Why you need only one injection isolator

Figure 1: The RIDLEY UNIVERSAL INJECTOR offers extraordinary wideband operation with either large voltage drive, or high current drive.

A Brief History

I have been working with frequency response analyzers for over 30 years. My first introduction was when working on switching power supplies during my undergraduate degree at Boston University. My senior project was to design an offline switching power supply with a 20 kHz sine wave output regulated at 6,000 VAC. During that project, it quickly became apparent that the problems encountered during the design would have to be analyzed in both the frequency domain and the time domain. My first frequency response analyzer in school consisted of a vacuum tube amplifier, function generator, and oscilloscope. All measurements were done manually.

When I graduated in 1981, I went to work at Prime Computer in Massachusetts, designing 1 kW power supplies for computers. I attended Dr. David Middlebrook’s course on structured analog design, and I was eager to try out his loop gain measurement techniques in the lab. This was a very new topic to power supply designers in those days, and most switchers were sent out into the field without any measurements. Of course, many of them were returned later for repair.

At Prime Computer, we used a narrow-band voltmeter from Hewlett Packard. This simple analog machine was capable of putting out a test signal, and measuring a single return signal at that frequency, with a switchable bandwidth. The output was purely analog and visual – a moving coil voltmeter that displayed the size of the measured signal. With a single input channel like this, Dr. Middlebrook taught his course attendees how to measure both gain and phase of a system.

How was this done? With a series of three measurements at each injection frequency. You measured the input signal, the output signal, and the differential injected signal. Then, by applying the cosine rule, you could calculate the phase angle between each of the signals. It was a wonderful teaching tool to learn how these measurements are made, and it gave incredibly clean results. The pure analog measurement technique, and the natural low-pass filtering of the moving coil voltmeter, eliminated almost all spurious measurements. We were lucky at Prime to have computers everywhere to at least speed up the data entry and calculations with very early versions of spreadsheets, but it was very time consuming.