Links  

     AJHorn
     LTspice
     FreePCB
     LibreCad

     EAC
     SoX
     Foobar
     Mp3tag
     Audacity

     diyAudio

Amplifiers

Introduction

All seems simpler with amplifiers compared to loudspeakers. Amplifiers have a flatter frequency response and fewer distortions than the best transducers or loudspeaker. So what is there to worry about?

In amplifiers, the correspondence between total harmonic distortion (THD) and what you hear is almost zero. The problem is not the listener, but the method of measurement and its interpretation.

Naked Facts

Simple distortion data THD ignores all of the complex behaviour of the upper harmonics and their fall-off under complex operational conditions. The behaviour of these harmonics makes the differences in tonal balance of various amplifiers with identical frequency responses and THD data. THD measuring instruments simply add all the data together, paying no attention to the order of distortion.

The widespread use of negative-feedback (NFB) reduces all of the harmonics in direct proportion to the NFB ratio, but unfortunately adds a lot of new harmonics of its own. In practice it replaces large percentages of 2nd and 3rd harmonics with small percentages of many high-order harmonics. Therefore, NFB has no general potential to improve the proportion of harmonics in amplifiers and the naked facts are:

  • Simple single ended circuits generate 2nd and 4th order harmonics
  • Push-pull operation adds 3rd, 5th, 7th and 9th order harmonics
  • Feedback may generate a series of harmonics up to the 80st order
  • Reactive loads to circuits add a frequency dependency to the harmonics
        (Analysed by Norman Crowhurst, reprinted in Glass Audio, Volume 6-7)

Another source of distortion is the EMI noise (electro-magnetic interference) generated by power supplies. They are responsible for a significant coloration due to the adding of high order harmonics in amplifier circuits. This is not so bad for pure Class A amplifier circuits, since the current demand is constant. For Class AB circuits however this is a disaster. The current demand for Class AB amplifiers always fluctuates and this means the noise spectra and harmonics of the power supply are always changing with the sound too.

Interpreting the Facts

So, how does all this relate to real music? Real music is dominated by many closely-spaced tones of which a choir or violins have the most dense spectra of all. It has been already shown in 1950 by D.E.L. Shorter, BBC that with a few closely-spaced fundamental tones, intermodulation (IM), sum-and-difference sidebands exceed the simpler harmonic series. As the number of tones increases, the number of IM sidebands increases at a much faster rate than the simple harmonics (2nd and 3rd order harmonics).

Therefore, amplifier distortion involving a significant number of upper harmonics has far more audible effects on reproduced music than lower harmonics (2nd and 3rd order harmonics) which, in other words, lower harmonics are potentially less audible (although they dominate almost any simple THD measurement)!

Design Considerations

Although all issues described above are important, most of them cannot be fully overcome. Nevertheless, some general considerations will help to keep problems within limits:

  • Continuous rise of total harmonic distortion curve versus output power with small variations within frequency
  • Continuous fall of even and odd order harmonics over a wide range of signal levels
  • Harmonics around 6th order and higher should ideally fall below the amplifier's noise-floor
  • Soft clipping characteristic of amplifier when driven into saturation

Meeting these criteria more or less means in practice including the following items in an amplifier design:

  • No negative feedback (NFB) or if inevitable, a local NFB ratio of no more than 20db (on an amplifier which already is as linear as possible)
  • Amplifier topology as simple as possible, but not simpler!
  • Single Ended or Class A push-pull operation
  • Stable power supply with (almost) no EMI noise and fluctuation according to the signal

Unfortunately, the difficulty here is determining how much deviation of any of these four items is tolerable.