Passives: HF first to clip?

[gnarly]

For passive speakers with one amp channel driving the speaker, don't the high frequencies have to be the first to clip? (assuming music with full spectrum content)

Looking at scope traces of summed sines, I see the higher frequency rides on top of the lower. Which establishes the peak voltage of the combined sines.
So when an amp hits the voltage limit of its rails, seems like the high frequencies have to be the first to get clipped.
Yes/no ?

 

[MarkS]

It's the sum of the sine wave that clips, not each individual sine wave. That has no meaning (unless the signal is a single sine wave, then of course that is the signal that clips).

When the clipped signal is expressed as a sum of sines, there are extra high-frequency sines that were not present in the unclipped signal. These get sent to the tweeter, which can be bad for the tweeter if their amplitude is too high.

 

[wwenze]

Why can't it be low frequency riding on top of the higher? Seen radio transmissions?

 

[gnarly]

It's the sum of the sine wave that clips, not each individual sine wave. That has no meaning (unless the signal is a single sine wave, then of course that is the signal that clips).

Yes, it is the sum of the sine waves that clips. That is what i meant to be saying. Superposition of frequencies has the higher frequency sine riding on the lower.
Seems to me that always puts higher frequency content at the peak voltage .of any sine summations.

Here's a 50Hz and 1000Hz scope shot of what I'm trying to convey.
Blue trace is 50Hz alone
Yellow is 1000Hz alone
Green is their summation.

Summation is of course a higher voltage than either single sine on it's own, but note how the 1000Hz sine rides on top the 50Hz sine. And that superposition is what increases the summed voltage....and the increase in voltage is all 1000Hz sine content.

This example starts with equal voltage sines. Clearly if gain is raised to both, the 1000Hz sine will be the first to hit an amps rail voltage limit.
So let me move on to unequal voltage sines..

Here I've 4X the voltage of the 50Hz sine, leaving the 1000Hz unchanged. Equivalent to +12dB SPL @ 50Hz...hey my sub needed some boost, lol !
1000Hz still rides superimposed, and will be the first to hit the rails, and clip.

To the extent music waveforms are a collection of sine waves ala Fourier, seems to me higher frequency content clips first, as amp gain is raised.

 

[popej]

Or maybe saying otherwise: clipping adds high frequency distortion.

 

[gnarly]

When the clipped signal is expressed as a sum of sines, there are extra high-frequency sines that were not present in the unclipped signal. These get sent to the tweeter, which can be bad for the tweeter if their amplitude is too high.

Hi, responding separately to this trying not to overload the previous post.

I know any clipped signal, like when a sine's peak gets flattened into a more square wave form, shows an FFT with increased high-frequency content.
There's always been debate about the significance of such HF in terms of tweeter overload. It exists sure, but is it significant.

I ran a bunch of tests on two-way passive crossovers, measuring the voltage delivered to the woofer and tweeter, as the amp was driven into clipping.
Idea being does the total power delivered shit in proportion between the woofer and tweeter once amp starts to clip.
It did shift , but so slightly the shift was completely inconsequential to the power the tweeter (and woofer) were receiving, relative to change in to overall drive level.

My strong conclusion is too much juice kills tweeters, not a shift in spectral content from amp clipping.
Also think it's worth noting, if the superposition analysis above is correct, that's a big effect on the voltage the tweeter gets.

 

[NTK]

From Erin's Feb 2024 measurements of the Wiim Amp.
https://www.erinsaudiocorner.com/electronics/wiim_amp/

Multitone (several dozens sine waves riding on top of each other) distortions without clipping.

Multitone distortions with clipping.

 

[wwenze]

Note that the extra HF content from clipping is of a phase and amplitude that is of a direction that counteracts the original LF signal, this is why when you sum them together you get the square wave. This is the same reason why you get overshoot from a square wave when you apply LPF to it.

I wouldn't call it "not present in the unclipped signal", the energy has to have come from somewhere. Instead think of it as a portion of the energy - a lot of it in HF - has been removed from the original signal. Which is why the remaining signal also has these HF content but negative.

 

[MAB]

Hi, responding separately to this trying not to overload the previous post.

I know any clipped signal, like when a sine's peak gets flattened into a more square wave form, shows an FFT with increased high-frequency content.
There's always been debate about the significance of such HF in terms of tweeter overload. It exists sure, but is it significant.

