HF receivers. I wonder how long it will take for commercial equipment to include this.
Scammers are a real threat to us all. Australians lost billions of dollars last year. Nightlife Tech Guru Peter Marks, a software developer and technology commentator from Access Informatics, joined Philip Clark on Nightlife to discuss the latest news in technology. https://www.abc.net.au/listen/programs/nightlife/tech-talk-with-peter-marks/104622796
Lots of useful parts for sale at good prices. I bought PL-259 crimp connectors for $3 each instead of the much higher prices I see them online. Tempted by one radio but the seller didn't accept my offer. Probably should have stayed until the end but it's a long drive back to Drummond. I did buy a few books and a nifty hot air gun.
It seemed a bit quieter than last year but there was lots to see.
Mostly, it was great to catch up with friends. Here's Peter, VK3YE being sold morse keys by Drew VK3XU:
Andrew, VK3BEK; Ralph VK3ZZC (who purchased yet more test equipment), Dave VK3ASE:
I bought a few books including the excellent "radio handbook" by William Orr.
I looked it up on Amazon and find that a new copy is on sale for $749!
The Apple Vision Pro is a product they've been working towards for many years. The software progress has been public, in the form of ARKit, and the hardware in private. It shipped early this year and, while I've read many reviews, it's something I wanted to try myself.
You can book a personal 30 minute demo at Apple stores and it was easy to get an appointment.
The first step is that they put your glasses (if you wear them) in a machine that figures out your prescription. I had mine with me but they preferred to let the machine do it. My sight requires astigmatism correction and the magnetic clip in lenses they came back with seemed good.
There are some new user interface gestures to learn. Basically you "point" at things just by looking at them and "click" by touching your thumb and forefinger. It's hard not to try to point with your fingers at first.
After putting the headset on and adjusting the straps to keep light out the magical thing is when the outside world appears thanks to the high resolution cameras and high resolution screens in front of your eyes. It really looks like you're looking at the world through glass. Perspective seems the same and it's high dynamic range.
The first thing I did, which perhaps puzzled my demonstrator, was turn my head side to side rapidly to test the refresh rate. It's good but not perfect, the scene got a bit choppy.
There's a bit of calibration to do, looking at dots in front of you etc. It worked well.
Then a bunch of demos to run through. 2D photos, 2D video, 3D photos, 3D video and "immersive" photos and video. I have no interest in sport but the 3D video of basketball was quite impressive.
Most iPad apps can be used in a floating window and I looked up my WSPR Watch app in the App Store but it wouldn't let me install it.
The most impressive thing to me was how solidly user interface elements seem anchored in place around the room you're in. You can surround yourself in "windows" and pull or push them away.
The headset is bulky but didn't feel heavy on my head. After wearing it for about 20 minutes it did feel a bit hot and was a bit of a relief to take off. I could certainly imagine wearing one on a flight to privately watch a movie but I think that would be a waste of the technology.
I saw a couple of minor glitches during the demo but overall it was a solid platform.
The jury is still out on the future of spacial computing. One day the big flat screens we put on our desks will seem as dated as the deep CRT screens we used to use.
It was great to try this out. If you have time, take the opportunity. There was zero pressure from the demonstrator to have me buy one. At AU$5,000 and up, I wasn't tempted. If the price comes down to something similar to an 8k monitor they'll sell like hot cakes.
When I wanted to calculate something in the radio shack, I used to google for the formula and then execute it on a scientific calculator. Just now I wanted to know "5dbm as peak to peak voltage across 50 ohms" so I asked ChatGPT which responded very helpfully:
This is really useful. I look forward to having it built in to my phone.
Having said that, this interesting video looking at whether high SWR causes common mode current on coax points out that ChatGPT isn't always correct but is useful.
I generally use superglued on MePads from QRPme but burn through them at a great rate. Recently I had a go at cutting insulating tracks on PCB with a Dremel cutter wheel.
I purchased a set of diamond (really?) tipped hole saw bits intending to use them to cut my own pads but this didn't really work as the thickness of the cut is so wide you make a big hole and get a very small pad.
By using the hole saw to just cut the copper layer I get a nice insulated pad and it's not hard to add extra pads to a circuit under construction. Here's a diode ring modulator currently on the bench:
MePads are great but it galls me to pay more for postage than the product. I wonder if a PCB manufacturer can make something like them?
Peter Marks, software developer and technology commentator from Access Informatics joined Dom Knight with all the latest in tech news. It's been 40 years since Microsoft Excel was developed, touchscreens are out and buttons are back! listen at:
Recently I've been working on an AM transmitter for 40m. It occurred to me that FM should be much simpler to produce. I had ordered some varactor diodes recently and first tried to FM a crystal or ceramic resonator. I can't get much deviation and the audio did not sound good.
Then I remembered a post on Soldersmoke about a largely unknown but surprisingly stable oscillator called a Franklin oscillator. It has two transistors and has a very light load on the tuned circuit.
Rather than the variable supply transistor I fitted a 5V three terminal regulator. Coupled to the tuned circuit is a varactor diode with an electrolytic capacitor through which line audio is inserted. There's a trim pot to set the DC to the diode.
Initial results are quite promising.
The brilliant Charlie Morris, ZL2CTM, has an excellent video on the Franklin oscillator.
