Slashdot Mirror

...the tube amp distortion might sound good to some people, but it is still distortion.

I am getting so sick of this "tubes sound good because they add distortion" meme that I'm ready to spit. A GOOD tube amplifier, (single-ended, built with triodes that have very linear specs), sounds better than just about any solid-state amp. The subject is way too complex to go into here, but have a look at Lynn Olson's investigations at http://www.nutshellhifi.com/li.... In short, The way THD is calculated makes it a very poor metric for the quality of audio reproduction. Here are a few hints: 1) Odd-order harmonics, and higher-order harmonics, are both much more audible and much more objectionable than low-order and even-order harmonics. This has been known since at least the 1940's, perhaps earlier. 2) THD measurements treat ALL spurious harmonics as if they contribute equally to degradation of sound quality. 3) Solid-state devices, pentode tubes, and push-pull amplifier topologies using any kind of amplifying device, require fairly large amounts of negative feedback. Although NFB results in better THD performance, it also results in more high-order and odd-order harmonic content. 4) Good triode tubes in a competently-designed amplifier have better intrinsic linearity than any other amplifying device available. Consequently, they require less NFB, and therefore generate less high-order and odd-order distortion.

In short, THD is a shitty figure-of-merit because its manner of calculation was chosen to make pentode-based push-pull amps look good on paper. Well-respected audio engineers at the top of their field argued for THD being calculated using either the square, or even the cube, of the order of the harmonic. They lost the argument, likely 'because business'. And since then, generations of techs have blindly believed that if it measures low in THD, it's all good. Distressingly often, that just ain't so.

And then there's the consideration of IMD, which adds a whole 'nother level of complexity.

... they produce more distortion... They do not sound better, given $X spent on whatever, presuming some reasonable amount of tech is returned per dollar.

This is simply not true; in fact, it's a line of BS almost as bad as the one that says thousand dollar cables will make your system sound magical. (BTW, something you hear is not necessarily true just because it's said by a bunch of engineers bathed in the fountain of 'knowledge' passed on by an earlier generation of engineers whose conclusions were 'helped' by marketing requirements).

First, let's look at the metric known as THD; it is calculated as the ratio between the total power of all harmonics and the power of the fundamental. It assumes that all harmonics are equal in their detrimental effects on the final sound. However, we've known since the 50's that: 1) odd-order harmonics are more audible, and more objectionable, than even-order harmonics; 2) audibility and objectionability of harmonics increases as the order increases. In fact, the best audio engineers of the day, (such as the BBC's D.E.L. Shorter), proposed that THD be calculated based on the square, or even the cube, of the order of the harmonic. They were overruled, probably because the amplifier manufacturers of the day had turned to push-pull pentode designs with lots of NFB; these had many more odd-order and high-order harmonics than earlier triode designs, and the unweighted THD calcs made them look better on paper.

Why does this matter? Well, first, negative feedback obtains lower THD figures at the expense of creating more high-order harmonics - the total power in the harmonics is decreased, but the number of measurable harmonics, (and they, or at least their effects, are audible, as it turns out), increases. Second, negative feedback can increase the number of odd-order harmonics, depending on the feedback scheme and the amplifier topology. What does this have to do with tube VS transistor? In short, triode vacuum tubes, (the 'triode' part is important), have far greater intrinsic linearity than any semiconductor device. (Intrinsic linearity is a measure of the device's distortion in a circuit that has no negative feedback). The best directly heated triode power amplifiers can obtain percent THD figures in the low single-digits without the use of any negative feedback. That means that they can avoid excessive odd-order and higher-order harmonics to an extent that no transistor amplifier can - transistor amps simply require too much NFB.

For a better view of this that is both broader and deeper than my explanation above, see Lynn Olson's excellent explanations here, and here

As for tubes' fragility: although they wear out and can break, they are also more likely to survive EMP's, solar flares, lightning strikes, and power-line surges... ;-)

... they produce more distortion... They do not sound better, given $X spent on whatever, presuming some reasonable amount of tech is returned per dollar.

This is simply not true; in fact, it's a line of BS almost as bad as the one that says thousand dollar cables will make your system sound magical. (BTW, something you hear is not necessarily true just because it's said by a bunch of engineers bathed in the fountain of 'knowledge' passed on by an earlier generation of engineers whose conclusions were 'helped' by marketing requirements).

First, let's look at the metric known as THD; it is calculated as the ratio between the total power of all harmonics and the power of the fundamental. It assumes that all harmonics are equal in their detrimental effects on the final sound. However, we've known since the 50's that: 1) odd-order harmonics are more audible, and more objectionable, than even-order harmonics; 2) audibility and objectionability of harmonics increases as the order increases. In fact, the best audio engineers of the day, (such as the BBC's D.E.L. Shorter), proposed that THD be calculated based on the square, or even the cube, of the order of the harmonic. They were overruled, probably because the amplifier manufacturers of the day had turned to push-pull pentode designs with lots of NFB; these had many more odd-order and high-order harmonics than earlier triode designs, and the unweighted THD calcs made them look better on paper.

