The horrible quality waveforms in the transformer primary do not yield clean sinewaves in the secondary. In fact, they yield dirty waveforms in the secondary.
Now, if the secondary "looks into" a well-regulated power supply, with good filtering prior to the regulators, including film capacitors to blunt the hash, you may be able to get away with lousy secondary waveforms.
But I would not connect this straight to an output stage in an amplifier, whether solid-state or tube.
The only high-fidelity circuit I know which uses a triac to control power transformer power is Bob Carver's magnetic field amplifier design. In this design, a circuit in the amplifiers samples signal size and provides a feedback pulse to a triac. The triac varies the duty cycle of a "bucking" winding -- for want of a better name -- in the power transformer. The triac is not directly connected to either the functional primary or the functional secondary.
* * * * *
If a lamp dimmer fires into a standard power transformer, and, for whatever reason, the secondary is unloaded, the primary will change into a choke. This will create especially large transients on the AC line.
But whether it is large or small, this hash affects AM radio reception and, to a lesser degree, television reception. Moreover, this interference can be so powerful that your neighbors next door will also experience this interference.
* * * * *
Engineering data sheets dealing with triac power control often refer to a method called Proportional Interval control. In this design, rather than chopping waveforms to pieces having varying duty cycles, a special controller circuit will remove entire cycles of the AC line voltage. If you remove every other cycle, for example, the power provided to a circuit will drop to exactly half; and the voltage generated in a transformer secondary, will appropriate changes in the filter constants, will also be approximately half. Since waves are not being chopped, hash is not generated.
Fred
).