Amplifier classes

This thread seeks to explain the different amplifier classes in a fashion so you can make a qualified decision about what your application requires.


  1. Overview of the amp classes
  2. Detailed explanation
  3. Bi-Amping, Bridged- & Parallel Amplifier
  4. TL;DR

Classes overview:

Class Construction Advantages Disadvantages
Class A Single transistor, full-cycle operation - Very high gain possible
- low distortion
- low efficency
- Requires clean power supply
Class B Double transistor push-pull, semi-cylce operation - Higher efficency than class A - Crossover-distortion possible
Class AB Double transistor, more-than half-cycle operation - Improved efficency over Class A or B
- Crossover distortion cancels out
- Inefficent at high output power
Class G
Class H
Variable Supply Serial transistors - Higher output power than Class AB
- Higher Efficency than Class AB
- Complex circuit
Class D PWM modulated Transistor or MOSFET - Very high efficency
- Lightweight
- Requires extensive filtering
- signal/noise ratio is load dependant
- low PWM frequency causes signal degradation

In Detail:

Vin - Signal input
Vout - Signal output
Vcc - Power Source, the + or - indicate power relative to Ground
Gnd/0v - Ground, 0 Volts

R - Resistor
Q / T - Transistor
C - Capacitor
D - Diode

A transistor:
C - Collector, input side
B - Base, controll
E - Emmiter, output
Just like a valve, it does not matter if you have the waterwheel before of after the valve.

The Current flowing from the Base to the Emitter enables a current to flow from the Collector to the Emitter.

For an amplifier, we want the transistor to regulate the flow, not just open and close. It does that, but only in the “saturation region”.

Class A
In a Class A amplifier, the input current is directly proportional to the output voltage of the transistor. Due to the transistor curve, the internal resistance is high during most of the operation causing high thermal losses (= low efficency).

Class B

Class B is similar to Class A, just that instead of one transistor doing all the work, it uses two transistors in push-pull. As each does only half the work, each is closser to their sweet-spot more often.

Each transistor still sees the complete input signal causing distortion.

Class AB

Class AB adds diodes before the transistors. Diodes only allow current to flow in one direction. This way the transistors only work when the other does not.
This eliminates the potential for distortion of Class B.

Class G & Glass H
Often mentioned together because they are the same concept, namely modulating the supply rail depending on output power requested, they are implemented differently.
This “modulated supply”-approach greatly improves efficency and available output power over Class A/B/AB.

To achieve this, multiple supply rails are used.

Class G:
Class G Amplifier are very similar to Class AB, they add a higher voltage power supply they switch in depending on the amplitudes of Vin and desired volume.

Class H:
class H
At low amplitudes, only T2 and T3 are active, the amplifier works as a Class A.
At higher amplitudes, D1 and D3 become conductive switching on Q1 and Q4, the amplfier “stacks” more power on top.
The Diodes D2 and D4 prevent backfeeding voltage into the +40V or -40V power supply.

Class D
In a Class D amplifier, a clock signal and Vin turn on Transistors or MOSFETs. As the supply voltage to the switching element on trigger is always out of the saturation region, switching losses are minimized.

The resulting PWM signal is then smoothed using low-pass filters and inductors to approximate a sine-wave.

Bi-Amping, Bridged- & Parallel Amplifier


  1. Several amplifiers in paralel
    – scales liner with power
    – signal to noise gets worse with every added amplifier
  2. One Amp per frequency
    – Frequency splitter before amplification
    – Tuned amplifiers depending on application

Bridged Amplification
In bridged amplification (or “bridged mode”), the amplifier essentially works in push-pull. This way the available voltage increases the power to a theoretical double.

Parallel Amplification
In parallel amplification, the available current is doubled. This setup is usefull when driving very low resistance speakers.
The danger is in backfeeding the “weaker” (= lower output offset) amplifier with voltage.


