These comments were the first thing I thought of as well. There was a lot of stuff I couldn't get into because of the word limit. But long story short, I tried various filtering techniques, but none would work. The schematic here is really unreadable, but I posted a link to the original article in a comment above. There, you can see the schematic and other images with better resolution. You will see that there are two opto-couplers converting the line frequency to a digital signal. One is the "phase", which outputs a 60 Hz 50% duty signal. The other is "ISR" which outputs a 120Hz 10% duty. I was able to clean the 50% with a .01 to ground. But when I tried that on the 10%, the computer wouldn't start. Everything I did to the ISR signal caused a boot failure. I'm not really sure why. Replacing the signal was the only alternative left.
My first thought would be a tuned LC filter, resonant around 120 Hz. That would have avoided the noise problem and been far more reliable, and a whole lot cheaper.
When I was designing test equipment for factory floor operation we almost always put a Sola Brabd constant voltage transformer ahead of the controls and instrumantation power input. It was always cheaper than even one service call to fix a machine, and it gave us a reputaion for producing very reliable testing machines. The noise on the waveform on the scope traces looks a bit like it comes froman inverter drive circuit for a medium horsepower motor.
THe machine;s been running for over a year now, WKetel. And no failures yet. I would think it's pretty reliable. I agree a filter would be best, but I could find none that would work. And when you're faced with "make it work or I want my money back", you do what you got to do to make the customer happy.
I think Andy might be talking about the formatting of the article done by EEtimes, and not the author, which is me. The text has clearly been revised, and very well I think. The problem lies in the resolution of the images. You can not make out the text in the scope images at all, and the schematic is unreadable as well. It's also difficult to take in how bad the signals really were with these low resolution images. They really make the article somewhat nonsensical due to lack of information.
The URL below is the article as I originally wrote it, and much more readable.
@gigavolt (and Andy) - point taken, and very valid. But a more polite request direct to Caleb would probably have got a better result - space in the columns is limited, but they can put click links to bigger versions of images.....