Amp class definitions

Class A amplifier - Class A operation is where both devices conduct continuously for the entire cycle of signal swing, or the bias current flows in the output devices at all times. The key ingredient of class A operation is that both devices are always on. There is no condition where one or the other is turned off. Because of this, class A amplifiers are single-ended designs with only one type polarityoutput devices. Class A is the most inefficient of all power amplifier designs, averaging only around 20%. Because of this, class A amplifiers are large, heavy and run very hot. All this is due to the amplifier constantly operating at full power.The positive effect of all this is that class A designs are inherently the most linear, with the least amount of distortion.

Class B amplifier - Class B operation is the opposite of class A. Both output devices are never allowed to be on at the same time, or the bias is set so that current flow in a specific output device is zero when not stimulated with an input signal, i.e., the current in a specific output flows for one half cycle. Thus each output device is on for exactly one half of a complete sinusoidal signal cycle. Due to this operation, class B designs show high efficiency but poor linearity around the crossover region. This is due to the time it takes to turn one device off and the other device on, which translates into extreme crossover distortion. Thus restricting class B designs to power consumption critical applications, e.g., battery operated equipment, such as 2-way radio and other communications audio.

Class AB amplifier - Class AB operation allows both devices to be on at the same time (like in class A), but just barely. The output bias is set so that current flows in a specific output device appreciably more than a half cycle but less than the entire cycle. That is, only a small amount of current is allowed to flow through both devices, unlike the complete load current of class A designs, but enough to keep each device operating so they respond instantly to input voltage demands. Thus the inherent non-linearity of class B designs is eliminated, without the gross inefficiencies of the class A design. It is this combination of good efficiency (around 50%) with excellent linearity that makes class AB the most popular audio amplifier design.

Class D amplifier - Class D operation is switching, hence the term switching power amplifier. Here the output devices are rapidly switched on and off at least twice for each cycle. Since the output devices are either completely on or completely off they do not theoretically dissipate any power. Consequently class D operation is theoretically 100% efficient, but this requires zero on-impedance switches with infinitely fast switching times -- a product we're still waiting for; meanwhile designs do exist with true efficiencies approaching 90%.

Class G amplifier - Class G operation involves changing the power supply voltage from a lower level to a higher level when larger output swings are required. There have been several ways to do this. The simplest involves a single class AB output stage that is connected to two power supply rails by a diode, or a transistor switch. The design is such that for most musical program material, the output stage is connected to the lower supply voltage, and automatically switches to the higher rails for large signal peaks. Another approach uses two class AB output stages, each connected to a different power supply voltage, with the magnitude of the input signal determining the signal path. Using two power supplies improves efficiency enough to allow significantly more power for a given size and weight. Class G is becoming common for pro audio designs.

Class H amplifier - Class H operation takes the class G design one step further and actually modulates the higher power supply voltage by the input signal. This allows the power supply to track the audio input and provide just enough voltage for optimum operation of the output devices. The efficiency of class H is comparable to class G designs.

Well yes but....

Wards,

You gave the class-A example of only the push-pull case. There is also class-A using a single ended output stage, tube or SS. This is the class-A that is most common in tube amps boasting Single Ended triode. It has a different sound entirely also than push-pull triodes driven class-A.

I cannot think of any driver circuits and voltage gain stage amplifiers not run in class-A except perhaps P-P AB1 used to drive an interstage driving transformer for a larger pair of output tubes, and of course in switching amps, sometimes incorrectly called digital amps.

I have a Panasonic HT receiver that is a Class H+ design. For a fairly inexpensive receiver, it's a very solid and well designed unit that sounds significantly better than its price would suggest. It does run fairly hot while at idle (powered up but with no input signal) that is of some concern.

Wardsweb,

How would you position a non-switching Amp (like a Pioneer SA-9800) in the classes that you have described? Class A?