I ran a bunch of tests on two-way passive crossovers, measuring the voltage delivered to the woofer and tweeter, as the amp was driven into clipping.
Idea being does the total power delivered shit in proportion between the woofer and tweeter once amp starts to clip.
It did shift , but so slightly the shift was completely inconsequential to the power the tweeter (and woofer) were receiving, relative to change in to overall drive level.

My strong conclusion is too much juice kills tweeters, not a shift in spectral content from amp clipping.
Also think it's worth noting, if the superposition analysis above is correct, that's a big effect on the voltage the tweeter gets.

Rod Elliott has a good discussion on tweeters and clipped amps, with the spectra of amps while clipping.

Why Do Tweeters Blow When Amplifiers Distort?

ESP - The Audio Pages. Why Do Tweeters Blow When Amplifiers Distort? Find out what really happens when amplifiers cause tweeters to fail.

sound-au.com

He makes a case for it coming down to the power, not harmonics from clipped lower frequency content.

 

[popej]

My strong conclusion is too much juice kills tweeters, not a shift in spectral content from amp clipping.

True, but is this the answer to OP question?

 

[wwenze]

Sometimes waveforms are unintuitive, but the math still add up intuitively. Just get used to ignoring visual intuition.

This is the FFT of a 1kHz sine, clipped 50%
The unclipped wave reads 0dB
The clipped wave reads -4dB for 1kHz

So, first takeaway: You do lose energy from the fundamental frequency. Around -4dB in this case. Note that if I simply reduced the amplitude by half I would lose -6dB. So there's still extra 2dB of energy due to the non-sine shape, that has to be countered by other frequency components.

Let's look at a combined 100Hz and 1kHz both at 0.5 amplitude.

It's supposed to be -6dB for each peak but you know, Audacity and low FFT size. Higher FFT size and the peaks don't show up.

Same wave, clipped 50%. Both original frequency content are now -8.6dB.

Thus takeaway: HF and LF are reduced equally during clipping, agreeing with what we mathematically already know.

 

[gnarly]

Rod Elliott has a good discussion on tweeters and clipped amps, with the spectra of amps while clipping.

Why Do Tweeters Blow When Amplifiers Distort?

ESP - The Audio Pages. Why Do Tweeters Blow When Amplifiers Distort? Find out what really happens when amplifiers cause tweeters to fail.

sound-au.com

He makes a case for it coming down to the power, not harmonics from clipped lower frequency content.

Thx for that. Good paper.
Wish I'd seen if before buying some passive crossovers to test. Could have saved me some money Lol...only reason I bought the passive xovers was to test what I thought had to be myth.

In addition to the harmonics myth, see the paper also appears to support the superposition idea that high-frequency has to clip first.
Saying significant portions of the tweeters waveform went missing when supply rails were hit, at moderate amp clipping.

 

[DVDdoug]

If you take typica music and high-pass and low-pass filter, the high frequencies are weaker, including the peaks (depending on the cutoff/crossover frequencies).

If you push it into clipping the high frequencies will still be weaker on average but both may still clip (in which case they will have identical peaks).

....Tweeters can't handle as much power as woofers (or midranges).

 

[gnarly]

True, but is this the answer to OP question?

Uh, you realize I'm the OP?

Sometimes waveforms are unintuitive, but the math still add up intuitively. Just get used to ignoring visual intuition.

This is the FFT of a 1kHz sine, clipped 50%
The unclipped wave reads 0dB
The clipped wave reads -4dB for 1kHz
View attachment 509285

So, first takeaway: You do lose energy from the fundamental frequency. Around -4dB in this case. Note that if I simply reduced the amplitude by half I would lose -6dB. So there's still extra 2dB of energy due to the non-sine shape, that has to be countered by other frequency components.

Let's look at a combined 100Hz and 1kHz both at 0.5 amplitude.
View attachment 509286
It's supposed to be -6dB for each peak but you know, Audacity and low FFT size. Higher FFT size and the peaks don't show up.

Same wave, clipped 50%. Both original frequency content are now -8.6dB.

View attachment 509288

Thus takeaway: HF and LF are reduced equally during clipping, agreeing with what we mathematically already know.

