Tracking down RFI

I read H. Ward Silver, N0AX's article Recording Signals which is Experiment 114 in his Hands-On Radio series in the July 2012 QST and building the primitive radio telescope for Jupiter and solar noise sounded interesting. I don't use Microsoft Windows of course, but it only took some minor changes to my audread program to get it doing some of the same basic stuff as Radio Skypipe, but under OpenBSD.

More about my stripchart program later, I still have some tweaks before it's ready for prime time. I got it working enough to see that I had a noise problem. Bursts of what sounded like arcing that were intermittent. I recorded it over about a week on and off, at night, in the daytime, when it was raining and not. It sounded (to my imagination) like a wire arcing against something and the wind would move it so it didn't arc for several minutes, then it would be back. I could sort of hear a 60 Hz component in there, so I didn't think it was an accumulation of distant lightning. It showed up in the waterfall of Fldigi as maybe 60 and 120 Hz lines but that's not what Fldigi's meant to be used for, so they were way to the left side with no calibration marks. I'm still trying to come up with a better FFT for looking at these.

I was preparing a case to call the electric company over this and trying to document as much as I could. I've made probably 20 1-hour recordings and plotted the data to gif files with Gnuplot. Each of these 1 hour recordings has about 35,000 data points, even downsampled from 8000 to 10 samples/second. I made a PDF with 6 plots on 1 page and printed it, but the PDF took up most of a CD, close to 700 megs. There's a screenshot of it below.

Just by coincidence there's some sort of skew in the baseline of the upper right plot which looks close to what I'm interested in measuring. But the noise was making that difficult, even if only by messing up the autoscale when I fed it through Gnuplot. I made the plot black and white to save on color toner in the printer.

Then there came a day when I was home alone so I could turn the power off without having to explain myself at length. My IC-7000 normally runs on a 12 volt deep cycle battery anyway, and my laptop can run on its internal battery for a couple of hours. So I got the laptop hooked to the radio, recorded a half hour, then turned off the main breaker and recorded again. The result is below (scroll the iframe horizontally.)

The vertical scale on the laptop is a little different because I'm going into a mic input with a homebrew attenutator but it worked well enough. This isn't a dual-trace scope: the top recording was about 45 minutes before the bottom one, so don't expect peaks to line up. These are 1/2 hour plots, not 1 hour like above. The horizontal scale is labeled in number of data points: there's no x data in the file I fed to Gnuplot.

So I turned off all the little circuit breakers while I was recording, then at the end of 1/2 hour turned the main breaker back on. While listening to the radio I turned the little breakers back on one at a time. One of them caused some noise, so I turned it back off and kept going. With all but one turned back on, my noise looks like below.

That's mostly innocent background noise above, but I see regularly-spaced positive spikes. This is back on my desktop computer, so the vertical scale isn't the same as the image above. It's also a 1-hour plot again, not 1/2 hour. The spikes are a project for another day and they don't cause that much of a problem for the recording I want to do anyway. They seem to be about 1.3 seconds apart.

The circuit breaker that's still turned off so far has a cheap Chinese touch lamp and an electric blanket plugged into it. My guess is that it's the touch lamp, which also turns itself on if I even so much as run an autotune cycle on my IC-7000. No FCC certification.

I had a couple hours of peace and quiet, then I had to turn it back on. It came back with a vengance. Instead of sounding like arcing, now it's a constant thin buzz. You can see it in the Fldigi waterfall below as 15 vertical lines. The frequency of 20.150 MHz is the same one as in N0AX's column. By the Jove website, frequencies much lower get reflected or absorbed by the ionosphere, much higher and transmission line loss among other things start to be a problem. I'm already at the end of 100 feet of RG-58, the antenna's a 3-band 15-17-20 meter fan dipole I made a few years ago. Unfortunately the lamp's within about 30 feet of it, off one end.

The frequency of the lamp's output isn't 60 Hz. It's somewhere around 193 Hz, with harmonics. One almost lines up with the 2500 Hz mark, and counting down 10 gives about 570, so (2500 - 570) = 1930 / 10 = 193. I just made an mp3 file of it.

In typical frugal Unix/OpenBSD fashion, my files that I record are text. That's what Gnuplot mostly wants, and it's easy to write other programs to read them in and use them. A 1-hour run makes a text file of 177200 bytes, which (all using -9 max compression) is 43778 as a bz2, 49928 as an xz, 55875 as a gz, and 56033 as a zip. The files don't have any time stamps in them, they rely on being given names like audio_2012-07-03_1729.txt and knowing there are 10 data points per second. Exact timing isn't critical anyway.

I know the problem will be back, but after having gotten itself a little attention it's being quiet a few hours after all the fuss. It faded out gradually over a couple hours. I know where I'll start looking next time. Like a lot of intermittent problems this can be tricky to find, and now I'm learning it has multiple personalities with almost totally different symptoms. It's been about 8 hours now since I shut it off for a few hours and look:

It could have been giving the lamp a chance to cool down. Or it could have been a circuit breaker that needed to exercised by turning off and back on to clean the contacts. Almost every breaker in 2 different panels got turned off and back on. Maybe one's getting ready to fail, they're all about 40 years old. All I really know is that now I can set up an old laptop and a different radio, maybe put up another antenna, and start a long term project of listening to Jupiter.

Oh, and I finally got the alligator clip leads out of the audio connection. Now a piece of RG-174 comes out of the computer's line in jack and is twisted onto the pigtail wires on the back of the IC-7000. Some day I'll knock it loose and solder it on when I put it back.

A week later the noise came back.

AB1JX / toys/