|
|
| Elliott Sound Products | Valves (Vacuum Tubes) - A Primer |
Main Index
Valves Index
When you look at the schematics for a transistor and valve amplifier, the valve amp looks to be much simpler. A perfectly functional valve power amp may only need 3 glass envelopes, one being a twin triode plus two output valves. It is implicitly understood that valve rectifiers are a dreadful waste of space, time and energy, and provide zero sonic 'benefit' (despite the outlandish claims you may find on the Net). In comparison, a similarly specified transistor amp will typically require at least 6 (but more commonly 7 or more) transistors to do the same thing. Similarly, the valve amp may need as few as perhaps 20 other parts (excluding the power supply).
Again, the transistor amp will usually require more passive components as well, and the circuit diagram will appear positively cluttered compared to the valve amp. However, all is not what it seems. The component count for a valve amp includes valve sockets and of course the big item - the output transformer. This is usually so costly that one can buy all the parts for a transistor amp for the cost of that one part alone. Then we have to get the valves which are also relatively expensive Even 'cheap' valves cost more than 'expensive' transistors), and to make matters worse they have a finite lifetime. This means that they will need to be replaced at some time in the future.
Where a nominal 50W transistor amplifier may need ±35V supplies, that's all they need. A valve amp of the same power may require several different voltages, including B+ of perhaps 450V, several lower and decoupled (bypassed with capacitors) supply voltages, a main power supply filter choke, a negative voltage for output valve bias (-40V for example), and 6.3V for cathode heaters. In some cases the heaters for preamp valves may have to be DC to prevent hum, and this is another complication. So, while the amplifier itself might seem simple, its power supply can become quite complex and will be far more expensive than for a transistor amp.
It is (theoretically) possible to make a transistor amplifier using the same topology as a typical valve amp. No-one does so because it's not sensible, and the way transistors work complicates matters somewhat. Using MOSFETs (which are a form of transistor after all) simplifies the process, but it's still not sensible because the topology for any amplifier should be complementary to the way the active devices work. This is the case with all valve amps, and most transistor amps as well. Working outside the acceptable parameters generally gives an end product that is sub-optimal at best. Examples are transformerless valve output stages and transformer output stages with transistors. Neither of these is ideal, because the topology is not optimised for the intended purpose.
With the possible exception of lateral (audio) MOSFETs, transistors have far higher gain than any valve, and it's become common to use a lot of global negative feedback with all transistor amps. A few hi-fi valve amps also used a lot of global negative feedback, but this requires an output transformer with extremely wide bandwidth to minimise phase shift. Such transformers are very difficult to wind, and were enormously expensive as a result. Performance of some of the best examples was comparable to a decent transistor amp today, but at many times the cost.
Contrary to popular belief, global negative feedback is not evil, and it certainly doesn't "ruin the music" as some will claim. Done properly and tested sensibly (i.e. within the scope of real audio signals), negative feedback will always give better results than any 'zero feedback' or 'low feedback' design. This applies regardless of the amplifying devices used.
Valves hold a special place in history of course. As the fist devices created that could electronically amplify a signal, they paved the way for all electronics that we enjoy today. Valves were first used in numbers for two things we take for granted today - telecommunications and radio (or 'wireless' as it was known at the time). These two apparently simple (by today's standards) applications changed the world. The place of the humble triode (the first device that could amplify) is firmly entrenched in history, and started the electronics revolution that we know today.
One 'valve' that shows no sign of fading into obscurity any time soon is the cavity magnetron - the microwave generator used in microwave ovens. There is no solid-state equivalent that provides the same affordability as a magnetron, but this will change eventually. With an output of 1-1.5kW of RF energy, the magnetron may be close to 80 years old by now, but it's simple, cheap and reliable. One down-side common to all valve-type circuits is the requirement for a high-voltage supply (typically 2-3kV) that is extremely dangerous - hence the many warnings on a microwave oven!
Valves (vacuum tubes) ... much as I've tried to ignore them in the (futile) hope that they'd go away, they haven't, and probably won't. Despite what you might imagine, I don't dislike valves, and in fact I still have a soft spot for them. Note that I will use the term 'valve' as opposed to 'tube' because that's what I have always called them. I was trained in electronics at a time when valves were still very much current technology, and although transistors were around (indeed, so were some early ICs), the courses I attended had not caught up with the times.
My reluctance to publish anything to do with valves is based on the simple fact that many of those available today are not to the quality standards that existed when they were being manufactured in the UK, Europe, US and Australia (to name a few). Some of those from Russia are very good, but the quality is variable, and too many cowboys seem to be involved in the wholesale and retail businesses that supply valves to the end users. Many of the Chinese valves are somewhere between dubious and useless, however there are exceptions. Even getting decent valve sockets can be an issue, to the extent that very well known valve guitar amp makers have been caught, installing sockets that lose their grip on the pins after only a few insertions.
