The “vacuum tube” or “thermionic valve” became a topic of interest for me only in the last year and a half, as a guitarist and ex telecomms/IT technician, because of it’s technical history (Bletchley Colossus):
in code breaking during WW2, and that a GPO engineer Tommy Flowers helped build it:
“The Enigma of Intelligence – Andrew Hodges
HarperCollins Publishers Ltd; New edition edition (28 Jan 1985)
Flowers was asked by Alan Turing (Turing Machine inventor, mathematician and Bletchley code breaking genius) to build the hardware for the first computer which utilised valves as switching elements.
As a musician, the valve has an almost mythical reputation as a “superior” tonal element in audio equipment, most famously in guitar amplifiers made by Fender and Marshall companies from the 1950s onwards, which are directly related to the way social culture in the West evolved, particularly with the emergence of “Rock and Roll” and its evolution through 1960s “hippie” culture into heavy rock through the 1970s and beyond partly because of amplifier design development for the requirement of higher power, louder amplifiers.
I will try to keep this document as easy to read and “layperson friendly” from a technical point of view as possible, as I explain a little from many aspects of the history, technical function, use and desirability in audio equipment of these magical little glass devices that have had such a massive yet generally underrated impact on our technical and social development over the last century, being THE key component to enable the existence of 1st generation radio and TV sets. Invention and Application
The Thermionic Valve has proven to be both enduring and enigmatic over the 110 years of its existence since 1904. It has survived a massive decline after the 1970’s in sales, application and popularity with the impact of the tiny and light transistor that emerged to replace it in all home appliances – TV and radio sets – as well as in industrial and military electronic applications – though to a lesser degree in Eastern bloc countries due to the realisation that it could survive electro-magnetic pulse phenomena in the event of nuclear war, that would destroy far less robust equivalent transistor based military circuits, and were reported in a Russian MIG 25 fighter that defected to Japan in 1976.
Since that initial decline it has become more popular again since the 1980’s as many guitar players returned to valve amplifiers after an initial burst of ownership of transistorised amps due to their low cost, preferring overall valve tone despite the general increase in cost, weight and power consumption.
Many early valve amp models are now expensive collector’s items and regarded as Classic in all senses of the definition – original tone, simplicity but effectiveness of design, handling ruggedness, build quality and component robustness.
In it simplest form this evacuated, glass housed contraption acts as a diode, which enables current to flow in an AC circuit in one direction only, and was invented by John Ambrose Fleming in 1904.
“he invented the two-electrode vacuum-tube rectifier, which he called the oscillation valve, for which he received a patent on 16 November.”
The vacuum tube evolved directly from research work done by many, such as Thomas Edison on the incandescent light bulb (a heater filament in an evacuated glass tube).
A simple equivalent schematic showing how this unidirectional current flow is achieved is shown below, using a silicon diode first, as it clearly shows the direction that current will flow only when the left side of the diode has a positive voltage so acts as a one way valve, so shows how an AC sine wave is “rectified” into a varying (unregulated) DC voltage and current (Direct Current – current that flows in one direction in a load only) from the top of resistor R to the bottom, but not in reverse:
The glass housed valve version works in the same way because the the current that flows through the evacuated space is made up of free electrons that have been energised sufficiently through the cathode connection being pre-heated so able to break away from the special electron rich metallic compound surface of the cathode, towards the (electron deficient) positively charged connection (the anode or plate). Even though this space is a vacuum, and there is no “conventional” conductor like a wire that most people may be more familiar with, as a concept required for electric current to flow between two points, electrons still flow, as they do in a cathode ray TV set (as it was experimentation with the Crookes tube as early as 1869 that also led to the invention of TV originally).
The pink glow is caused by high energy electrons colliding with residual air molecules in the bulb before striking the green phosphors at the screen end. The screen “cross” image is created by the shaped anode casting the electron shadow.
A similar blue glow can sometimes be seen in present day power valves in guitar amps such as in my Fender Champ as the electrons strike water vapour absorbing chemical gas in the tube.
Schematically, a diode may be shown as below with the anode (a) and cathode (k):
The equivalent circuit as the silicon diode above would be:
As the anode (a) goes positive, electrons (negatively charged) from the cathode are attracted to it flowing in the opposite direction to Conventional Current (positive to negative). The Conventional Current is still from top to bottom through the load resistor R.
It is important to understand this basic principle at this point so that further developments in the design that were made to create the triode (three connections) can be understood, because it is that important discovery that enabled the current amplifier to be realised – the triode valve.
