The DIY Chronicles, Part Two
As I mentioned in my last article, I decided to concentrate on an analog approach to my amplifier project. Digital had many advantages, but amplifer design is not like designing computer chips. Sound quality was my final goal, and only a few digital amps are of high-end class, even in the year 2000. The main reason I (briefly) gave up the digital way was a money problem. The simulation software cost an arm and a leg in those days, so I built all the prototypes I designed—a lot of time, money, and disappointments!
Analog stuff was a bit easier to design because fewer components were involved. The prototype boards were quicker to build, the results quicker to get, and the components much nicer to burn! I can tell you that now I immediately recognize the subtle smells of transistors, resistors, capacitors, and so on.
Before going ahead, I think it would help to refresh the reader about some often misleading concepts: negative feedback and balanced technology. The goal is not to confuse you with technical terms, but to help you understand.
Negative feedback consists of inverting the phase of a portion of an amplifier's output signal and connecting it to the amplifier's input. Global negative feedback is usually high in level, and the result is very low THD. However, the resulting sound is often described as "cold," "hard," or even "transistor-like" compared to the softer sound of tubes, which usually use smaller amounts of feedback, even when it is global feedback.
Local negative feedback, as used by Matti Otala 30 years ago, consists of applying a small amount of negative feedback in a local loop around each stage of an amplifier. The result is much better control over the signal transients. This greatly removes the "cold," "hard" sound associated with transistor gear. Today, most amplifiers use local negative feedback.
A third type of negative feedback is the degeneration associated with every active device and can't be avoided. Every tube or transistor is generally connected to the power supply via resistors. However small, the voltage drop across this resistor is of inverted polarity to that of the current flow going through the device, tube or transistor. This voltage drop is a kind of negative feedback because it acts to oppose the amplifying action of the device. It is called "degenerative" because of its origin: inverted polarity.
No amplifier manufacturer in the real world can claim that no feedback at all is used from input to output. In practice, only global feedback can be entirely avoided, while local feedback can be partially omitted in particular stages (eg, output stages). Any electronics engineer will tell you exactly the same thing—no matter what the manufacturers claim in their ads.
Balanced or Symmetrical?
These terms are often confused because . . . they're confusing. But an explanation is easier than you might think.
In "balanced" operation, because the signal is amplified in both phase directions simultaneously, you need a third wire to carry the second signal. One wire, called "hot," carries the "normal" signal, as usual. The second wire, called "cold," carries the same signal as the "hot" one, but inverted in phase. The third wire is the ground reference.
As balanced cables transport simultaneous positive and negative signals, it's easy to amplify the difference between them: Positive ("hot") minus Negative ("cold") gives a signal twice the amplitude of the positive one. It's easy to understand: If the signal is, say, 2 volts for the hot side, it's –2V for the cold side. So, 2 minus –2 gives 4.
But here's the magical part: If a parasitic click or a magnetic hum or RFI is present around the cable, it will be picked up equally by the whole cable, in both the "hot" and "cold" sides. When you take the difference between both, the noise is canceled! It's easy to understand: If the noise is 1V, it will be 1V for both the "hot" and "cold" sides. (The noise doesn't "know" which wire is "hot," which "cold.") So: 1V minus 1V gives 0V, or signal but no noise.
Balanced designs are mainly used in professional studios, where very big lengths of cable are used. A lot of high-end gear uses the balanced topology.
In "symmetrical" operation, half of the signal is handled by the positive portion of circuit, half by the negative one. Transistors amps these days commonly use complementary devices (NPN and PNP transistors) and symmetrical (plus and minus) power supplies. As tubes are of one kind only (equivalent to the NPN transistor), their circuits are generally asymmetrical—or, in my opinion, mislabeled single-ended (another story).
Next week: Hervé Delétraz decides exactly what type of amp he'd like to build. Readers can contact Mr. Delétraz via e-mail at firstname.lastname@example.org
Below are some drawings that are not actually functional in the real world but that explain the growing complexity when we try to fully balance a signal. I think that fig.2 is the best compromise, and gives the best-sounding result in power applications. Fig.1 is the absolute theoretical best—ie, the only true class-A topology—but requires too many compromises in practice: such as bandwidth vs efficiency vs impedance matching vs stability, and so on.—Hervé
Fig.1: Single-ended, asymmetrical topology
Fig.2: Single-ended, symmetrical topology
Fig.3: Balanced, symmetrical topology