|Elliott Sound Products||Amplifier Classes|
Rod Elliott - Copyright © 2014
There are already many articles on the Net that cover this topic, some quite well (but often without enough information), some badly and some that are largely wrong. It's usually not the descriptions that are incorrect, but the comments about alleged sound quality. For example, some Class-A amplifiers are very good indeed, but others are terrible. It's not only the class of operation that makes an amplifier good, bad or indifferent, but how the circuit is designed and how much effort has gone into minimising problems. Many 'boutique' amplifier makers will make outlandish claims for their chosen topology, but advertising hype is not fact and should be ignored.
Many Class-AB amplifiers are far better than the vast majority of Class-A amps, despite being far more efficient and lacking the gravitas of being called 'Class-A'. There are also some obscure classes, some of which are not defined, and others are useable only with (some) radio frequency signals. There are others where there is no 'official' definition, so there is often confusion about whether an amplifier is one or the other (Class-G and Class-H are the main examples of this).
Amplifier classes that are used exclusively with radio frequencies will not be covered here, only classes that are directly related to audio.
While Class-C is generally thought to be purely an RF technique, it was (at least technically, and if taken to the extreme of normal definitions) used by Quad in their 'current dumping' amplifiers. Output transistor conduction was not quite 180° as required for Class-B. The difference is really academic, so the output stage can just as easily be called Class-B because the conduction angle really is very close to the full 180° for each device in normal operation. Close analysis of the Quad system shows that it largely behaves like a more 'traditional' amplifier, but with unexpectedly low distortion - especially considering the relatively poor power transistors available at the time.
All classes of amplifier (except Class-D) can be made using bipolar transistors, MOSFETs or valves (vacuum tubes). If used in a linear circuit, MOSFETs should be 'lateral' types which have lower gain but are more linear than 'vertical' MOSFETs (the most common types are generally known as 'HEXFETs' because of their internal structure). These types are designed for switching applications, and even the manufacturers don't recommend them for linear use. HEXFETs and other switching types are not linear. Although it's possible to make linear amplifiers using HEXFETs, careful device matching is needed and there are some interesting traps that await the unwary. Naturally, vertical MOSFETs are ideally suited to Class-D amplifiers, where they are used exclusively.
Amplifiers can also be hybrids, meaning that they use a combination of valves, transistors and/or MOSFETs. When we talk of hybrid amplifiers, it is usually taken to mean a combination of valves and semiconductors. Hybrid amps can be any class, but are most commonly either Class-A or Class-AB. While there's no real reason that a valve front end can't be used with a Class-D amp, that is a rather unlikely combination and serves no useful purpose. There are many combinations that serve no useful purpose, but that hasn't stopped advertising people from extolling their (alleged) virtues.
In amplifiers where negative feedback is not used to provide correction and increase linearity, the distortion produced will affect the sound. Harmonic and intermodulation distortion products are created that can seriously reduce an amplifier's performance. This applies regardless of the amplifying device, class of operation or topology. Despite claims by some, negative feedback is not evil, and properly applied in a competently designed amplifier using any of the available devices (valve, transistor or MOSFET) it will almost always improve sound quality overall. Very few amplifiers with no negative feedback will qualify as hi-fi. There are exceptions, but the additional complexity is such that there is little or no overall benefit.
Class-A Output device(s) conduct for complete audio cycle (360°) Class-B Output devices conduct for 180° of input cycle Class-AB Output devices conduct for more than 180° but less than 360° of input cycle Class-C Output device(s) conduct for less than 180° of input cycle (RF only) Class-E, F Sub-classes of Class-C, RF only Class-D Output devices switch at high frequency and use PWM (pulse-width modulation) techniques (Note that Class-D does NOT mean 'digital' Class-G Make use of switched power rails, with amplifiers typically having multiple power supply rails Class-H Use modulated power rails, where the supply voltage is maintained at a voltage slightly greater than required for the power delivered Class-I A proprietary variant of Class-D (it appears that this is not officially recognised) Class-T Another proprietary amp class, and also a variant of Class-D (this is also not officially recognised) BTL Bridge-Tied-Load. Not a class of operation, but sometimes thought to be. Can be applied to any class amplifier
The above is a very basic summary of the different amplifier classes, and all (non RF related) classes are covered below. Note that Classes G and H suffer from great confusion, with the terms regularly used interchangeably. They are quite different techniques, and should be treated as such. No-one appears to have made any effort to categorise them, despite their popularity - especially for high power public address (sound reinforcement) applications.