Having spoken at some length with a couple of friends and done a bit of preliminary research, it seems that there is something of a dearth of good information available on the Net - there is any amount of info, but much of it is apocryphal, misleading or just plain wrong. There is also a significant body of work that is none of these things, but it can be very difficult for readers to pick the difference between the good, bad and indifferent. There is also a fair amount of 'magical' thinking - attributing mystical properties to valves, or implying that valve amplifiers achieve things that transistorised amps simply cannot. For the most part, this is untrue - there are certainly differences, but they are not as great as many people seem to think.
Having said that, there are some things that valves do naturally that may be difficult with transistor amps. In any serious analysis though, it becomes obvious that most of these characteristics are not the things that make or break the sound. Of course, guitar amps are a somewhat different animal altogether, in that they are operated outside the linear region for much of the time. Where the linear regions of almost any amplifier are surprisingly similar, once pushed into deliberate distortion, things can change rapidly. However, many of the claims for valves over transistors even in this region are often greatly overstated.
It is extremely important to be aware that there is no magic in valves compared to transistors. A valve amp is no more or less capable of reproducing 'micro-dynamics' (whatever you imagine that might mean) than an amp using semiconductors. Any amplifier that can reproduce the full audio range - and do so without adding appreciable distortion - will generally be indistinguishable from any other similar amp in a double-blind test. There may be exceptions, but measurements will always reveal the reason(s) for any differences ... if performed competently.
This is one of several articles about valves. There is a great deal to discuss, and even more to be learned. Valves are interesting, not just from the historical perspective, but because they have attained almost cult status despite the fact that they are essentially a dead technology.
This article is not a history lesson - I will not be covering the many inventions and inventors who gave us the valve as we know it today. There is a vast amount of information on the Net for those who really want to know the historical progress of valves, and I will (at most) give a very brief account of developments as they relate to the function of each valve type.
Note that the valve diagrams that follow show indirectly heated cathodes, but there are some valves that use a directly heated cathode - commonly called a filament. This is most common with rectifier valves, but there are some old designs (the 300B is an example) that also use a filament.
This section is somewhat minimalist - it's intended as a brief overview only.
Valves rely on thermionic emission to function, hence the early term 'thermionic valve'. Almost all valves are operated with a 'hard' vacuum - there are few molecules of gas, and a system called a getter is used to collect molecules of gas that escape from the metal electrodes. This is commonly seen as a silvery section of the glass envelope, and it gradually degrades as gas molecules are absorbed. Gas molecules become ionised and will be attracted to the cathode because they have a positive charge due to an electron being displaced. Ion collisions with the cathode cause damage to the coating, and dramatically reduce the life of the valve. A valve can be considered at end-of-life when the edges of the getter start to turn brown. If the entire getter is a light colour, the valve has 'lost' its vacuum and is no longer usable.
The cathode is the source of electrons, and in (almost) all valves it's heated to ensure there's a sufficient number of 'free' electrons to make up what is known as the space charge - a cloud of electrons surrounding the heated cathode. As electrons leave the cathode, it is left with a small positive charge, and this attracts the electrons back to the cathode. At any given time (and with no other forces in evidence), there will be a very large number of electrons (the space charge) surrounding the cathode.
Some early valves used what were known as 'bright emitters'. These were usually filaments (as opposed to indirectly heated cathodes) and operated at a much higher temperature than we now consider normal. Because of the high temperature, they glowed quite brightly, hence the term. To get sufficient emission from a pure tungsten filament (bright emitter) requires a temperature of around 2,000-2,500°C, where a modern cathode may operate at around 750-800°C. This improves life and reduces power consumption. (Note that no two sources seem to agree on the typical operating temperatures for valve cathodes, and the figures given are rough estimates only.)
To improve the emission characteristics and allow operation at lower temperatures, the cathode is coated with materials having a low 'work function'. This means the material requires comparatively little energy (heat) to cause electrons to 'boil' off the surface. Typical materials used are barium oxide, strontium oxide, calcium oxide and thorium oxide. There are several others, and if you want to know more you can look it up on the Net - there's a vast amount of information available.
A few electrons leaving the cathode will have enough energy to pass through the space charge and migrate to other elements in the valve, mainly the grid(s) or the plate (anode). In the absence of any other supply or low resistance, any element that catches free electrons will become negatively charged and will then tend to repel other electrons. This technique is sometimes used to bias a valve - it's commonly known as grid-leak bias, and the control grid is connected to the common (ground/ negative) supply via a very high value resistor. The tiny current created by those electrons that strike the grid is sufficient to bias the valve into a usable state for normal operation.
When the anode is made positive with respect to the cathode, the space charge is attracted to the anode, so there is an easily measurable current flow. Should the anode be made negative with respect to the cathode, electrons are repelled and current flow is reduced to close to zero. Including a control grid allows the current flow between the cathode and anode to be controlled (hence the name). If the control grid is made positive or just less negative (with respect to the cathode), the electron flow to the anode is increased and vice versa.