Thanks for that. And yeah, I've looked at such FFT analysis too, with pretty much the same conclusion.

But I'm now thinking that looking at clipping in the frequency domain misses the boat. I think it's a time domain phenom, that needs time domain measurements.
Check out the Rod Eliot paper MAB linked....Rod does a bit of both freq domain and time domain analysis.

The way I think the frequency domain might measure superpositions effect, is with transfer functions of full-range pink as an amp is pushed into varies degrees of limiting, particularly looking at Coherence across the spectrum. (expecting the upper end of the spectrum to drop coherence first.)

 

[marin.weigel]

Yes, it is the sum of the sine waves that clips. That is what i meant to be saying. Superposition of frequencies has the higher frequency sine riding on the lower.
Seems to me that always puts higher frequency content at the peak voltage .of any sine summations.

Here's a 50Hz and 1000Hz scope shot of what I'm trying to convey.
Blue trace is 50Hz alone
Yellow is 1000Hz alone
Green is their summation.

Summation is of course a higher voltage than either single sine on it's own, but note how the 1000Hz sine rides on top the 50Hz sine. And that superposition is what increases the summed voltage....and the increase in voltage is all 1000Hz sine content.

This example starts with equal voltage sines. Clearly if gain is raised to both, the 1000Hz sine will be the first to hit an amps rail voltage limit.
So let me move on to unequal voltage sines..

View attachment 509281

Your intuition has been deceived by looking at an overlay of the individual signal components with their composite sum, such that, visually, the higher frequency component's peaks seem to make for the increased amplitude relative to the amplitude of the components.

But it's the sum of both components causing the higher amplitude, so, unintuitively, if you now clip the signal peaks, both components of the composite waveform will be affected in relation to their relative amplitude.

In general, overdriven clipping amps suffer from the abrupt discontinuities in the signal's rate of change, effecting odd order harmonic components with roughly equal energy distribution per octave in most cases.

 

[MarkS]

In general, overdriven clipping amps suffer from the abrupt discontinuities in the signal's rate of change, effecting odd order harmonic components with roughly equal energy distribution per octave in most cases.

Yes, but the amount of energy going to those higher harmonics turns out to be tiny, at least theoretically.

Here is a plot of a single frequency, both full amplitude (orange line, peak = 1) and clipped at √(3/4) = 0.866 (blue line).

A pure sine wave with this peak amplitude would have the power (proportional to the square of the amplitude) reduced by a factor of 3/4, that is, lowered by 25%.

But the clipped amplitude turns out (after calculations) to have a power reduced by a factor of 0.891, a smaller loss of 11%.

Furthermore almost all of this power is still in the fundamental, 0.888. Just 0.003 goes into all the higher harmonics.

I found this result quite surprising.

Even clipping at half amplitude (which would be considered very severe clipping) reduces the total power by a factor of 0.670 with just 0.012 going into the higher harmonics.

 

[wwenze]

But I'm now thinking that looking at clipping in the frequency domain misses the boat. I think it's a time domain phenom, that needs time domain measurements.

Time domain measurement = frequency domain measurement

The whole time domain and frequency domain is just different ways of showing the data. Btw that Fourier transform? It just takes time domain data and shows it as a sum of sines, which is still time domain when you sum the sines in time domain.

Fourier series is telling you that this wave has the both the 1kHz and 100Hz components reduced by 2.5dB.

Ok, I think I might have gotten the answer to the hidden question

You are thinking that HF gets removed more than LF despite mathematics say both get reduced by the same amount

The answer is, both are true. But the two statements are rules following different trend lines and they only intersect in one location: When HF voltage is the same as LF voltage.

In your example, your LF voltage is noticeably lower than the HF voltage.

Ok, so what happens when we do the same with a wave that is 0.8 LF and 0.2 HF? (Or around -1.94dB LF and -14dB HF)

Audacity returns -4.4dB LF and -20.8 HF. Thus giving us a reduction of -2.5dB LF and -6.8dB HF respectively. Or some might say in English, the HF got reduced more than the LF in this case.

But this is due to the relative original amplitude. If you repeated the process with a wave that is 0.2 LF and 0.8 HF, the reduction follows the pattern.