The term 'Class-A' means that the amplifying device (transistor, MOSFET or valve) conducts for the complete audio cycle (360°). It does not turn off at any output voltage or current below clipping, where the output voltage would otherwise exceed the supply voltage. Since it is not possible for a device to remain linear if the amplifying device is turned off or fully conducting, the output level must be low enough to ensure that neither extreme is reached. In the case of amplifiers that use an output transformer or inductor, the upper limit is actually double the supply voltage, as the inductive element adds an extra voltage that would otherwise not be available. Note that biasing circuitry is not shown in the drawing below. DC flowing in the inductor or transformer winding causes additional problems, and they are related to some of the issues faced by single-ended designs.
By definition, all single ended audio amplifiers are Class-A. They may use inductors, transformers, resistors, active current sources, the loudspeaker itself (bad idea) or even a light bulb as the current source. With all Class-A amplifiers, the amplifying device current must be slightly greater than the peak output current. For example, if the load (loudspeaker) can draw up to 4 amps, the amplifying device requires a quiescent (no signal) current of slightly more than 4A. Where the loudspeaker is used as the 'current source' output power will be limited to a few milliwatts because DC flows in the voicecoil.
Figure 1 - Single-Ended Inductor And Transformer Output Stages
Note that in the two examples shown, the voltage across the amplifying device approaches double the supply voltage. While this might seem unlikely, it is quite normal and is due to the stored energy in the inductor/ transformer. This is added to and released under the control of the transistor or valve. The DC current flow through the inductive element must be at least as great as the peak current demanded by the loudspeaker load (but reduced due to transformer action for the valve example).
Without feedback, both transformer and inductor output Class-A amps tend to have a higher than normal output impedance, and this may also apply to other designs where feedback has been eliminated or minimised. Where transformers or inductors are used, the amount of feedback that can be used is usually quite modest due to high frequency phase shift in the inductive component. Increased output impedance causes colouration in most speakers, especially an increase in apparent bass and extreme treble. This is not because of Class-A, it happens with any amplifier of any class if the output impedance is greater than (close to) zero. Most amplifiers are designed to have an output impedance of less than 100mΩ (0.1 ohm), but 'low' and 'zero' feedback designs can have an output impedance of up to several ohms. Speaker systems are invariably designed to suit amplifiers with very low output impedance.
Amplifiers can be single-ended as shown above, or push-pull. Single-ended valve Class-A is popular in some circles as the so-called SET (single-ended triode) amp as shown in Figure 1. Despite being Class-A, these amplifiers generally have low power (as expected) and often very high distortion. This distortion (both simple harmonic and intermodulation) is due to the basic non-linearity of all valves, and is also partly due to the output transformer. Push-pull operation improves matters, and is described in more detail below.
There are also single-ended transistor (or MOSFET) amplifiers. Those having an inductor load used to be common in early transistorised car radios (almost always using a PNP germanium transistor), but are very uncommon today. Examples of more conventional single ended amps (by today's standards) are the Zen (by Nelson Pass) and the 'Death of Zen' (DoZ) described on the ESP website. These amplifiers are very inefficient, typically managing a best case of 25% (meaning that 75% of all power supplied to the amp is dissipated as heat).
Push-pull Class-A amplifiers use two amplifying devices, and as one conducts more, the other conducts less (and vice versa of course). At no time does either transistor or valve turn completely off, nor do they saturate (turn fully on). By definition, they must conduct (hopefully but rarely linearly) for the full 360° of each and every cycle of audio they amplify. Efficiency is still poor, but distortion is reduced dramatically because the devices are complementary, and second harmonic distortion in particular is cancelled. In fact, all even-order harmonics are cancelled, leaving only relatively low levels of odd-order harmonics. There is no fundamental difference between push-pull amplifiers of any class, other than the bias current. For Class-A, the current through the amplifying devices never falls to zero at any point during the signal waveform, or at any power level.
