1970s Design Indulgence

The alchemy of solid-state

My career direction had to go solid-state because tubes were firmly on their way out at the time.

A tube can be built in a normal room. Except for the vacuum forming "burn-off" stage, it's all done with temperatures the average person can comprehend, not so with transistor formation. Temperature control of the chemical process is critical, and at furnace heat, or you don't get a transistor.

Perseverance in perfecting transistors (hahaha) has led to making chemical structures imitate the mechanical and make miraculous atomic level machines. Excuse my simplicity of explanation.

Mankind (sorry, not PC). Humankind can now turn sand into Femto machines and have thus achieved success in alchemy. They may not yet be able to turn lead into gold, but they can turn silicon group chemistry (of which lead belongs) into loadsamoney, which is close enough.

Did all this need to happen to escape the planet? Or was it just so teenagers were able to bop to pop?

In my humble opinion, the transistor might be likened to the tower of Babylon. "OK, if we can't build a tower, we'll build a spacecraft." I am convinced the transistor and all that's followed is a case of post-flood madness!

We have bass!

How long will it last?

t = (1 / 2pi f) cos-1 (voltage - ripple voltage/ voltage) {use radians!}

This equation tells us the current pulse width of the rectifier diode conduction angle, which happens 100 times a second and is applied to a power supply's smoother capacitors.

If we do not use any power, the rectifier doesn't switch. There will always be leakage, though, and so the rectifier will always switch. The ripple voltage produced will be tiny, and the equation tells us the current's pulse width.

It tells us that the smaller the current being drawn, the narrower the pulse of current is.

If the ripple voltage is 100mV and the voltage 70 volts, the pulse width is 170uS, appearing every 10mS. Translated into frequency, we have a massively asymmetrical 5800Hz.

Consuming full power, assuming 10,000uF of smoothing, the current is 2.5 amps, and from V = IT/C, we get 2.5 volts of ripple.

From the above equation, the current pulse is 860uS per 10mS, or a massively asymmetrical 1160Hz.

Simply switched-on and idling, the smoother capacitors, designed for 100Hz ripple current, have this asymmetrical mid-audio frequency current pulse. When in use at domestic listening levels, the frequency will not differ much because there isn't much current demand.

Intuitively, we should see that the smoothers must draw the inverse ratio in current multiplication x load current. Therefore, even if the load current is tiny, the charge current might be 100 times greater. If the quiescent load of two amplifiers (a stereo pair) was 100mA, what is the asymmetrical pulse frequency, and what is the current in the pulse?

Through V(ripple)  = 0.1 x 0.01 / 0.01 we find it is 100mV, which conveniently coincides with the earlier calculation which gave 5800Hz.

The current required is 0.01 x 5800 = 58 times 100mA = 5.8 amps. What the capacitor sees is nothing for 98.3mS, and then 5.8 amps for 170uS.

Show me an electrolytic capacitor designed for such conditions, and I will build a conventional transformer supply. But it might not just be the capacitor. What about the reaction inside the transformer?

The workings in a switched-mode power supply must work similarly by the laws of physics, so are we merely kicking the can down the road? In some ways, we are.

Although we can obtain an SMPS which produces a stable output voltage from zero current to maximum, the SMPS will emit noise below a certain load current, which the above equations have hopefully explained.

To reiterate, at low load-current charge pulses get shorter, but the charge current is still significant. The shorter pulses relative to a typical SMPS operating frequency of 100kHz, make them considerably higher than 100kHz. The smoothing capacitors of an SMPS being rated for 100kHz ripple will be working outside their rating, and the massively asymmetrical waveform sets up a reaction in the SMPS transformer. The harmonics produced result in high-frequency mechanical vibration.

By placing a load on the SMPS output, the noise reduces, and at roughly 10% of its power rating, audibility ceases. It is wasteful of power, but you don't get anything for nothing.

Going back to the conventional power supply, we can see that the smoothing capacitors get less of a massively asymmetrical waveform by increasing the load. On full-load, the capacitors might fare better. A class-A amplifier can be considered a full load, and so a class-A amplifier might not suffer this problem. I am sure this is a contributing factor to class-A sound. Having said that, I must warn that pure class-A is not true class-A but high biased class-AB, and its power supply will suffer to a similar degree.

Graham Sir