That is taken to extremes, and quite out of the range we deal with.
Person 1: "My car doesn't handle well"
Person 2: "it might help if you put tires on those wheels"
dave
THD is next to meaningless because the collapsing of the HD spectra into a single number removes the important information and has little correlation to what the ear/brain perceives. We need to know where the distortion is, but that is still subject to interpretaion.
It only takes a small amount of higher order HD (4+) to be more painful than a whole lot more than 2 or 3.
dave
a simple THD vs Level plot will give strong clues as to the dominant order of the nonlinearity
even the LM3886 datasheet has THD vs Level, Frequency, Load graphs
only those dissing "THD" try to make it about a single number at a single level, frequency
even the LM3886 datasheet has THD vs Level, Frequency, Load graphs
only those dissing "THD" try to make it about a single number at a single level, frequency
My take is that some people posting here are somewhere else involved with some audio snake oil peddling.
Any opinions contrary to their business interests will start a storm of snarky remarks with no other purpose than burying the thread. They're easy to recognize though since their remarks are empty of any substance or reasonable counterpoint - a typical one would be like "this thread is a monumental display of ignorance".
Last edited:
Yes, saying THD tells nothing at all is silly. However as you point out it's also true that within a spectrum it's also useless for being informative towards preference; typically depending on harmonics more than amount.
And actually you're illustrating it is about data. The theories exist for a directive of information gathering. "I bet if this than this" and data is collected. Otherwise we'd be busy writing down everything including about us writing down everything... It's just not an attractive way to say it. Science doesn't look for right and wrong, but it does try to predict (theories) where it's going to find information (significant predictable variances); and hope it's useful. And of course this helps explain why there's moral oversight seen so much, since science at it's heart isn't seeking morality/right&wrong.
We're drifting away from “respect for science” and “THD is useless / useful” … which kind of misses the OP's point.
Yet, it also is a valid talisman of the larger “science respect” discussion: so much of what we do in the audiophile / DIY world is centered around abstractions of complex systems with simplified numbers.
When I read, “2nd order distortion is sweet” (or other such stuff), all it means to me is that the amplifier either has very soft clipping, or that its linearity is fine, but its transconductance curve is S shaped, amplifying small signals somewhat more, larger ones less, up to the clipping point, where it all breaks down in a subjective sort of way.
When I read, “3rd order distortion”, I see visions of oscilloscope traces that visibly (often copiously) droop off smooth peaks. 4th order, zero-crossing distortion. 5th order, weird stuff in the waveform, everywhere. "All Odd" distortion just seems to show up as a really non-straight transfer curve, with odd low-level attenuation, make-up for that at mid levels, then compression of peaks.
The point? The point is that THD is just as useful as a single characteristic of all that stuff could possibly be: qualitatively useless, but quantitatively precise. Good science. Little empirical steerage.
Indeed - even the maxims which various really smart, informed people here spout regarding the presence of certain harmonics is just patently pointless, when in fact, what's really affecting the input-to-output waveform transfer is NOT exquisitely well crafted harmonics, but simply waveform amplification nonlinearities. Nonlinearities that vary with the gain and/or input level. Nonlinearities that might not even be time-independent (power supply condenser droop at louder levels of output, and relative to program material itself!)
In a scientific sense, to call signal transformation nonlinearities “harmonic distortion” is a gross misstatement.
But that's the point, isn't it? A whole lot of science (and engineering) is simplifying stuff which can be measured to varying precision, with computed results that make systems comparable. Distilled down to 1 number. Or to a handful of them. EVEN IF the representation is hugely misleading. EVEN IF everyone knows its rather misleading. Because when it takes 40 numbers or god forbid, a graph to represent the transfer characteristics of an amplifier, the weak flee. They gather at shows, and stand around pontificating about this, or that, or the other amplifier's new numbers, compared to the ones made in the 1970s. Or the effect of using thoriated tungsten Russian pentodes versus the Telefunkens made before 1955.
Or the advantages of tube amplifiers over sand-state. Or vice versa. Or the horrible / exquisite sound stages created (or tickled out) by Aikido line stages, yada, yada, yada.
Let's face it, goats… all we really want are amplifiers with nice smooth transfer curves, without odd little playback level bumps and wiggles. Some people swear by the nearly ideal flatness of well-made sand-state topologies. Others lose their grip on reality when they sit before a hugely radiant infrared vacuum emitter, with its natural S shaped transfer curve. None (with one key exception: rock musicians) of us want transfer curves that are time dependent, varying with program material, line fluctuations, position of the input volume controller. We don't. We don't really want amplifiers that introduce a lot of waveform artifacts, some of which are “harmonic” and others which clearly are not. We don't want our amplifiers fighting to control the speakers so tightly that that in turn creates a whole army of new artifacts. We don't.
And therein is the segue to return to the OP's point: respect for science. I could be entirely off, but I think the previous paragraph kind of epitomizes why SCIENCE is involved with the design, building and tuning of amplifiers. And why "THD" is both very useful, and sometimes awfully bone-headed as single-number measures go.
GoatGuy
Yet, it also is a valid talisman of the larger “science respect” discussion: so much of what we do in the audiophile / DIY world is centered around abstractions of complex systems with simplified numbers.
