About Power Amplifiers
My view on Power Amplifiers has changed during the years. My work with loudspeakers is the main reason. Long ago, with passive crossovers, I liked the mono (single cabinet) approach, (I still do!!) mainly because I felt that short wires to the woofers in particular was important, and that 2 power amps in the same cabinet (and maybe with the same power supply) would compete with each other in terms of voltage and current. This is still correct, but today, I go a couple of steps further. The main reason for this is that requirements for woofer, midrange and tweeter are very, very different, both from an electrical AND listening viewpoint. I will try to explain this, and as always, this is my personal opinion. I am not saying that my solutions are perfect for everyone, I just want to explain things from my present viewpoint.
Let's look at the different loudspeaker units, one by one.
Woofer: A woofer needs power, meaning both voltage and current, and the last may prove seriously important. Often the amplifier looks into a quite complex impedance, this is the case with many bass reflex systems. We combine with a big woofer, large voice coil with maybe 4 layers presenting a solid inductance. The resulting impedance curve has 2 peaks, and steep slopes with a phase difference between voltage and current. This places heavy demands on the amplifier. Damping is another issue – any sort of resistance between amplifier output and the speaker degrades damping, meaning less control of woofer impulse behavior – can easily be heard and measured. Moving on to transistors – power transistors have to deliver lots of current at substantial supply voltage level, they are typically quite big, cooling area is important. They are also quite slow (compared to small signal devices) and have lots of capacity here and there. Gain (beta) figures are often low, placing serious demands on the driving circuit. Linearity and complementary data are not exactly top notch. As a consequence, to reduce distortion, power amplifiers are/were often designed with large amounts of global feedback – good for steady-state signals, but not always good for transients and over-all stability versus complex loads.
Depending on room size and listening habits, we are talking about 100 – 200 Watts of power into 8 ohms, and preferably doubling up into 4 ohms – the load presented by the woofer + cabinet can easily swing from 3 ohms to 50 or even more, all this inside a very limited frequency window.
What about listening? Humans are not very sensitive to bass – just take a look on the Fletcher-Munson curves. We are more sensitive to transients, they go much higher in frequency, and we are able to detect short-time events in music. Distortion in the bass area is not critical, especially if (and this is often the case with woofers) we are dealing with even-order distortion components.
And, with high loudness levels, the “internal” distortion generated by our ears and related nerve systems, tends to rise.
A short list of properties for a good woofer power amplifier:
Very low output impedance
Stability to cope with complex loads
Speed – meaning bandwidth / slew rate: moderate – often the woofer is only used between 10 – 500 Hz, mostly sine-wave type signals – fast transients are distributed to midrange and tweeter
Distortion: moderate demands – but linearity is important – distortion should stay within reasonable figures, also at high power output levels.
- - -
Midrange is for me a very different world. Midrange speakers are typically located in a closed cabinet, one impedance peak only, and seldom with any extreme size. A typical case is a gentle rise with a peak height of maybe 10 to 20 ohms for an 8 ohm midrange speaker. Power level is clearly lower than for a woofer, due to the energy content distribution in music. Damping is inherently better in a closed cabinet compared to a bass reflex cabinet – and yes – I have measured!!
Transistors for midrange amps: smaller types due to reduced power demands, lower capacity, faster, better complementary properties, higher beta and so on.
Distortion: This is critical!! We are able to detect very small amounts of distortion in the midrange area, again – look at the Fletcher-Munson curves. Speaking evolutionary, all scientists agree that the human language is the most contributing factor to our hearing sense, and we can detect the smallest of change in terms of frequency, phase, stereo image / placing and so on. Distortion must therefore be VERY low in a midrange amplifier. Bandwidth and slew-rate is also important, together with intermodulation distortion and transient behavior. So, all in all:
Moderate power needs due to easier load demands – 50 to 100 Watts max depending again on room size and listening habits.
Low output impedance
Very high stability and linearity
Good bandwidth and slew rate
VERY low distortion of any type
Low noise – high signal to noise ratio, noise is easily audible in midrange speakers
- - -
Many of the things said about midrange also applies for the tweeter amplifier. We just turn the power demands a bit down (most tweeters are an even easier load compared to midrange, smaller voice coil, and also typically closed-cabinet type) and speed demands a bit up.