Why does this matter? Well, first, negative feedback obtains lower THD figures at the expense of creating more high-order harmonics - the total power in the harmonics is decreased, but the number of measurable harmonics, (and they, or at least their effects, are audible, as it turns out), increases. Second, negative feedback can increase the number of odd-order harmonics, depending on the feedback scheme and the amplifier topology. What does this have to do with tube VS transistor? In short, triode vacuum tubes, (the 'triode' part is important), have far greater intrinsic linearity than any semiconductor device. (Intrinsic linearity is a measure of the device's distortion in a circuit that has no negative feedback). The best directly heated triode power amplifiers can obtain percent THD figures in the low single-digits without the use of any negative feedback. That means that they can avoid excessive odd-order and higher-order harmonics to an extent that no transistor amplifier can - transistor amps simply require too much NFB.

For a better view of this that is both broader and deeper than my explanation above, see Lynn Olson's excellent explanations here, and here

As for tubes' fragility: although they wear out and can break, they are also more likely to survive EMP's, solar flares, lightning strikes, and power-line surges... ;-)

ignore the DAC the amp the source and everything... ...except the speaker drivers themselves. even the best in the world are wildly non-linear.

and then there's the air between your ears and the speakers

another non-linearity

Best source? .0001% THD. best amp? .0001% THD. Speakers? 1% THD haha good luck.

There is a fundamental problem with your argument, and that is the failure to take into account the nature and type of the distortion. It's not your fault - you share the misconception with most audio engineers, (who ought to know better), that THD figures correlate well with listening tests.

Quoting from my own comment in an earlier Slashdot story:

"THD measurements are taken as the ratio of the total power of all harmonics to the power of the fundamental, with no weighting of any kind applied. The trouble is, human hearing doesn't respond to harmonic distortions in this linear fashion - our ears find higher order harmonic distortions much more apparent and objectionable. This deficiency was noted by prominent BBC engineers D.E.L. Shorter and Norman Crowhurst in the 40's and 50's, when they proposed weighting harmonics by the square or the cube of the order; but their voices were drowned out by market forces that wanted a simple, flattering figure of merit that made the newer, more powerful pentode-based amps, (with lots of negative feedback), look better on paper than their lower-powered triode predecessors. The market won out over scientific and technical accuracy, (it usually does), and today engineers the world over, ignorant of this history, mistakenly believe that low THD is the gold standard for measuring and defining audio amplifier quality. (For a good technical analysis of distortion and the sound of an amplifier, see Lynn Olson's excellent investigation)."

Yes, speakers are hugely non-linear - but their non-linearity doesn't make distortions earlier in the reproduction chain inaudible, even though those distortions can be several orders of magnitude smaller. And that applies to all earlier distortions, whether they originate in an amplifier, a DAC, the digital encoding, or the recording equipment itself. Also, an amplifier with 1% THD can sound much better than one with 0.001% THD, not because the distortion in the 'poorer' amp sounds good, but because the distortion in the supposedly 'blameless' amplifier sounds bad. Then there are Intermodulation Distortion and Transient Intermodulation Distortion, which are difficult to measure thoroughly and seldom appear in amplifier specs, yet are often audible.

Audio quality isn't nearly as simple as THD figures imply, nor as simplistic as most manufacturers of audio equipment would have you believe.

ignore the DAC the amp the source and everything... ...except the speaker drivers themselves. even the best in the world are wildly non-linear.

and then there's the air between your ears and the speakers

another non-linearity

Best source? .0001% THD. best amp? .0001% THD. Speakers? 1% THD haha good luck.

There is a fundamental problem with your argument, and that is the failure to take into account the nature and type of the distortion. It's not your fault - you share the misconception with most audio engineers, (who ought to know better), that THD figures correlate well with listening tests.

Quoting from my own comment in an earlier Slashdot story:

"THD measurements are taken as the ratio of the total power of all harmonics to the power of the fundamental, with no weighting of any kind applied. The trouble is, human hearing doesn't respond to harmonic distortions in this linear fashion - our ears find higher order harmonic distortions much more apparent and objectionable. This deficiency was noted by prominent BBC engineers D.E.L. Shorter and Norman Crowhurst in the 40's and 50's, when they proposed weighting harmonics by the square or the cube of the order; but their voices were drowned out by market forces that wanted a simple, flattering figure of merit that made the newer, more powerful pentode-based amps, (with lots of negative feedback), look better on paper than their lower-powered triode predecessors. The market won out over scientific and technical accuracy, (it usually does), and today engineers the world over, ignorant of this history, mistakenly believe that low THD is the gold standard for measuring and defining audio amplifier quality. (For a good technical analysis of distortion and the sound of an amplifier, see Lynn Olson's excellent investigation)."

Yes, speakers are hugely non-linear - but their non-linearity doesn't make distortions earlier in the reproduction chain inaudible, even though those distortions can be several orders of magnitude smaller. And that applies to all earlier distortions, whether they originate in an amplifier, a DAC, the digital encoding, or the recording equipment itself. Also, an amplifier with 1% THD can sound much better than one with 0.001% THD, not because the distortion in the 'poorer' amp sounds good, but because the distortion in the supposedly 'blameless' amplifier sounds bad. Then there are Intermodulation Distortion and Transient Intermodulation Distortion, which are difficult to measure thoroughly and seldom appear in amplifier specs, yet are often audible.