The addition of an extra connection called the “grid” – placed physically much closer to the cathode than the anode, enabled this third connection to become a control point or voltage controlled regulator for the main current flowing between the anode and cathode. This works because an inverse square law affects the strength of attraction between electrically charged objects, and as the grid is much closer to the cathode than the anode is, then a proportionately much smaller change in voltage at the grid has a much larger influence at attracting electrons away from the cathode than the anode does. Once broken away, these electrons are then swept up in the overall current flow between anode and cathode, but can pass through the mesh structure of the grid.
Conversely, if the grid is made very negative relative to the cathode, electrons at the cathode can be repelled away from the grid so strongly as to prevent any electrons leaving the cathode at all, so preventing any current flowing across the valve at all. This is called “cut off” as opposed to when the maximum current flows at “saturation” when the electron emission from the cathode to anode (or plate) is at maximum before too much heating of the anode or cathode causes physical damage and burns out the plates. This point dictates the maximum power handling or wattage of the device.
Further research led to different valve designs for different applications – especially through World War Two when high frequency radio valves were required with the invention of radar. These designs included use for audio, radio and power applications, and many manufacturers made many equivalent spec. but differently named types, many of which still exist as “New Old Stock” and some are still produced today.
Probably the most common is a 9 pin (nonal), mini dual triode type called the 12AX7 or ECC83.
56 mm x 20 mm diameter
This is widely used both as a dual channel or two stage audio pre-amp in many guitar amps, mixing desks, microphone preamps, and other audio gear, due to its relatively small size and frequency response characteristics. It is widely available from many manufacturers for around US$20 or less.
As a current amplifier, it works due to the principles explained above, and in the simple common (almost) practical circuit below, as a common first stage guitar amplifier example (showing one triode only of the two that exist in one glass housing):
Note that it is a signal inverter when an anode resistor is present – as the grid signal goes up (positive), more (conventional) current flows from anode to cathode, so more of the available 300V is dropped across the load resistor, so the voltage measured at the anode goes down (negative). The amplified output signal is 180 degrees out of phase with the input.
Valve Constants – Gain, Anode Resistance and Transconductance
The control that the grid has over current flow may be say, up to 100 times that of the anode, depending on the triode type and circuit design, so a change of 1V DC at the grid may cause the same amount of current to flow through the valve as if the anode alone had increased its voltage by 100V DC. This proportion (100V/1V example) is a defining characteristic for particular valve types, and is called the gain of a valve and its maximum value is constrained by another valve characteristic – the anode plate resistance (62.5k ohms for an ECC83) – and so the maximum power dissipation ability and breakdown voltage for a specific valve type. For an ECC83 the gain cannot be above 100.
The third defining characteristic of a valve is its “transconductance” or the inverse of resistance, measured in Siemens.
This is a strange concept to grasp at first, as it implies that more current flows as valves “resistance” increases, but it is more a question of voltage observation, as there is no resistance to electron flow in a vacuum from the anode (plate) to the cathode – only through the plate.
As electron flow is proportional to the field strength between the anode and cathode, as the voltage increases, more current flows. The more current that flows means that more of the available HT is dropped across the load resistor – the 100k in the diagram. This means that there is less voltage now available across the valve, yet more current flows through it – an apparent paradox at first sight when compared to the behaviour of other common components in electronic circuits – most of which have a resistance of some form to either AC, DC or both.
Current has increased per voltage drop across it, so the ratio is measured in milliAmps per Volt (the Siemens) so the current to voltage ratio across the valve has increased.
It is the increase in grid voltage that caused more electrons to boil off the cathode in the first place, which is the main reason for the increase in anode to cathode current, not because the anode voltage changed markedly first. This grid control gives the marked increase in current flow for a decrease in voltage across the valve overall, so its resistance appears to decrease, but resistance as a usual concept does not actually exist here.
I view transconductance as the efficiency of a cathode to emit electrons for a given temperature and voltage difference across the anode and cathode. This view enables easier understanding of why the electron flow increases, as the grid has more electrons to attract away from the cathode to be swept up in the anode to cathode field in the first place.
Ohms Law states V = IR for a physical conductor, where:
V = Voltage across the component (a resistor in some form)
I = Current flowing through the component
R = Resistance of the component to the flow of electrons through it.
So, V/I = Resistance. Mathematically inverting this gives I / V or 1/R = Transconductance, so it can be seen that a more freely conducting valve (greater current flow) has a higher Transconductance for a given voltage across it, so how the definition is derived from Ohm’s Law.
In most practical circuits such as the first stage of a guitar preamp, the gain of an ECC83 is optimum around 60 or so, when amplifying AC, which is usually achieved with a 100k load resistor with a 300V High Tension voltage source. Note that valve can still function adequately as a small signal amplifier or distortion stage with HT voltages as low as 80V DC depending on requirements, as explained in the Load Line section later.
