I does seem like we have a different understanding, I think.... I know my understand is correct, not because I am an EE, but also because I have studied communication theory that involves Fourier series, transforms (such as the FFT).
I have read lots of harmonics related articles, and there is nothing in those links that contradicts my understanding..
The "deformed" (I take it you meant non sinuisodal), waveform is in fact an infinite series of sinusoidal waveform of the fundamental, and harmonics of the fundamental. If you add the point on wave values of say a 1000 Hz, 2000 Hz, 3000 Hz sine wave together you will get a deformed looking wave. For musical instrument, I am sure you know, has a lot of harmonic contents but they would not sound objectionable. When reproduced by the audio signal path of electronics amplification, the electronic devices (loudspeakers too obviously) add their own and are not not pleasant sounding to our ears because while they are still harmonics of the fundamental, they are not what the same harmonics the instrument produced, but are extra ones.
There is good reason for THD to be stated in - X dB, or SINAD (that includes noise) in X dB, it is to make it easier to relate to threshold of audibility.
Please take a read of Benchmark's below:
https://benchmarkmedia.com/blogs/ap...preting-thd-measurements-think-db-not-percent
and note it says:
"When distortion reaches 1% it is just 40 dB lower than the music. When distortion reaches 0.1% it is 60 dB lower than the music. When distortion reaches 0.01% it is 80 dB lower than the music. Obviously the 40 dB, 60 dB and 80 dB figures are easier to understand than 1%, 0.1% and 0.01%. "
As you mentioned before, this may be off topic though I think it is still "on" topic until it isn't if we keep going and we may be getting there
. So if we disagree then I am totally okay with agreeing to disagree.
But let me try one more time before we move on:
In your first link, if you go to chapter 13 (yours link to chapter 11), and go right to the bottom of the page:
http://www.dspguide.com/ch13/4.htm under the last section the "The Fourier Series" you will see the table below that show the component harmonic frequencies and their magnitude. You can see how the pulse train (rectangular wave) can be reconstructed with harmonics, the example use only the first 14 harmonics, to get to the original it need a lot more, but anyway, in the table, you can see that the 3rd harmonic is only 0.2 V and this is a very ugly waveform, of a 1 kHz fundamental.
A music signal waveform could look very complex/ugly but it is still just a series of harmonics of fundamental sine wave. Obviously the harmonics of a pure square, triangular or whatever "deformed", i.e. non sinusoidal waveform are not "distortions" if they are produced by musical instruments of electronics if produced perfectly, otherwise they would be distortions if created by the electronics such as amplifiers. But harmonics are still harmonics, that are harmonics of the fundamental frequency of the original signal, as such they are sine waves at multiples of the fundamental frequency.
View attachment 72407
Thank you for an interesting discussion.
I wonder if anyone will really read this monster post. Let's see. And let's see who learns what here
@peng . Hopefully we all do.
Agree with ~all you say above, it's all good info. But I 50% disagree with that Benchmark quote.
"When distortion reaches 0.01% it is 80 dB lower than the music. Obviously the 40 dB, 60 dB and 80 dB figures are easier to understand than 1%, 0.1% and 0.01%. "
For the N part of the THD+N or SINAD distortion, I agree that the dB value is definitely the most intuitive/useful.
However, for the HD part I pretty much disagree completely: to me the % value is most intuitive and the dB value (potentially) confusing.
And IIRC, THD was originally thought as a % number.
At this time, I assume most people actually got a good understanding of my points and
@peng 's.
So I'll just throw a few more examples/analogies/views, just to see what sticks. That is, I will just repeat the ~same thing over and over. Sue me
I'd like to know which analogy people find most useful. And should be useful for me too, to kinda re-organize the HD-info in my own head. And there is still a chance that I got some things wrong. At least some of the stuff below might be just "my view" and not necessarily 100% correct.
One view: about the THD number as either dB or %
The THD dB value I see as confusing because it creates the impression that the THD dBs are related the sound level dBs (similarly to the way noise floor dBs are related to the sound level dBs). The % value does not carry that dBs-to-other-dBs confusion. But many also interpret a value of 0.01% THD to mean that some 'lower' 0.01% part of the signal is distorted and the rest is ok. I think it's much a better analogy to see 0.01% THD as a 0.01% 'deformation' of the entire signal.
A potentially
useful graph: Green is the original signal, Blue is a sample 3rd harmonic and Red shows how the original signal will look after adding the 3rd harmonic
into it: it's 'deformed'. And it's not deformed at the top or bottom, it's fully deformed.
And
another one showing how a pure sine-wave is 'deformed' by (idealized, even-order) HDs until it becomes a square-wave. Fully deformed again.