Audio quality isn't nearly as simple as THD figures imply, nor as simplistic as most manufacturers of audio equipment would have you believe.

Yes, transistor audio amplifiers are "way more linear" with gobs of negative feedback applied, if THD, (Total Harmonic Distortion), is your measurement criterion. ANY amplifier is more 'linear' with correctly applied negative feedback. The basic premise is that added harmonics are bad - if you feed a pure sine wave into an amplifier, you want a pure sine wave at the output. The problem is that in audio, THD is a fundamentally flawed measurement with very poor correlation between lab measurements and listening tests.

THD measurements are taken as the ratio of the total power of all harmonics to the power of the fundamental, with no weighting of any kind applied. The trouble is, human hearing doesn't respond to harmonic distortions in this linear fashion - our ears find higher order harmonic distortions much more apparent and objectionable.. This deficiency was noted by prominent BBC engineers D.E.L. Shorter and Norman Crowhurst in the 40's and 50's, when they proposed weighting harmonics by the square or the cube of the order; but their voices were drowned out by market forces that wanted a simple, flattering figure of merit that made the newer, more powerful pentode-based amps, (with lots of negative feedback), look better on paper than their lower-powered triode predecessors. The market won out over scientific and technical accuracy, (it usually does), and today engineers the world over, ignorant of this history, mistakenly believe that low THD is the gold standard for measuring and defining audio amplifier quality. (For a good technical analysis of distortion and the sound of an amplifier, see Lynn Olson's excellent investigation.

By the way, in the 'tubes vs transistors' debate, good triodes have the advantage of being more intrinsically linear than transistors. This means that they require less negative feedback to tame their distortion, and often sound wonderful with NO negative feedback. The THD figures of amps built this way are often quite poor, but look at their spectra and you'll see predominantly second- and third-order, with a smooth and rapid falloff of higher order harmonics. Occasionally solid-state amps can give this kind of performance, but tubes have an easier time of it. Designing a good-sounding, (as opposed to good-measuring), audio amp, requires a lot of skill, and a lot of knowledge about distortion mechanisms and how to counter them. Unfortunately the prevailing practice in HiFi is to add more gain, throw most of it away with additional NFB, get a nice low THD figure, and call the job done. Amps designed this way generally sound like shit, if not initially, then after 20 minutes or so of listening, at which time listening fatigue sets in.

Yes, transistor audio amplifiers are "way more linear" with gobs of negative feedback applied, if THD, (Total Harmonic Distortion), is your measurement criterion. ANY amplifier is more 'linear' with correctly applied negative feedback. The basic premise is that added harmonics are bad - if you feed a pure sine wave into an amplifier, you want a pure sine wave at the output. The problem is that in audio, THD is a fundamentally flawed measurement with very poor correlation between lab measurements and listening tests.

THD measurements are taken as the ratio of the total power of all harmonics to the power of the fundamental, with no weighting of any kind applied. The trouble is, human hearing doesn't respond to harmonic distortions in this linear fashion - our ears find higher order harmonic distortions much more apparent and objectionable.. This deficiency was noted by prominent BBC engineers D.E.L. Shorter and Norman Crowhurst in the 40's and 50's, when they proposed weighting harmonics by the square or the cube of the order; but their voices were drowned out by market forces that wanted a simple, flattering figure of merit that made the newer, more powerful pentode-based amps, (with lots of negative feedback), look better on paper than their lower-powered triode predecessors. The market won out over scientific and technical accuracy, (it usually does), and today engineers the world over, ignorant of this history, mistakenly believe that low THD is the gold standard for measuring and defining audio amplifier quality. (For a good technical analysis of distortion and the sound of an amplifier, see Lynn Olson's excellent investigation.

By the way, in the 'tubes vs transistors' debate, good triodes have the advantage of being more intrinsically linear than transistors. This means that they require less negative feedback to tame their distortion, and often sound wonderful with NO negative feedback. The THD figures of amps built this way are often quite poor, but look at their spectra and you'll see predominantly second- and third-order, with a smooth and rapid falloff of higher order harmonics. Occasionally solid-state amps can give this kind of performance, but tubes have an easier time of it. Designing a good-sounding, (as opposed to good-measuring), audio amp, requires a lot of skill, and a lot of knowledge about distortion mechanisms and how to counter them. Unfortunately the prevailing practice in HiFi is to add more gain, throw most of it away with additional NFB, get a nice low THD figure, and call the job done. Amps designed this way generally sound like shit, if not initially, then after 20 minutes or so of listening, at which time listening fatigue sets in.

Yep. Almost like you said, BUT you remebered them the wrong way. Tubes produce even harmonics and solid state amps (where are not really talking about digital amps here..) produce odd order harmonics.

Here are some nice looking articles:
Herron Audio: TAS tube article
Vacuum Tube Primer
The Sound of the Machine - The Hidden Harmonics behind THD