So for example, if a 200mV peak to peak Humbucker guitar pickup signal was input at the grid, then the output voltage at the anode may be about 60 x 0.2Vpp =12Vpp or 6Vpeak or 4.2V RMS (Root Mean Square).
Root Mean Square
This is an important maths related concept that is not too difficult to understand and occurs a lot in electronic audio and power circuits that deal with sine waves due to their relationship to circular motion – for example when rotating electric motors and alternators are involved, which generate a fluctuating AC sine wave voltage when spinning in an electric or magnetic field. When this sine wave voltage is applied to a resistive circuit, current flows in the same sinusoidal way, at an “average” between 0V and the maximum peak of each half of the AC cycle.
The value of the square root of 2 is about 1.414…and worth remembering for audio engineers to calculate peak to peak, peak or RMS values for both power and voltage in audio circuits, and relates the amount of equivalent electrical energy expended in an AC or DC circuit.
In a DC circuit the current and voltage is constant, but in an AC circuit it fluctuates between a maximum and 0V, in both directions.
There is a point in the ½ cycle at about 71% of the peak AC value that drives the equivalent average current through the circuit as a fixed DC voltage would to expend the same energy. This value is 1 / Square Root 2 = 1 / 1.414… or 0.707…71%
The RMS value of the power used in the AC circuit is the equivalent of that used in the DC circuit.
240V AC RMS mains = 240V DC
As the RMS AC value is equal to the equivalent DC in voltage terms, then it means that there is a peak AC voltage value above this RMS value that is 1.414 times greater than the DC or RMS AC value.
This means that for the example of UK AC mains at 240V, it has a peak of its sine wave at 240V X 1.414 = 339.5V peak. This is important for valve amp technicians as it is usually the minimum value that mains AC voltage gets rectified to DC (using a 1:1 mains transformer) to run the valves DC power supply and is high enough to be lethal from electrocution.
Some 100W valve amps can have DC supply rails at 550V DC or more by using mains transformers with a greater step up ratio secondary winding.
These can be drawn graphically on a manufacturers Datasheet “grid curve” plot that defines the characteristics of a valve for a range of grid voltage values relative to the cathode, and is available from the Internet in most cases:It can be seen below that for a small change of grid voltage, a large change in anode voltage occurs – more or less depending where on a grid curve you measure the change.
As the grid becomes more positive relative to the cathode, the more the valve is “switched on” (more current flows but the anode voltage drops as explained above) and vice versa.
There are more linear parts of the grid curves higher up and toward the left of the plot, and less linear sections lower to the right. This means that depending what voltage value curve the valve is “biased” at, e.g from 0V to -4V relative to the cathode, dictates how linearly amplified an input signal at the grid will be represented at the anode. For example:
If the grid is biased at -1.5V relative to the cathode as above, a 1V peak to peak AC signal (-1V to -2V) applied at the grid will cause a 60Vpp change at the anode for a 100k load resistor. This shows the AC gain would be 60V/1V = 60.
The logic for this Load Line being drawn for a 100k load resistor is like this:
Imagine an AC signal at the grid that causes a 0V to -4V difference between the grid and cathode.
At one extreme, this value causes all current to be cut off completely as the valve is effectively an open circuit. This means that the total 300V HT voltage is now present at the anode, and no current is flowing. This is the first point that can be drawn on the X axis for this circuit condition.
The next extreme is when a maximum grid input swing to 0V relative to the cathode is input. This causes maximum current to flow in the circuit, so now the valve is effectively a short circuit (no resistance) to the load resistor from ground, so all the 300V HT is across the load resistor with 0V at the anode.
Ohms Law shows that in this circuit condition, 300V across a 100k resistor means a current of 300V/100000 ohms = 3mA flows in the circuit between anode and cathode. This is the second point that can be plotted on the Y axis and the two points joined together by the blue line to show the Load Line for a 100k load resistor at 300V HT.
Now different performance characteristics can be predicted depending on what voltage the grid is set at relative to the cathode under no signal or “quiescent” conditions.
How the input signal at the grid is amplified depends on where the grid is biased, and its effect on the symmetry of an AC signal can be seen below, for example when the grid bias point is set at -2V relative to the cathode:As seen above for a bias of -2V at B, when a symmetrical 4Vpp sine wave signal is input at the grid, and makes the grid negative relative to the cathode, the valve is switched off at maximum swing, and this negative half of the signal gets amplified as a compressed peak at the anode (the voltage is going up here), and when the input signal goes positive from -2V to 0V at the grid, the signal gets amplified as a an elongated trough at the anode (the voltage is going down here).