Note that (thermionic) emission decreases as a valve ages, so a valve that can draw (say) 10mA under a set of defined parameters will show progressively less current under the same conditions as the valve heads towards its 'end of life'. This is the basis of many valve testers - especially those that simply give a 'good' or 'bad' reading on a meter. The effect of this can be reduced voltage swing on the anode of an aged valve, so it may increase distortion. For output valves, the available output power will fall. The gain provided by the stage will usually not be greatly affected, it just won't be able to provide the same (undistorted) voltage swing as a new valve.
Once you understand these (very basic) concepts, it becomes possible to understand how valve stages function. I don't propose to go into any more detail about the fundamentals of emission, but (as expected) there's a great deal of info available. Make sure that any reference material you rely on is from a trusted source - not all articles will be factual, some may even be quite wrong.
The first valve invented was a diode (John Fleming, 1904), and since a diode passes current in one direction but not the other, the term 'valve' was applied - the diode acted like a one-way valve. The name has stuck for Australian, British and New Zealand residents, and is in sufficient usage that it's accepted (albeit reluctantly) in the US. There, the term 'vacuum tube' (or more commonly, just 'tube') is preferred.
Regardless of what we call it, a diode valve has two elements or electrodes ('di', meaning two, plus the end of the word electrode). These are the anode (A) and cathode (K). The cathode is heated, and tends to 'boil' off electrons. When the anode (also commonly known as the plate) is made positive with respect to the cathode, the electrons travel across the vacuum and complete the electrical circuit. Should the anode become negative with respect to the cathode, no current flows. The negative anode charge repels electrons, and any current that does exist is extremely small. After some time in the 1920s, most valves used increasingly specialised coatings on the cathode material itself, in order to improve the emission characteristics.
The symbol for a (dual) diode is shown to the left. The version shown here uses an indirectly heated cathode, but many diodes use a directly heated cathode - that is to say that the heater and cathode are one - they are not separate. It is traditional to refer to such a cathode as a 'filament', and it is generally believed that this term came from the fact that the earliest diode was a filament (incandescent) lamp, with an extra electrode (the anode, aka 'plate') added. Why exactly anyone would add an electrode to a lamp is a short history lesson in itself, but it was the beginning of electronics as we know it today. The earliest diodes were used as a 'detector' - able to detect the presence or otherwise of a radio frequency signal.
Diodes are available as small signal types (commonly included in the same envelope as a triode or pentode) and as rectifiers to convert the AC output from a power transformer into DC (after filtering) to operate the equipment. Valve rectifiers were the only option in the early days, but are (or should) now be considered to be of historical significance only.
Other variants followed, and the most common version used today has two anodes (or plates), allowing a full-wave rectifier to be made with a single 'tube'. Like the one pictured, these may use either an indirectly or directly heated cathode. A major disadvantage of using a filament (directly heated cathode) is that a separate winding is needed on the power transformer to power the filament, because the cathode is the positive output terminal.
Indirectly heated cathodes have their own problems though, especially if the same heater winding (on the power transformer) is used for the rectifier and input valves. Hum can easily be injected into the heater circuit, which then can cause serious hum problems due to heater-cathode leakage in the input circuits. The insulation quality of indirectly heated diode valves is often not sufficient to withstand the high voltages used, so it may be necessary to use a separate transformer winding anyway.
A major limitation with valve rectifiers is the allowable capacitance following the rectifier. This is provided in datasheets, and can be as low as 20μF (5Y3GT - Tungsol). The maximum inrush current is the main limitation, and exceeding that can result is 'cathode stripping' - the cathode coating is torn away bit-by-bit until the valve is no longer usable. This limitation often results is sub-optimal filtering, so supply hum is superimposed on the output.
There is a great deal of nostalgia about valve rectifiers, but they are grossly inefficient compared to semiconductor diodes. They do have one advantage though, and that's the slow heating time. This allows other valves in the circuit to get to operating temperature before the full HT is applied. Filter capacitors are less stressed, because there is no sudden current surge, and the voltage never rises above their normal operating voltage. If silicon diodes are used, series resistors will help mimic the valve rectifier's rather soggy regulation and limit the switch-on surge current.
In order to prevent the HT from being applied before the valves warm up sufficiently, the input AC can be supplied via a relay with a time delay circuit. This is a far better option IMO, but not one that valve purists will usually adopt. There are a great many well known valve guitar amps that use silicon diodes for rectifiers, and this is one compromise that is often accepted (albeit reluctantly by some users).
Adding a third element to a valve (Lee De Forest's 'Audion', 1906) was the breakthrough that finally allowed us to amplify a signal. Prior to the triode (tri - three), there was nothing in the new field of electronics that provided amplification. Adding the grid allowed a small voltage to control the current passing between the cathode and anode. The grid is most commonly a fine wire spiral, wound so that it is close to the cathode. It is insulated from other elements within the valve.
