Power Amplifiers: Introductory Comments

     Power amps are reviewed and rated within categories based on their design approach (e.g. single ended vs. push-pull, OTL vs. transformer coupled, tube vs. solid state, etc.). In this way, apples are compared to apples and not to oranges.
     A power amp has a complex task to perform, and the various amplifier design approaches have differing pros and cons as to how well they perform the diverse aspects of this complex task.
     Consider first the most obvious aspect of a power amp's task, putting out power. A power amp featuring a design approach with a class AB1 push-pull output stage can put out more power than one with a class A push-pull output stage (for a given cost level), and the latter in turn can often put out more power than one with a class A single ended output stage. Likewise, a push-pull tube amp with pentode or tetrode tubes in the output stage can put out more power (again, for a given cost level), than one with triode tubes in the output stage.
     Why not simply go for the design approach with the most power? Because a power amp is also supposed to sound transparent and clean and pure and neutral over a wide bandwidth. And design approaches with less power capability often tend to have intrinsic advantages in some of these other performance desiderata.
     For example, design approaches which are other than pure class A (over their full power range and into any load impedance) face the problem that some of their output devices will frequently turn off temporarily during the flow of the music signal, and then have to be turned on again. This poses a problem because most output devices evince time and amplitude distortions when they are first turned on, and in the region near turn-off where their signal is very small (further discussion in IAR Journal 1-2). The pure class A design approach does not face this problem because it keeps all output devices turned on and in their linear regions all the time, so it has a leg up in providing clean purity of sound over the class AB1 design approach, the tradeoff being that it can't provide as much power (for a given cost level). Note though that this problem can be specifically addressed and minimized by special engineering tactics in class AB1 designs (for instance the cross coupling employed by Audio Research power amps).
     Consider next tonal neutrality. A power amp is ideally supposed to behave like a voltage source, with zero source impedance. If the power amp's source impedance is higher than zero, then its frequency response into real world loudspeaker system loads won't be flat, but instead will vary in sympathy with the fluctuations in the loudspeaker system's load impedance curve. So the power amp will have a propensity for tonally colored frequency response into most loudspeaker systems, with the particular nature of the non-neutral tonal colorations depending on the nature of the impedance curve fluctuations (how wild they are, and at what frequencies they peak and dip), in the particular loudspeaker system you have chosen as your favorite. And the higher the power amp's source impedance is (perhaps as a function of frequency), the worse will be its propensity for tonally colored frequency response. Thus, power amps with higher source impedance will be worse tonal balance chameleons, assuming a different non-neutral tonal balance personality with each different loudspeaker system load impedance curve they encounter. Further discussion of this effect, together with measurements, can be found in Hotline xx. In this aspect, solid state power amps have the advantage, with the lowest source impedance. Transformer coupled push-pull tube amps are next best, and OTL tube amps are the poorest.
     Consider next clean purity, another obvious desideratum. Amp designs with push-pull or balanced topologies can achieve lower overall levels of distortion than single ended amps. On the other hand, they achieve this overall distortion reduction chiefly by canceling out even order (2nd, 4th, etc.) distortion byproducts, which leaves behind a disproportionate share of odd order (3rd, 5th, etc.) distortion byproducts. There are many schools of thought about distortion byproducts, with various psychoacoustic theories to support each. One school holds that it sounds less musically natural when odd order distortion byproducts predominate over even order, and that small amounts of odd order distortion, especially orders higher than 3rd (e.g. 5th, 7th, 9th, etc.), are less tolerable than larger amounts of even order distortion. Another school goes further, theorizing that the most natural sounding (benign, tolerable, euphonic) distortion is one in which the particular even and odd distortion byproducts are distributed in a progressive manner that simulates the amplitude distribution of musical instruments' natural harmonic overtones. And some psychoacoustic experiments have suggested that adding extra amounts of second order distortion can actually make music sound euphonically richer.
     Some of these distortion theories are used to explain or justify the musical charms of power amp designs featuring single ended output stages. With a single ended output stage, even order distortion (especially 2nd order) plays a large role in the distortion picture. The distortion in these amps consists of a large amount of 2nd order distortion, somewhat less 3rd order distortion, and a progressively declining distribution of both even and odd order byproducts above that. This simulates the natural harmonic overtone distribution of musical instruments, and perhaps also provides a richer sound by adding all that 2nd order distortion.
     On the other hand, as a matter of principle we all might want our amps to be neutral reproducers, not honey toned manipulators (see discussion of accuracy below).
     More importantly, with most single ended amps the overall distortion level is so high that it creates objectionable other problems, making the music sound grundgy, smeared, dirty, or fuzzy. Thus, although the single ended concept may indeed provide a type of distortion profile whose tonal balance sounds charming (albeit colored), the absolute level of this overall distortion is often so high that it imposes other undesirable (and highly audible) distortions on your music. We find most single ended power amps to sound objectionably grundgy, smeared, dirty, or fuzzy, due to their high overall distortion levels. There are only a few exceptions, and these single ended amps deserve special praise for bringing overall distortion problems down to acceptable levels (though still detectable, and still higher than push-pull or balanced designs).
     A power amp should have a power bandwidth covering the entire musical spectrum, which actually extends beyond the nominal 20-20,000 Hz range. Single ended power amps also face a special challenge in this arena. Their topology forces their output transformer to do some things which dictate some severe engineering tradeoff considerations, and these in turn usually limit the amp's power bandwidth capability, often to being less than even the minimal 20-20,000 Hz, rather than more as is actually required. Again, single ended amps that can deliver wide power bandwidth deserve special praise.
     Push-pull or balanced topology tube amps with output transformers can achieve much better power bandwidth, usually comfortably wider than the minimal 20-20,000 Hz (how much wider depends on choice of output tube, exact topology, transformer design and cost, etc.). OTL tube amps in principle could achieve even wider bandwidth, since they don't have an output transformer to limit bandwidth. But in practice, OTL tube amps may face problems in actually delivering power over this bandwidth, especially in the bass where high current is required, since OTL amps lack the impedance matching advantage of an output transformer (which effectively turns the high voltage coming out of a tube into the high current required by most speakers, especially for bass drive).
     You might think that solid state power amps can deliver the widest power bandwidth of all. However, most of them achieve this only by engaging in what could be viewed as a kind of cheating. They start out by using solid state devices which have a low (narrow) inherent (open loop) bandwidth, with very poor high frequency response (perhaps only up to 2 Hz, or 20 Hz). They then employ large amounts of negative feedback to extend this bandwidth to adequately cover the musical spectrum (or well beyond if they choose). The trouble is that this spectral extension is achieved via feedback rather than directly, and this feedback brings with it doses of ugly sounding high order odd order distortion byproducts, as well as other ancillary problems (transient and phase discrepancies, susceptibility to transient overload, vulnerability to disturbances at the output port [e.g. RFI, reactive speaker kickbacks, etc.]). The more a solid state amp seeks to extend its bandwidth via feedback, the worse these troubles can become.

back to table of contents