This effect causes Harmonic Distortion in the output signal and in guitar amps this is sometimes desirable (but not for Hi-Fi amps) as it causes 2nd order harmonics to be produced and can make a richer musical tone, or add warmth to vocals for instance.
The valve has become a tone generator as well as an amplifier.
From the graph, it would seem a bias point set at -1V would give the most symmetrical output for input signals up to 2Vpp before distortion and/or clipping would occur, because the grid curves are more evenly spaced apart in this region. This would give undistorted output values for a 2Vpp input signal (0V to -2V) of about (225V-95V = 130Vpp anode swing) / 2V = 65 gain.
On the same scale as the above diagram, it is of use for the guitarist to know that a guitar output signal is about 100-200mV (single coil-Humbucker resp.) from the average guitar pickup, so when applied to the grid of the first triode in the first preamp valve it only outputs about 5-10Vpp at the anode for a gain of 50:
This is why it is possible and common to use “stomp box” boost type pedals or other amplifiers as an extra preamp to the first triode because there is input voltage headroom available here before distortion occurs.
This effectively shifts the gain stage amplitudes along by a factor of up to ten, which will generate more volume or distortion earlier in the gain stage chain for a smaller turn of the Master volume control. It may also affect the tone, as the first triode – which is normally designed to be a clean (undistorted) amplifier stage to reproduce the natural guitar tone – can be pushed to distortion levels also, effectively adding another layer of distortion and all the associated harmonics that go with it.
These Load Line observations imply that different combinations of load resistor values, bias points and input signal sizes will affect how a signal is reproduced when amplified, and of course this is true, and is used in many preamp stage designs to tailor a particular tonal quality, from enhanced jangly clean tones to full blown clipped, distorted harmonically rich signals that may be “THE” characteristic of an amplifier manufactures tone, such as those often sought by experienced tone fussy guitarists, with Fender and Marshall being those companies whose amplifier tone made them the worldwide successes they still are today.
The simplest way to set the grid to cathode voltage difference – the bias point – is by fixing these connection point voltages relative to ground (0V) and so to each other with resistors. This can be done adding two more resistors to the basic circuit with the anode load resistor already seen above:
The grid resistor Rg acts as an input impedance (usually 1M ohm) for AC signals applied here, and fixes the grid to ground (0V) as no DC current flows through it.
The value of Rk is chosen so that the grid is more negative relative to it, so the triode is toward – but not at – cutoff, rather than at the same voltage (0V) toward saturation.
Reading off the Load Line, if the grid to cathode bias was set at -1V as in the plot above for a clean, symmetric amplification, then the cathode voltage would be at 1V for a grid at 0V, and the current would be 1.5mA on the Y axis.
From Ohm’s Law, this would mean that a resistor of 1V/0.0015A = 1500 ohms had been used for the cathode resistor.
This also shows another rough rule of thumb for both valve and transistor amplifiers, which is that the ratio of anode to cathode resistors approximates the gain of the stage:
100k/1.5k = 66
This is why these anode and cathode load values are seen in many real amp circuits that use an ECC83 preamp valve such as this input from a Laney 100W Klipp.
Negative feedback is a side effect at a single triode stage of a valve amp just with the addition of a cathode resistor. I’ll explain why this happens after looking at the concept of NFB itself.
This concept and circuit design feature was first successfully introduced to the public in a commercially available Hi-Fi audio amp – the Williamson amplifier – in 1947.
Named after the designer, DTN Williamson, after he laid out a set of achievable specifications for Hi-Fi, whereby Harmonic Distortion should be no higher than 0.1% at full rated power, it was shown that with careful design to take inductive and capacitive frequency dependent phase shifts into account, this figure could be achieved.
Ninety degree phase shift can occur at different frequencies in each of these components (inductors and capacitors) relative to a resistor – one component has a voltage that leads the current flow through it and one that lags it. This doesn’t happen for a resistor.
This means that when used in a feedback circuit then there is a displacement along the time axis of more and/or less than a perfect 180 degrees phase shift for most but not all frequencies, so inexact in or out of phase (positive or negative) feedback occurs with complex waveforms, and new waveforms can be produced as a result of this mixing process which naturally distorts the original signal instead of cancelling it.
For example, here are two identical (red and blue) sine waves mixed together when 90 degrees out of phase with each other, to create a 3rd resultant that is identical (another sine wave in this case) but larger – (by a factor of the square root of 2 or 1.414…) – than the two parents (the carrier and modulator in radio jargon), and 45 degrees out of phase with those that created it:
The basic principle of NFB involves appreciating that when an identical signal of the same amplitude is added to another (amplitude modulation) in anti-phase (180 degrees out) then they will cancel each other completely.
If they are added in phase then they add together to create an identical waveform that is twice the amplitude as before.