Amplifier Class Signal/Noise Distortion Crosstalk Efficiency Complexity Power
Class A :headphones: :white_check_mark: :white_check_mark: :face_vomiting: :white_check_mark: :tired_face:
Class B :slight_smile: :tired_face: :white_check_mark: :neutral_face: :neutral_face: :neutral_face:
Class AB :white_check_mark: :white_check_mark: :neutral_face: :slight_smile: :frowning: :slight_smile:
Class G :white_check_mark: :neutral_face: :neutral_face: :white_check_mark: :nauseated_face: :metal:
Class H :white_check_mark: :neutral_face: :neutral_face: :white_check_mark: :confounded: :metal:
Class D :neutral_face: / :nauseated_face: :face_vomiting: :face_vomiting: :white_check_mark: :slight_smile: :100:

Nice resource :+1: Much appreciated


I heard an enormous difference going from Inexpensive Class A (SMSL sAp-1, 100$-ish) to inexpensive Class D (JDS Atom Amp, 100$-ish).

To me, “Class A” is what sounds the most natural (although the sAp-1 got really laid-back treble).

I’m asking everyone here, do you think “Class A” sounds the most natural/realistic?
Or, it just does not matter as long as it’s done right?
Or maybe the best way to ask this is… have you heard any “Class A” amp that sounded weird, unnatural, soulless?

To my ear Schiit Asgard 3 (class A/B) sounds far more natural than Atom. Now some of that is $200 vs $99. But, the Asgard also sounds way more natural than SP200. I’ve found Micca OriGain speaker amp to sound quite good, though.

1 Like

If Class A has lower distortion, how come all the best measured amps are Class D?

I believe the answer is here:

Class “A” amplifiers are considered the best class of amplifier design due mainly to their excellent linearity, high gain and low signal distortion levels when designed correctly.


Well, then I’ve yet to see any Class A amp which is designed correctly lol.

I still don’t know what exactly makes amps sound different.

For example, THX 789 and Topping A90 both measure excellent. But 789 (or any THX circuit) sound sharper.
The graphs for frequency response shows both are linear.

Or something like RebelAmp. It measures good enough not to be audible. Yet the Class A magic happens and it sounds different.

1 Like

The filtering in the stage after the transistor/MOSFET can be tuned to catch the distortion, so to speak.
In other words, Class-D is cheating.


Numbers like SiNAD and THD+N, tell you only where the noise floor is, the characteristics of the noise have a huge impact on how an amp sounds.
Any noise floor greater than -80dB’s (And I’ve seen people argue the threshold is closer to -60dB’s) will be inaudible.
Any slight variation in Frequency Response in an amp will be completely swamped by by the noise introduced by the FR errors in the headphone.
So for all it matters pretty much every amp has a flat frequency response and a low enough noise floor.

Amps however clearly sound different, some sound brighter, or darker and yet for the most part they all measure flat. Some of this is the fact that measurements are done into constant resistive loads, and most headphones aren’t constant resistive loads. Some is the fact that measuring frequency sweeps or single constant tones is measuring a single aspect of an amp.

In the measurement that commonly do get taken what does matter is the nature of the noise, a -60dB 50 or 60Hz mains hum is detrimental to the sound, a -60dB second harmonic spike can be considered euphonic.

As to the why class A amps sound more natural, some of it will be the mix f harmonics in the noise, but I suspect a lot at the entry level is the reduced complexity.


Because people don’t know how to measure properly.
I do not know how to play guitar, but I have never ever seen someone tune theirs by going guitar hero. But apparantly that is how you measure an amp :neutral_face:

How to class A:

  1. Aquire electronics text book
  2. Open chapter about amplification circuits
  3. Copy the first example
  4. Follow “application circuit” from the spec sheet of the parts you plan to use
  5. Profit

While a single 1khz tone is not good for measuring amps, multiple tones is a good way to show the distortion for the entire length of the frequency response.

But it isn’t necessarily, It capture steady state response to a few frequencies.
That means for example you miss all of the high frequency ringing that surrounds impulses.
An FFT of pure white noise, would probably be better, But that would have it’s own issues.

The problem fundamentally isn’t the measurements it’s the interpretation of them.
Most people listen to headphones between 75 and 85 dB’s, the ambient noise floor in most rooms is over 30 dB’s so any noise below -60 dB’s is well below that floor, he’ll let’s be safe and call it -80dB’s, so why even bother comparing noise floors of -115dB vs -118dB, it’s not relevant.
The measurements have their place to ensure there isn’t something seriously wrong with the design of the Amp or DAC, but while they can provide some indication of how an amp might sound, they aren’t sufficient in that regard, and in some cases the design compromises that improve the noise floor to the insane levels we see today (excessive feedback for example) can have a negative effect on the perceived quality.