Distortion is still important, but to a fraction less degree than for the midrange, partly depending on the intended frequency range for the tweeter: a low crossover frequency like 2000 Hz increases distortion demands compared to a 5000 Hz crossover point – for me, again, the Fletcher-Munson curves are important. Or, to put things into another perspective – try and see, if you can hear the difference between a 12 Khz sine wave and a 12Khz square wave – note that the first overtone of the square wave is found above 20 Khz.....................
Tweeter voice coils are small, impedance load is easy, with some ribbon types with flat membranes it resembles a simple resistance, not hard to drive. But, control is still important – low output impedance is still fine!!
Transistors: even smaller and faster, more delicate, lower power demands – could be 20 – 50 W, maybe 2 or 3 in parallel – distributes power and allows you to use smaller and faster and sometimes better devices.
To sum it up:
Low output impedance
Very high stability and linearity – no high frequency tendency to oscillate!!
Bandwidth and slew rate is very important
Very low distortion
Low noise – high signal to noise ratio – noise is VERY easy to hear
- - -
As you can see, my choice today is clear: one amplifier for each loudspeaker type. It makes sense in many ways, as outlined above. When it comes to practical building, advantages remain. Take for instance the power supply. For a well-grown woofer power amp, transformer and capacitors are normally big, and since power demands are high, a mono amp solution is the best, you do not want heavy demands in one channel to have influence on performance in the other channel!
But you do not need a 500 V/A transformer to drive a tweeter!! A tweeter amplifier is not happy with low-frequency kickback from a woofer, a midrange amp is not happy with sudden current surges from the woofer amp, when using a common PSU setup. And so on. During all my years of listening, and experimenting with amplifiers, I have only encountered sound improvement, when I switched from single amp amplification to dual amps – on a 2-way system, also with a passive crossover. Of course, you have to split the crossover in 2, and make a gain match on the amplifiers you use – I am fairly hysterical about this – gain should be accurate within millivolts – measured on the same load.
- - -
Components for Power Amplifiers.
During the years, I have used more or less everything – valves, transistors, meaning bipolar, fet's, mosfet's and IC:s, sometimes in one single amplifier. Today, my focus is on the job that one particular amp is supposed to do. For woofers, power IS important – meaning a large transformer and a solid bank of powerful capacitors. Circuit performance is crucial, expensive components can never fix a bad basic construction. All parts must work well together, components must work well inside their SOA, and tolerances tight when required. I do not like hot transistors – prefer working temperature to stay below 50 – 60 deg. Celsius. Don't underestimate cooling requirements – power transistors are often expensive, and if and when they get fried, they can create serious injury to other components, speakers and everything else connected to the amp. No fun. As DIY, you can allow yourself good components, that ensures a long and healthy life for your amps.
Bipolar or MOS or IGBT power devices?? Depends on the job, today I find it increasingly difficult to find first class audio components, very much is switch or automotive or power related, but not audio – annoying. Bipolar transistors has fine complementary properties, but need base currant. A MOSFET needs less currant (but observe internal capacitors – must be charged/discharged) but complementary properties are often not impressive. IGBT – hard to find audio types today, at least for me. Some people talk about mosfet sound, or class A sound, in my opinion this is a simplification. Running a bad circuitry with lots of distortion in class A does not make it better – it just gets warmer!! During my years with testing equipment for High Fidelity Magazine, I had the privilege to test some seriously expensive gear. I could not see anything that indicated that, for instance, mosfet amps sounded better (or worse) than bipolar amps. Think instead of the job the transistor is supposed to do, sometimes a FET is the obvious solution, sometimes a bipolar, or an IC. Today, I am happy to work with IC circuitry, some of them (like the LM4562 in my SOTA LJ Line Amp) are really close to perfect – can't hear them, can't measure that they are there.
Free PC calculation software can be a great help – see TINA from Texas Industry or MicroCap from Spectrum Software – once expensive – now free software. Naturally, they do not match today's expensive professional packages, but checking a circuit for voltages, currents and power can save lots of time, instead of trying, soldering and de-soldering (burnt) components. Power amps are DIFFICULT, compared to pre-amplifiers, everything is so to say stepped-up, placing heavy demands on circuit layout and grounding – be very careful with PCB layout. Destructive self-oscillation is another matter – at some point you will need an oscilloscope to check this. Check the thermal condition for ALL transistors, sometimes it is an advantage to put transistors on the same heat-sink, reducing DC drift and maybe risk of thermal runaway.
There will be more specific information with each power amp in their respective section.