To summarize:
- If you start with 0.5 LF and 0.5 HF and clip 0.5, you get -2.6dB in both LF and HF
- If you start with 0.8 LF and 0.2 HF and clip 0.5, you get -2.5dB in LF and -6.8dB in HF
- If you start with 0.2 LF and 0.8 HF and clip 0.5, you get -6.8dB in LF and -2.5dB in HF

The conclusions that get drawn from the data:
- Mathematically, HF and LF behave identically
- For the case of lower HF amplitude than LF, the lower signal gets reduced more. Here the visual intuition works, because when you cut the wave, the percentage of LF removed is less than the percentage of HF removed.

The same amount of energy is being removed from both waves, but removing 1 unit of energy from LF versus 1 unit of energy from HF, the HF looks like it got the worse part of the deal.

Yea, science makes things boring.

 

[gnarly]

Your intuition has been deceived by looking at an overlay of the individual signal components with their composite sum, such that, visually, the higher frequency component's peaks seem to make for the increased amplitude relative to the amplitude of the components.

But it's the sum of both components causing the higher amplitude, so, unintuitively, if you now clip the signal peaks, both components of the composite waveform will be affected in relation to their relative amplitude.

Hmm.....I'm tying to leave intuition alone, and look at measurements. I'm being taught in prosound measurement classes to look at both the time and frequency domain, for what each has to offer.
This superposition case, and what it may mean for an amp to drive a passive speaker, is such an example of using the time domain I think.

Anyway, continuing on with how/what to measure to determine if it holds water....here's latest experiment.
Using the same two sines, 50Hz and 1000Hz, I scoped their summed signal with REW in two ways:

First channel is where their sum lightly clips the amps output...the amp being a line stage device.
Second channel is the same summed signal , adjusted (-4dB) to get output below clipping, and then with an equal and offsetting +4dB gain in the soundcard input.

Here's the processor schematic used. Note the red clip light on Ch1's output, and no clipping on Ch2.

Ch1, the clipped is green.
Ch2, non-clipped is yellow.

Think I've figured out how to scope and compare the 50Hz sine alone, when the sum is clipped vs when the sum is not clipped.
I'll post that tomorrow, to see if it decreases output too, along with the HF 1000Hz..... like you say you think it will.
I'm doubting it will decrease...but we'll see.

 

[gnarly]

Time domain measurement = frequency domain measurement

Yep I so wish everyone half-way serious with audio & measurements fully embraced that fact.
I picture a first class in Fourier measurements teaching that equivalency over and over.

Building on that, I picture a second class teaching that the equivalency is for the snapshot window in time that the measurement spanned.
Explaining that the measurement assumes steady state conditions over the measurement's time window, and goes on to explain the inherent time - frequency resolution tradeoff.
Third class, where I hope I'm gaining qualification, is learning when to use frequency domain measurements, and when to use time domain measurements.

The same amount of energy is being removed from both waves, but removing 1 unit of energy from LF versus 1 unit of energy from HF, the HF looks like it got the worse part of the deal.

I'm not measuring that.

Had a fun time this morning picking up from my last post, making a bit of a unique test bed (unique for me at least).

Experiment went like this....
first measure the 50Hz and 100Hz sine individually that are each below clipping.
(I used equal level sines to address your conclusions)

Set voltage cursors and leave locked for rest of experiment)

Next, sum them and measure. Voltage increased and clipping observed as expected.

Then , I extracted the 1000Hz sine component from the above clipped summation. Note the regions of clipped output.

And extracted the 50Hz component from the clipped summation.. Looks identical to its individual measurement, with no indication of a decrease in level.

So at this point I'm getting more convinced that superposition will cause higher frequencies to clip first.

I'm not an amp expert by any means, but I do know amps can clip either due to a lack of current capacity, or from a rail voltage limitation.
I'm picturing superposition as imposing a rail voltage limitation, mucking with HF, ....even before the well discussed power (current driven) limitations begin to kick.

Just thinkin...hopefully halfway correctly

 

[wwenze]

Notice your clipping is not at a fixed value which is what is expected or rather defined as clipping.

Meaning the mathematical function of the clipping is not "if value above >1, set to 1", but rather something else.

How are you generating the combined wave? For all we know the equipment could be doing "If ( channel 1 + channel 2 ) > 1, clip channel 2"