When the shape of a current or a voltage wave do nor completely match with a pure sinewave, then harmonics are present on it. Generators do generally produce sinewave shaped voltages, and when in the circuit non linear elements are presents, then distortions appear on currents, voltage drops, and finally, voltage itself.
That is how harmonics are generated Julian
The common terminology that "harmonics are generated" by certain equipment can be somewhat confusing. In most systems, harmonic voltages and currents are the result of non-sinusoidal current flow into a non-linear load, such as a VFD, DC rectifier, etc. This non-sinusoidal current can be represented as a series of waveforms consisting of the fundamental and various harmonics of the fundamental (Fourier analysis). These non-sinusoidal currents flowing through the impedances of the system cause distortion of the voltage waveforms. For analysis of this problem, it is common to model the non-linear load (VFD, etc.) as a "harmonic source", even though, strictly speaking, these loads are not "sources".
Synchronous generators can be a "source" of harmonic voltage. It is common for generators to produce enough 3rd harmonic voltage to cause problems on some systems. Higher order harmonics are also present in most generators, but at such low levels as to not be an issue.
Lastly, induction motors are not sources of harmonics.
jwerthman - good answer. I believe that induction motors and transformers can be sources of harmonics in cases where the unit is overexcited (voltage/freq too high), resulting in iron saturation and non-sinusoidal excitation currents.
Solid state devices generate harmonics by taking nice lovely sinewaves and chopping them into pieces. The smaller the piece the more it looks like a square (or rectangle if you prefer). A square wave is composed of all of the odd mulitiples of the orginal. Thus if you take a nice pure sinewave and add to it a sinewave of 3 times its frequency but 1/3 the amplitude, and then add to that a sinewave of 5 times the frequency but 1/5 the amplitude, and keep going, you will get a perfect square wave.
I thought I would just add a point about odd and even harmonics - just because i never knew why we only dealt with odd hamonics.
If an even harmonic component is added you will see (it is difficult to explain without drawing) that the resultant waveform has an interesting symmetry. The start of the postitive half-cycle is not what happens at the start of the negative half cycle. In fact the start of the positive half cycle is reflected in the end of the negative half cycle.
If the harmonic content is odd, what happens at the start of the positive half-cycle is always what happens at the start of the negative half-cycle (even if you adjust the phase shift of the harmonic). Since pratically all electrical loads behave in exactly the same way in both the positive and negative half cycle, (even if they are distorting the waveform) we only ever have to consider the odd harmonics.
