Know Audio: Amplifiers And Distortion

As we’ve traced our no-nonsense path through the world of Hi-Fi audio, we’ve started with the listener, understood the limitations of the human ear, and thence proceeded to the loudspeaker. We’ve learned a bit about speaker cabinets and their design, so it’s time to venture further down the chain to the amplifier that drives those speakers.

The sharp-eyed will be ready to point out that along this path also lies the  speaker cables, but since we’ll be looking at interconnects at a later date we’ll be making the dubious and simplistic assumption for now that the wires between speaker and amplifier are ideal conductors that don’t have a bearing on listening quality. We’ll be looking at amplifiers in enough detail to warrant more than one piece on the subject, so today we’ll start by considering in a slightly abstract way what an amplifier does and where it can fall short in its task. We’ll be introducing probably the most important thing to consider in any audio system, namely distortion.

The job of an audio amplifier is to take an audio signal at its input and present the same signal on its output at a greater amplitude. In the case of a preamplifier it will usually be designed to work with high impedances in the order of 50 kΩ at both input and output, while in a power amplifier designed to drive speakers or headphones it will drive a much lower impedance. Commonly this will be 4 Ω or 8 Ω for loudspeakers, and 32 Ω for headphones.

All Amplifiers Are (Slightly) Broken

In the ideal amplifier, what appears at the output should be a completely faithful replica of what is presented at the input. In reality, this perfectly linear device does not exist, because of the flaws of electronic components. Though we can make amplifiers that are astoundingly good, there will always be a slight difference between the output and input signals beyond merely their amplitude.

My single-frequency THD+N analayser flowgraph in GNU Radio
Measuring single-frequency THD+N through a flowgraph in GNU Radio

If we were to imagine for a moment an extremely non-linear amplifier which when fed with a pure sine wave returned a square wave on its output, most of us are probably aware that the square wave would be the sum of that original sine wave and a series of its multiples, or harmonics. Any audio amplifier we are likely to encounter will be far closer to a perfect linear amplifier than that, but even the smallest non-linearity will result in detectable harmonic levels on the output.

It’s these harmonics generated within the amplifier that are referred to as harmonic distortion, and they are measured by feeding a pure sine wave into the device and measuring the ration of fundamental frequencies to harmonics on the output (Our 2020 April Fool may have been poking fun at high-end cables, but included a method for doing this with GNU Radio.) This is expressed as Total Harmonic Distortion, usually as a percentage of the output which is composed of harmonics. A good quality Hi-Fi amplifier will typically have a %THD in the order of a fraction of a percent. THD is measured at a single frequency at any one time and the quoted figure is usually at 1 kHz, but it’s worth considering the effects across multiple frequencies. It’s usual for an amplifier to have a slightly worse THD at higher frequencies, for example.

The power supply transformer anc capacitors at the front of this amplifier are the largest of its components. Christian Herzog, CC BY 2.0.
The power supply transformer and capacitors at the front of this amplifier are the largest of its components. Christian Herzog, CC BY 2.0.

When we think of distortion it’s harmonic distortion that comes to mind, but THD is not the only type of amplifier shortcoming that can make a difference to the output.  For example, if you deconstruct the task an audio power amplifier performs, from one perspective it amplifies the input to drive the speakers, but from another it forms the bridge between its power supply and via the speaker cables, those speakers.

Put simply, its task is to turn high power DC from the power supply into high power AC for the speakers. Thus the amplifier is only one part of the circuit, and speakers, amplifier, and power supply must be matched in their capabilities and up to the task in hand.

Should you open up a high-end linear Hi-Fi amplifier you’ll find that its case is mostly filled up with power supply components; a very large mains transformer and a block of smoothing capacitors. This is because the power supply is designed to be able to service much higher instantaneous loads than its normal constant load, to adequately handle peaks in the audio stream such as drum beats. The point of exploring this avenue is to understand that the distortion of an amplifier depends not only on the amplifier itself but also the  quality of the components that surround it, in that a seemingly distortion-free amplifier can still be exhibit distortion at times if it is paired with an inadequate power supply.

Another example of distortion from an unexpected source could be found in the stereo tube amp that was my youthful folly. It suffered from an appalling phase response that gave a spectacularly wooly sound, yet it still had a pretty decent frequency response and THD that would otherwise have labelled it as a good amplifier. Fortunately it’s not likely that you’ll be troubled by that in any reasonable amplifier, but we hope to have shown that there are many potential sources for distortion in amplifier design beyond a simple THD measurement of the amplifier block itself.