I think I have just found the answer. Pioneer states (regarding the SA-9800):


"This amplifier has the impressive feature which is the NSA (Non-Switching-Amplifier), created specially by Pioneer. This is a type of amplifier that adopts the merits of both the class A and class B amplifiers, and it excels in keeping down the heat loss and allows the transistors to operate all time in the active region. For this reason, a high output power with a very low distortion factor is yielded all the way up to the high frequencies."


In the case of a Pioneer NSA, is it safe to assume that it is a class AB? The part that bothers me is "adopts the merits of both class A and class B".

So, is a NSA class A or class AB?

"It does run fairly hot while at idle (powered up but with no input signal) that is of some concern."

Jimi,

Not to worry, although the class H tracks the input signal, which it then uses to vary the current over the output devices, there is always a little quiescent current on the output devices even if no input signal is present. This is not a flaw, it's part of the design, keeps the trannies nice warm and improves the rise time of the output device.

My cousin has a Panny 2-ch amp and it was always interesting to me that when the amp was left on for some time with no signal and then a signal is introduced, the temp will actually fall slightly for a minute or so before it gradually rises above the idle temp. Do you also find that or do I need to go take my medicine?

Best Regards

gerrit

So, is a NSA class A or class AB?

Click to expand...

It's just Pioneer's way of avoiding saying that it's a Plane-Jane Vanilla Class A/B amp.

Wardsweb said:

Class A amplifier - Class A operation is where both devices conduct continuously for the entire cycle of signal swing, or the bias current flows in the output devices at all times. The key ingredient of class A operation is that both devices are always on. There is no condition where one or the other is turned off. Because of this, class A amplifiers are single-ended designs with only one type polarityoutput devices.

Click to expand...

Imprecise bordering on incorrect - yes, in class A output device(s) are always conducting to some degree, however, it does not follow by any stretch of immagination that therefore class A amps are single-ended and designed with just one type/polarity of output device.
In fact, there are several distinct variants of class A topologies, some fairly straightforward, some much more ingenious. A few simplified descriptions:

Classical single ended forward amplification: in general, the output device directly varies the load current (developing a voltage across it) - for a transistor or tube design, there is a resistor, choke or transformer in the collector/drain/plate circuit. This circuit can also be used in a differential arrangement where it does not require an output coupler (transformer or capacitor), as well as in a single ended unity coupled transformer variant.

Current shunting single ended: There is an (active) current source in the gain element's collector/drain/plate circuit. The current source sources current to the load, the gain element effectively shunts the 'excess' current away. The current source has a value equal to the maximum load current. Similar to the above, a differential variant is possible.

Current steering push-pull - two output devices, each biassed at slightly higher than half maximum load current. As one device current increases towards maximum load current, the other's decreases towards the 'slightly higher' part of 'slightly higher than half load current'. Neither device is ever fully turned off. This approach uses direct coupled output.

hybrid/bridged forward variants, example: two forward amps with the collector/drain/plate resistor being used as the load, have independant power supplies and share that load so that bias currents through it cancel out. The result is a push-pull amp which does not require transformer coupling but uses a single type/polarity output device. This is suitable, among other uses, as an OTL tube output. Of the solid state variety, IIRC the original Sumo Andromeda used this approach.

Undoubtedly there are many more...

Wardsweb said:

Class B amplifier - Class B operation is the opposite of class A. Both output devices are never allowed to be on at the same time, or the bias is set so that current flow in a specific output device is zero when not stimulated with an input signal, i.e., the current in a specific output flows for one half cycle. Thus each output device is on for exactly one half of a complete sinusoidal signal cycle. Due to this operation, class B designs show high efficiency but poor linearity around the crossover region. This is due to the time it takes to turn one device off and the other device on, which translates into extreme crossover distortion. Thus restricting class B designs to power consumption critical applications, e.g., battery operated equipment, such as 2-way radio and other communications audio.

Click to expand...

Again, inacurate - after the point where it says: output devices each conduct exactly one half of a sinusoidal signal.