The variation in the plate current can easily be applied across a resistor to convert it back to a voltage, but in the early days the nice stable resistors that we take for granted today were not common, so a transformer was often used. These have the advantage of being able to convert impedances as is still done with output stages, but were (and still are) expensive.
Now, we use a resistor load for all preamp stages, and a transformer only for the power amplifier stage. The resistor 'current to voltage converter' has been the method of choice (for audio at least) since the 1920s or thereabouts. Transformers are expensive and have a limited bandwidth - two issues that are neatly solved by using a resistor. Using a resistor is very inefficient though, but this is not generally a problem for low frequency preamp stages.
Valves can be though of as voltage to current converters. The voltage on the grid controls the current through the valve (not the voltage on the plate as you may have thought). The current change is converted into a voltage change by the plate resistor. If the resistor is (say) 47k and the current changes by 100μA, there is a voltage change of 4.7V across the resistor (Ohm's law ... V = R x I). Although the resistor load is very inefficient, it is convenient - a transformer ensures that almost all of the current variation is converted into a voltage with fewer losses.
Towards the end of the valve era, many valves were given a gain figure in mA/ V, where the voltage (V) was applied to the grid, and the current (mA) was the change of plate current for a 1V change of grid current. A more common gain figure was transconductance (Gm), which is in μmhos (mho is ohm spelled backwards). The use of the mho is now pretty much gone in all fields except valves - the shiny new unit is the Siemens (S), but the measurements themselves are identical. A valve with a Gm of 1,000 μmho has a Gm of 1mS.
Note: 1 Siemens (1S) is equal to 1 Ampere per Volt, so 1mS is the same as 1,000μmhos, which is 1mA/ Volt.
If a valve has a transconductance of 1,100 μmhos, this is exactly equivalent to 1.1 milli mhos, 1.1 mS (milli-Siemens) or 1.1mA/ V. These terms are therefore fully interchangeable. The transconductance for triodes is generally within the range of about 0.8 to 8mS. This makes it easy to convert from one to another. Another common specification was 'amplification factor' or 'mu' (μ). This is the theoretical maximum possible gain obtainable from a particular valve, and is determined by the cathode to grid spacing and the pitch of the grid spiral. Without some specialised circuitry, no common valve will ever have the gain implied by the 'amplification factor'.
Regardless of the names given to the conversion factor measurement of a valve, the end result is identical - a change of grid voltage causes a change of plate current, and this is converted back to a voltage using a resistor or transformer. Now that we have some control over the behaviour of a valve, a new measurement sneaks in - plate resistance. This isn't a real resistance - it's simply a convenient way to express the dynamic relationship between the change of plate voltage to plate current (with the grid held at a constant voltage).
Plate resistance varies with plate voltage (as does transconductance), so a measurement taken at a plate voltage of 200V will be different from that taken when the plate is at 100V. During the design phase of any valve amplifier section, it's important to know (or at least estimate) the plate resistance and transconductance for the voltage that exists on the plate. Since a valve is (or attempts to be) a voltage controlled current source, one would like the plate impedance to be infinite, but a triode has too little gain to even remotely approach that. The plate resistance is effectively in parallel with the load (the combination of plate resistor and any circuitry following the stage), so a low plate resistance reduces gain to well below that which we might expect.
Another parameter you often see is so-called 'amplification factor' (abbreviated to μ ... pronounced mu). The amplification factor of a valve is the theoretical maximum gain that can be obtained. It is based on the variation of anode voltage to grid voltage, but is measured with the anode current held constant. The only way a triode can achieve its quoted gain (based on μ) is if the plate load resistance (as well as any following stage) is infinite. A valve with the grid very close to the cathode has a high amplification factor. The typical values for μ fall between 10 and 100 for most triodes. μ is largely a physical parameter, so it is (theoretically) not affected as the valve ages. While this is a simplistic approach, in practice it is quite close to reality, although a small change will occur as a valve ages. In many cases, even though a valve may have poor emission, be noisy and/or microphonic, it may still provide (very close to) the gain expected. What it perhaps can't do is provide the normal output voltage swing without serious distortion, but at low levels the valve appears to function normally.
μ = ΔVa / ΔVg Where Δ means incremental change, Va is anode voltage and Vg is grid voltage.
Note that plate impedance, transconductance and amplification factor are small signal parameters, and only work when the variation in plate voltage is very small - typically less than 10% of the steady state voltage.
Since this is just a primer, the actual design of valve stages will be left for another article. However, it is very important to understand the parameters and their interactions with the real world, because these are the things that influence the performance of the final circuit. The descriptions given here are not the last word by any stretch of the imagination, so the next instalment will cover the parameters and their effect on the final design in more detail.
The low gain and limited bandwidth of early triodes led to a great deal of experimentation in the early 20th century. One of the most important areas in the early years was radio, or wireless as it was known then (the term is now back in vogue for networking). Communications were limited to wire transmissions before the valve, which was very restricting. The problem with a triode is that it can have considerable capacitance between the plate and control grid, and combined with high impedance circuits this allowed some of the high frequency signal on the plate to be coupled back to the grid. This is feedback, and it reduces the gain at high frequencies due to the stray capacitance. Not generally a problem for audio, but a major issue for radio frequency use.