While it is often claimed that Class-A distortion levels are always lower than Class-AB amplifiers, this is not necessarily the case. A well designed Class-AB amp can often achieve lower distortion and better frequency response overall than many Class-A designs - especially those claiming 'low' or 'no' feedback. Despite claims to the contrary, there is no intrinsic improvement in sound quality from Class-A in any form. Perceived differences are often due to output impedance or perhaps the listener preferring the 'wall of sound' created by higher than normal distortion. There are countless claims that Class-A sounds 'better' than other classes, but this is not necessarily true.
Prior to the widespread use of opamps in small-signal applications, low-level stages were always Class-A, and that remains the case for valve preamp designs. Very low distortion is possible in well designed circuits, but as with power amplifiers there is no 'magic'. It's not commonly accepted, but in general any two preamplifiers of equivalent performance (with equally low distortion and noise, and having the same bandwidth) will sound the same, regardless of the technology used - but only if tested using proper double-blind techniques.
Figure 2 - Power Device Operating Current And Typical Device Gain Vs. Current
In the above (left graph), it is obvious that the current never falls to zero, but it is very important to understand that it is not constant. Because the current varies (from 56mA up to 4.7A), so does the gain of the amplifying device, also shown (right). Valves and transistors are capable of very linear output if the current remains constant, but their gain always varies with current, and this leads to distortion. The gain vs. current graph is taken from the datasheet for a 2N3055, but nearly all devices have the same issue. Note that the typical gain of the 2N3055 varies from over 100 at 200mA down to less than 30 at 5A. There are some bipolar transistors that have remarkably flat gain vs. current graphs, and these give higher performance (and lower distortion) over their operating range, but very few have useful gain at only a few milliamps. Note that most valves have far worse behaviour in this respect - claims that they are "inherently linear" are unfounded.
It might not look like it, but the waveform shown in Figure 2 has over 7% THD. The second harmonic is dominant, but the third isn't far behind. As always, there is a full spectrum of harmonics that diminish smoothly with increasing frequency.
In reality, there are very, very few 'true' Class-B amplifiers. The term 'Class-B' dictates that each amplifying device conducts for exactly 180° of the signal waveform, which implies that they will not conduct at all if there is no signal. While this can certainly be done, the penalty is distortion, which will always be worst at low levels. The above graph showing the gain of a 2N3055 demonstrates that it falls with decreasing current. What is not shown is that at very low current (a few milliamps) the gain falls to almost nothing. While some power devices are a little better, it is unrealistic to expect that any device capable of 100-200W dissipation will have acceptable gain at perhaps 20mA. This applies to all known amplifying devices - including valves.
Low gain at low current means that there must be a region of low overall gain through the amplifier, and that means that negative feedback cannot remove the distortion because the amplifiers open loop gain is very low and little feedback is actually available. The result is what is commonly known as 'crossover' distortion, because it occurs as the signal crosses from one output device to the other.
Figure 3 - Crossover Distortion With Class-B Amplifier
In the above, the crossover distortion around the zero volt point has been deliberately exaggerated so it's easy to see. In reality it can be quite subtle, but is almost always audible, even if a distortion meter shows that overall distortion is quite low. The total harmonic distortion of the amplifier I used to simulate the above was about 1.4% at full power(120W), but because of the nature of the distortion it would be judged (quite rightly) as "bloody awful" by any passably competent listener. True Class-B is virtually impossible with valves, because their gain is too low at very low current. Almost without exception, valve amps are Class-AB - even if described as Class-B.
Because Class-B is not generally considered to be a viable option, it will not be discussed further. However, it should be obvious that Class-B can only be used with a push-pull topology.
To eliminate the objectionable crossover distortion, almost all amplifiers (whether valve or 'solid state') use Class-AB. A small quiescent current flows in the output devices when there is no signal, and ensures that the output devices always have some overlap, where both conduct part of the signal. Some manufacturers claim that their amp operates as Class-A up to some specified power, and this can certainly be true. However, most amplifiers only operate at very modest quiescent (no signal) current, often as low as 20mA. For an 8 ohm load, that equates to a couple of milliwatts of 'Class-A operation' - hardly worth getting excited about.
