Clipping & Underpowered Speakers

[Aerith Gainsborough]

How does clipping in the source fit into the "blow your tweeters" problem?

View attachment 201336

From non relevant to potentially problematic, depending on the amp power used and the implementation of the tweeter, would be my guess.

Though one variable would be eliminated: potential instability of the amplifier when running out of spec.

PS: the above would sound rather horrible, even on low volume. I hope that is a track you overamped for demonstrative purposes.

 

[Deleted member 46664]

How does clipping in the source fit into the "blow your tweeters" problem?

View attachment 201336

Amplifier clipping only happens at maximum power ... because you're demanding more than it can deliver.
The clipping in your graph above will just sound a bit raunchy at any volume.

 

[RayDunzl]

PS: the above would sound rather horrible, even on low volume. I hope that is a track you overamped for demonstrative purposes.

Nope, it's solo acoustic guitar on CD...

Stephanie Jones Open Sky

It experiences some momentary lapses of reason.

I don't hear a problem with it, unlike some late model possibly self-engineered Al di Meola, which, while never clipped in the same manner, has been aggressively limited and continually crushed into its container and sadly sounds it when played above casual levels. This is acoustic guitar and hand percussion.

• Al Di Meola – guitar, handclapping, djembe
• Hernan Romero – handclapping, djembe

The light blue in the middle of the traces is indicative of the average level, for those not familiar with how Audacity presents the data.

I note Al's overall average is the same or even louder than the loudest average of Stephanie's track.

 

[BDWoody]

I just love how much I continue to learn here.

I have to agree. There is a depth of knowledge here, on such a wide range of subjects, that it's hard not to be humbled. New members accustomed to being the 'smartest one in the room' tend to get a bit of vertigo until they find their footing...if they find their footing.

The willingness of our members to patiently educate those willing to learn makes this a place I'm proud to be associated with.

I don't know anywhere quite like it. It's not perfect, but what fun would that be!

 

[gnarly]

I hope you don't mind if I jump in here....

Not at all, the more the merrier, thx for jumping in. Here to share and learn ...always learning still

Clipping is not an average or integration ... it is an immediate effect, right down to half cycles of the waveform. On music with some dynamic range, an average might show you a low reading, even though serious clipping is already under way.

Of course. Clipping maps to True Peak SPL, whereas time integration maps to a RMS SPL average over the integration period.
Different animals, different tools, that i think can probably be used to detect the onset of clipping.... that was the point i was trying to make...comparing the spread between time averaged SPL and highest peak.

You need an oscilloscope directly connected to the outputs of your amplifier so you can see the output waveform. This shows what a scope will show you... This only happens at maximum amplifier power. Notice that the tops of the waveform a simply not there anymore...

Sure, i guess probably everyone involved with this thread is familiar with textbook sine wave clipping.

But I doubt how many folks realize that clipping often doesn't have such a clean clipped look. Like NTK pointed out in post #93, and Amir shows with measurements.
Also wonder how many folks realize that sine waves ride on top of each other, which leads me to think the HF can get clipped without clipping the LF.
Here's a quick example:
First screen is a single 250 Hz sine. Second is 250 Hz and 8000 Hz.

The sine waves were analog sent through a UMC404 being used just as a mixer.
The gain was set just below clipping using the single 250 Hz sine, and the scope voltage cursors were then fixed in place (first screen).
Second screen adds the 8000 Hz sine with 1:1 ratio, keeping voltage cursors intact.

Interestingly, the UMC404 clipping is not symmetrical.....just the top halves are clipped.
(Kind of a case in point about clipping not always looking like textbook truncation.)

The bottom halves show HF riding on LF very well, and is the voltage output needed for unclipped amplification.
The top halves have clipped the HF, but it seems the underlying LF remains unclipped,

Of course since music is not simple sine waves, i think we can expect complex waveforms to have transient peaks riding on underlying waveforms in much greater proportion that this simple example.

You're not going to track this with tone bursts or a microphone or a level meter. You need an oscilloscope to see it.

Must disagree. This is where a true peak reading SPL meter can be used, ime. It all comes down to how fast is the meter in capturing SPL peaks.
REW's peak meter is pretty fast....snap your fingers, or drop a fork on a plate, and witness over 120dB.

