60 Hz Filter for the HP 6920B

The HP 6920B Meter Calibrator uses the waveform of the input AC power to derive the waveform of its calibrated AC output. The HP 6920B controls the average value of the output AC. Consequently, if there is much distortion on the power line, there will be a difference between the RMS and average value of the meter calibrator AC output. 

My shop reference is my HP 3468A DMM, which measures the RMS value of AC, but I often need to calibrate average-reading AC meters. Now the power line at our house has horrendous amounts of distortion because our town is the terminus of the DC power line between Hydro Quebec and the New England power grid. That terminus has a giant inverter that connects the DC power line to the AC power grid, and that inverter has quite a bit of distortion on the AC output. So my shop reference isn’t useful in standardizing the average AC voltage from the meter calibrator, unless I take additional steps.

Now the HP 6920B has an external 60 Hz reference input designed to be driven by a low-distortion oscillator, so that it will output a non-distorted sine wave on its AC ranges. I’ve driven this input from my General Radio 1310-B audio oscillator to generate a pure sine wave on the HP 6920B output, but I’ve discovered that the oscillator needs to be synchronized with the power line frequency. If it is not synchronized, then there will be glitches at the zero-crossings of the meter calibrator output due to the difference in phase between the power supply ripple and the inverter switching.

The General Radio 1310-B does have an external connector that is both a sync signal output and a sync input. I used a 6.3 volt filament transformer to drive the sync input of the audio oscillator. This removed the glitches and gave a clean low-distortion sine wave output from the meter calibrator.

I described all this, in less detail, in a blog post about my HP 6920B.

However, it’s a pain to set all this up when I want to use the calibrator. It is also a bit worrying, because the HP manual cautions against applying power to the 6920B when it is set for an external reference input without also applying the external reference signal. Doing so would risk damaging the 6920B inverter. So I must take care that everything is connected properly and that the oscillator is putting out a signal of sufficient amplitude to drive the inverter in the HP 6920B calibrator.

To remove this difficulty, I’m in the process of designing and building a 60 Hz filter to clean up the harmonic distortion on my very dirty power line. I’ll then feed output of the filter to the HP 6920B external 60 Hz reference input. 

The filter is a fourth-order Butterworth bandpass with a Q of 10. I'm implementing it with two second-order state-variable bandpass filter sections. Don Lancaster's Active Filter Cookbook covers all this quite nicely. 

The piece that I was missing was information on how to tune the filter once it's been constructed. Lancaster's book is silent on this subject, as is my copy of Williams and Taylor's Electronic Filter Design Handbook. I'll be using 5% polystyrene capacitors I have on hand. I plan to make the frequency-controlling resistances tweakable, to compensate for variations in the capacitance. But how do I know what direction and how much to tweak?

I asked about this on the Homebrew section of the Antique Radio Forum, hoping to find someone who had experience in tuning active filters, but I didn’t get much in the way of useful advice. So, I decided to investigate this myself.

I wrote some code in the Racket programming language to generate root-locus plots of the second-order state variable filter as the values of resistors R1 (the resistor in the first integrator) and R2 (the resistor in the second integrator) are varied around their ideal value. The results are pretty interesting. 


The red lines are the R1 root-locus and the blue lines are the R2 root-locus. Let’s zoom in on the positive pole.


You can see that R2 affects only the frequency of the filter, while R1 mostly affects the damping or Q of the filter.


Zooming in on the R1 root-locus, we can see that it also has a small effect on frequency of the filter, but in practice it should be negligible.

So, what have I learned? To tune the second-order bandpass state variable filter, one should first tune R2 to the desired center frequency of the filter. Then, one should tune R1 to set the 3 dB bandwidth for the desired Q. 

© Steve Byan 2011-2019