Technical specifications and detailed information on the valve audio amplifier, including its development history.

Circuitry and performance Characteristics of valves

The characteristics of valves as gain devices have direct implications for their use as audio amplifiers, notably that power amplifiers need output transformers (OPTs) to translate a high-output-impedance high-voltage low-current signal into a lower-voltage high-current signal needed to drive modern low-impedance loudspeakers (cf. transistors and FETs which are relatively low voltage devices but able to carry large currents directly).

Another consequence is that since the output of one stage is often at ~100 V offset from the input of the next stage, direct coupling is normally not possible and stages need to be coupled using a capacitor or transformer. Capacitors have little effect on the performance of amplifiers. Interstage transformer coupling is a source of distortion and phase shift, and was avoided from the 1940s for high-quality applications; transformers also add cost, bulk, and weight.

Basic circuits The following circuits are simplified conceptual circuits only, real world circuits also require a smoothed or regulated power supply, heater for the filaments (the details depending on if the selected valve types are directly or indirectly heated), and the cathode resistors are often bypassed, etc.

The common cathode gain stage

The basic gain stage for a valve amplifier is the auto-biased common cathode stage, in which an anode resistor, the valve, and a cathode resistor form a potential divider across the supply rails. The resistance of the valve varies as a function of the voltage on the grid, relative to the voltage on the cathode.

In the auto-bias configuration, the "operating point" is obtained by setting DC potential of the input grid at zero volts relative to ground via a high-value "grid leak" resistor. The anode current is set by the value of the grid voltage relative to the cathode and this voltage is now dependent upon the value of the resistance selected for the cathode branch of the circuit.

The anode resistor acts as the load for the circuit and is typically order of 3-4 times the anode resistance of the valve type in use. The output from the circuit is the voltage at the junction between the anode and anode resistor. This output varies relative to changes in the input voltage and is a function of the voltage amplification of the valve "mu" and the values chosen for the various circuit elements.

Almost all audio preamplifier circuits are built using cascaded common cathode stages.

The signal is usually coupled from stage to stage via a coupling capacitor or a transformer, although direct coupling is done in unusual cases.

The cathode resistor may or may not be bypassed with a capacitor. Feedback may also be applied to the cathode resistor.

The single-ended triode (SET) power amplifier

A simple SET power amplifier can be constructed by cascading two stages, using an output transformer as the load.

Differential stages

Two triodes with the cathodes coupled together to form a differential pair. This stage has the ability to cancel common mode (equal on both inputs) signals, and if operated in class A also has the merit of having the ability to largely reject any supply variations (since they affect both sides of the differential stage equally), and conversely the total current drawn by the stage is almost constant (if one side draws more instantaneously the other draws less), resulting in minimal variation in the supply rail sag, and this possibly also interstage distortion.

Two power valves (may be triodes or tetrodes) being differentially driven to form a push–pull output stage, driving a push–pull transformer load. This output stage makes much better use of the transformer core than the single-ended output stage.

The long-tail pair

A long tail is a constant current (CC) load as the shared cathode feed to a differential pair. In theory the more constant current linearises the differential stage.

The CC may be approximated by a resistor dropping a large voltage, or may be generated by an active circuit (either valve, transistor or FET based)

The long-tail pair can also be used as a phase splitter. It is often used in guitar amplifiers (where it is referred to as the "phase inverter") to drive the power section.

The concertina phase splitter As an alternate to the long-tail pair, the concertina uses a single triode as a variable resistance within a potential divider formed by Ra and Rk either side of the valve. The result is that the voltage at the anode swings exactly and opposite to the voltage at the cathode, giving a perfectly balanced phase split. the disadvantage of this stage (cf the differential long-tail pair) is that it does not give any gain. Using a double triode (typically octal or noval) to form a SET input buffer (giving gain) to then feed a concertina phase splitter is a classic push–pull front end, typically followed by a driver (triode) and (triode or pentode) output stage (in ultra linear in many cases) to form the classic push–pull amplifier circuit.

The push–pull power amplifier

The push–pull output circuit shown is a simplified variation of the Williamson topology, which comprises four stages: a SET input stage to buffer the input and give some voltage gain. a phase splitter, usually of the cathodyne or "concertina" type. This generates exactly equal but opposite drive signals for the following push–pull circuitry, but gives no gain. Note that as shown, the Williamson topology concertina phase splitter is direct coupled (with a resistor) to the input stage. This requires careful design of the input stage since the nominal voltage of the input valve anode will define the operating point of the concertina as well. Other topologies include the paraphrase, floating paraphrase and differential (long-tailed pair). a driver stage. This gives further voltage gain for each of the push–pull signals, and depending on the output stage valve's requirements may be a type selected for higher voltage or lower Z drive capability. The output stage, where the load is the transformer rather than an anode resistor. The original Williamson used KT66 pentodes "triode strapped" (operating as triodes). The majority of later push–pull amplifiers used the ultralinear connection instead.

