Why are we, as audiophiles, entranced by the reproduction of music using vacuum tubes?
As it turns out, noise is traditionally thought of as something to be minimized in all high end audio systems. However the proper reproduction of noise is actually the key characteristic that makes tube amplifiers sound so good. This article explores why noise is so important to the “tube sound.”
There are volumes of theory regarding how tube amplifiers saturate and produce harmonics and how their distortion characteristics affect listening. Here, we will explore how tubes reproduce noise and how this particular characteristic represents the basis of why tube amplifiers have a unique, characteristic sound that defines audiophile listening.
In order for us to explore the reproduction of noise, lets first define what we mean by noise. Noise, as applied to audio listening, is defined by three separate characteristics. For the purpose of this article, lets separate those characteristics and define them and their sources. Instead of using the generic term, noise, lets call anything that is not part of the desired soundtrack an artifact.
The first artifact we concern ourselves with is interference. Interference is hum, cable modem chatter, cell phone crosstalk, microphonics, etc. It’s outside of the soundtrack, electrical sound in our gear. In our world today, we are bombarded by these sources of interference that largely work their way into our audio equipment through radio frequency (RF) coupling. They are primarily of man made origin and are controlled by good engineering practices that reduce these artifacts from being picked up.
The second artifact is distortion. Distortion is a byproduct of the nonlinearities of our audio equipment and add directly into the audio soundtrack. Distortion can be either harmonic in nature, ie, overtones of the desired soundtrack or it can be intermodulation effects–mixing products of multiple tones contained in the desired the soundtrack. There also can be non-harmonic distortion such as clipping. Clipping is where the amplifier just runs out of power to reproduce all of the content of the soundtrack as each instrument is recorded and therefore cannot produce the sound amplitude as intended. All audio gear has linear and nonlinear operating regions and hence will produce some form of distortion. It’s the task of the design engineer to increase the linear operating regions and design them around the primary listening regions both in frequency and amplitude to reduce these sources of distortion.
The third artifact, and the one we will focus on for this article, is thermal noise. Thermal noise is a phenomena of nature where ambient heat causes electrons in conductors to vibrate and hence cause electrical noise. Thermal noise is the limiting factor for amplifier design and starts at the pickup and is amplified and added through the entire amplification process. Because, by nature, noise is a very low amplitude, high gain stages such as phono stages are much more susceptible to corruption by the effects of thermal noise. Due to it’s process of creation, the vibration of electrons in conductors, thermal noise is statistical in frequency and amplitude and it’s overall level is directly a function of the bandwidth of the amplifier. Audiophile grade amplifiers tend to be very broadband and therefore amplify a significant amount of thermal noise to a level where it can interfere with listening levels if not properly controlled by design. For discussion purpose, thermal noise is defined by the equation:
Noise Voltage = KTB (EQ 1) Thermal Noise Equation
Where K is a constant, T is the ambient temperature in Kelvins and B is the bandwidth (in Hertz) of the amplifier.
Noise is an important factor because it is always present in audio reproduction. Noise is the sound we hear between the musical notes. It’s the background. The quietness we hear when there is a break in the soundtrack. Vacuum tube amplifiers have some unique physical characteristics that reproduce this noise that we audiophiles describe as warmth. We also describe the non musical intervals as having a black background and these characteristics are unique to the physics of how vacuum tubes actually amplify the soundtrack as well as the noise. We will discuss the difference of how tubes amplify the signal and noise and compare these properties to how solid state devices—transistors and integrated circuits–amplify the desired signal and noise.
As we learned in EQ1, noise power is derived from thermal agitation of electrons in all of the equipment conductors and hence is a thermionic effect that generates a noise voltage with a very low, almost immeasurable current. In order to achieve a black listening background, we want this naturally occurring noise to be reproduced just as it is created. We want to maintain its bandwidth and statistical properties. Now with this concept in mind, lets go into how our devices actually amplify—the tube versus the transistor.
A vacuum tube amplifies because of it’s thermionic control characteristics. Vacuum tubes modulate a stream of electrons that flow from the tube cathode to the tube plate by impressing a control voltage on the tube grid. The grid is a fine wire mesh inserted between the cathode and the plate that controls the flow of electrons. The grid responds to an impressed voltage and draws no current from the input source. So in mathematical terms, the tube plate current is a function of the grid voltage. There are no current demands from the source by the tube grid.
So, in a practical application, a phono stage with 65 dB of gain, will amplify the desired signal plus the thermionic noise without changing the statistical characteristics of the noise. Because the tube grid draws no current, the thermal noise presented to it is amplified just as it is created–statistical in amplitude and frequency. Hence we get the full bandwidth of the natural noise. It is not changed by requiring the noise to have any current content or minimal voltage threshold before it is amplified. In short, because the tube is a thermionic device and noise is a thermionic effect, the tube, by the nature of it’s physics, amplifies the noise just as it is created in nature.
In the case of a solid state device, the amplification is defined by the current that flows between the emitter and the collector. The current flow between the emitter and collector is controlled by the semiconductor junction current. So the transistor current is controlled by the transistor junction current. This control feature is not a thermionic effect but rather a junction effect controlled by current rather than voltage. So the key difference between solid state amplifiers and tube amplifiers is that solid state devices require a current rather than a voltage to cause amplification. Remember from EQ1 that thermal noise is a voltage rather than a current so when a transistor amplifies noise, it requires that noise to create a junction current before it is amplified hence the noise no longer retains its statistical properties in amplitude and frequency. Some components of the natural noise are changed before they are amplified. The noise sounds different!
This explanation is not to say that tube amplification sounds better or worse than a solid state amplifier, only that it sounds different. The difference we hear is the warmth—the sound between the notes—the noise.
The noise characteristics of amplifiers are defined by the earliest stages in the amplification chain because the noise is a very low level and impacted by the succeeding gain of the amplifier stages. So, if the tube warmth and blackness is a desired characteristic, it is best to use tube amplification early in the gain process. In short, the phono stage characteristics are more important to the background sound than the power amplifier characteristics.
In summary, the tube amplifier has a sound that is different than a solid state amplifier largely due to the way it reproduces background noise. The warmth and blackness in the background of the tube amplifier is due to the fact that the tube amplifier is a thermionic device and reproduces background noise more faithfully than a solid state device. While it is always the desire of the design engineer to minimize any noise that will interfere with the desired soundtrack, noise is always present and it defines what we here when there are pauses in the music–the sound between the notes.
One final thought on noise. Noise, because of its statistical qualities, is an analog effect. Digitization will also profoundly affect the sound and nature of background noise and hence analog amplification more accurately reproduces naturally occurring background noise.
At Rogers High Fidelity, we design our analog audio gear using exclusively vacuum tubes. Our PA-1A phono stage has the industry’s highest signal to noise ratio.
[Roger Gibboni is President and Chief Engineer of Rogers High Fidelity—Ed.]