In circuit design, things aren't always as they seem. That's especially true when it comes to linear circuits.
Yes, I know that analog simulation is now quite sophisticated. Equation-based analog circuit descriptions and EDA synthesis aside though, sometimes it's the simple things that trip you up.
Let me cite an example. I've been using Electronics Workbench's MultiSIM simulation package for quite a few years, so when I needed a circuit for a recent RF design project, I turned to my trusty software package (MultiSIM v6.11 PowerPro running on a modest Pentium II PC).
SPICE is nice, I thought to myself, as I launched the software---knowing that MultiSIM's schematic-capture front-end would free me from the tedium of a SPICE node annotation task.
Maybe I was over-confident. Over the years MultiSIM helped save considerable time that otherwise would've been spent breadboarding. The software did a yeoman's job of running Bode plots for RF bandpass filters, for example, and it did a good job simulating closed-loop AGC performance. I was in love with MultiSIM's virtual instrumentation, calling up plotters, DMMs, and dual-trace oscilloscopes to my heart's content.
Then it happened. I defined and captured a very simple FET-based Colpitts oscillator for an IF strip that needed a stable crystal-controlled conversion oscillator. The 9-MHz FET oscillator would feed a shunt-feedback direct-coupled buffer comprised of a pair of bipolar transistors.
My plan was to let MultiSIM tackle the LO (local oscillator) simulation first. As it turned out, the LO tackled MultiSIM.
When I toggled the Simulate button in MultiSIM, nothing happened. My virtual oscilloscope just sat there, with a recurring trace on its screen, and no vertical deflection whatsoever. My simulated LO was dead in the water!
No amount of tinkering with the circuit's parameters helped get the virtual oscillator running. I changed models for the circuit's crystal, and for its capacitive divider. I tried various FET models. I tinkered for hours. To no avail.
Discouraged, I turned to melting solder, lashing up the actual components and hooking them up to my lab supply. As soon as I applied power, I was relieved to see a clean sinewave on my bench scope.
A few days later I described my plight to my friend Mike Carter. In his capacity as a professor of electronics engineering at the University of New Hampshire, Mike uses MultiSIM in his classroom.
"Oscillator simulations are the bane of my students," Mike said. "Most oscillators don't take off. The simulator's parameters are perfect, and there are no real world imbalances to get the things running."
Mike suggested pulsing the oscillator from Vcc to shock the simulation into operation.
Back at my PC, I called up my LO and added a wire from the power source to the crystal. I poked in a momentary pushbutton from MultiSIM's library.
Sure enough, Mike was right. As soon as I hit the space bar on my keyboard to close the pushbutton and apply a pulse, the LO started oscillating! Like I said, in analog design, things aren't always as they seem.