High frequency distortion manifestation

[mike7877]

Very often, amplifiers have increased THD+n in the 5/10/15kHz spectrum once power increases past 1-5 watts.

Two things:

If you've got one watt of 10kHz on top of twenty watts of 1kHz, do you get 10kHz @ 1 watt performance for the 10kHz tone, or
Is the distortion of the 10kHz modulated at 1kHz, increasing to a peak of 10kHz @ 20 watts at the crest of the 1kHz wave?

That was the first thing - it had two parts...

Second thing:

Typically
What does the 10kHz FFT look like@ 1w, 2w, 5w, 10w, 20w, 50w?
Are the harmonics the same, or do they change?
Does noise increase?

If these things don't follow a pattern, let me know.
Also, does class AB typically behave one way, D another?

 

[solderdude]

A 10kHz FFT will only show the fundamental and the harmonics which all will be above the audible band as the 2nd harmonic is 20kHz.

Distortion at higher frequencies increases because the open-loop frequency response of an amplifier with feedback is much smaller than at lower frequencies.
Thus there is less feedback and thus more distortion. But... for higher frequencies distortion is less important. Also the amplitude of high frequencies is not very high so distortion products will also be smaller as a small signal uses a smaller portion of the transfer characteristic thus is relatively more linear.

Yes, A-B and class 'D' behave very differently as they work on very different principles.

 

[Philbo King]

There are many types of distortion. Distortion is defined by any difference between input (as scaled by gain) and output. It is a big topic, so I'll address only IM distortion here.

Intermod distortion is the source of the frequency sum and difference products, and appear in various amounts when the system becomes non-linear. They do follow a pattern, defined by the aspects of system non-linearity. But it is not a simple pattern; it is affected by frequency response, slew rate, amplitude distortion and a host of other factors. The math of predicting intermod distortion products is briefly addressed here, in the subtopic 'Causes of Intermodulation':

Intermodulation - Wikipedia

en.m.wikipedia.org

 

[mike7877]

A 10kHz FFT will only show the fundamental and the harmonics which all will be above the audible band as the 2nd harmonic is 20kHz.

Distortion at higher frequencies increases because the open-loop frequency response of an amplifier with feedback is much smaller than at lower frequencies.
Thus there is less feedback and thus more distortion. But... for higher frequencies distortion is less important. Also the amplitude of high frequencies is not very high so distortion products will also be smaller as a small signal uses a smaller portion of the transfer characteristic thus is relatively more linear.

Yes, A-B and class 'D' behave very differently as they work on very different principles.

Say that 3kHz distortion in the chart, looks like 5kHz. Instead of my question being about 10, it's about 3kHz.
As power increases, is the only thing that happens is that the harmonics increase in level? Could 2nd be dominant at 5w and 3rd at 20? Are the harmonics going down in level as the power level increases, but the "grass" increasing in level?

Is some of the distortion then due to slew rate slightly deforming the sine wave as the signal ever so slightly leads the output?


About AB/D - of course they're different. I'm asking what the difference is in the distortion. The amounts look similar, and I'm wondering - one possibility might be that AB has more harmonics, while D has more ultra high frequency switching noise

 

[mike7877]

There are many types of distortion. Distortion is defined by any difference between input (as scaled by gain) and output. It is a big topic, so I'll address only IM distortion here.

Intermod distortion is the source of the frequency sum and difference products, and appear in various amounts when the system becomes non-linear. They do follow a pattern, defined by the aspects of system non-linearity. But it is not a simple pattern; it is affected by frequency response, slew rate, amplitude distortion and a host of other factors. The math of predicting intermod distortion products is briefly addressed here, in the subtopic 'Causes of Intermodulation':

Intermodulation - Wikipedia

en.m.wikipedia.org

I'll read this tonight, thank you.

 

[solderdude]

Say that 3kHz distortion in the chart, looks like 5kHz. Instead of my question being about 10, it's about 3kHz.
As power increases, is the only thing that happens is that the harmonics increase in level? Could 2nd be dominant at 5w and 3rd at 20? Are the harmonics going down in level as the power level increases, but the "grass" increasing in level?

Design dependent.
The transfer characteristic determines the signal fidelity (distortion).

The 'grass' you can see (with multitones) is IM distortion mainly (in the higher frequencies mixed with HD as well).

About AB/D - of course they're different. I'm asking what the difference is in the distortion. The amounts look similar, and I'm wondering - one possibility might be that AB has more harmonics, while D has more ultra high frequency switching noise

Both designs have a transfer characteristic. With class-D there will be more distortion due to the HF filter and the way the feedback is created.
Also IM products are bound to occur by 'mirroring' against the switching frequency. Class AB doesn't have that.

Both design topologies can have inaudible amounts of distortion. As with all designs it is more design dependent than topology dependent but for sure the different topologies simply differ in several aspects.