It's worth mentioning that with valve amplifiers, there are two sub-categories, Class-AB¹ and Class-AB². It's generally accepted that Class-AB¹ means that output valve control grid current does not flow at any time, and with Class-AB² there is some grid current - typically only at maximum output. This means that the control grid becomes positive with respect to the cathode. As with Class-B, push-pull operation is a requirement for Class-AB, which cannot work linearly in any other mode.
Figure 4 - Basic Push-Pull Output Stages
The above stages are highly simplified, but are equally suited to Class-A, Class-B or Class-AB. The only difference between the operating mode is the quiescent current (Iq), which can vary from zero (Class-B) up to 50% of the maximum peak speaker current (Class-A). A valve output stage requires each device to be driven with the opposite polarity, so as one device is turned on the other is turned off. Valves have no complement (opposite polarity device), so they require that each is driven with an opposite polarity signal. The current through each valve must be the same to prevent a net DC from flowing in the transformer windings because that will cause premature core saturation. With the transistor stage, a single polarity signal is used because the transistors themselves are complementary (NPN and PNP), so as one turns on the other automatically turns off.
Transistors (or MOSFETs) can also be used with a transformer output in the same way as the valves shown, but this is very uncommon today. It may still be used for some specialised applications, but is a far from a mainstream technology. Several early transistorised power amps did use output transformers.
In all cases, and regardless of the class of operation (other than Class-B), the quiescent current must be carefully controlled to account for temperature variations. The bias control networks shown need to be adjustable in most cases, and additional measures taken to prevent a phenomenon called 'thermal runaway'. This happens when the transistors get hot, and draw more current than they should. This causes them to get hotter still, so they draw even more current and get even hotter ... until the output stage fails. Thermal runaway is also possible (but uncommon) with valve stages, especially if the control grid bias resistors (not shown) are a higher value than recommended.
Figure 5 - Idealised Current In Output Devices for Class-AB
The above is typical of the current measured through each output transistor for Class-AB operation. We see the transistor current vary between zero up to the full output for one ½ cycle, then do the same in the other transistor for the second. Each transistor is turned on for very slightly more than half the waveform, and the load is shared between them. The upper part of the current waveform is provided by the NPN transistor (see Figure 4), and the negative part is provided by the PNP transistor. Any discontinuity as the signal is passed from one device to the other shows up as crossover distortion, so the bias current (Iq) must be high enough to avoid problems, but not so high that it reduces efficiency or causes excessive heat.
It's only at very low levels that we can see that there is a small area where the amplifier operates in Class-A. As noted above, this is typically only a few milliwatts. The current through the output devices still varies, but over a limited range. In a valve stage the same thing happens, but there's a larger area of 'overlap' where they operate in Class-A. This is not because valves are 'better' - in fact it's because they are far less linear than transistors and need more Class-A area or distortion will be intolerable.
Class-C is only used for RF (radio frequency) applications, because it relies on a tuned (inductor/ capacitor (LC) 'tank') circuit to minimise waveform distortion. Operation is only possible over a very limited frequency range where the tank circuit is resonant. Output device conduction time is less than 180°, but the drive signal is (more or less) linear over the conduction range.
Classes E and F are similar to Class-C, and also use RF amplifier topologies that rely on LC tank circuits. Where class C amplifiers are common below 100 MHz, class E amps are more popular in the VHF and microwave frequency ranges. The difference between Class-E and Class-C amplifiers is that the active device is used as a switch with Class-E, rather than operating in the linear portion of its transfer characteristic.
Class-F amplifiers resemble Class-E amplifiers, but typically use a more complex load network. In part, this network improves the impedance match between the load and the switch. Class-F is designed to eliminate the input signal's even harmonics, so the switching signal is close to being a squarewave. This improves efficiency because the switch is either saturated or turned off. [ 5 ].
First and foremost, Class-D does not mean digital. There are several Class-D amplifiers that accept a digital input (S/PDIF for example), but the class designation was simply the next in line after A, B and C. The first commercial Class-D audio amplifier was produced by Sinclair Radionics Ltd. in the UK in the 1964, but it was a failure at that time because of radio frequency interference and the lack of switching devices that were fast enough to work properly. This was before high-speed switching MOSFETs were available, and bipolar transistors of the time were far too slow. Although the MOSFET was invented in 1962, it took some time before they were commercially available and HEXFETs didn't arrive until 1978. The earliest reference I found to something resembling Class-D was the subject of US Patent 2,821,639 in 1954, but that was a servo system for motor control and was far too slow for audio. There was also a patent taken out in 1967 for what is claimed to be a Class-D amplifier , and many others followed.