When I read, “2nd order distortion is sweet” (or other such stuff), all it means to me is that the amplifier either has very soft clipping, or that its linearity is fine, but its transconductance curve is S shaped, amplifying small signals somewhat more, larger ones less, up to the clipping point, where it all breaks down in a subjective sort of way.
When I read, “3rd order distortion”, I see visions of oscilloscope traces that visibly (often copiously) droop off smooth peaks. 4th order, zero-crossing distortion. 5th order, weird stuff in the waveform, everywhere. "All Odd" distortion just seems to show up as a really non-straight transfer curve, with odd low-level attenuation, make-up for that at mid levels, then compression of peaks.
The point? The point is that THD is just as useful as a single characteristic of all that stuff could possibly be: qualitatively useless, but quantitatively precise. Good science. Little empirical steerage.
Indeed - even the maxims which various really smart, informed people here spout regarding the presence of certain harmonics is just patently pointless, when in fact, what's really affecting the input-to-output waveform transfer is NOT exquisitely well crafted harmonics, but simply waveform amplification nonlinearities. Nonlinearities that vary with the gain and/or input level. Nonlinearities that might not even be time-independent (power supply condenser droop at louder levels of output, and relative to program material itself!)
In a scientific sense, to call signal transformation nonlinearities “harmonic distortion” is a gross misstatement.
But that's the point, isn't it? A whole lot of science (and engineering) is simplifying stuff which can be measured to varying precision, with computed results that make systems comparable. Distilled down to 1 number. Or to a handful of them. EVEN IF the representation is hugely misleading. EVEN IF everyone knows its rather misleading. Because when it takes 40 numbers or god forbid, a graph to represent the transfer characteristics of an amplifier, the weak flee. They gather at shows, and stand around pontificating about this, or that, or the other amplifier's new numbers, compared to the ones made in the 1970s. Or the effect of using thoriated tungsten Russian pentodes versus the Telefunkens made before 1955.
Or the advantages of tube amplifiers over sand-state. Or vice versa. Or the horrible / exquisite sound stages created (or tickled out) by Aikido line stages, yada, yada, yada.
Let's face it, goats… all we really want are amplifiers with nice smooth transfer curves, without odd little playback level bumps and wiggles. Some people swear by the nearly ideal flatness of well-made sand-state topologies. Others lose their grip on reality when they sit before a hugely radiant infrared vacuum emitter, with its natural S shaped transfer curve. None (with one key exception: rock musicians) of us want transfer curves that are time dependent, varying with program material, line fluctuations, position of the input volume controller. We don't. We don't really want amplifiers that introduce a lot of waveform artifacts, some of which are “harmonic” and others which clearly are not. We don't want our amplifiers fighting to control the speakers so tightly that that in turn creates a whole army of new artifacts. We don't.
And therein is the segue to return to the OP's point: respect for science. I could be entirely off, but I think the previous paragraph kind of epitomizes why SCIENCE is involved with the design, building and tuning of amplifiers. And why "THD" is both very useful, and sometimes awfully bone-headed as single-number measures go.
GoatGuy
At exactly same measured dimensions, a scratch on your spare tire's rim is less annoying than one on your dashboard's plexiglass.
A single number like "total length of scratches on the car" won't tell much unless it's 0.
A single number like "total length of scratches on the car" won't tell much unless it's 0.
All true, but JCX is also correct. You can gain insight into the orders of distortion in THD by looking at how it increases with level.
The reason why is in the math.
Agreed.
I think I've made this point on this forum lately as well.
I like to say that no amplifier has THD or IM, because THD and IM are not components or even properties of amplifiers, but are rather means for characterizing (as you just alluded to) the nonlinear distortion of audio gear.
Nonlinear distortion or if you will a curved transfer function is what the gear has.
The analogy breaks down because THD is not the least all-inclusive but only includes a fairly well-defined specific short list of things. For example, it is neither faceplate color nor knob color, nor frequency response nor phase shift.
Yet another expert who has no real-world experience with loudspeaker measurements, or measurements of amplifiers that are clipping.
And if you get into the music, things get really hairy. Many musical instruments produce more harmonics than fundamentals.
The proof is here:
"That is taken to extremes, and quite out of the range we deal with."
What you call extremes are things that are all around us. By dismissing them you obviously don't think they are even present.
As with scratches, none discernable is a possible and highly relevant condition.
I covered this once in my recent post, I presume you are just deflecting to avoid admitting an uncomfortable truth.
No, just driven loud enough.
And no need to get into the Guitar Amp world, just go to an average Party with and average DJ .
As of the OP: my respects for stating that the Emperor is actually naked, instead of showing the sheeple mentality and agreeing with the masses no matter what they say, and I'm not surprised that the "tailors" which profit one way or another from "making" the imaginary "Emperor's suit" instantly jumped to bite your throat.
No, THD is specified at a frequency. Example: THD = 0.1% at 1 kHz for 1W. It is the sum of all HD components at a frequency. There is no such thing as a single THD for all frequencies. Harmonic distortion by its very name implies there is a single fundamental frequency.
To the OP, this forum is like a microcosm of the world. There are all kinds of folks here, just like out in the real world. Whom you listen to and what you take away is up to you.
- Status
- Not open for further replies.