logo2 The Electronics Pages | DITFS

Distortion In Amplifiers Due To Transformer Power Supply Misunderstandings

The following not only applies to single supply power amplifiers, but also to single supply preamplifiers.

The reservoir capacitor is seen as being the source of power of the power supply to its load.

Power is voltage times current. Perhaps current has something to do with it?

Current is the flow of charges around the circuit (which must include the circuit being powered by the power supply), and the reservoir capacitor charge is being used up by the load being that circuit.

As such its voltage falls, and each half mains cycle the diodes turn on to recharge it.

Capacitive input filter tables tell us the maximum ripple current, therefore the time the rectifiers connect the capacitor to the transformer must be 1/6.3, roughly a sixth of the time. And so will be roughly 6.3 times the load current.

That portion of the current taken by the load is weakly connected to the load by the wiring from the transformer to the bridge rectifier, and by the wiring from the bridge rectifier to the reservoir capacitor.

The connection is both resistive as well as inductive, and the LCR circuit formed is completed by the reservoir capacitor. Resonance will find it hard to be established, but the conditions are in place.

The diodes do not switch cleanly because of those conditions, and that results in the usual RF interference caused by these power supplies.

The "cure" is to bypass the diodes of the bridge rectifier with small film capacitors.

However, all these do is lower the interference frequency so that it doesn't cause noise in RF equipment. So the resonance must still exist somewhere.

Measuring a transformer's primary inductance with secondary open circuit, and then closed circuit, the ratios of inductance will be at least 10:1.

The effect of switching therefore results in transformer harmonics. Harmonics affect the magnetising current, and a low magnetising current (which is the vogue) allows a greater degree of harmonics.

Therefore, it is seen that the diodes in the bridge rectifier can't be switching cleanly and precisely.

Some of this can be reduced using a zobel network (snubber) on the transformer secondary. Increasing magnetising current also helps.

Another help is to take off power such that it removes the reservoir capacitor wiring current pulses (1/6.3) which is done by "starring" off the bridge rectifier just above the reservoir capacitor connections.

A further help would be to connect the bridge rectifier directly to the transformer.

All told, this reduces transformer winding harmonics helping the bridge rectifier diodes switch cleanly.

charge currents

Asymmetry Due To Output Capacitors

It's interesting to watch load current from the HT's perspective. Look at the output load current and you'll see it's symmetrical. Therefore you'd expect the HT current to be two positive half cycles per full output cycle. You might be amazed and perplexed to see that it's one positive half cycle followed by an equally long flat.

The reason is that the output capacitor supplies load current on negative half cycles, not the HT.

Obviously, the output capacitor cannot supply that demand continuously, but the top up is such that the output capacitor maintains equilibrium. Therefore, the equally long flat has some slope.

At f0 (the turnover frequency of the output capacitor), the slope is 45 degrees, but that flattens with increasing frequency.

100Hz HT current

500Hz HT current

scaling differences due to frequency response of current probe

Now, consider the ground wire from the amplifier back to the power supply. It has half cycle versions of the output signal on it - like a mark-space ratio of 1:1.

Those half cycles are up-scale, down-scale, all over, because it's music, and are mixing with the charge half cycles (remember, this is current, not voltage), and it is going to result in a comb effect, and if really bad, the signal is going to break up badly near f0 (at the bass end).

By making the power take-off at the bridge rectifier, the capacitor charge pulses are not included in series with the half cycle versions of the signal.

However, they will then mix with the pulses from the transformer, but these are bi-directional, and so won't result in as much damage to fidelity.

Why will they mix there? Because current has to flow back to its origin, but here, it is only 1/6.3 of the time.

By making the transformer to bridge rectifier wiring as short as possible, the mixing effect is reduced.