For more info and a detailed description of Class-D amplifiers, see the ESP article Class-D that has far more detail than can be included here.
The unfiltered output of a Class-D amp superficially resembles a digital (on-off) signal, but it is purely analogue, and requires high speed analogue design techniques to get a design that works well. It's as far from traditional TTL or CMOS logic ICs as a valve amp design! The output of a Class-D amplifier must be filtered (using an inductor and capacitor) to remove the high switching frequency from the speaker leads and (hopefully) eliminate RF interference. Many Class-D amplifier ICs operate in 'full bridge' mode, and neither speaker lead may be earthed. See BTL (bridge tied load) below for a description.
Class-D amplifiers utilise PWM (pulse width modulation), with a perfect squarewave (exactly 50% duty cycle) representing zero output. A representation of the creation of a PWM signal is shown below. A comparator (literally an IC that compares two signals) is used, with one input fed by the desired signal, and the other fed with a high frequency triangle waveform. If the blue trace shown is filtered using a low-pass filter, the original sinewave will be restored.
The output filter is not actually necessary for a Class-D amplifier to produce sound in a speaker, and the high amplitude switching waveform will (usually!) not 'fry' the speaker's voicecoil because the impedance at the switching frequency is very high. However, without the filter, the harmonics of the PWM waveform will create substantial radio frequency interference over a wide frequency range. This is obviously unacceptable, as it could easily swamp broadcast radio (especially AM) and would cause havoc to other radio frequency bands as well.
Figure 6 - Generation Of PWM Waveform For Class-D amplifier
Notice that for a correct representation of the signal, the frequency of the PWM reference waveform must be much higher than that of the maximum input frequency - usually taken to be 20kHz. Following the Nyquist theorem, we need at least twice that frequency, but low distortion designs use higher factors (typically 5 to 30 - 100kHz to 600kHz). The PWM signal must then drive switching power conversion circuitry so that a high-power PWM signal is produced, switching from the +ve to -ve supply rails (assuming a dual supply topology). The use of BTL allows single supply operation without an output coupling capacitor, simplifying the power supply.
The spectrum of a PWM signal has a low frequency component that is an amplified copy of the input signal, but also contains components at the switching frequency and its harmonics that are removed in order to reconstruct the original (but amplified) modulating signal. A high power low-pass filter is necessary to achieve this. Usually, a passive LC filter is used, because it is (almost) lossless and it has little or no dissipation. Although there must always be some losses, in practice these are usually minimal.
Class-D and its derivatives are the most efficient of all amplifier technologies at medium to high power output. Switching losses mean that Class-D may be less efficient than Class-AB at low power. Early efforts had limited frequency response because very fast switching wasn't easy to achieve. The availability of dedicated PWM converters and MOSFET driver ICs has seen a huge increase in the number of products available, ranging from a few watts up to several kilowatts output.
As with all types of amplifier, there are many claims made about Class-D amps. Descriptions range from "like a tube (valve) amp", to "hard and lifeless" and almost anything you can think of in between. Some claim they have wonderful bass while others complain that the bass is lacking, flat, flabby, etc., etc. Very few of these comparisons have been conducted properly (double blind) and most can be discounted as biased or simply apocryphal.
I have tested and listened to quite a few Class-D amps (as well as 'Class-T' - see below), and most that I've tried are at least acceptable - bass performance in even the cheapest implementations is usually very good indeed, with some able to get to DC easily. There may be cases where the DC resistance of the output filter inductor causes a lower than expected damping factor, but this seems fairly unlikely for most of the better designs.
Some definitely have issues with the extreme top end - I can't hear above 15kHz any more, but I can measure it easily. The output filter has to be designed with a particular impedance in mind, because this is necessary with passive filters. As a result, if the loudspeaker impedance is different from the design frequency above 10kHz, then the response of the filter can never be flat. There is a trend towards using higher modulation frequencies than ever before so the filter can be tuned to a higher frequency, but there will still be some effect.
