Generating Bad Signals And The Story Of The Poorly Tested Butterworth Bandpass Filter

(New follow-up post here!)

How hard could it be to construct a little band pass filter, throw a signal through it, and see if it properly attenuates? A breadboard, a small handful of components, an online filter design tool. Fifteen minutes, tops? I was longing for the pleasurable jolt of satisfaction that one only gets by watching an illuminated wavy line grow and shrink.

I found the Marki Microwave LC filter design tool, chose a 2nd order Butterworth bandpass filter design, and centered it around the 40M ham band (7.1MHz or so). Hopefully they don’t mind that I cut and pasted the results of their lovely tool below.

I was not fortunate enough to have all of the correctly valued components, so I needed to improvise. Three capacitors wired in parallel added up to the 11nF value of C1. A homemade winding of 22awg wire using the guidance of another online calculator, gave me an inductor somewhere around 43nH. The picture below shows how I measured the DIY inductor; just ignore the low inductance reading, it needed another loop.

Below is the circuit on the breadboard. The component labels match the NI Multisim circuit layout shown below, rather than the Marki diagram above.

I used the Koolertron 60MHz signal generator programmed to a 7.1MHz sinusoidal wave with an amplitude of 5 volts peak-to-peak (V_pp). To get started, I connected the output of the signal generator directly into my beloved Tektronix 2465 300MHz oscilloscope. This scope apparently retailed for $5350 in 1986, which was pretty close to the price I paid for my first NEW car in the late 90s (that’s a fun story for another day!) Fortunately I paid just a sliver of that for a gently used version a few years back. You can see the signal generator alligator clips attached to the oscilloscope probe below.

And sure enough, the scope confirmed the 5V_pp output of the Koolertron.

And likewise, the scope showed the period of the generated waveform to be 140.8ns. Take the inverse of the period and you now have the frequency, and unsurprisingly, it is 7.1 MHz, again confirming the output of the signal generator.

I now attached the signal generator to the circuit. And the oscilloscope across R2. Don’t worry if the circuit looks different then the one I showed you; I went through a few changes before I settled on the correct one (which is the one labeled above). If you are real observant, you’ll notice that I put the positive side of the signal generator on the ground of the circuit…oops! Again, I fixed this, but did not take a picture.

So, take a look at the minute-long video below. This is more or less what the output of the circuit looked like on the oscilloscope. There was no real strong increase in signal strength seen until the sweep of the generator reached over 20MHz. At that point, the signal continued to grow to a maximum of 1.2V_pp near the end of the sweep starting around 40 to 50 MHz. This certainly DID NOT look like the anticipated response of the bandpass filter that I thought I was constructing. I expected a signal with very low amplitude (voltage) which then rose quickly to around 5V_pp around a frequency of 7MHz and then dropped back down to a very low voltage again. I wasn’t too sure how high this amplitude would get, given the potential for drops in voltage throughout the circuit. Take a look at the video below to see what a sweep from 1MHz to 60MHz (and then back down to 1MHz) through the circuit looks like on the oscilloscope. Let’s just say, its not what I expected.

I rechecked my design, and re-ran the circuit a few more times. The results looked similarly bad each time. Certainly not the predictable outcome of a working band pass filter.

I decided to simulate the circuit in NI Multisim just to make sure the circuit itself would work. I constructed the model circuit exactly as I had made it in real life. I remeasured each component to make sure I had the exact values for each one. I tried to model it so it was exactly what I had on the breadboard, including the three parallel capacitors. In Multisim, I attached a virtual signal generator, virtual multimeter, and virtual oscilloscope to the circuit. Note that the voltage is set up on the signal generator to be 2.5V_p , which is the same as 5V_pp.

As you go through these remaining images below, the voltage of the generator remains the same, and I alter the frequency. You can watch the changing waveform compared with frequency in the oscilloscope window in the top right. And the middle floating tool represents a digital multimeter. I have written on the screenshots below two things: 1) the DMM’s voltage readout (in RMS) for each frequency, and 2) the frequency of the test. Keep in mind, this was designed to be a 7 MHz band pass filter, meaning frequencies above or below somewhere around 7MHz should be attenuated.

At 6 MHz, the sinusoidal wave’s amplitude began to increase.

And the amplitude reaches it’s maximum in between 7 and 8 MHz. EXACTLY AS IT IS SUPPOSED TO!

By 9 MHz, the amplitude of the signal is decreasing again.

And the signal is negligible by the time 21 MHz is reached. This is the exact pattern expected, with the signal attenuated until reaching the passband around 7MHz, where the amplitude then increases, and then decreasing again once the frequencies are out of the passband.

So now I am scratching my head…

The simulation works exactly as was expected, although I did open both a resistor (R1) and an inductor (L1). I had to increase the power and amperage capacity of both of them in order to run the simulation. Perhaps this is a clue as to why my circuit isn’t working?! I am however, suspicious that there is something wrong with the signal generator.

For my last hurrah with trying to figure out the problem with this circuit, I check the output of the signal generator against the newest member of my test instrument fleet — the beautiful Bell Labs analog multimeter!

At 7MHz with a 5V_pp (1.77V_RMS ) nominal output by the signal generator, only 200mV_RMS is detected by the analog voltmeter.

By 24MHz however, 7.2V_RMS is detected. This is the equivalent of 20.4V_PP !!! In essence, nearly 3 times the amount of voltage that I had programmed in to the signal generator.

And the voltage drops back down again at 35MHz. Clearly, the signal generator is not holding a stable amplitude as it moves from frequency to frequency. What is odd to me however is the fact that the initial set up of the generator showed a very reliable reading on the oscilloscope of 7.1MHz and 5V_pp . Did my circuit “break” the Koolertron? Is it just a cheap, unreliable piece of consumer test gear? Is there some other explanation for why it cannot properly hold its amplitude? Regardless, it looks like I am in the market for a new signal generator…

There you have it. My latest foray into the interesting world of band pass filters. I probably should have started with a simple 1st order RC circuit, low or high pass, to get started. Regardless, this pushed me to learn even more about the various analyzers I have. And like the infamous “they” always say: When you have one piece of test equipment, you are always right; when you have two pieces of test equipment, you are always wrong.

You made it to the end!
KM1NDY