McIntosh unity-coupled In 1949, Frank H. McIntosh and Gordon J. Gow introduced the “unity-coupled” push–pull output stage, in which the primary of the output transformer was divided so that half of the load impedance appeared in the plate circuits of the output valves and half in their cathode circuits. The tightly coupled bifilar winding caused the valves to operate partly as cathode followers, introducing local negative feedback within the output stage and reducing output impedance and high-frequency distortion associated with transformer leakage reactance. Later descriptions and patent disclosures generalized the arrangement, including forms employing a trifilar transformer with a separate, tightly coupled screen winding, preserving the AC feedback characteristics of the distributed-load topology while allowing independent DC screen supply.

Cascode

The cascode (a contraction of the phrase cascade to cathode) is a two-stage amplifier composed of a transconductance amplifier followed by a current buffer. In valve circuits, the cascode is often constructed from two triodes connected in series, with one operating as a common grid and thus acting as a voltage regulator, providing a nearly constant anode voltage to the other, which operates as a common cathode. This improves input-output isolation (or reverse transmission) by eliminating the Miller effect and thus contributes to a much higher bandwidth, higher input impedance, high output impedance, and higher gain than a single-triode stage.

Tetrode/pentode stages The tetrode has a screen grid (g2) which is between the anode and the first grid and normally serves, like the cascode, to eliminate the Miller effect and therefore also allows a higher bandwidth and/or higher gain than a triode, but at the expense of linearity and noise performance.

A pentode has an additional suppressor grid (g3) to eliminate the tetrode kink. This is used for improved performance rather than extra gain and is usually not accessible externally. Some of these valves use aligned grids to minimise grid current and beam plates instead of a third grid, these are known as "beam tetrodes".

It was realised (and many pentodes were specifically designed to permit) that by strapping the screens to the grid/anode a tetrode/pentode just became a triode again, as such making these late design valves very flexible. "Triode strapped" tetrodes are often used in modern amplifier designs that are optimised for quality rather than power output.

Ultra-linear

In 1937, Alan Blumlein originated a configuration between a "triode strapped" tetrode and normal tetrode, in which the screen grid of a tetrode is connected to a tap on the output transformer primary between the anode and the supply voltage. This electrical compromise provides characteristics intermediate between triode and pentode operation. In a 1951 engineering paper, David Hafler and Herbert Keroes determined that when the screen tap was set to approximately 43% of the anode voltage, an optimized operating condition within the output stage occurred, which they termed ultra-linear. By the late 1950s, this configuration had become a dominant design for high-fidelity push–pull amplifiers.

Output transformerless Julius Futterman pioneered a type of amplifier known as "output transformerless" (OTL). These use paralleled valves to match with speaker impedances (typically 8 ohms). This design require numerous valves, run hot, and because they attempt to match impedances in a way fundamentally different from a transformer, they often have a unique sound quality. 6080 triodes, designed for regulated power supplies, were low-impedance types sometimes pressed into transformerless use.

Single-ended triode (SET) power amplifiers Some valve amplifiers use the single-ended triode (SET) topology that uses the gain device in class A. SETs are extremely simple and have low parts count. Such amplifiers are expensive because of the output transformers required.

This type of design results in an extremely simple distortion spectrum comprising a monotonically decaying series of harmonics. Some consider this distortion characteristic is a factor in the attractiveness of the sound such designs produce. Compared with modern designs SETs adopt a minimalist approach, and often have just two stages, a single stage triode voltage amplifier followed by a triode power stage. However, variations using some form of active current source or load, not considered a gain stage, are used.

The typical valve using this topology in (rare) current commercial production is the 300B, which yields about 5 watts in SE mode. Rare amplifiers of this type use valves such as the 211 or 845, capable of about 18 watts. These valves are bright emitter transmitting valves, and have thoriated tungsten filaments which glow like light bulbs when powered.

See paragraphs further down regarding high-power commercially available SET amplifiers offering up to 40 watts with no difficulty, following the development of output transformers to overcome the above restrictions.

The pictures below are of a commercial SET amplifier, and also a prototype of a hobbyist amplifier.

One reason for SETs being (usually) limited to low power is the extreme difficulty (and consequent expense) of making an output transformer that can handle the plate current without saturating, while avoiding excessively large capacitive parasitics.

Push–pull (PP) / differential power amplifiers