Figure 7 - Effect Of Output Filter At Different Impedances
All Class-D amplifiers need the output filter - it is essential to prevent radio and TV interference. We know that a passive filter must be designed to suit a particular impedance, but what is the ideal? The problem is that there isn't an ideal, and loudspeaker makers make no attempt to standardise on a designated impedance at (say) 20kHz. A nominal 8 ohm speaker may well measure 32 ohms (or more) at 20kHz, due to the semi-inductance of the tweeter's voicecoil (this usually does not apply to tweeters and horn compression drivers).
In the above graph, I've shown a more-or-less typical filter circuit, along with the response with different load impedances. Should a reviewer's (or customer's) speaker happen to be 16 ohms at 20kHz, then there will be a boost of 3dB at 20kHz with the filter shown. The response isn't deliberately done that way to look bad - it's a simple filter that's fairly typical of those used on commercial Class-D amplifiers. Some listeners will report that the amplifier has 'sparkling' high frequencies, and another will complain that it's 'harsh' and/ or 'ear piercing'. It's neither, it's simply a matter of an impedance mismatch. Some Class-D amps use a Zobel network at the output in an attempt to provide a predictable load impedance at 20kHz and above.
In the past we have never had to worry too much about impedance. The power amp has a very low impedance, speakers have a variable impedance that has a nominal quoted value, and no more needed to be said. Class-D has changed that, but no-one is taking notice. If speaker makers were to add a network that ensured a specific and standardised impedance at 20kHz and above, many of the disparaging claims about Class-D amps might just go away. This would also ensure that some of the (IMO 'lunatic fringe') esoteric speaker cables don't cause amplifiers to oscillate (but that's another story, described in Cable Impedance). Don't hold your breath.
More recently, many Class-D designs have included the output filter in the overall feedback loop, so output level remains constant regardless of load impedance and (signal) frequency. In some cases, the phase shift of the filter is used to set the oscillation frequency (i.e. essentially a 'self oscillating' design). While this should stop reviewers from griping, it almost certainly will do no such thing. There are many Class-D designs that measure (and sound) every bit as good as Class-AB amps, but it's very difficult to remove prejudice from any sighted (i.e. non-blind) listening test.
This type of amplifier is now very common for high-power amplifiers used in sound reinforcement applications. The amps are often very powerful (2kW or more in some cases), but are more efficient than Class-AB. At low power, a Class-G amp operates from relatively low voltage supply rails, minimising output transistor dissipation. When required, the signal draws current from the high voltage supply rails, using a second set of transistors to provide the signal peaks. See the ESP article that describes Class-G amplifiers in detail for more information.
Class-G amplifiers may have from 4 to 8 power supply rails (half used for the positive side and half for the negative). Four rails are quite common, and might provide ±55V and ±110V to the power amplifier as shown below.
Figure 8 - 4-Rail Power Supply Class-G Amplifier Voltages
In the above, you can see that the upper (higher voltage) supplies are used only if the output signal exceeds the lower supply rails (±55V in this example). Lower dissipation means that the heatsinks and transformer can be smaller than for a Class-AB design with the same peak power output. The output signal is shown dashed when it's being provided by the higher voltage supply rails and added output transistors.
Class-G is reasonably easy to implement (less complex than Class-D, but more complex than Class-AB), and because of the increased efficiency, the heatsinks and power transformers needed are somewhat smaller than one might expect for an amp of the quoted power rating.
There are concerns (raised all over the Net) that there will be switching noises as the supplementary supply rails are switched in and out of circuit, but there is no evidence that this is audible with programme material in any competent commercial products. While some noise may be audible (or at least measurable) with sinewave testing, it's doubtful that it will cause any identifiable distortion with speech or music. This is largely because the supplementary supplies are not switched in until the output power is already quite high, and any effects will be insignificant compared to the sound level of the signal. This isn't something I've had the opportunity to test, but major manufacturers would have many complaints if their amps made 'untoward' noises where otherwise equivalent amps did not.