And
another one showing how a pure synth signal looks after you put a guitar's HDs into it (this one is just a visual aid, not necessarily 100% correct).
A different view: sort of a big simplification/abstraction which might work visually. And yes, I'll try to describe a graph in words
Let's take an original signal at 90dB volume and plot it on a THD-freq graph as a sort of "music floor" at 90dB. It would be a straight music-floor-line at 90dB.
The N itself would be another straight line at say 20dB (i.e. -70dB under the original). Because the N is best seen as a separate signal and (hopefully 100%) unrelated to the music-floor signal.
I guess
@peng that we fully agree up to here.
Here's where we (potentially) part ways.
The question is: how would one plot/draw the THD on this graph? Would it be an extra THD-floor-line, somewhere between the music-floor line and the noise-floor line? Most people do that. Even some audio experts do that (or seem to think that way).
And I think that's just wrong. On such a graph, my best 'visual' representation of the THD would be as 'something' that makes the 90dB music-floor line undulated instead of straight ... whatever that 'undulated' looks like is a another discussion and let's forget it for now.
Mostly because of the above, I think that putting HD and N together in a single number is a pretty bad idea. An opinion I seen a few times from expert EE & acoustics people (sorry, no link/quote). THD+N is ok from a measurements/engineering perspective but if you want to discuss audibility/acoustics, that number it's a useless mashup.
Another view: sort of a gedankedexperiment that should help with the psychoacoustics/neuroscience of HD.
I think the psycho-acoustics give a much better view of HD than the view you get from math & THD graphs. And you made a very helpful mention
@peng of the harmonics that are included in any note played by a musical instrument.
So let's say we grab a piano and play a single, 'pure' note like A4 (i.e. a frequency of ~
432Hz). And lets assume that we recorded that signal perfectly with no extra distortion. If we analyze/plot it, we'll see the fundamental at 432 Hz and multiple harmonics at 2X, 3X, etc.:
This might just be one of those proverbial "image better than a thousand words". So let's see some of those words:
- First of all, the above graph might be a surprise for some: a 'simple' piano note looks exactly like a THD measurements graph. You have the fundamental at 432Hz and it's HDs at 2x, 3x, etc. It's almost 100% the same as a THD graph. And there is no metaphor/analogy in here, those two graphs are indeed mathematically ~100% same (that's mostly under a big 'AFAIK', anything to correct here @peng?!?).
- The HDs of this simple piano note are seriously high, its THD would be a gruesome >50%. And btw, the HDs of some instruments go much higher than that, even above 100% (for the trumpet IIRC).
So, if we play that recorded piano note, how will it sound with that giant, >50% THD in it?
To figure out how it'll sound, we can apply the HD-as-a-floor-distortion model to the above piano-note graph. Here are its predictions:
- at >50% THD we should hear a seriously distorted, "aaa, my head hurts" sound. Still somehow recognizable as a piano A4 note.
- we should hear that THD-floor as a sort of high level 'mud' in the A4 sound.
But we don't! What we will actually hear is a crystal clear piano A4 note.
In the case of the trumpet with >100% THD, the HD-as-a-floor-distortion model says that we shouldn't even hear the A4 note fundamental, just some fully distorted HD-floor garbage. But again we don't, we just hear a crystal-clear A4 trumpet note.
Generally, no matter how many HD spikes and no matter how big they are, the human ear will still hear an A4 note. And the (432Hz) pitch of the A4 note will not change at all and it will still be recognizable to our ears as an A4 note/pitch.
This would be my base for saying that the HD-as-a-floor-distortion model is simply Wrong: it predicts that we'll hear lots of 'garbage' but there is none to be heard.
And here's something even more 'strange': with a bit of math and signal-processing it is possible to take the A4/432Hz fundamental out of that recorded piano A4 sound. And we will have a recording of a piano A4 note which only contains its HDs.
What will you hear if you play that HDs-only recording?
...
Take a guess... really curious if someone guesses right here. I surely didn't the first time.
---
Here's the 'strange' answer: you will hear the ~same piano A4 note. We took out the A4 fundamental but our ears still hear it from just its HDs.
This is pretty much my base for saying that the HDs are 'imprinted' into the signal. Actually, according to the above experiment, the HDs are not just 'imprinted' into the signal,
the HDs are the signal!
Not sure exactly what is the math/EE view on this but the human ear says it quite clearly: the HDs are the signal.
.....
No idea if anyone will read this to the end but anyway, no more time, gotta stop. Should be enough for a while. More to follow about the real "money question":
how do those HDs actually sound to our ears? ... Sometime.