Drawbacks of Class A push pull audio power amplifier?

Question

I am willing to develop an Audio power Amplifier myself in a Push Pull configuration of two complementary transistors (NPN,PNP).

On the internet, it is found that most of the conventional Class AB amplifiers are designed in Common Collector configuration, hereby described as CC. The CC configuration is known for the amplification of the current while keeping the voltage either same or below the base input, which seems to be a major drawback to me because this is not how one should harness the full power. CC is also known for driving a load of Lower Impedance from an input signal of Higher Impedance.

To harness most of the power, I have developed my own amplifier circuit in Common Emitter or CE configured as a push-pull circuit like this...

^{simulate this circuit – Schematic created using CircuitLab}

In the circuit, the transistors are configured as Class A or Common Emitter Push Pull Amplifier. I have also played this amplifier and this is quite loud and distortion free.

This circuit is not found anywhere on the internet. As a new approach, I am quite unaware of the drawbacks of this circuit in the long run. So I thought that it would be better to discuss the long term drawbacks (if any) in this forum and why this configuration is not used in industry grade audio power amplifiers?

Thanks and regards.

Answer 1

Please notice you have already drawn the circuit into a simulator so you can just simulate it yourself how suitable it is for any purpose. Please notice the simulator is ideal and any real-world non-idealities you must add yourself.

This circuit is not seen much because usually when transistor circuits are introduced, their properties and applications are discussed there, before going to more complex circuits like these kind of simple audio amplifiers.

One such rule of thumb is that you should never build a circuit which depends on transistor current gain, i.e. Hfe or beta, because due to manufacturing tolerances. A transistor is only guaranteed to span a specific range like 100 to 400 of gain, so even if you buy 1000 transistors, you are not going to find two transistors that have Hfe matching within given tolerance to be able to build this circuit.

To be useful, the circuit must have 0V output with 0V input, as speakers don't tolerate DC and there is no AC coupling on the output. Maybe a few millivolts is acceptable tolerance.

And because the transistors are run in open-loop, and biased with just a resistor between their bases, the output is unknown.

As the PNP and NPN transistors have identical base current, they must have identical current gain so that they have matching collector current and there will be matching voltage over both transistors so that output is 0V.

Both transistors have 80..160 range for Hfe, so average is 120, or approximately 100 as a rule of thumb.

Assuming they have 0.5V Vbe, there's 35V over the 10k resistor and both have 3.5mA of base current. Now, it means that the idle bias collector current must be 350mA. The circuit wastes 6 watts per transitor doing nothing, assuming identical current gain.

But the current gains could be 119 and 121. The other transistor would try to pass 416mA through it, but the other would try to pass 423mA. That means 7mA must be going into speaker, or 56mV into 8 ohm speaker.

Please note that I only covered how the putput will vary according to transistor beta as an example. There are also other things that should be considered how poor the output will be if one or both supply voltages are non-ideal, how the speaker impedance affects it, how the transistor Vbe non-idealities might affect it. And then the AC signal analysis.

So the circuit is not very useful as a practical circuit. The circuit is usually found in the examples of poor ciruits or as a task to analyze why some circuit is not very good.

Answer 2

Class A: In idle state there's DC-current = half of the peak current or more through the transistors.

Push-Pull: The nonlinearity is compensated by having 2 devices with opposite nonlinearities feeding the same load.

Both of the concepts may be fulfilled to some degree if this works without smoking and bad distortion.

The circuit is not common, because it has no control of the DC current directly through the output transistors. You may say: There is, R1 is selected for good DC idle current.

It isn't. Transistors may start a thermal runaway because there's no feedback to stabilize the DC operating point. Besides transistor individuals have different current gains. You must select a close enough NPN-PNP pair. The selection is needed also for good non-linearity compensation by the push-pull operation.

Predicting the gain of the output stage is difficult. It depends radically on the used transistors. Feedback stabilization of the gain is essential for successful high yield industrial production. A hobby circuit can work well when the total gain including the preamp stage too, is good.

Some hifi-gurus preferred the class A, no feedback-idea for audio amplification because early transistors were so slow that the common feedback stabilized (see NOTE1) class AB amp, which worked fine with sinusoidal input 1 kHz or less could create bad distortion at 10kHz. With practical signals transient sounds (for ex drums) became muddy. Your output stage doesn't have that harm. But it can be well neutralized by the fact that you have a normal feedback stabilized stage in front of it. A modern high speed opamp should be used. In that sense your opamp is quite good. It unfortunately can be a little overloaded by the 2 parallel (for AC) transistor BE joints and that can increase the distortion. A series resistor between the opamp stage and the transistor stage could help.

NOTE1: The hifi gurus didn't blame DC operating point stabilization with feedback and even gain stabilization feedback was acceptable if it happened inside an amp stage. But gain stabilization feedback over several amp stages was seen crap because the intermediate stages could easily get saturated when the slowness of the later stages prevented fast enough output voltage changes.

Not asked, but you may have "look mama, I built this! Listen how beautifully it sounds!" -syndrome. I have had it. I was so proud of my 1st working radio receiver that I was totally deaf for its awful distortion. Others asked me to shut it down immediately. I thought the others were only dumb or evil.