As this process involves simple addition and subtraction, then it is possible to add a smaller identical wave to a larger one in anti-phase to reduce but not completely cancel out the larger one for any amount up to 100% to give total cancellation. This is proportional negative feedback.
This principle has side effects that have benefits when used in electronic circuits that include:
Diagrammatically this can be shown where a proportion of the output that represents up to 100% of the input signal size is fed back either in phase (positive feedback) or anti-phased (negative feedback):
In the case of positive feedback, if no limit to the maximum possible output is set then the output will continuously grow and add back to the input until circuit oscillation and/or component destruction occurs. Microphone to speaker howl is a common example of runaway PFB.
Understanding the process itself a little better, it may now be understood how it occurs with the addition of a cathode resistor as mentioned above.
When a signal is applied to the grid and makes it more positive, more current flows from anode to cathode, which instantly drops a higher voltage across the cathode resistor than before because of greater current now flowing through it.
This reduces the difference between grid and cathode so opposes the initial effect of the signal grid increasing the difference between grid and cathode in the first place. This has a self balancing effect on the quiescent DC (no signal condition) values of the stage due to temperature changes, component aging and component changes listed above so may now be better understood why this is so.
In the first stage triode, this NFB and its associated reducing effect is often undesirable, as its role is to raise as much of the raw guitar signal as possible to a higher level to not lose any high or low frequency component within it, until it is at a sufficiently strong level to pass to the next stage.
NFB can be negated here with the addition of a relatively large “bypass” capacitor (e.g. 250 microFarad) in parallel with the cathode resistor to allow the lowest guitar frequencies (bottom E is about 82 Hz) to pass to ground at the cathode and so not cause NFB to occur at these frequencies, so they are amplified at this stage and not reduced with NFB. As the capacitor Ck passes AC but blocks DC, the quiescent bias values of -1V DCgk for the above example is not affected, so its gain remains constant.
As a capacitor and a resistor in series and/or parallel act as a tone filter, this means that particular values of capacitor used here can act as a simple form of low pass filter and so cause NFB to occur for some frequencies but not others, thereby boosting those that are bypassed and cutting those that are fed back negatively. This is the principle of the Presence control found on many valve amps though it is usually applied to the grid of a particular type of valve stage called a Long Tailed Pair (see below) because it is variable for frequencies around the upper mid band about 400-6kHz.
A switch can be added in series with the capacitor to make an optional simple tone control to the above stage – to go between a negatively fed back sound or not – a form of “bright” or “top” switch.
There is a lot going on in the simple first triode of a gain stage!
It is an amplifier through current gain, a DC to AC converter, a tone generator through harmonic distortion, a high input impedance stage, a DC decoupling stage if blocking capacitors are used at its input and output, and a low pass filter and/or possible tone control.
Now you can imagine the tonal control and signal increase and/or decrease potential for a whole amplifier that stacks stages and looks like this classic Marshall JMP 100W from the 1960’s:
Note the component values and placement for those of the first triode, mains transformer and DC rectifier stages I have discussed already, in red, and the NFB circuit from the output transformer speaker tap to the presence control in blue that affects the opposite side of the signal grid of the Long Tailed Pair stage. This stage is designed as a “Differential Amplifier” as one input side is grounded via the presence pot and the other accepts the audio signal. The voltage difference between both inputs is amplified.
Note the 1959 reference is not the year it was designed or built, but possibly the model (year) of the Fender Bassman amp from which it was cloned using UK components after the Fender patent expired for the Bassman.
Where Hi-Fi requires minimal distortion of a reproduced signal, to keep as faithful to the original sound as technically possible, it is often desirable for a guitar amplifier to have some distortion to “enhance” the guitar tone, and this can be up to 12% Harmonic Distortion or more compared to the 0.1% in a good Hi-Fi system.
This does not include deliberately designed guitar amp distortion stages where the guitar signal can be converted to near square wave shapes beyond all recognition of the original except for the frequency.
Because valve amps amplify higher frequencies more, generally, then NFB can be very useful in taming ear splitting, dynamic “toppy” amps, by damping a speaker cone’s travel and reducing and/or limiting the size of the signal fed to it.
There are some simple common ways it can be applied to an amplifier across one or more stages simultaneously aside from the local stage NFB that occurs with an unbypassed cathode resistor.
The easiest requires no maths or signal level readings to ensure neither too much nor too little NFB is applied (that can still cause oscillation due to frequencies in the higher or lower sound spectrum being shifted away from a “perfect” 180 degree synchronisation and causing oscillation (PFB) at these frequencies as explained above).
