Filter Project Proves Once Again: “Stuff” Happens!

This is one of a series of “Notes” I published on Facebook. Since Facebook has discontinued the Notes feature, I am publishing that series here on my blog.

I’m pretty self sufficient when it comes to problems; I create my own. Every once in a while, though, I get a little outside help…

I set out to build a bandpass filter for the 136 kHz band. Fortunately my junk box yielded all of the necessary parts and it didn’t take long to construct the simple circuit. As an interesting side note, when seeking capacitors for RF projects that aren’t in my “new parts” stash, I usually have to dig through RF boards out of old receivers and/or transmitters. This time, I had to get suitable components from audio boards! That says something about just how low in frequency this 2200 meter band is! In fact, some hams have successfully used high end audio amplifiers as RF power amplifiers for this band!

Since I now have test equipment, I avail myself of it whenever possible. The first thing I did is put my new filter on the tracking generator/spectrum analyzer, expecting to see a lovely bandpass response matching what I had seen in the filter design software. Whoa! Something was wrong. Instead of a flat response in the 100 to 200 kHz range, dropping steadily and steeply above and being nearly 70 dB down at 600 kHz, there was a “peaky” response in the desired range with a very sharp secondary peak at 465 kHz and a moderately steep roll-off above that. Response was down only 43 dB at 600 kHz. I checked return loss (SWR) and that was awful too. What could be wrong?

After verifying the circuit layout and marked component values, I realized I must have a bad component or perhaps the type of capacitors I used wasn’t working well in a filter circuit. I poked and prodded, changed out all of the critical capacitors, but there was no improvement. In desperation I resorted to something which probably isn’t an approved troubleshooting technique: bypassing components with a wire one at a time while watching filter response on the spectrum analyzer. Bypassing any part changed the response, but only one caused the spurious response at 465 kHz to vanish. I wasn’t at all sure this meant anything, but it did seem curious. That part was one of three inductors in the circuit. It, like the others, was wound on a small FT37-43 toroid core. Thinking somehow this inductor was messed up, I wound a new one using a different type of wire. This still didn’t change anything.

After much head scratching and pondering, I was running out of ideas. The only thing that made any sense was that somehow that inductor was the wrong value. But how could it be? Since nearly all of the different ferrite mixes (materials) used in these cores look alike, it was theoretically possible that a core of some material other than #43 had worked its way into the bag. That didn’t seem likely since this was a still sealed bag of cores I bought from a major supplier some time ago but hadn’t used until now. Nevertheless, it was worth investigating the possibility since I had no other clues.

I went back to the filter design software and began experimenting with different values for this one inductor. It was supposed to be 47 microhenries. I found that if I changed it to 8 microhenries in the filter designer, the calculated response was almost exactly what I was seeing in the real world on my test equipment. Interesting! The question then became whether the same number of turns it takes to get 47 uH on a FT37-43 core would yield 8 uH on an FT37 size core of some other material. Lo and behold! The same number of turns on a core of #61 material gave exactly that much inductance! It was starting to look as though I might be on to something with this “wrong core” theory. Most of my theories aren’t as promising.

I wound a new inductor using another core from the same lot and put it into the circuit. Nothing changed. I tried another with the same result. Then I decided to get serious. I calculated how many turns it would take on a #61 core to get 47 uH, and wound one accordingly. I had to use really small wire since a much larger number of turns was required. With this inductor the filter response changed to what it should be! I did indeed have some #61 cores in this lot of supposedly #43 cores! Further investigation (involving a lot of coil winding and soldering) revealed that the lot of cores were about a 50/50 mix of the two types. I swear this was a sealed bag from the supplier until I opened it to make this filter. Either the supplier or the manufacturer must have got some #43 and #61 cores mixed up.

In the end I have a perfectly working filter with flat response and less than 1 dB insertion loss across the 100 to 200 kHz passband, rolling off steeply on either side and reaching -70 dB at 600 kHz. It continues to roll off steeply, reaching almost -90 dB at 800 kHz. Return loss in the passband is good at -20 dB or better. The circuit board and construction technique is crude but at this frequency it just doesn’t matter. One of the nice things about building and experimenting in this part of the radio spectrum is that you can get away with a lot of sloppiness. You cannot get away with a 600% error on inductor values in a filter, though. Well, maybe you can if you don’t mind using a relatively poor filter.

I questioned the wisdom of investing in test equipment a few years ago. I really couldn’t afford it. There are still times I wonder if I can afford to keep it. But it has saved by sorry butt more times than I care to admit! I often wonder just how I got by all those years without it. Skill? Luck is more likely! I don’t see how I would have spotted this problem without the test equipment. Undoubtedly I would have placed the filter in service assuming all was well, and never known that it was not performing as intended.

Leave a Reply