And i think a tone bursts thru a speaker, captured by a microphone, can show clipping.
I haven't tried that directly, but i have been capturing tone bursts with a mic, for the purpose of locating times-of-flight to drivers acoustic centers.
Here is an example of how well the speaker bursts can be captured with a mic.
Given the clean start and stops of the 1.5 cycle bursts, i have to believe clipping would show up clearly too.

150 Hz and 700 Hz 1.5 cycle bursts.

The more severe the clipping the more of the musical wave form is driven into pseudo-square waves until finally all of the signal is clipped... at which time you're very likely to exceed the power handling of your tweeter, even with a lower powered amplifier.

Have to disagree again...i've put that into the myth bucket......that like many, i believed for a while.

 

[puppet]

Wonderful. Below is your post #87, and now you reference an article that uses clipping of single frequency sine waves in the "analysis"... and I was being misleading in using test tones?


View attachment 201285

That article is (6) pages ... did you get a chance at reading all of it?

 

[Deleted member 46664]

Of course. Clipping maps to True Peak SPL, whereas time integration maps to a RMS SPL average over the integration period.
Different animals, different tools, that i think can probably be used to detect the onset of clipping.... that was the point i was trying to make...comparing the spread between time averaged SPL and highest peak.

Actually clipping maps to an electronic overload... it is actually a failure mode of the device.

Air because of it's spongy nature will not carry the information without serious degredation.

 

[danadam]

How does clipping in the source fit into the "blow your tweeters" problem?

It is still 44.1k sampling rate, so I'd think that whatever is produced from those samples, will be low-passed at 22k before it reaches speakers.

As for what's below 22k, I'm not sure if that's a proper way, but here is FFT of 1 second at 1:34 of Quiet Winter Night / Blågutten:

• converted to 16/44 with normal (flat) dither,
• converted to 16/44 with shaped dither,
• the first one amplified by 6 dB, resulting in ~1'700 samples clipped, then attenuated by 6 dB,
• same as above but 9 dB, resulting in ~7'000 samples clipped,
• same as above but 12 dB, resulting in ~17'000 samples clipped.

I don't know how to quantify the increase in the required power for those additional frequencies.

 

[Aerith Gainsborough]

I don't know how to quantify the increase in the required power for those additional frequencies.

Looks like a 15-20dB rise between 18- 20 KHz (bright blue to orange).

Since you amped it up by 12dB, we see a 3-8 dB increase due to the clipping (?). That's roughly 2 - 6x the power.
If you operate the tweeter near it's limit before the clipping ,that can push it over the edge.
Though with any half decent tweeter design, your ears will probably die first, given how low the highs are in regards to the rest of the spectrum even after clipping.

 

[danadam]

Since you amped it up by 12dB, we see a 3-8 dB increase due to the clipping (?).

The FFT is taken after attenuating back by 12 dB, so everything on the graph should be due to clipping. Light blue is FFT of top track, orange is FFT of bottom track:

 

[Aerith Gainsborough]

The FFT is taken after attenuating back by 12 dB, so everything on the graph should be due to clipping. Light blue is FFT of top track, orange is FFT of bottom track:

Oh, sorry, I misunderstood.

So we're talking about real 15-20dB gains in the high end from light blue to orange..
That's 32x - 100x the power routed into the tweeter in the extreme high end.

We're still 60dB down from the peak though. If the peak is 400W in the bass, the tweeter would need to handle 400W/100000 = 4mW, so the super high frequencies are not a problem at all.

It's very difficult to judge the 4-5KHz range, I think that one could be more problematic. It looks like a ~6dB difference here, so 4x the power over non clipped.

Though we are still 25dB below the peak. Again 400W peak -> 400W/316 = 1,27W. Nope, not much of a problem either.

At 2.6KHz (crossover point of my Aria's tweeters), the power is a lot though. We're talking only 10dB down from the bass area.
That would be 400W/10 -> 40W into the tweeter.

If I interpret these numbers correctly, the additional high frequency content due to clipping should be a complete non-issue to the tweeter.
Looks to me that if any tweeter would be roasted by this, it would be roasted due to excessive volume itself and not due to the clipping.

Only remaining danger would be any instability/oscillation, if this were amplifier clipping instead of digital clipping.

 

[dc655321]

I don't know how to quantify the increase in the required power for those additional frequencies.

Could you not plot ratios of fft bins (or bin value squared)?
eg: +12dB bins / flat-dither bins

 

[Spkrdctr]

Even with a 35 year electronics career (literally) behind me, I have (until this thread) just believed the "accepted wisdom" that clipping is dangerous for speakers.

