1970s Design Indulgence

I am not going to criticise anybody for not challenging me about the frequency equivalents of the time periods. Instead, I will chastise myself by pointing out my error.

We have been discussing conduction angle in the form of time t in relation to the rectified mains waveform.

By simply inverting t, I gave the equivalent frequency, now obviously, in a rectified system, that is wrong (but also right as we will come to).

A t of 100uS inverted gives the number 10,000. That being a positive pulse, when we add its negative pulse to produce a full cycle, it must be 5,000 cycles per second. Therefore, when I stated 5800Hz, it should have been 2900Hz.

Now, a capacitor might be ripple current rated for 100Hz, but that ripple current is from a rectified 50Hz. Surely the rating should be for 50Hz?

And therefore, a rating at 20kHz should mean a rectified 10kHz. Therefore, I have just stated that it was right too.

Capacitor ripple current ratings are useless to a degree. Their only use is where the ripple voltage is close or equal to the HT voltage, where the conduction angle is close to half the frequency being rectified (do you follow?).

Can we therefore establish a smoother capacitor's lifetime? I fear the answer is no.

Assuming we now have a "perfect" amplifier through the DC being good at all signal frequencies, we discover what should be obvious - that the whole thing gets hot!

After reading this, you might think I'm making it all up to sell more Proprius amps, but I'm afraid to say I'm not at all that clever.

Amplifier thermals need to be carefully considered. Quiescent conditions are established to obtain the amplifier's best performance, not for the amplifier's quiescent conditions to suit the amplifier's temperature.

If we put a heater of some sort in the case with the amplifier, the case will become hotter than if it only contained the amplifier. The amplifier will then run hotter. Because it is a bipolar junction transistor amplifier, its output stage draws more current due to thermal runaway behaviour.

Even if the heatsinks are large enough to survive apocalyptic thermal runaway, for the case/heatsink assembly to operate at a safe temperature, the temperature rise must be countered.

Ambient temperatures of operation are a nightmare of sorts for the designer to try and workaround, but what if we put a heater in the case as well?

The amplifier uses thermal "feedback," in which a heatsink mounted sensor opposes the increasing current, which the positive temperature coefficient of the output stage gives rise to.

This is supposed to ensure the optimal quiescent setting, which dictates the low volume distortion is maintained so that performance doesn't differ.

Suppose we install a heater that doesn't have any thermostatic control. In that case, the amplifier's thermal see-saw will need to cater for that as well.

Perhaps you cannot think of an occasion of putting a heater inside the amplifier case, here is a suggestion: its power supply!

The transformer is made of iron (or particular steel to be more accurate), and iron has low thermal conductivity. At switch-on, the transformer is cold (at ambient room temperature), but it produces heat with work.

The heat eventually spreads out to all parts of the case, including the amplifier output stage, mounted on the heatsink(s).

Now, what is the optimum quiescent setting of output stage bias? Should it be set at switch-on, or should it be set once the entire case has reached its thermal equilibrium?

But what if the case cannot take the temperature rise of both the heater and the output stage? The amplifier quiescent bias must then be fudged to assist the "equilibrium."

If we also added another heater to bleed-off current to widen the power supply reservoir capacitor conduction angle, we have worsened that problem.

Perhaps we must consider an outboard power supply (as per the Proprius). Otherwise, we must consider a thing called ventilation?


Here, the do-it-yourselfer is disadvantaged, as the amount of casework required becomes prohibitive.

To look at another design where all the information is available, I am studying Bob Cordell's Designing Audio Power Amplifiers book. He, too, argues for the 25mV emitter resistor voltage, which is the maximum allowable before gm-doubling. I note what seems to be a preference for 0.22-ohm emitter resistors - this sets an output stage current of 114mA.

The rest of the amplifier consumes around 20mA for the drivers, 10mA for the VAS, and let's guess 5mA for the IPS.

The total current is 298mA per stereo pair, and for the sake of roundness, we can say 300mA.

The Cordell power supply illustrations include bleeder resistors across reservoir capacitors, value 2k ohms, and on a voltage rail of 55 volts, consume 27.5mA.

However, when it comes to working-out the conduction angle, we are met with a surprise in that the value of capacitance he uses is huge! 50,000uF or 2 x 25,000uF are the values illustrated.

Running this through V ripple = I T / C , we find for 325mA and on 60Hz USA mains, 0.325 x 0.0083 / 0.05 = 0.054 V, or just 54mV ripple voltage.

To obtain the conduction angle time, we again use cos-1 (HT - V ripple / HT) x 60Hz time constant, which is 0.0443 x 0.00265 = 0.000117, or 117 uS.

Straight away, the opposite direction can be seen. The charge current pulses are of shorter duration than Self's and are half that of my design (excluding the bleed resistors). 117uS per 0.0083S is 1.4% of the time, where on 50Hz mains, I had 200uS per 0.01S, which is 2% of the time.

The current drawn on idle is 325mA, which means a 23 amp ripple spike. On my design, the current drawn was 140mA, which meant a 7 amp ripple spike.

I can only conclude that with 50,000uF capacitance, the ripple current might match that available in a capacitor's specification. The use of two Kemet 22,000uF ALF20 capacitors in parallel indicates 22.68 amps availability at 10kHz, with a minimum lifetime of 18,000 hours.

To date, I have not found any designer pay much attention to this phenomenon. Therefore, I can only deduce that Self and Cordell work on an "it's always been like that" basis, "so why change it?"

I have only found one designer who pays attention to the conduction angle, that being Morgan Jones.