This is simply connecting the cathode (and bypass cap if there is one) from a preamp or power stage (in the correct phase) to one of the 4, 8 or 16 ohm speaker output connections of the output transformer:
The amp stage (power stage in the above case) is still grounded sufficiently via the very low DC resistance output transformer secondary which is grounded at the chassis, so the bias value of Rk (R15) is unaffected. In the above diagram, along with the anode, control grid and cathode is another connection (pin 4) called the screen grid, to create a “tetrode” valve.
Tetrodes and Pentodes – Screen and Suppressor Grids
The screen grid was developed to reduce capacitance between electrodes to aid higher frequency operation.
It had a problem with “secondary emission” of electrons bouncing off the anode causing oscillation, so the pentode was developed from this.
“A suppressor grid is a grid used in a thermionic valve (also called vacuum tube) to suppress secondary emission.”
This aids the suppression of high energy electrons bouncing back off the anode when they collide at very high speed, so preventing them interfering with grid current, which contributes to the cause of oscillation in triodes. This is achieved by connecting the suppressor grid to the cathode so it is very negative with respect to the anode so repels the electrons back to the anode.
The ECC83 valve has 9 physical pins so is used as two triodes (6 pins) as mentioned above, including the three heater filament pins 4,5 and 9. *
This differs from a 9 pin pentode such as an EL84 that fits the same holder, but three of the nine pins are unused and are functionally numbered differently except the heater pins 4 and 5:
Electrodes from top to bottom:
anode (plate) pins
control grid pins
Valve and Transistor Differences
Usually, optimum valve behaviour (DC efficiency and maximum signal swing) is when 1/3 of the available source voltage (300V say) is across the load resistor leaving around 200V across the valve itself for a quiescent state (no AC signal to amplify). This is different from a silicon transistor, when it is usually more efficient to set ½ the available source voltage across the transistor, partly due to the way input and output impedance matching to other stages works. The amplification principle of both devices works similarly though their equivalent connections are named differently.
The “anode” of a transistor is called a “collector” because it is the recipient of the excess electrons that exist in the “cathode” or “emitter”. The grid equivalent, or voltage controlled current regulator is called the “base”. Schematic equivalence between connections is shown below:
The MAJOR difference to realise between these components at this point is that the valve is a current amplifying device primarily, as it can handle much larger current flow than the transistor, so the transistor is a regarded as voltage amplifier only for small signals, as there is still a large proportional voltage change at the collector of a transistor when used with a load resistor, but for a very small current flow.
This difference is key to later understanding of why there are major audio tonal differences between silicon (or “MOSFET”) power transistors, and power valves, when used in power stages to feed loudspeakers directly, as they behave very differently in this mode of operation.
“Frequency Dependent Power Feeding”
This is partly where a lot of the myth and confusion lies in the debate between audiophiles, technicians and/or other musicians as to which type sound “better” for what reason. Personal subjectivity is always debatable but the physics of operation is not, and valve amps sound louder in general because they try to feed more current to an increasingly resistive (AC reactive) speaker load when frequency increases, in a square law proportional increase. Silicon power transistors do not.
This means that more current – therefore power – IS fed to a speaker the higher the frequency that is amplified with a power valve system. Perceived volume IS therefore louder with increased frequency.
Transistor power devices have a more linear response in this respect.
This does not mean that one is “better” than the other, (though from a “perfect” signal reproduction viewpoint, a transistor power stage IS technically more accurate), just that they behave differently under the same conditions, and it turns out that the human ear prefers the increase in volume and the corresponding shift in tonality (generally) that occurs with that overall in an audio system that uses valves in either or both the preamp or power amp stages.
I have personally built or repaired about 11 valve guitar amps in the last year or more and have seen the evidence for this “frequency dependent power-feeding” behaviour of all the valve amps I have taken frequency response curves for, which typically shows an increase in output voltage at the speaker – usually from about 1 kHz up – unless very heavy Negative Feedback is used in the design of the circuit:
Above: FR curve for a 100W Laney Klipp head
Above: FR curve for a 2 channel + boost 40W Ashdown combo
Above: FR curve for PP18W valve kit amp with no NFB
Now compare those plots with the one below when NFB is introduced:
Above: FR curve for PP18W valve kit amp with heavy NFB applied
The effect of NFB is plain to see in the overall flattening of the frequency response of the system to within +/- 3dB from 20Hz to 7kHz.
I modified the PP18 kit amp to have a variable NFB control which at minimum NFB (at a comfortable testing volume) gives a voltage at the speaker for a 1 kHz test tone of 20Vpp and showed 0.5Vpp at the LTP grid.
This is a ratio of 40:1 output to input.
When turned nearly full up the control causes a large drop to 3V at the speaker for 0.75V at the LTP grid.
This is a ratio of 4:1 output to input.