Arguably, pure class b is only possible if the output devices had ideal turn-off when the signal is going through zero. This has less to do with the time it takes for the devices to conduct or stop conducting, than the fact there are no such ideal devices. Class b is a mathematical idealisation - mathematically, the signal passes through EXACTLY zero an immeasurably small amount of time, therefore any real device is incapable of working in 'pure class B', as, in contrast to all other classes, class b defines an EXACT angle of conduction (180deg). In other words, there is no bias current that will turn off or on an output device EXACTLY and IMMEDIATELY at one half sine, because the devices are not capable of EXACTLY or IMMEDIATELY turning on or off that way. That being said, here we are really discussing semantics, rather than practical amps.

The linearity of designs approaching pure class B is actually HIGHER than that of class AB designs - not to do with switch on and off times (except in limit cases) but with the fact that when two devices with gain operate at the same time on the same common node, the gain is double that of a single device - since gain is supposed to be constant, this translates to distorsion. This would be the case where class B is biassed 'towards' class A - i.e. both devices conduct at some point in the signal magnitude spread. In the case where the output devices are biassed the other way (towards class C, i.e. there is a part of the signal where neither device conducts), a part of the sine wave around the zero crossing is 'missing' entirely, or heavily distorted. Curiously, while the first case is described as class AB, the second is not described as class BC

In reality, the behaviour of this 'crossover region' is always a compromise. Amongst other things it is dependent on the output device technology, temperature, accurate pairing, load. For small signals, it is arguably the largest source of distorsion and therefore one of the factors that most contribute to the 'sound' of an amplifier.

Also, class B designs, though signifficantly more efficient than class A, definitely do not deserve to be chosen for battery applications only on that merit - at best, class B (that is, 'ideal' class B) has an efficiency of about 70%, and in practise, 50-60% is more common. That being said, until the advent of usable class D, if you wanted less heat (and/or more power from a given output device), you used class B in audio.

Wardsweb said:

Class AB amplifier - Class AB operation allows both devices to be on at the same time (like in class A), but just barely. The output bias is set so that current flows in a specific output device appreciably more than a half cycle but less than the entire cycle.

Click to expand...

Since output devices do not turn off ideally, they are biassed so that a 'quiescent current' passes through both at zero signal input, like in class A, but only in the 'crossover region' around the point where the switchover from one to the other device is located, i.e. where the devices are most nonlinear.
The great problem here is what this current should be so that most of the non-linearity cancels out, wich of course depends largely on what kind of nonlienarity we are talking about. In theory, symetric circuits can supress second order distortion - which is the only kind that devices with a second order characteristic have. The latter would be (In THEORY!) triodes and VFETs. Since complementary push-pull circuits are symetric as far as the scope of this argument, it follows those devices used in class AB would yield superior results. Note that I sais IN THEORY.
As a famous personality would have it, in theory, there is no difference between theory and practise, but in practise, there is - therefore, things are not that clear cut. Suffice to say that the biggest problem here is exact (and I do mean EXACT) matching of two devices. So yes, this is why VFET amps sound great, but also why the VFETs are (were) so expensive

It should be noted that all of the above concerns output stages of a power amp - in an overwhelming number of cases, input and driver stages operate in class A. Currents there are far lower (often several orders of magnitude), and class A inefficiency is not a great concern.

Class C, not discussed in this thread, has limited and indirect use in audio, see below. This is the case where the input signal has to rach some level before the output devices even start conducting. Fed a sine wave, such an output case would be 'missing' entire portions of the sine wave around the zero crossing, resulting in rather horrendous distortion - which is why this is not how it is used in audio. It does have widespread use in RF power amps, though.

Wardsweb said:

Class D amplifier - Class D operation is switching, hence the term switching power amplifier. Here the output devices are rapidly switched on and off at least twice for each cycle. Since the output devices are either completely on or completely off they do not theoretically dissipate any power. Consequently class D operation is theoretically 100% efficient, but this requires zero on-impedance switches with infinitely fast switching times -- a product we're still waiting for; meanwhile designs do exist with true efficiencies approaching 90%.

