Jim is right, and we are getting to it only partly so in the "VFD" naming remnant from the past.
The term VFD for "Variable Frequency Drive" is a hangover from the very early method of getting a measure of speed control by changing the frequency of the AC supply, especially with
synchronous motors where the internal magnetics are supplied current to have the motor rotate at a rate locked to the AC line frequency. Most industrial AC motors were the induction type, which operated with a necessary
slip, the motor rotating slower than the AC supply alternations as part of the way the rotor gets magnetized.
Providing a sort of AC, basically power switched on and off in sequence to the windings, with energy storage chokes, and frequency slowing down, does have the desired effect, over a partial range, and quite drastic drops in torque. There is a limit to how far you can "slow down" an AC motor in this way. Lowering the switched voltage, and letting it operate with huge amounts of slip also slows it down. None of this is good for efficiency.
Pulse widths and "carriers"
Motor speed control moved on to the
pulse width modulation scheme Jim mentioned. This is where the coils of the motor are switched on and off by transistor electronics at much higher frequencies - way too high to have any relationship to complete shaft rotation. The switching on and off ranges from about 2kHz to 20KHz or more. This is known as the
carrier frequency, which is a reference to radio frequency modulation for communication, where the term comes from. Higher power motors, and those with very long cables (30m or 50m) need the lower frequencies, but it comes with an very unpleasant buzzing squeal which only goes away when the carrier rate is set to higher than about 8kHz. So if the "noise" coming out of the motor drives you nuts, it may be the VFD menu configuration only needs to have the frequency set higher.
I have found that even with higher power servo motors (70kW) with cables as long as 20m, 12.5kHz was perfectly OK, and much nicer to live with.
How is the speed changed?
This is by altering the energy delivered during a high frequency cycle by changing the ON time proportion compared to the OFF time. It might change (say) from 10%ON 90%OFF to 50%ON, 50%OFF. This is PWM, Pulse Width Modulation, and you hear the term "duty cycle", which is another way to describe that fraction. If going at (say) 12500Hz, the whole period is 80 microseconds, of which some time is ON, and the rest is OFF
Feedback
It's not enough to just vary the energy in that way. To get a good speed control that will deliver a good high cutting torque, even when turning slowly, or to keep the speed constant, regardless of the load (within limits), needs a way to ramp up the duty cycle as needed. This is done with a feedback control loop. You set the speed demand, and the system compares that to the measured speed, and adjusts accordingly until the loop is satisfied. The ways the speed, acceleration, braking, etc. are measured can vary. Precision servo motors use sensor encoders, but for straight speed control, it is sufficient and accurate to get the information from the currents going to the coils, which have to be measured anyway. This is the "sensorless vector mode" that
@JimDawson refers to. It's not magical, but I allow it can seem that way.
Inside the electronics, there are more feedback loops. A Torque loop, which controls at MHz, Position loops, and more. Current limits, tripouts, braking, etc. You can imagine what happens if all power is cut because a torque limit is reached. The system might let go of a few tons load, instead of at least keeping on delivering what it can. Even when you switch off, the braking can snap shafts unless the energy is delivered over the correct milliseconds into a resistor to bring the motor to a swift, yet non-violent stop.
The cost of this kind of kit has fallen dramatically over not many years. The term "VFD" is now generic, like "Hoovering". Not to be take literally. Yes, it still uses frequencies, but they are much higher, and we don't vary them much anymore. Instead, we vary the mark-space ratio of the PWM waveform, and we sense the speeds from ferrite toroid coils monitoring the currents.