The output amplitude has decreased by 40/4 or 10 times from before to after NFB.
The trough in the curves from about 100-600 Hz is common to all amps I have tested also, and I assumed this was a characteristic of the ECC83/12AX7 preamp, but just realised that speakers have their minimum (rated) impedance (e.g. 4, 8 or 16 ohms) around 450 Hz according to Aiken Amps so maybe this dip is a combination of both components, but mainly the speakers.
Valve and Transistor Distortion Differences
Another major difference to the tonal perception between transistor and valve amps is the way in which clipping occurs. Transistor clipping occurs abruptly, and valve clipping occurs softly, and this difference is the main reason most guitarists choose valves over transistors for finer control of their overdrive settings.
Transistor clipping is often described as harsh or brittle, and a main reason is that a transistor acts more like a switch once it gets to the point of transition between linear current conduction and saturation, so goes more instantly from a clean to full distorted tone, where a valve is far more robust and can handle much more subtle current changes so has a much wider boundary between clean amplification and saturation. If plots are drawn of the amplification or Transfer curves for both a transistor and a valve, this transition boundary is plain to see and so the reason for the way a sine wave is distorted in each case:
By definition, any waveform that has straight horizontal or vertical edges denoting fast transitions has components of a square wave, so contains many odd number harmonics which give a more subjectively non-musical aspect to the tone to most people. This does not necessarily mean it is unpleasant overall, and has a place in heavy metal/thrash type guitar tones for instance.
There is a lot more to an amplifier tone than just valve or transistor characteristics though, as the tone stack design itself (bass, middle, treble controls etc.), component values and quality, speaker models and cabinet design all contribute massively to the final timbre of the sound output from a whole system, with the speaker probably being the most effective though under-estimated single component in the whole amplification chain that affects tone the most, due mainly to how quickly the speaker cone and body can respond to dynamic changes in the output signal. This depends greatly on the mass of the speaker as its inertia and momentum will be larger the bigger the speaker.
The Tremolo Circuit
This sound effect was one of the first used in electronic audio amplifier circuits and is still popular in guitar amps today. The first Fender amp with this feature was the Tremolux in 1955.
Tremolo is the effect of the periodic rising and falling of volume (amplitude modulation or AM) of a sound signal, and electronically it can be achieved by using a positive feedback oscillator. To understand how this works then look at the single triode again but with a feedback loop from the anode to the grid added:
A capacitor is required linking the anode to the feedback resistor Rf, to block the high voltage present at the anode from connecting to the grid and damaging it, let alone messing up the grid to cathode bias voltages already set there. The cap still allows the AC output to feedback to the input via Rf. The cap must also be able to stand the large DC and AC signal voltage swings across it (maybe 300V+) because one side is at DC ground via Rg and Rf, and the other is at the anode.
As the AC signal at the anode is maybe 60 times that of the grid and 180 degrees phase shifted, a suitable value feedback resistor is chosen so the input signal is not squashed out of existence. If this value was chosen to give say ½ the value of the input signal to be fed back then this would reduce the input signal by up to half its initial value. This is another method for adding NFB to a valve or transistor stage.
However, this is not quite what happens with a capacitor link, that alters the phase of the anode signal by up to 90 degrees from the already 180 degree phase shifted signal at the anode, as this makes the resultant slightly larger than its original size!
This would actually cause PFB, as the signal and NFB have added, not subtracted. The resultant is the same whether the NFB is “in phase” or “antiphase” by 90 degrees, just the resultant gets shifted in front (leads) not behind (lags) the signal by 90 degrees instead.
In fact, if you study these sines and cosines in Excel, you find you can’t reduce a signal at all with NFB when it is 90 (=270) degrees out of phase, you only get addition as a result.
As you approach a zero value NFB, it gets closer to leaving the original signal level as it was.
You can see why this gets more complicated with complex waves of different frequencies and so resultants, and it gets more complex the more capacitors involved as they interfere asymptotically, which basically means they converge in non-linear ways, so 2 capacitors in series does NOT produce 180 degrees phase shift just because 1 gives up to 90 degrees. Actually it takes at least 3 capacitors to achieve this phase shift as seen in real circuits below.
This is also why it takes a little time for an oscillator to settle into its resonant periodic cycle – it doesn’t happen instantaneously, which can be audible in some old amplifiers where component drift has thrown out the original component values to a degree that slows down this convergence, and so the effect takes time to “build” sufficiently quickly.
Also, these two additional components act as a tone filter again, so not all of the frequency range of a complex signal may get fed back depending on the value of these components, so this is also an area where exact 180 degree NFB drifts out.
In the case of an tremolo though, PFB is required to make this stage oscillate, but controllably.