There are no surprises among the parts.
Is this £10 Chinese audio kit better than any transistor amp simply because it has a couple of tubes? We don’t think so.

There’s one more moment in which you’ll hear distortion mentioned in discussion of Hi-Fi that’s worth exploring here, and it’s unexpectedly in a positive light even if it’s in a very subjective context.

It’s a long-running thing in audiophile circles as to whether a tube amplifier is better than a transistor one, and one of the arguments mobilised in the defence of the tube amplifier is that behind the much-vaunted “tube sound” is that tubes produce more even harmonic distortion while transistors produce more in the odd harmonics.

It’s one of those things that almost certainly has more to do with legend than reality, and probably originates in the days when transistor Hi-Fi was in its infancy and tube amplifiers were a mature technology. Many transistor amplifiers from the 1960s had noticeable flaws, and it’s possible that some of their legacy has rubbed off through the decades. It’s safe to say though that the art of making a very good transistor amplifier was cracked a very long time ago, and aside from the bragging rights there should be little reason to favour one over the other of equivalent quality in 2021.

Trust Only Your Instruments

In the previous paragraph we’re coming close to the original motivation for this series and the conclusion of our distortion overview. Saying that a good tube amp is no better than a good transistor amp will doubtless have some audiophiles exclaiming in disbelief that I clearly have no idea what I’m talking about and the difference is obvious in listening, to which I would like to remind them of the first article in this series about the human ear. Given the choice between a subjective assessment from a human ear and a reproducible measurement from a calibrated instrument, the only trustworthy information comes from the instrument. A hallmark of audiophile writing is often preposterous claims about audio components, so this series is presenting an introduction to audio based upon real engineering. Thus any discussion of distortion should end with this: only take notice of verifiable readings from trustworthy instruments.

We’ll be returning to the subject of distortion in a later instalment when we look in more detail at its measurement. Meanwhile we will return in the next article of the series to look in more detail at the amplifiers themselves. We’ll cover their different types, and their relative merits and disadvantages.

31 thoughts on “Know Audio: Amplifiers And Distortion

  1. Why isn’t predistotrion used with audio amplifiers ?

    If a very high quality microphone was placed correctly, you could even use predistortion to correct for some non-linearities in the speakers.

    (I’m thinking of this from the RF world – plug “Open Source Adaptive Digital Predistortion for High Power Amplifiers” into your search engine of choice).

    1. Use your favorite search engine for “servo subwoofers”. Woofers in general operate within a low enough frequency range that it is not hard to correct for their non-linearities and other undesirable characteristics (undershoot, overshoot, resonance, etc). They usually sense actual cone position vs ideal and correct.

    2. It is not just one part of the signal chain,but a sum total of the signal from source to speaker that makes a quality sound signal for reproduction ,Tube or transistor, I prefer tube chain is only as strong as its weakest link…I have a question for. Audioheads,Longer interconnect or shorter speaker cable??? What do you. think???.

      1. Check out Kii Audio, Bruno Putzeys is in audio engineering known for his breakthrough in Class D amp research (Hypex), but how he designed the Kii speakers is exactly what Elliot said: CAT5 interconnects and 24 amps directly connected to their driver. I love the Kii BXT, so accurate (hearing them, but also measuring them) makes my work effortless. Also less second guessing about which link is the weakest (there are less links).

  2. HIFI – when we were students about 1995, one of the tasks for a weekend was to compare different amplifiers and speakers.
    Measurement tools – none – we had our ears.
    How can tools reflect the complexity or your ears?
    TAKE FIVE was one of the test records at the time where you could compare.
    We built a few boxes, including a large horn and some horns with full range speakers
    – all had their advantages related to the music you like.
    Even built a few amplifiers and pre-amps and compared them.
    I have not seen a measurement toolset of whatever cost that could come close to our ears.

    1. The problem being your ears are your ears, not really portable or reproducible as a standard. If you have “good” ears you might be as asset to someone building or buying equipment around you. That will not be of much use if someone wants to bring something in from a different area. So the idea is that the measurements are not perfect for describing the subtleties but they do give you an overall idea of what you are dealing with.

      1. I find your post rather silly,
        as you do not show an alternative
        – it seems you have none.
        What is your measuremment set?
        It was funny to see
        that the ears of different people in our tests came to the same conclusion.
        What do you have to offer ?

        1. I’ll trust an audio analyser over anyone’s ears, including my own. Calibrated, reproducible, and able to detect far more than we ever can. Have a search for “Audio Precision”.