The added screen grid is so-called because it 'screens' the control grid from the plate, reducing the capacitance and increasing bandwidth. Although it's connected to a positive supply, for AC (the wanted signal) it's most commonly effectively at earth (ground) potential by virtue of a bypass capacitor. The positive DC supply dramatically increases the gain, because the screen acts as an accelerator to the electrons that have been liberated by the hot cathode. This greatly increased gain comes at a significant cost though, due to a process called secondary emission. The electrons are accelerated to such a degree that when they hit the plate, they have sufficient energy to dislodge electrons from the plate's surface. Some of this secondary emission is simply attracted back to the anode from whence it came, but some is captured by the screen grid. This increases the dissipation in the screen, causes distortion, and leads to a negative resistance characteristic at some point in the operating range. This is known as the 'tetrode kink'. Most tetrodes produced soon after their introduction (in particular the KT66 and, later, the KT88) were described as 'kinkless tetrodes'. While this implies that there is no kink, these valves do have a kink in their plate characteristics, but it is dramatically less severe than 'ordinary' tetrodes. I suspect that the term was primarily used as a marketing tool, but it's also a reasonable description. These are beam tetrodes, and have become one of the most popular valve types ever produced for power amplifiers.
The beam power tetrode is an interesting variation of the tetrode that became (and still is) extraordinarily popular. These were initially developed to bypass the Philips patent for the pentode (next section), in around 1933. Although the greatest benefits weren't realised for some time after the tetrode (tetra - four) was introduced, the screen grid proved that valves could have very high gain and, more importantly for radio applications, a wider bandwidth than previously thought. The gain of these valves is far less dependent on the plate voltage than is the case with triodes. The screen grid current is also much lower than a power pentode, typically around 5-10% of the plate current (a pentode screen typically draws about 20% of the plate current). As the plate voltage varied with signal, there is very little gain change - provided the screen grid is held at a constant voltage. This also means that the effective plate resistance is much higher. Plate resistance is effectively in parallel with the theoretical 'voltage controlled current source' model for a valve, and the higher the value the greater the available gain - at least with a resistive load.
Selection of the screen grid operating voltage is important. If it's too high, there will be excessive current flowing in the screen grid, raising its temperature - possibly to destructive levels. Except for a few specialised topologies, the current in the screen is completely wasted, in that it doesn't contribute to the plate current to produce useful output. The lower screen current with beam tetrodes was obviously a great benefit. Overall, the tetrode was a giant leap in performance, having much higher gain and better high frequency response than could previously be obtained from these new but very expensive vacuum tubes.
Two beam confining plates (commonly referred to as 'beam forming' plates) are connected to the cathode, and these force the electron beam to follow a specific path, bypassing the grid support wires in particular. They also help to suppress secondary emission from the plate. The 'beams' that give the valve its name are formed by careful alignment of the control and screen grids, which focuses the electron beams just before the plate surface. This forms a 'virtual cathode' (aka space charge), and since it has a relatively strong negative polarity due to the focussed electron beams, it acts as a convenient means to suppress secondary emission as it acts as a virtual cathode.
Virtually all of the tetrodes available today (and indeed since the late 1930s) are beam types. Beam confining plates are used primarily to keep the electron 'beams' away from the grid supports, and the control and screen grids are aligned to form the beams. It is common to direct the electron beam(s) onto that part of the anode mechanical structure where there is the most metal (typically at the seam where the two halves of the plate are joined). This provides improved heat radiation because of the increased surface area, raising the plate dissipation and the power the valve can handle.
The pentode (penta - five) was developed in 1930, by Philips in the Netherlands. Because of the problems of the standard tetrode (primarily secondary emission and the 'tetrode kink'), a third grid was added, and connected to the cathode.
This suppressor grid did what its name suggests - it suppressed the secondary emission from the plate, by repelling electrons. High velocity electrons pass straight through the relatively open suppressor grid, but the negative potential is sufficient to prevent secondary emission electrons from migrating back to the screen grid.
The development of the pentode was a very significant improvement over anything that came before. Having much higher gain than a triode because of the screen grid, along with greatly reduced secondary emission thanks to the suppressor, it became the valve of choice for high gain applications. Pentodes were also made as power output valves, and (along with power beam tetrodes) are the most commonly used output valves in guitar amplifiers.
Like the tetrode, the screen grid in a pentode both accelerates the electron beam and shields (screens) the control grid from the anode. This provides the high gain and extended high frequency response needed for radio, radar and (later) television receivers, and in the larger versions provided more power than was ever available before. Because of the high gain, it became possible to make amplifiers that had a significant current swing in the plate circuit, but with grid drive voltages that were achievable with relatively simple circuits. Each and every step in the development of valves has led to applications that were never possible before.