This is achieved by the addition of two more capacitors to shift the phase of output to input (which is already 180 degrees out of phase remember) back IN phase again, which works to produce PFB, with a variable pot to change the time period of the oscillations, and a quite simple and clear circuit to help understand this is found in a Marshall Mercury that uses a transistor. I have coloured the components required to understand the PFB principle in blue if it seems too complex overall:
R18, R21 and R22 (Rg) form part of the HT voltage divider chain so that only 26V DC maximum is across the transistor collector to emitter, so it is not destroyed. The periodic output it taken from C24.
The main thing to understand here is that each cap is the same 47nF value and each is paired with a resistor of 220k value to discharge to ground through, except one (C20) that is paired with a 270k (R19) in series with a 100k variable pot (VR3). These last three components enable the speed (frequency) of the PFB oscillations to be varied.
This section functions because a resistor and capacitor in series allows charge to flow on and off the cap at a rate that depends on the value of the cap and resistor, the larger the values the slower the electric charge can flow on and off so the time period is longer. If all three cap/resistor pairings were the same value then the PFB oscillations would settle into a low frequency cycle equal to the charge period of one/all of the pairs as they are the same. By altering one cap’s resistor values with a pot then the oscillations can be changed manually within the maximum and minimum time periods for a 47nF cap and 270k alone or a 270k + 100k = 370k when the pot is at minimum speed (the discharge time is longest).
These values also ensure the system does not go to runaway positive feedback and get out of control, but stay within fixed limits of both period and amplitude, and are pleasing to the ear.
These steady oscillations are amplified at the anode and are passed on by the “Depth” control (VR4) to be added to the power stage grid signal as a low frequency modulator. This controls the degree of swell to the guitar signal. The valve version below is from a Tremolux and has more components than the transistor example as it uses a “cathode follower” design stage that joins two triodes together in an inversely symmetrical way – a Push-Pull stage:
You can still identify the three NFB capacitors circling back round from the anode to the grid as in the transistor diagram, with a Speed pot to ground via a 100k resistor. The effect On/Off footswitch is in the input circuit here instead of the output as in the Marshall.
These work in two slightly different ways – the Marshall just shunts to amplified oscillations to ground when the footswitch is closed circuit, and the Tremolux breaks the PFB loop by shunting it to ground, so the oscillations cannot build in the first place.
An easier to understand and close equivalent to the Marshall circuit above using only one triode of a 12AX7:
For an in depth understanding and component choice for the design of a tube oscillator see here:
The sound of tremolo can be very atmospheric and moody, and an example can be heard of this from a Fender here:
Amplifier Classes – A, B, and AB
The operation of both valve and transistor amplifiers fall into particular classes to compare their behaviour. These extend beyond class B to C, D, G and H but those are beyond the discussion scope of this document.
In analogue circuits the two most common are class A and B as well as a hybrid of the two, class AB.
Class A operation has been described already throughout this document as the main example in a centrally biased valve (or transistor) that amplifies equally both positive and negative halves of an AC signal (for an undistorted output). This design can produce the most accurate reproduction of an input so is most desired for quality HiFi, but it is the most inefficient as it uses 50% of its maximum potential current swing keeping the amplifier biased at the quiescent operating point, so much energy is lost through heat in this design.
Class B operation is where amplification of only one half of an AC signal occurs, and the amplifier is “off” for the other half. This means it can be as much as 78% efficient – as little energy is wasted as heat when the amplifier is not conducting for 50% of the time, and only then at maximum current for a short period of time at the cycle peak.
Class AB designs try to incorporate the best traits of class A and B with compromises made between energy efficiency and accuracy of signal reproduction.
This class has been hinted at with the Long Tailed Pair section which uses two class B triodes working in opposition in a Push-Pull fashion with each triode amplifying one side of the AC cycle each to reproduce the whole input signal overall at the output.
This means there is a point where each triodes “hands over” from an “off state” to the other which is an overlap point that can produce “cross-over distortion” so is less accurate in signal reproduction, but more efficient than class A designs as minimal DC bias current flows through each amplifier as they act in opposition and are both “off” unless a signal is present.
The Future of Vacuum Tube Technology?
“A gate-insulated vacuum channel transistor was fabricated using standard silicon semiconductor processing. Advantages of the vacuum tube and transistor are combined here by nanofabrication. A photoresist ashing technique enabled the nanogap separation of the emitter and the collector, thus allowing operation at less than 10V. A cut-off frequency fT of 0.46 THz has been obtained. The nanoscale vacuum tubes can provide high frequency/power output while satisfying the metrics of lightness, cost, lifetime, and stability at harsh conditions, and the operation voltage can be decreased comparable to the modern semiconductor devices.”