Answer 3

I found a thread on dyiAudio that discusses this. User ubergeeknz asks:

In most older (pre 1980) designs that I've seen, mostly British or British based designs, the output stage is either a totem pole arrangement or in the later designs, some combination of complementary szilikai/Darlington arrangement, but where the outputs are complementary they tend to be common emitter with the collectors driving the load. But at some point, it seems consensus shifted towards common collector complementary output stages in class AB amps. Can anyone explain the rationale behind both arrangements? And why the preferred arrangement seems to have changed over at some point in the late 70s/early 80s?

User steveu answers:

Speed/bandwidth. Power transistor are slow, or at least historically they were, and still are relatively slow. In order to make a stable amplifier with a predictable gain, you need feedback and in order for that feedback to be stable, the amplifier has to be "compensated", ie the slowest stage has to dominate the total phase lag so that the gain falls to zero dB (unity) before the total phase lag reaches 180 degrees and makes the negative feedback into positive feedback. Common collector / emitter follower stages are driven by a low impedance voltage source that provides more current at higher frequencies and therefore can drive the power transistor on and off faster than a current source and with less phase lag at any given frequency. Speed translates into less distortion because the feedback controls the output at higher frequencies, ie more closed loop gain at that frequency. Also, a follower has a local feedback and low impedance so the amplifier output buffering is faster and less dependant on the global feedback. Note that a Szilikai pair requires a Zobel network to stabilize the local feedback, and a follower does not. Note that older amplifiers used "quasi-complimentary" outputs (one side using a Szilikai pair) because power PNP transistors were rare and expensive.

Your design doesn't seem to have any feedback, which would make the final gain hard to control, especially with two different transistors.

Answer 4

I was looking for a quick reference to the push-pull design and stumbled upon this post. Couldn't help but notice that your design isn't a true push-pull amplifier.

Consider that you want to drive a very powerful speaker (nominal impedance is <10 Ohms) but your starting point is an op-amp with high-impedance output (>1k Ohms). The point of making a push-pull between the two is that both transistors are connected in the common collector configuration and thus, they provide very small output resistance (~1/hfe or ~1/hfe^2 with Darlington design) and near-to-unity voltage gain (because the b-e junction is in forward high-injection). On a minor side, through careful (but achievable - see below) biasing, you can remove any non-zero DC bias at the output port, which will destroy the speaker in the long run by heating it up constantly; and, you can remove the cross-over distortion due to the required base-emitter voltage drop. Overall, a push-pull design provides high driving capability without a transformer, which is bulky and has a pass-band limit.

In a two-stage amplifier with an op-amp providing voltage gain (high output impedance, nearly zero output current) and a push-pull giving current gain (low output impedance, unity voltage gain), you can use the final stage (push-pull) output as the feedback to the op-amp, thus making the output stay at zero all the time. To bias the base nodes, you can use a pair of diode-connected PNP/NPN transistors of the same part number; Further optimization can be done to give you a true class-AB amplifier.

In your circuit, the two transistors are connected in common emitter configuration. This is not getting to the point of a class-AB amplifier. The output resistance of a common-emitter amplifier is rather high (indeed the "better" the transistor, the higher the output resistance, as it's determined by the base width modulation which is more significant when a higher hfe is desired). It cannot be used to drive a low-impedance load, or at least with decent efficiency. Since you have heard it loud and distortion-free, I suspect that your load isn't powerful enough to expose the limitation of a common-emitter design and thus warrant a push-pull. Try a 4-Ohm or 8-Ohm power resistor instead, and use an oscilloscope to look at quantitative measures, spectrum & distortion as such.

Furthermore, when the two collectors are tied, there is no guarantee that this node stays at 0V at rest (quiescent point). First of all, the collector voltage in forward active mode is nearly constant versus current, and so a small change in the biasing resistor will cause a large shift in output DC level. Compounded with process variation (extrinsic factor) and electron-hole mobility difference (intrinsic factor), it is nearly impossible to reliably achieve 0V DC bias. If you feed it back to the op-amp stage, which is feasible in the previous case, it may require an op-amp output beyond the voltage rail. With the non-zero quiescent voltage, once you connect the load, a constant DC current will be heating it up. Even if it's not that extreme, your total voltage swing without distortion, and therefore maximum output power, is limited due to the reduced headroom. These process-dependent results may not be easily observed in academic/theoretical simulations, as all models/parameters are too idealized to reflect the reality. So, just use an oscilloscope.

There are practical use cases for tying up the collectors. Actually, it's more common nowadays to use CMOS transistors. One side serves as the actual amplifier and the other side provides high impedance (from the drain port) with a relatively low drain-source voltage drop (lower than a transistor-resistor design with comparable voltage gain).

In other answers, @oneprivate and @Adam Haun mentioned power transistors being slow, at least in the past. I'm not knowledgeable enough in audio amps to understand the historical reasons for these design choices, but IMHO you can improve the frequency response by transfer function engineering. Power transistors are generally "slow" as the input-output coupling capacitance is generally large, which is further amplified by the Miller effect (forming an additional low-pass/shunt capacitor at the input port). So I guess .. you will want to separate the voltage and current amplifying stages?