Click to expand...

Class D amps fall into a cathegory called 'remodulation amps'. Class D amp research is very much a work in progress. Class D output stages are constructed in such a way that they have 'infinite gain' characteristics, i.e. they go into hard clipping for any input signal. This is used to turn input signals into their PWM modulated equivalent using an auxiliary triangle wave generator (although other subtly different approaches exist as well). The modulated signal is then converted back using an output filter.

Because devices in the output are not ideal, i.e. are not ideal switches and do not turn on or off immediately, designers have to satisfy themselves with choosing devices that have these characteristics at least largely predictable.
Dealing with the 'unpredictable part' of that, yields many variations of the basic design. One that is worth noting is a topology which has 3 states on the output rather than two (+, -, zero rather than + or - only). This, together with some creative signal processing, is used as a basis for Tripath class D designs, which they call 'Class T'. Class T is a trade name and does not fit in any way the established definitions of classes - it is merely a subgroup of class D.

One important note here: the reason that class D amps can sound really good, despite the relatively high distortion factor when expressed as a percentage, is that there is no inherent mechanism in class D amps akin to crossover distortion, so in this manner they are similar to class A. They do, however have other mechanisms of distorting, but the distortion produced tracks very well with the output signal magnitude - if done well, there is no sudden rise either at the low or high end of the output power range, discounting clipping, of course. One of the reasons for this is the nature of a class D amp - it is a modulator. The good thing here is that most intermodulation artifacts get shifted around the modulation frequency (100kHz at the very least, more commonly 200 or more), where they subsequently get filtered out by the output filter. This provides a very different distortion spectrum.

Wardsweb said:

Class G amplifier - Class G operation involves changing the power supply voltage from a lower level to a higher level when larger output swings are required.

Click to expand...

Class G and H (and up?) are output stages that are in essence hybrids of two different class output stages. This is where, for instance, class C is used in audio, as one part of the hybrid.

The approach here is to make use of the fact that class B amps have the highest efficiency when they are close to clipping. This is very important because the full power output of an amp is only needed very rarely - the average power of an audio signal is MUCH lower than the peak power. The idea is therefore to keep efficiency close to the maximum for class B by giving the class B amp just enough power to keep it always a bit under clipping. Given the character of audio signals, an appreciable increase in efficiency is possible.

In a class G design, a class A, AB or B amp is driven from power rails that are 'the larger off' a relatively low constant power supply or an output of a another power amp, driven by the same input signal, but off a higher power supply.
When the output signal from the class A/AB/B amp approaches the constant power rails, the other amp provides the 'extra' power supply to the class A/AB/B amp, usually via diode switching networks. The other amp is typically class C, but other approaches have been used as well - full modulation of the power supply using a class D amp (distributed into two switching power supplies) has also been done, by Bob Carver, for instance, in his Sunfire amps.

Wardsweb said:

Class H amplifier - Class H operation... modulates the higher power supply voltage by the input signal.

Click to expand...

The class H designs are a hybrid of a class AB and class D output stage or similar switching amp. Unlike the above example with class G where the power supply signal 'follows' the output once it reaches a certain amplitude, via another power amp (typically working in class C), in class H, the power supply is discretely switched from one power supply to the other (more than 2 are possible but not common) when the output reaches close to the lower of the two. Typically the two rails have a 1:2 ratio. Sanyo uses this approach in their new output hybrids, and this is a technique popular in cheap HT receivers too, either done discretely or using said Sanyo hybrids. Another case are some philips chips used as power amps in car stereo, the higher power supply is derived from the lower by 'bootstrapping' a capacitor charged to the lower power supply, on top of the lower power supply, using bipolar or MOSFET switching transistors.

There seems to be a great deal of cunfusion regarding class G and H power amplifiers. Very often the two are explained exactly oposite than above - with definitions of G and H reversed. In either case, the distinction tends to be blurry depending on which kind of power amp is used for power supply management.

