1970s Design Indulgence

If you can justify and describe each part of a circuit's function in that it fits-in with real physics (and not pretend), then the circuit must perform properly, mustn't it?

Unfortunately, we sometimes find even that doesn't work, and it has to be because we haven't spotted a vital factor.

The essential factor here is the amplifier's use. When this is understood, we can then find the anomalies in the design. We can then also see how much we don't understand about amplifiers, perhaps?

We ought to understand these factors based on what we already know - the common sense - but marketing isn't about common sense.

Did you ever own a small portable radio? Perhaps you have a clock-radio? Operated at a comfortable volume, can you hear it in other parts of your home? I guess you can.

How many watts does the little radio output for you to hear it as described above? I will answer that for you; it is probably one-third of a watt, maxed-out.

Now, using common sense, what is the wattage you listen to using your stereo amplifier? It won't be much louder, and even if twice as loud, it will only be three watts for the equivalent of maxed-out.

After placing averaging meters on my prototype amplifier, I can report that for my 12'6 x 10' office, 1/3rd of a watt per channel is quite enough.

Now, transients of say five times the average amplitude exist. And the square power law means you'll need 25 times the power for them, and that means the amplifier must be able to deliver eight and a bit watts.

The clock-radio needn't do as much, because the signal being dealt with has been compressed by an "Optimod" so as not to overmodulate the airwaves. That is the thing that separates clock-radio from hi-fi.

Now, I have shown myself the optimum listening conditions in watts. Perhaps I need to look again at my amplifier design to see what it is doing at low-level rather than at rated power.

The Zobel network of 0.1uF and 10 ohms, starts to load the amplifier at 160kHz, meaning by 1.6MHz the load is the speaker and this 10 ohm resistor, and correct me if I'm wrong, but doesn't that make the load 4.44 ohms?

Even before 1.6MHz, the load is edging towards it. What is the beta at 1.6MHz? It is Ft/1.6MHz, which according to the datasheet typical performance graphs, is 5.

So, the drivers now see a 20 ohm load. Being 40MHz their beta being 25 at 1.6MHz, the VAS load is 500 ohms.

The VAS collector load, by bootstrap, is 1.2k. 1.2k driving 500 ohms results in attenuation. The bootstrap cannot override that.

In actual fact, the output stage gain (beta) cannot drive the 8 ohm load at that frequency, and it ought to be an output triple to get the beta up!

But, in this design, I am pushing for the maximum I can get out of the old "early 70s / late 60s" configuration.

Now, if we were using such as 2N3055H's we'd be a decade lower in frequency, but 0.1uF and 10 ohms was in use then, and it worked, but one wonders why, as it would only have been tickling it.

The Zobel is there in case the load gets unplugged, which means the amplifier phase margin and gain margin, runs out, and can oscillate under those conditions, causing damage to the transistors and smoking resistors.

But does it need to be so critical? It should be sized to save the amplifier - not to screw its performance.

The guru Self simply says it's a rule of thumb that always seems to work, so why change it? Obviously, this amplifier cannot drive it, for reasons set out above.

Simulation shows 0.047uF and 33 ohms will save the amplifier. It also shows a much improved phase margin.

Now to the output inductor. 5uH gives a break frequency of 255kHz with an 8 ohm load.

F break = 8 / 2pi 0.000005 = 254648 Hz

The usual thing is to bypass the inductor with a 10 ohm resistor. Therefore the amplifier will see a load of 18 ohms eventually, once past 255kHz.

I'll let you play about with the mathematical computations, but it means that my 4.4 ohms is slightly wrong - it might be nearer 6.

Even so, with falling beta, the load on the VAS is defeating its bootstrap, and the rate of fall is steeper than 90 degrees.

The resulting sound when the rate of fall exceeds 90 degrees is vile. It rings, and it is audible, and the audibility is due to distortion generated up high, modulating the highs we can hear. Eventually my hearing suffers.

This old design configuration also suffers the many other things I have written about. I have been able to follow most of these things on studdying commercial circuit diagrams through the era, and their solutions. Some of these solutions resulted in particular sounds, which some obviously liked, and others didn't.

My purpose here, is to try and make it sound right, and there is only one right, which is what you get by complying with the laws of physics. Being able to get at those laws of physics is the difficult part, and by going through the motions, it can be seen where other designers have thrown in the towel.

Edited by Graham Slee - 02 Oct 2020 at 6:39pm