The line between Class-G and Class-H becomes more blurred as more articles are published and more designs are produced. The original Class-H amplifier (which was referred to as Class-G at the time) used a large capacitor that was charged and then switched into the circuit when needed to generate a higher supply voltage to handle transients. Other variants use an external modulated power supply (usually switchmode) that provides a voltage that is just sufficient to avoid clipping, or a supply that's 'hard' switched to a higher voltage when required.
When a Class-H amp uses a switched supply, it doesn't track the input, but is switched to a higher voltage to accommodate signal peaks that exceed the normal (low voltage) supply rails (this is shown in the light green and light blue traces below). There may be situations where the output signal is fairly constant (highly compressed audio for example), and just above the switching threshold. In this case, the amplifier can conceivably dissipate a great deal of power, but it seems that it's not a major problem because thousands are in use and failure reports are fairly uncommon. Because of the switching, a higher voltage to the output transistors is applied only when needed, so output devices are only subjected to a relatively low voltage for much of the time, and receive the full voltage only if necessary. This reduces the average power dissipation, and increases overall efficiency.
Some external supplies are 'tracking', which is to say that they use the audio signal to modulate the supply voltage in 'real time', so it follows the audio signal closely. Another system uses switching, so the supply voltage is raised (from a low voltage to high voltage state) when required to reproduce a peak signal. The amplifier stage itself is linear - usually Class-AB. While making use of one or more separate supply rails for each polarity does increase total output stage dissipation at the transition voltage (it may be dramatic with some signals), the theory is that it will only happen occasionally.
When a power supply modulation principle is used, it's often done using switchmode supplies, and there are two - one for each supply voltage polarity. The quiescent supply voltage is only ±12V, but can increase up to ±110V as needed by the output signal. The tracking supply is shown below in dark green and dark blue.
Figure 9 - Tracking Power Supply Class-H Amplifier Voltages
Do the above qualify as Class-H or Class-G? According to my classification system it's Class-H, but if you prefer to think of it as Class-G then be my guest. Either way, this can be a complex scheme to implement, but can provide the 'sound quality' of Class-AB and close to the efficiency of Class-D. Most switchmode tracking supplies are deliberately slow, so they track the audio envelope rather than individual cycles. This reduces efficiency but makes the supply far easier to implement.
One of the first amps that could be classified as Class-H was the Carver (so-called) 'magnetic field amplifier'. This used switching in the AC mains supply to vary the voltage to the main power transformer. The design was let down by the use of a transformer and heatsinks that were far too small, so sustained high power could cause the 'magic smoke' to escape and the amp wouldn't work any more.
It is commonly accepted by technicians and engineers that all electronic devices rely on 'magic smoke' held within their encapsulation.
Should anything cause this smoke to escape, it means that the device can no longer function. Yes, this is facetious, but the principle is sound .
Because the lines that separate Class-G from Class-H are so blurred (they are really non-existent), it's probably fine to use either term for either type of amplifier. However, it would be nice if some convention was applied so we would know exactly what technology is used in any given amp. My preference is to classify Class-H as any design where the power supply voltage is externally modulated, such as with a tracking switchmode power supply. There is little or no agreement anywhere as to the true distinction between them though, so it's really a moot point. Feel free to consider them differently from my description, or consider them to be the same thing with different names.
Proprietary to Crown Audio, the BCA (balanced current amplifier) is a patented form of Class-D [ 2 ]. It uses a BTL (bridge-tied load) output stage, with two PWM signals in phase with each other. With zero signal, the two switching outputs cancel to give zero output, and with signal each is modulated so that one part of the switching circuit handles the positive portion of the signal, and the other handles the negative portion (allegedly!). It's claimed that the output switching signals are 'interleaved' (symmetrical interleaved PWM), hence Class-I.
It has also been claimed that little or no output filtering is used or needed, but that seems rather unlikely because of RF interference problems. Great and glowing (but largely unsubstantiated) claims are made as to how it is superior to 'ordinary' Class-D amplifiers, but the documentation is sparse and quite unhelpful from a technical standpoint. There are some other claims that don't really stand up to scrutiny as well, but I don't intend to cover this in any more detail.
Intriguingly, there is also a Class-I amplifier described in a Chinese publication [ 3 ] that is completely different from that used by Crown. It's a Class-AB amplifier with an 'adaptive' power supply, which really makes it Class-H (although that depends on the description of Class-H that you might think is the least inappropriate).