Other variations of both techniques are also possible, usually involving bridging and feedforward techniques.

The Yamaha B-2 incorporated Yamaha's ingenious HCA class A circuit which was a topic of extensive discussion on the Wireless World magazine in the 80s, that circuit used a control input stage amp to keep the signal at class A operational level, the Yamaha delivers cool running yet true class A operation which is truly a luxury.

EchoWars said:

It's just Pioneer's way of avoiding saying that it's a Plane-Jane Vanilla Class A/B amp.

Click to expand...

Actually, there is merit to the 'non switching' claim. There is absolutely no merit in claiming this was exclusive to Pioneer, in fact, the idea was in widespread use as early as 1971, and that's as far as I know of - so probably well before that.

It is a simple modification of class AB - normally two resistors! - that keeps both output devices AT LEAST conducting the bias current. In a classical class b design, especially in MOSFET amps, it is possible for one output device to go completely into the off state (virtually zero current) when the other conducts. This is not per se a bad thing - for instance, it is almost a negligible problem in MOSFET output stages, however, it can be a problem in bipolar output stages. The distinguishing factor is the speed and type of the output devices.

With MOSFETs, driving output devices is essentially like driving a capacitor, which is nonlinear with respect to the output voltage, with very little to do with the output current. In other words, the non-linearity is localized to the output device. Turn-off times therefore remain predictable, and further, the turn on and off region is far more gradual than with bipolars (you pay for this with lower gain and, without an auxiliary power supply, lower eficiency).

With bipolar transistors, there is something called diffusion capacitance, which is a linear approximation of a very nonlinear charge storage mechanism not present in unipolar devices (FET, MOSFET and of course tubes), which is a function of the base current. As such, this is dependant on the output current, which in turn is dependant on the load. Keep in mind that a loudspeaker is a highly nonlinear load. The nonlienarity of the output stage is therefore not confined to the output stage, but depends on many external factors which the designers cannot control, but can, to an extent predict and account for.

High current output devices, such as used in high power amps of that era, tend to be slower and have a lower gain. In other words, their intrinsic turn on and turn off times are high, plus, due to large output currents and lower gain, the base currents tend to be high as well, aggravating the turn off problem due to the diffusion capacitance issue.
Since in a typical class AB solution, the appropriate output transistor turns off once the signal is out of it's angle of conduction, this may become a signifficant problem, because the device continues to conduct even with noi signal for a short while - manifesting as increased distortion at high frequencies, due to the sensitive crossover region being polluted by such switching artifacts. With heavy reactive and nonlinear loads, this problem can become so pronounced that it can lead to oscillation and even failure of the output transistors.
To solve this, the issue of device turn-off characteristic is largely avoided by not alowing it to fully turn off at all - at the expense of slightly lower efficiency.
The technique was commonly used during the 'power wars' of the early 70s, ultimately resulting in semiconductor advances aimed at avoiding the root cause. This is also why it is not in widespread use any more - on the lower power end, LAPT, PowerBase and similar high Ft transistor technologies render the technique largely obsolete, and on the high power end, MOSFETs are used which do not have this problem at all.
Additionally, fast bipolar output devices can actually produce better results if alowed to cleanly turn off completely as the 'Gm doubling' effect present in the crossover region, where both devices conduct, is minimized.

Pioneer did manage to trademark the term "Non Switching Amp". However, I do not know when it was trademarked. I do know that it showed up on Pioneer amps as early as 1979, and maybe earlier.

Yamaha1 said:

The Yamaha B-2 incorporated Yamaha's ingenious HCA class A circuit which was a topic of extensive discussion on the Wireless World magazine in the 80s, that circuit used a control input stage amp to keep the signal at class A operational level, the Yamaha delivers cool running yet true class A operation which is truly a luxury.

Click to expand...

