1970s Design Indulgence

Another way of reducing toroidal transformer saturation is to listen really loud!

Isn't efficiency great? Best not waste anything, and toroidals are really efficient, and not dissimilar in some respects to a ferroresonant transformer, in the way of tight regulation.

Lightly loaded, the toroidal easily saturates, so you might wonder why they're so popular, especially in hi-fi. If one thing really contributes to the metallic sound of hi-fi, it is a saturated toroidal transformer.

Perhaps we should use an underrated transformer? If we did, the saturation would decrease before things got really loud. An underrated transformer would not be able to deliver the full power of a design. Still, it could be designed to provide more power and never use it. Perhaps the marketing department could call it headroom?

Earlier in these ramblings, we looked at underrating to rid ourselves of the buzzing due to distorted mains ("DC on the mains"). With the help of Mr. Self, we found 70% was relatively safe.

The amplifier started out as being around 50 WPC, but its power has reduced to meet the specification. 300 VA was just right for 50 WPC, but we're now at 40 watts. It should need 360 VA, and 70% of that is 252 VA.

Even then, at normal domestic listening levels, the load on the transformer is relatively tiny. Using the same maths for a ten-watt amp, we only need a 126 VA transformer, covering all domestic listening levels. The transformer might heat to destruction during a full-power test, however. Still, the load at normal listening levels might be sufficient to "unsaturate" the transformer and reduce its harmonic output.

Lift the lids of several hi-fi amps, and the transformers look a bit small compared with the bulk of the 300VA unit used in this amplifier. Perhaps their designers have crossed this bridge?

However, reliability matters, and you never know if the amplifier will be used with 4-ohm speakers at disco sound levels. So what is the answer? It might sound silly, but how about switching a dual secondary transformer to remove a secondary, making it 150 VA? The switch could be labeled so the user would be in no doubt why it exists.

It must be really confusing and challenging to understand audio-electronics. At one moment, you're being told about PSRR. The next, you're being told that an amplifier's output is signal-modulated power supply.

PSRR (power supply rejection ratio) isn't about rejecting the steady direct current (DC), which the amplifier needs to "chop up" into a continuously sampled facsimile of the source (analogue). PSRR exists because power supplies are basically sh*t.

After abusing down the toroidal supply, recently made measurements showed distortion bumps at 100Hz (the ripple frequency) and 300Hz. I understand the 100Hz, but not the 300Hz.

I am not even going to analyse these power supply anomalies. Instead, I am scrapping the toroidal supply altogether and reverting to the laminated transformers.

The PSRR of this amplifier is primitive, but in their day, they worked. Therefore, they should still work today. If they don't, then the power supply must be different. It is different - it is a toroidal transformer-based power supply (or was).

Magnetising current increases with voltage, and the increase in magnetising current increases core saturation. Designed to pull 11.5mA at 240 volts, the transformer pulled 16mA on 240 volts - the reason being the mains frequency. The transformer was tested on 60Hz but is being used on 50Hz. It is the equivalent of increasing the voltage 17% (which incidentally agrees with Rod Elliott's article). It increases core saturation by 48%.

Now, this is a good transformer. Matters get much worse using an off the shelf toroidal transformer from the distributors. Core currents become incredibly high, which is not only wasteful but places the transformer well into saturation. This might be OK for a circuit with a large PSRR but will put saturation harmonics well and truly in the signal modulation of primitive amplifier design.

Such harmonics might contribute in some small way to the overall THD of the amplifier, but mainly, it contributes to the transient harmonics, where we are unable to measure, which leaves it up to the designer's ears.

In some ways, it is akin to clutching at straws. It is, in fact, just one of many variables that can adversely affect the sound. However, each of these variables must be discovered and dealt with, but in modern amplifier design, were obviously not and became the equivalent of op-amps, which care little about their power supply - the reason being PSRR. The circuit complexity required takes us straight to today's pattern-part designs.

The reasoning above suggests that the laminated transformer and primitive amplifier are inextricably connected.


Edited by Graham Slee - 20 Oct 2020 at 9:48am

Of course, it may be me who's out of step. It's known that adding 2nd harmonic distortion makes for a more punchy - power-chord - sound. The problem with that is it obliterates whatever is in the space it occupies.

By shutting some part of the global NFB resistor at a high frequency, the shelving down effectively adds to the loop-gain by 3dB at that frequency. This additional loop gain reduces THD, and is very valuable at high frequencies, where conventional methods result in rising high frequency distortion.

This technique simulates as giving additional negative feedback between 20kHz and 800kHz, with double what it was at 100kHz.

The additional "control" is within the parasitic instability range, and might help there too.

The frequency response flatness suffers only by another -1dB at 20kHz.