Making a FET Phase Light -- The Learning Circuit 32

In case it's of any interest, the circuit has its roots in this phase-shift oscillator:

https://en.wikipedia.org/wiki/Phase-shift_oscillator

That has three RC networks and just a single transistor to give enough gain to make up for the losses through the passive

components. The transistor gives 180 degrees of phase shift (an inversion) and the three RC networks contribute the other 180

(ideally they'd each contibute 60 degrees, but for the BJT case it doesn't divide that neatly because of the way that the

transistor loads the last stage).

In Karen's circuit, each RC network is followed by its own gain stage. The two additional stages invert twice, leaving

things as they were. Each gain stage can then drive an LED as its load.

Here it is in a simulator.

I've made the first capacitor a slightly different value (2.15uF) to the other two to encourage the simulation in getting the

oscillation started - in a real circuit they'd never be identical (and the gate thresholds and resistors would be different

too). The real circuit probably starts quicker than the simulation. [I've also used a voltage generator rather than a voltage

source for the power. That just enables me to have the power applied after the simulation start time (at time = 1s).]

First plot is all the waveforms on top of each other which shows how it gets going with one stage getting a slight edge over

the others and then forcing the others into the sequence. The second one is with the waveforms separated and shows more

clearly how the resulting phases drive the three LEDs.

In case it's of any interest, the circuit has its roots in this phase-shift oscillator:

https://en.wikipedia.org/wiki/Phase-shift_oscillator

That has three RC networks and just a single transistor to give enough gain to make up for the losses through the passive

components. The transistor gives 180 degrees of phase shift (an inversion) and the three RC networks contribute the other 180

(ideally they'd each contibute 60 degrees, but for the BJT case it doesn't divide that neatly because of the way that the

transistor loads the last stage).

In Karen's circuit, each RC network is followed by its own gain stage. The two additional stages invert twice, leaving

things as they were. Each gain stage can then drive an LED as its load.

Here it is in a simulator.

I've made the first capacitor a slightly different value (2.15uF) to the other two to encourage the simulation in getting the

oscillation started - in a real circuit they'd never be identical (and the gate thresholds and resistors would be different

too). The real circuit probably starts quicker than the simulation. [I've also used a voltage generator rather than a voltage

source for the power. That just enables me to have the power applied after the simulation start time (at time = 1s).]

First plot is all the waveforms on top of each other which shows how it gets going with one stage getting a slight edge over

the others and then forcing the others into the sequence. The second one is with the waveforms separated and shows more

clearly how the resulting phases drive the three LEDs.

That's great additional information. Thanks, Jon!

Awesome stuff!

Do the LED's being different colors affect the startup timing distinctly?  If so, is that due to a difference in forward voltage?

Without changing the capacitor value, did it not get to oscillating at all?

-Sean

I've got three identical LEDs in there which I picked at random from those in the simulator. Don't even know what they are (it looks like it might be a Kingbright part number). I was just going for the general behaviour to show how it's actually trying to be a sinewave oscillator (with very heavy clipping).

I assumed that it wouldn't start unless I changed something on one of the stages, but I've just tried it with identical capacitors and it does eventually start, though it takes its time. Here's what it does

I assumed that it wouldn't start because I've often seen that behaviour with oscillators in the simulation where they need a kick of some sort to get them going, but here I'm wrong.

I can't explain that but, if you asked me to guess, I'd say that it might be something to do with the way the simulation proceeds. The simulation has to be in discrete steps of time and with feedback like this, with a signal circulating round a loop, I suspect it might be to do with the order in which the calculations proceed round the loop when it first starts. Just guessing, though.

Different LEDs would certainly make a difference. Partly because of the drops limiting the signal at different voltages, but also because the gain of a stage is related to the load. So, all in all, my simulation isn't that good as a model of the real circuit.

I wonder if the difference could be in the real parts. Since the components are made to a certain spec, but there is only a certain level of accuracy possible, there is a slight enough variance from component to component that the minute differences could eventually trigger the oscillation. That's my theory.