Ingeneous or not, no where in either the owner's or service manual for the B-2 does the use of HCA circuitry get mentioned. The first mention of same is, I believe, in the MX-1000 amp. Someone with the owner's manual for some of the earlier amps may correct this, but is certainly not mentioned in the B-2 literature.

If ilimzn has access to the MX-1000 circuitry I would be very interested in his view as to its operation. While Yamaha claims that the HCA floats the bias up, as needed, to maintain class A operation without the inefficiencies (and heat) of 'normal' class A operation, I am guessing that it more akin to class H or something like that (like all of the newer Yamaha Pro amps).

ilimzn - I have the schematic for the MX-1000 if you would like it e-mailed to you for comment.

Sorry it is the B2x, it wasnt called HCA by Yamaha at that time, I have Japanese magazine test of HCA operation with the MX-10000 and but it is in Japanese language and has the pertinent graphs which clearly indicates class A operation, loosely translated, it describes the working of class A design in detail. The P series Yamaha amps were never meant for HCA operation and are typically class AB as per Yamaha brochures, they dont feature the input control circuit nor APS which is the package for the HCA operation.

Dr. Strangelove said:

Pioneer did manage to trademark the term "Non Switching Amp". However, I do not know when it was trademarked. I do know that it showed up on Pioneer amps as early as 1979, and maybe earlier.

Click to expand...

Among other notable amps, the Sansui AU 20000 has a similar circuit... and that was way before 1979. You would not believe what people trademark...

Yamaha B-2 said:

If ilimzn has access to the MX-1000 circuitry I would be very interested in his view as to its operation. While Yamaha claims that the HCA floats the bias up, as needed, to maintain class A operation without the inefficiencies (and heat) of 'normal' class A operation, I am guessing that it more akin to class H or something like that (like all of the newer Yamaha Pro amps). ilimzn - I have the schematic for the MX-1000 if you would like it e-mailed to you for comment.

Click to expand...

Yes please - don't have much Yamaha literature save for the parts I reverse-engineered myself. I still owe you the mods for the B-2, I'll get to it, I promise.

As far as variable bias goes, it has been done before, many times. In fact, most class A amps of old had the 'power save' bias circuit that would reduce bias for small signals to lower heat, but once the amp was driven full, you would still get the full heater effect, so either way the heatsinks have to be dimensioned for the full output power. I have seen other approaches, which indeed have a lot to do with compound class amplification. While not exactly class H, it was a bridged feedforward design employing a class B power amp coupled with a feedforward class A correction amp. Since correction is just a small fraction of the full signal, the A class portion works of low voltage rails, signifficantly reducing the generated heat.

It should be noted that in a wildly varying signal like audio, the class A definition does offer some lattitude. The definition states that no output devices will get turned off at any point in the full 360 degree signal cycle (full sine), which is obviously dependant on what t he amplitude of the sine is. As long as the amplitude is such that the output current from the amp is appreciably lower than the bias current, you are working in class A, as per definition. However, a 50W amp with a bias current set to run in class A at 1W will sound quite different than if the bias was set for class A operation to the full 50W power output, at the same power, well below 1W. This is because the characteristics of the output devices vary with output current. In other words, it's easy to stamp 'Class A' on the front panel or invent new monikers for amplifier operation, but without providing the details, you really don't know what you are dealing with.

Thanks for the reply. And no hurry on the B-2 mods. You've all ready mentioned a couple of things that make good sense, like the resistor per device rather than per side. I plan to send the amp to EW in the next few weeks.

I'll get the MX-1000 schematic scanned and off to you in a couple of weeks, as well. Again, no promises on timing, as it is too big for my scanner to handle.

I plan to send the amp to EW in the next few weeks.

Click to expand...

Oh o...what am I in for here...?

You know. The stuff we discussed a couple of weeks ago......solid gold wire, diamond encrusted knobs, platinum RCAs. All the standard stuff. But the good news is you get a full 24 hours for turn-around and $10 down and $10 a month for six months. Oh, yes. And a refoam the woofers.