Isn’t this what to rebel Amp did? lol :laughing:


I had a look at the Rebel Amp. There is a shot of the PCB on the website.

There is nothing I can see from the trace layout I (with my hobbyist knowledge) that stands out as bad or questionable.
My biggest gripe with it is that he went not fully pedantic and made it completly symmetric over one axis (it is so close to being symetric).
Look at it!


That’s what THX does, right?

So how can I recognize a good amp? Should I go for Class A? Should I try every single amp?

For example, Aune BS1 is a discrete Class A amp (as the company says), and iBasso DX160 is an op-amp Class D player (not sure) which measures far better than BS1.
Which one would be a better choice?

I’m not a measurement whore. I just don’t know what to look for other than that.

As far as I can tell from the PCB and the datasheets, the THX-style use very powerful high speed Op-Amps. In case of the SP200 (based arround the TI OPA564), the Op-Amp is very fast and accurate, which in its main application (motor driver, etc.) is required, could be the cause of the “harshness” associated with THX-amps.

All Amplifier classes have their strenghts and drawbacks.

For delivering audio on a concert, Class G, H or D are the better choices due to higher efficency and available power.

For Headphones, Class G and H make no sense (as far as I can tell), Class B would give you audible distortion. Class D needs some good output filtering, but will do fine.
Class A or Class AB would be my choice (or Op-Amp based because simplicity).

At home in the listening room I would say “depends”. For the Subwoofer, Class D is common, Class G or H would also work.
For the speakers, Class A or AB are desireable, a good Class D will also lift its own weight.


There have been good Class D designs over the years, but they aren’t common at the higher end of the speaker market.
Depending on the efficiencies of you speakers, it’s usually A or A/B, with pure class A usually reserved for more expensive gear because of the heat or power requirements.
I run 2 x 130W Class A Valve monoblocks into my speakers, they each pull over 800W from the wall and generate enough heat to be used as space heaters at a push. They also cost $8000 when new.
But for a while I also used to independently amp the woofers on the speakers with a class D amp.

At the end of the day Class A/B/D or whatever, isn’t what’s important, it’s how it’s sounds. I know for example Jason over at Schiit has done some messing with class D designs, he just hasn’t hit on one he really likes.


Can imagine that being hard. Too much filtering gives you just buckets of mud, not enough filtering grinds your ears to pieces.

It has a lot of feedback, and what’s supposed to be a novel feedback design, though feed forward as they describe it has been around since at leas the 80’s, and has it’s own trade offs.

Like the rest of us assuming you buy that better measuring isn’t fundamentally better, your down to reading/watching reviews and listening.

No idea there is more to an amp than just topology, and I haven’t heard either.

Yes and I agree this is an issue, and I think why a lot of people get pulled into the objectivist camp, it’s a significant investment, and measurements give you an easy way to pin a number on “quality”.

The way this used to work, was you went to a brick and mortar store, discussed what you wanted, then listened to what they had. At any given store you might only hear perhaps 5 or 6 different brands, but it was at least something.

Having local friends who are also into audio is a huge win, a lot of stuff I heard early on was what friends owned.

Honestly at the entry level I wouldn’t over think it, pick one of the respected entry level Amp/DAC’s, and don’t worry about how the others differ too much, most of them sound more similar to each other than different. And you have to take a fairly large step up in price to really get a worthwhile difference.

E30 → Liquid Spark → most any lower priced headphone of your choice is a good start for anyone, and the same is true for almost any of the other popular options around the price. If metrics make you feel better then buy what measures better in that price bracket.
I think you should be looking to spend Bifrost2 money for a significant DAC upgrade over entrylevel (unless you need some feature), and probably similar for a decent Amp upgrade.
A lot of people get bogged down side grading, I mean I own 9 amps under $1000, and a couple more expensive than that, and I probably use 3 or 4 of them on a regular basis.

These are my opinions, and I know there is a lot of hand wringing on entry level gear here, but that’s pretty much the purpose of the forum.


Yup, for a beginner, look at this thread.

1 Like