Subject of patents, registered trade mark and much hoo-hah, Class-T is simply a slightly different form of Class-D, and still qualifies as Class-D, regardless of alternate claims. Tripath was the original maker of Class-T amplifiers and dedicated single ICs that usually only needed a few external passive components. Despite all the claimed benefits and a fairly wide customer base, Tripath filed for bankruptcy and was bought by Cirrus Logic in 2007. Where Class-T differs from 'classic' Class-D as described above is that the modulation technique does not use a comparator, and the switching frequency is dependent on the amplitude of the signal. As the amp approaches clipping, the frequency falls. It is claimed to be 'different' from other modulators, but there doesn't appear to be much evidence that the difference is significant - despite claims to the contrary. The modulation scheme is sometimes described as Sigma-Delta (Σ-Δ).
Class-T and several other Class-D amplifier makers share similar modulation methods, which at it's simplest simply means adding positive feedback around an amplifier so it oscillates at between 200kHz and 600kHz or so. Naturally, if you were to apply positive feedback to a conventional Class-AB amplifier, it would fail very quickly. The output devices are not nearly fast enough and the remainder of the circuit is not optimised for switching. This means that the actual circuitry is quite different from a conventional amp, but the principle is the same.
When an amplifier is made to oscillate to 'full power' with no input signal, when the signal is applied the duty cycle of the switching waveform will change. As it changes, the amplifier produces PWM by itself, without the need for a triangle waveform generator or signal comparator. A great many claims are made - especially by the now defunct Tripath and devotees - that this method is supposedly much better than all fixed frequency switching, and glowing reports of sound quality can be found all over the Net.
Despite the fact that Cirrus Logic does not appear to have done anything at all with Tripath technology, Class-T ICs are readily available all over the world, with the source of the ICs being China. Whether they are 'genuine' or otherwise is unknown, but one would think that any existing stock would have been depleted in the years since they stopped manufacture. It seems probable that those ICs currently available are not 'genuine' Tripath devices, but they seem to work well enough (yes, I have tried out a couple).
Overall, I doubt that there is really much real difference between a decent 'traditional' Class-D amp and a Class-T, and most of the comments about high frequency 'sweetness' (for example) are simply the result of the output filter interacting with the loudspeaker load. As always, unless comparisons are made using double-blind methodology and are statistically significant, then the 'results' have no value and are meaningless.
This is not a class of amplifier, but a method of using two amplifiers (of any class) to effectively double the available supply voltage. Almost all automotive sound systems use BTL amplifiers in the head unit, and each amplifier can deliver around 18W into 4 ohms from a nominal 12V supply. A single amplifier is only capable of a little over 4W under the same conditions. The only reason that BTL is included here is to dispel the myth that it's a class of operation.
Many commercial amplifiers use the BTL connection as normal, while others (particularly professional equipment) offer BTL as a switchable option to get the maximum possible power (often far more than any known loudspeaker can actually handle without eventual (or even immediate) failure. A basic diagram of a BTL amplifier is shown below, in this case it's a pair of the same amps that were shown in Figure 1 - Class-A inductor load. I used this amplifier because it's the most unlikely - solely to prove a point.
Figure 10 - BTL Connection Based On Class-A Amplifiers
As already explained, using an inductor give you a voltage swing of almost double the supply voltage. The peak-to-peak voltage from each amp is 56V (19.8V RMS), but when connected in bridge the output is 39.6V RMS. Power into an 8 ohm load is 196W, but each amplifier sees an equivalent load impedance of half the speaker impedance. If the individual amps are only rated for 8 ohm loads, then the speaker must be 16 ohms and power will be 98W.
The main thing to remember here is that BTL is not an amplifier class, it can be used with any class of amp.
|Copyright Notice. This article, including but not limited to all text and diagrams, is the intellectual property of Rod Elliott, and is Copyright © 2014. Reproduction or re-publication by any means whatsoever, whether electronic, mechanical or electro- mechanical, is strictly prohibited under International Copyright laws. The author (Rod Elliott) grants the reader the right to use this information for personal use only, and further allows that one (1) copy may be made for reference. Commercial use is prohibited without express written authorisation from Rod Elliott.|