My view is that IMD is more important in listening quality. If you listen to complex music, amplifiers with modest IMD sound indistinct, and I have been able to hear more of the separate muscial instruments with amplifiers having lower IMD.
IMD can be worse than THD, but amplifiers with low THD generally have low IMD as well. Therefore, whle THD may not be the best indicator of musical quality it is a measure of how an amplifier's IMD will perform. Therefore, I suggest THD is important for that reason. FOr if you have high THD you won't get good IMD.
IMD can be worse than THD, but amplifiers with low THD generally have low IMD as well. Therefore, whle THD may not be the best indicator of musical quality it is a measure of how an amplifier's IMD will perform. Therefore, I suggest THD is important for that reason. FOr if you have high THD you won't get good IMD.
Yes THD and IMD are different measurements of the same nonlinearity.
It seems like there could be cases where linearity appears good with a pure tone, but IMD still crops up with multiple tones.
If tone 1 alters gain symmetrically about the middle, HD may not show up because the tone is only modulating itself with perfect linearity. But if a 2nd tone is added, more complex interactions may then become apparent.
If tone 1 alters gain symmetrically about the middle, HD may not show up because the tone is only modulating itself with perfect linearity. But if a 2nd tone is added, more complex interactions may then become apparent.
Put two signal sources in series at the two frequencies you want to evaluate. Make sure the combined peak does not exceed the amp input voltage limit and cause clipping. You need to run the simulation for longer than a single frequency in most cases so that the subharmonics (f1-f2) are appropriately calculable, with a fine time step (at least on the simulator I use, which does not have a built in IMD function either), then perform the usual FFT on the results. For example while I might run a 1kHz signal for perhaps 20 cycles, with 100ns time step, a 19kHz/20kHz signal needs to be run with the fineness of 20kHz (10-20ns) but the duration of 1kHz (20ms) to give a sharp response in the FFT at 1kHz.
Last edited:
No. If the gain varies with amplitude, the transfer function will not be linear and that will show up as harmonic distortion.
You don't need to consider IMD separately from THD. Just make the transfer function a straight line. I will award bonus points if you do that without a lot of global feedback.
Ed
THD matters BUT if there is distortion, I rather it be coming from the amplification and not to be intrinsic to the speaker driver. The former can be controlled and 'tuned', the latter can not and tends to be very unpleasant, exp IMD.
Also, if there are audible differences between an amp of the same power and 0.005% or 0.0005% THD, there are likely other factors that need to be considered.
For me, the lowest distortion from the simplest circuit possible (ala Nelson Pass) has yielded the best sonic results. Of course that requires high eff, low distortion drivers.
Also, if there are audible differences between an amp of the same power and 0.005% or 0.0005% THD, there are likely other factors that need to be considered.
For me, the lowest distortion from the simplest circuit possible (ala Nelson Pass) has yielded the best sonic results. Of course that requires high eff, low distortion drivers.
Believe what you like.
Your ears will not hear 0.0005% or 0.005% THD distortion.
Yes speakers produce far more distortion, largely limited to 2nd and 3rd HDs.
For grins, driver X-max is defined and measured at 10% distortion.
Last edited by a moderator:
That doesn't matter anymore. Both distortion and noise are miles away from being audible. In the 80s, high-end was reached, so to speak, with the first very good transistors. Since then there have of course been improvements, but they are no longer audible and are only cosmetic for the brochure.
I try to believe what I hear, not what I like.
Yes, drivers contribute 1 to 2 order of magnitude higher distortion than electronics, and that such distortion is output level dependent. That's why I like high efficiency drivers with the lowest possible distortion.
I agree that if you take an 0.0005% THD opamp circuit and inject 0.005% distortion in your input signal you probably will not be able to hear any diff. However, if you swap two opamps with different design but equal 0.0005% distortion chances are good you will hear a difference. I don't know why. 🙂
I don't think you'll hear a difference if your speakers or headphones alone are 3% THD. 1,000-10,000 times higher.
Yesterday, I posted a test. You can answer the question with your own ears.
Take the THD Challenge
https://www.diyaudio.com/community/threads/take-the-thd-challenge.395242/
Ed
Take the THD Challenge
https://www.diyaudio.com/community/threads/take-the-thd-challenge.395242/
Ed
Good point. Listening tests are not always 'preference tests.' For example, listening tests could be 'discrimination tests' and or 'descriptive analysis tests.' In short there are various well known sensory test techniques to choose from.
Please see the attached for a pretty good review article on the subject. Interesting table on page 6.
Am I saying that again for the fifth time : ) ?
THD means total HARMONIC distortions.
If you hear it clearly, it's not too high and it's low orderH2...H4, that means -40...-60db level, you might like it depending on the musical content.Tube amplifiers are in that area.
If it's below -85db you can't hear it...so its components don't matter.
If it is in between -60...-85db it's BAD unless it's only H2 and H3 cause it's found in that subliminal gray area where you hear it, but can't differentiate it from the noise or faint sounds when passages have very wide dynsmics.With modern low dynamics music you can't hear anything in the subliminal range, not even at -40db...Dolby and DBX(THAT) defined all you need to know about subliminal perceptions of noise and distortions.
Treat THD as noise and you'll be happy!
THD means total HARMONIC distortions.
If you hear it clearly, it's not too high and it's low orderH2...H4, that means -40...-60db level, you might like it depending on the musical content.Tube amplifiers are in that area.
If it's below -85db you can't hear it...so its components don't matter.
If it is in between -60...-85db it's BAD unless it's only H2 and H3 cause it's found in that subliminal gray area where you hear it, but can't differentiate it from the noise or faint sounds when passages have very wide dynsmics.With modern low dynamics music you can't hear anything in the subliminal range, not even at -40db...Dolby and DBX(THAT) defined all you need to know about subliminal perceptions of noise and distortions.
Treat THD as noise and you'll be happy!
I hope I am not too offtopic. (This is copied from another forum)
Can anyone interepret these two THD+N charts vis-a-vis to Distortion graph. They seem unique so just learning.
Link : (1) Audio Reveal (2) Atol
Audio Reveal : The noise figure is not bad i guess. but distortion is high and rises with power. But can not interpret /decipher/chaff out noise characteristics as it is THD+N.
Atol : almost flat distortion graph (both harmonics at high level) BUT with rising power it does not taper down. what does it mean ?
Thanks
Can anyone interepret these two THD+N charts vis-a-vis to Distortion graph. They seem unique so just learning.
Link : (1) Audio Reveal (2) Atol
Audio Reveal : The noise figure is not bad i guess. but distortion is high and rises with power. But can not interpret /decipher/chaff out noise characteristics as it is THD+N.
Atol : almost flat distortion graph (both harmonics at high level) BUT with rising power it does not taper down. what does it mean ?
Thanks
I don't think it's a case of either THD or IMD snapshots having full enough information about a transfer function as with an FFT vs IFT.
Take for instance, 2 amplifying stages in series where the 2nd one cancels the HD of the first. The THD could look really good (at least for loworders like H2 or H3) but the IMD products would be further modulated with the input signal as they pass through the 2nd stage. At some point the complexity of the IMD would begin to resemble a raised noise floor, and addition and subtraction would be essentially random.
To be able to achieve low distortion at 20K a lot of things in an amplifiers design have to come together. So in my opinion the lower an Amplifiers THD is the better it will sound because all the components must be preforming optimally to achieve this. This was the design philosophy for the Wolverine amplifier. Optimise as many of the building blocks that make-up an amplifier as possible.