Even today, there are some applications that rely on the use of valve technology. The magnetron (as used for radar, and of course the microwave oven) is a valve, and there is no solid state equivalent. Very large radio frequency transmitters generally use valves, because they are easily scaled and are comparatively easy to keep cool enough to prevent self destruction.
During the heyday of valves, some very clever variants were developed. Pentagrid valves that were used as both an oscillator and RF mixer stage were common in radios, and reduced the number of individual envelopes needed to produce a receiver with acceptable gain and selectivity to be useful to the public. Many valves contain several different elements - triode-pentode valves could be thought of as a very early attempt at an integrated circuit, having two independent structures within the same glass envelope. These usually shared the heater connection, but all other electrodes were available as normal. Another common function was to combine a triode (or pentode) with a dual diode, enabling the one valve to be an RF detector and first audio amplifier.
Twin triodes are very common, and are the most popular preamplifier valves in use today. The ubiquitous 12AX7/ ECC83 is quite possibly one of the most successful valve designs ever, being the mainstay of almost every valve guitar amplifier ever made, as well as being popular for hi-fi amps, instrumentation and other industrial applications. Most conventional valves are classified as 'sharp cut-off', meaning that there will be some value (around -5V or so) of negative grid voltage that will reduce the plate current to a very low value. The cut-off current and grid voltage are sometimes quoted in datasheets.
Following from the above, there is one valve type that deserves a brief comment, namely the 'vari-mu' or 'remote cut-off' RF amplifier. These were designed to allow radio ('wireless') receivers to apply automatic gain control (AGC, sometimes referred to as AVC - automatic volume control). This allows the sensitivity of the radio frequency stage(s) to be changed to suit the incoming signal strength, so close by or powerful transmissions don't result in distortion in the RF stages or excessive volume changes when tuning between stations. Instead of the control grid being a continuous spiral of closely and evenly spaced turns of wire, the spiral is closely spaced at one end, and comparatively widely spaced at the other. Closely spaced grid wires give a high mu ('amplification factor'), and widely spaced grid wires give a low mu, so providing a progressive transition between the two gives a variable mu.
As the grid is made more negative, a progressively smaller area of the space charge can be controlled, since with only a slightly more negative grid, the area controlled by the finely spaced grid wires will be cut off. Only those areas of the grid that have a wider spacing will allow an electron flow, and this is a progressive change over a fairly wide voltage range. As the grid is made more negative, the gain is reduced and vice versa. AGC is designed to apply a negative grid voltage that's proportional to the signal strength, so a weak signal allows the stage(s) to run with maximum gain. Conversely, a strong signal creates a greater (more negative) grid voltage and reduces the stage gain. This allowed the gain to be varied over a fairly wide range without excessive distortion. A remote cut-off valve can handle a signal that's up to 30 times stronger (for the same distortion) than an equivalent sharp cut-off valve. Nearly all valve radios used a vari-mu pentode in the circuit to allow AGC. There is no semiconductor equivalent to a remote cut-off/ vari-mu valve, but transistors can still work well by reducing their collector current to achieve much the same results.
As noted above, the magnetron is a valve, as is another ultra high frequency amplifier, the travelling wave tube. The TWT is a highly specialised valve, specifically for high output power and very high gain. Operating frequency extends to ~50GHz. Another high power RF valve is the Klystron, which was common until fairly recently for UHF and microwave transmissions. There are literally hundreds of different types of vacuum tube, and up until very recently, most readers would have been reading this article with the help of a valve - the cathode ray tube (CRT). The invention of the CRT allowed radar systems to show the position of detected planes, ships, etc., and of course there was the CRO - cathode ray oscilloscope. The CRT was also instrumental in giving us television. Hmmm. Perhaps not such a good idea after all 
.
There are a great many other valve types of course, but it is outside the scope of this article to go into any detail. The majority of readers are interested in audio applications, either for guitar (including bass) or hi-fi applications. Even with the scope narrowed to those applications alone, there are still many, many valve types that are (or appear to be) suitable. Future articles will examine the most popular of these, but I do have to point out that if you expect information on truly ancient technology (single-ended triode amplifiers using 300B valves for example), then I'm afraid that you'll have to look elsewhere. That technology had an extremely short life in the very early days of audio, until it was found that push-pull operation was ever so much better in all respects.
The power supply is one of the least endearing parts of valve design, even for preamps. For power amps it becomes a very expensive undertaking, requiring a power transformer with multiple voltages. For example, you may require 300V [main HT supply], 5V [rectifier filament], 6.3V [valve heaters] and possibly a 40-50V winding for output valve biasing. To obtain acceptable filtering (low supply ripple) you'll need at least one inductor of around 10H or so, with a second inductor sometimes used to get a smooth supply for the preamp stages. Of course you also need one or two output transformers, and the cost of the magnetics alone can exceed the total build cost of a transistor amplifier of similar output power.