          1. They always worn people who get hearing aids, or change them, to give it time. The body adapts, so what initially seems strange, is “normal” after a while. And when it changes, the same getting used to has to happen.

            People can say “this sounds better than that” but there are lots of variables.

            At least with test equipment, there is a consistent standard.

        2. As the old story goes…. I measure distance by MY foot length. I would like a bed 10 of my feet in length please. Or in your case, I would like an auditory experience that sounds good to my ears. I can hear higher frequencies to my brothers and sisters (some younger, some older) measuring harmonics in higher frequencies by my ears are different to theirs. You might have come to some conclusions together (or separately, I don’t know your methodology in testing, double blind testing etc…).

          It’s subjective. You can’t measure subjective (except in my case you can measure my foot). It’s the whole reason for an atomic clock or SI having standard measurements for a gram. These items are (almost) universal, and instruments of measurement are calibrated (within certain tolerances) to specific standards of measurements.

          Here is a good (but not perfect) way to test speakers

          1. OMG that guy doesn’t know anything about sound reproduction, I always use his flawed speaker design to illustrate the Dunning-Kruger effect (the camera mic actually catches all the phase issues his design introduces so it is amusing to ‘hear’ him fail). Testing of speakers should be done in open field at multiple angles. So only a few labs can do it properly (NWAA, old but huge nuclear bunker, is one).

    2. Your ears and brains are suboptimal for measuring, especially when you are above 18 years of age (hearing aids anyone?). Sure they are complex but also very limited and very focused on danger queues.

    1. (oops – the forum software ellided my comment between pointy brackets. it should have read:

      In 1985, Stereophile Magazine bet amplifier designer Bob Carver that Bob could not make a transistor-based amplifier sound as good as (author’s favorite high-end audiophile tube amplifier). Bob won the bet…

  3. condnsder the following. Say solid state was the first active electronic devices. After decades of becoming accustomed to that audio, chances are may would have found, tube audio horrible, if that where newtechnology. Keeping in mind it’s possible,to construct solid state equipment, that audiophiles can’t detect,this is one P’n contest I avoid. Unless I see an oppurtuniy, to wind, someone up. :)

    1. Early transistor amplifiers were straight copies of tube amps. It took a while before solid state dumped audio transformers.

      If you want small and cheap, the output transformer of a tube amp was a limit to high fidelity.

  4. Very interesting and promising introdution. Looking forward for next posts.

    As for Tube vs transistor, who cares? Most interesting is, how to get astounding results from each technology? From my experience, amplifier topology is the key point.

  5. I wish you would have mentioned what “power” is (I’m no engineer but believe it is current times voltage). I also wish you would have mentioned what a typical input current and voltage are versus their outputs.

  6. love tube amps, im not going to tell you its better, its not, i fire up that thing up once every few years, but it makes modern music sound like garbage, so it really is just the harmonics, or dare i say; how it handles recorded distortion of guitar and vocal, ditto for music recorded with solid-state and modern mics. its a dream to use to listen to floyd or joplin, a dream, not a spreadsheet, it makes me feel young again, i also have an obsession with anything that lights up… so i made sure to get ahold of one with “space-charge” so it glows a very dim (pale?) cayan(light-blue) between the cathode (emitter) and the anode/plate (collector), freaked me out at first, thought soldering directly to the pins nuked it but it survived hours of playing with no lack of gain/freq.resp. and no increase of blue-ish glow.

    i also agree with michaelblack, if it goes through an audio/output transformer, its going to sound different.
    i had a tv from the 80’s that did not quite sound solid-state OR vaccum-state, turns out it used a large transistor in class-A mode connected to an audio/output transformer less then 2 inches big. well THEREs your design error! it should be twice that size or bigger. then again it was a tv, not meant as hifi.

      1. Class D has come a long way. In pro audio production my guestimate (judging by looking at my colleagues) is that at least 75% is using Class D for their main system. If you look at live event audio (PA), it is closer to 100%.

  7. Measurements are useful to engineers, they tell us that the design is working as intended.
    Customers buy with their ears, or the ears of their friends or reviewers. Within this ecosystem distortion figures are meaningless.
    A smart engineer will respect that disconnect and strive to build products that satisfy the customer.

    1. Figures aren’t meaningless! In blind listening tests, people prefer the flattest frequency response. Note that off-axis response is important – off-axis sounds bounces off things and into your ears, and your brain can filter it out better if it more closely matches the on-axis sound. Check out this great lecture:

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.