You only need to look at early hi-fi valve circuits to see the complexity of the supplies. Ok, they are not complex per se, but they are filled with expensive parts. Output transformers are very expensive to make, because they require interleaved windings (both primary and secondary divided into sections, with the secondary embedded into the sectioned primary. This is needed to minimise leakage inductance that reduces high-frequency response and adds phase shift such that feedback can only be minimal before the amp becomes unstable. I encourage the reader to look up the complete design for the Williamson valve amp. As originally published it was mono, and it used a power transformer with three secondaries, two inductors (chokes) and the output transformer. The specification for the output transformer was very 'tight', and it required be best steel available and very careful attention to detail.
Given that the design goal for the Williamson was an output of just 15W, it was an expensive amp to build in the heyday of valves, and much more so now. I would hazard a guess as to the current cost (for a mono system) at around AU$700 - just for the magnetics (power and output Tx, two inductors). There's probably at least another AU$200 for a set of valves (assuming KT66 as originally specified), and you still have to buy a chassis and all the passive components.
While I would love to have a Williamson in my collection, for the better part of AU$2k for the bits needed to build one it's not going to happen.
To be perfectly honest, I am of the opinion that 1930s (or earlier) technology belongs to the era where it was popular. Huge advances were made in the late 40s through to the early 60s, with the important parameters (such as distortion) reduced to far lower values than were possible before, along with sensible and usable output power and improved efficiency. The (then) new valve types and major increases in our understanding of output transformers made big differences to available bandwidth. The (almost) complete elimination of single-ended triode power amplifiers relatively early was a direct result of improved topologies, coupled with very good output transformer designs that were also far more efficient by virtue of push-pull power amplifier stages.
While I am somewhat reconciled to the fact that valves won't go away, this doesn't mean that all amplifiers using valves were 'good'. The truth be known, many were awful, and engineers of the day were delighted at the prospect of transistors - greater reliability, more power, and improved efficiency. When combined with lower distortion and generally improved technical performance (which is important, regardless of the opinions of some of those pushing the esoteric SET agenda for example), there is no comparison. It also follows that many of the early transistor amps were bloody awful, and to an extent the stigma has remained - over 50 years later, and some people who have never heard a bad transistor amp still think they're 'bad'.
All of the major manufacturers of quality valve hi-fi equipment used push-pull amplifiers, generally rated at between 10 and 50W, since it was determined that this was a very satisfactory power for domestic sound reproduction. Many of the designs used were very innovative, with highly specialised output transformers being common. Performance of valve equipment reached it peak just prior to the introduction of transistor amplifiers. Further development came to a standstill after decent transistors became available for relatively low cost.
This was to be expected, since the advantages to both manufacturer and end user alike were so great that all major makers of consumer equipment switched almost overnight, thus ending further valve innovation for the most part. Quite possibly one of the very last valves of any significance was the Nuvistor, the first of which (the 7586) was released in 1959. The more commonly known 6CW4 came a couple of years later. There were others such as the 'Compactron' - a multi-function valve designed for TV sets, but the list is short.
It's interesting to observe that manufacturers such as Leak, Quad, McIntosh, Fisher, etc., never used single-ended triode output stages. All output stages were push-pull because of the huge improvement in all of the parameters that were deemed to be important - frequency response, distortion (harmonic and intermodulation), hum and noise, output power, etc. These makers did not use push-pull designs to reduce cost or weight - many of the best amps at this time were 'cost-no-object', and could only be afforded by a small few consumers. One of the most famous amplifier designs for 'home construction' was the Williamson, which used a pair of KT66 valves wired as triodes and operated in push-pull.
Single-ended designs were restricted to mantel radios, record players and small PA systems. From around the mid 1930s on, these almost exclusively used small output pentodes, and were typically rated at about 1-5W output, with a restricted frequency response that matched the loudspeakers used in these applications. Even single-ended pentode guitar amps were common - mainly as practice amps. Most were dreadful (I know this because I had one when I was a teenager). In 1933, Stanley Mullard even made the point that pentodes were preferred over triodes for this application, because they have a very high output impedance that allowed the speakers of the day to perform better, with 'improved' low and high frequency response.
Having said this, it must be admitted that the SET (single ended triode) amplifier has a place in the world. It is a very convenient way to prevent doors from closing uninvited due to wind gusts, small children and pets. Needless to say, for continued reliable service in this rôle (and for the safety of others), it is best left disconnected from any power or signal source. An alternative valid use is to allow small boats to remain tethered to the ocean floor to prevent drifting about and causing themselves a mischief.
It's worth mentioning that the favourite valve for SET applications is the 300B, but few people would be aware that it was first made by Western Electric (part of AT&T and Bell Labs), and was intended for use as a telephone signal amplifier. As far back as 1922 or thereabouts, the power amplifier of choice was push-pull, and the Western Electric datasheet for the 300B describes the recommended operating conditions for SE and PP operation. Push-pull operation gave far more power and lower distortion than single-ended then, and nothing has changed since.
As near as anyone can tell, valves will remain with us for some time to come. Not only for their nostalgia value, but because there is a simple elegance in well designed valve equipment. Yes, such designs are comparatively inefficient and require the use of fairly fragile glass bottles that get hot, but that's considered a small price to pay by the great many valve enthusiasts. Transformers are both hard to get and expensive, although there are a few around that should perform quite well. A potential problem is getting a transformer that suits your favourite output valves. This can be a problem, as you could either drive the valves too hard (reducing their life expectancy) or be unable to get the expected power because of the impedance mismatch.
One area that often causes confusion is the power rating of valves vs. transistors in power amplifiers. The power rating for a valve is the average power dissipated, and there is no theoretical limit to the peak power (provided voltage and current remain within the datasheet limits of course). Transistors are rated for the peak dissipation, which is subject to the die temperature and instantaneous power. Exceeding the peak rating (at the worst-case operating temperature) can result in a phenomenon called 'second breakdown', where the transistor enters a negative impedance state. The result is usually instantaneous, catastrophic failure of the device. Where a pair of 30W valves can provide up to 80-100W output, the same output with transistors requires that their dissipation limit is at least double the expected power output (for example, 2 × 200W devices for 100W output).
This may seem like a serious limitation, but with modern devices it's easy (and relatively cheap) to achieve. A well designed transistor amplifier will run for years without any requirement to adjust the bias or change the transistors - they do not 'wear out' like valves do. When used in guitar amps and subjected to near constant clipping, the dissipation of a transistor output stage is reduced, but in a valve stage it's increased, leading to reduced valve life.
Preamps are much simpler, with the only issue being to get the proper heater voltage and a high tension supply that will give you the output swing you need. The heater voltage is far more critical than some people imagine, and if too low, the result can be cathode poisoning - a condition where the cathode materials are contaminated by trace amounts of gas. Should the heater be run at too high a voltage, its usable life expectancy is reduced, perhaps considerably. Where a valve calls for a 6.3V heater supply, this should be as close as possible to 6.3V AC (allowing for normal mains variations), or 6.3V DC, which can be maintained very accurately by using a regulator.
In many respects, there's a lot to be said for using valves and transistors together in a hybrid design. While not a purist approach, hybrids can give (what some might consider) the best of both worlds, using valves as voltage amplifying devices, and transistors as current amplifiers (for example). Transistors with the necessary voltage and peak) power ratings are now readily available, and the hybrid approach also permits the use of PNP transistors (or P-Channel MOSFETs), something for which there is no equivalent with valves. Valves come in one 'flavour' - the equivalent of an N-Channel FET, and no complement exists in the world of the valve.
Because this is only a primer, there is a vast amount of information that's not been included. The idea here is to introduce the basics, and to familiarise the reader with some of the concepts of valves. It's important to understand that a signal amplified by a valve is no different from the same signal amplified by a transistor or an opamp, provided signal levels are kept low enough to ensure that each device operates in its most linear region. In one respect, a valve preamplifier is potentially more linear than a transistor preamp without feedback, because the output voltage swings over such a small range compared to the supply voltage. If you need 2V RMS of signal, this may be only 2% of the valve's normal plate voltage, but the same swing from a basic transistor circuit could easily exceed 20% of the nominal collector voltage. With small variations, the device can easily remain within its linear region, but as the output level becomes a larger percentage of the available supply voltage, linearity suffers. This increases distortion (both simple harmonic and intermodulation), and can easily become audible.
One (of many) claims found is that valves are linear, while transistors are not. This is flawed thinking - valves are not linear. If they were, then valve amps would have no distortion at all. As noted above, valves do operate at high voltages by comparison, but a transistor operated with the same voltage and current, and having the same gain and output level, will beat a valve hands down for distortion. Does this mean that transistors are more linear than valves? No. It simply means that such comparisons need to be treated with some suspicion because the devices are very different from each other. In order to get a transistor stage to have the low gain of a valve, it is necessary to apply local feedback (using a relatively high value emitter resistor) and this changes the comparison completely.
Remember - there is no magic involved with valves. They don't do anything that can't be done with a carefully designed transistor stage, and for sheer performance, valves don't even come close to opamps or well designed transistor circuits. There are certainly some good reasons to experiment (especially with preamps), as the cost is relatively low and the experimenter will learn a great deal. Whether this knowledge is ultimately useful is another matter altogether.
Several references were used for this (and will be used for subsequent articles) - see below. Of these, the primary source of information is The Radiotron Designer's Handbook (of course). There were also many websites that I have looked at (and will visit for later articles), including Wikipedia. In some cases, sites visited only reinforced the fact that a depressingly large amount of the available information is either misleading or wrong. Others have some useful information, although in some cases it's only useful if one already knows the details. Quite a few sites did nothing more than jog my memory, but if only for that, they were useful.
It is inevitable that I too will make errors during the compilation of information, and these are regretted in advance. If any such errors are found, please let me know.
Main Index
Valves Index