I just love how much I continue to learn here.

Yes, but it depends. I think the point is that you end up clipping the sine wave thus creating a nice square wave which creates a doubling of power. I read that big article referenced in a post and it "seemed" to me that the takeaway was that. The clipping in of itself does not fry speakers as it may be such a small amount of clipping it doesn't matter. But, if you turn it up to overcome the clipping and now you have a lot of clipping the increased power through the voice coil will cause a "poof" and you have fried the coil.

That was my understanding. I could be totally wrong and if I am, I am open to you guys letting me know where I missed the point of the article.

 

[gnarly]

That article is (6) pages ... did you get a chance at reading all of it?

I've had the chance to read the whole article. Thx for the link.

One thing i think the author misses right from the gitgo, is once voltage is at the clipping point....you simply can't double voltage and push theoretical numbers around based on the doubling. The amp has maxed out its rails, and its available power.
Once clipping starts here is no more voltage available, and there is also no more wattage available beyond the amps rating.
Even if a square wave is created by clipping, wattage cannot climb above the amps rating.

Interestingly, of page 5 of the article when it analyzes the effects of severe clipping on real music, i pull out this quote...
"In principle we could expect both harmonic generation due to clipping and any “compression” of alternating LF/HF sections to result in a rise in the relative HF power level. However in this example any such result seems quite modest."

Hey, the real moral of the story, to avoid all this confusion .....is don't use passive xovers....multi-amp rules !!!

 

[antcollinet]

...

That was my understanding.

That's what I took from it also.

 

[antcollinet]

I...
Even if a square wave is created by clipping, wattage cannot climb above the amps rating.
...

Yes it can.

An amp is usually voltage limited. And is rated based on a sine wave output. So the max output voltage limits the peak amplitude of the sinewave at maximum power.

But if you output a square wave, you can output a higher RMS voltage without the peak voltage getting any higher. Peak voltage of a sine wave is sqrt(2) x rms. Peak voltage of a square wave is equal to the RMS voltage, which is why we get a doubling of power for a square wave.

Only if the power supply runs out of steam and cannot deliver the higher continuous current will the output power be limited to less than that 2x.

 

[10khz-lpf]

I am going to respond directly to the OP without reading most of the rest of the thread.

What we really need to consider is the peak to peak potential of the voltage (Vpp) driving the amp.
Clipping is when we exceed what the Vpp of the power supply can provide.

So technically you could not encounter true "clipping" if you have enough power. You could be playing back a really bad recording where clipping occured during the recording process, but in that case the "clipping" has become part of the source audio so it is not quite the same.

Examples of where you have enough of one kind of power but lack enough of another kind of power:

• Proper voltage but not enough current
• Proper current but not enough voltage
• Faulty circuit layout leading to the signal being forcibly brickwall limited by an underspecced component on the amp (this would surely lead to some fault in the amp sooner or later)

 

[Deleted member 46664]

One thing i think the author misses right from the gitgo, is once voltage is at the clipping point....you simply can't double voltage and push theoretical numbers around based on the doubling. The amp has maxed out its rails, and its available power.

Only at the frequency that is clipping ... with the normal musical decay demonstrated in message #128 the bass will be clipping first but as gain is increased it moves into the mid to treble range, above the crossover point for a tweeter. Clearly that amplifier still has a lot of extra amplitude it can produce.

It is a mistake to think that an amplifier simply stalls when it hits the voltage rails ... it won't stall until the entire output hits the voltage rails.... and given the change in the angle of the top orange trace (17,000 samples clipped) it should not be hard to see how increasing output with that going on can easily exceed the tweeter's power limits.

Once clipping starts here is no more voltage available, and there is also no more wattage available beyond the amps rating.
Even if a square wave is created by clipping, wattage cannot climb above the amps rating.

Not true ... until the waveform is 100% clipped.

 

[NTK]

That article is (6) pages ... did you get a chance at reading all of it?

Very tricky. I missed the left/right arrows at the bottom of the page, which seem to be the only indication (and aside from a careful read of the article) that there are more pages.

So, going to the conclusion page, the author started with these hypotheses:

​

The main hypotheses being compared being:​

• Harmonic generation due to clipping altering the waveform shape. Thus damage occurs due to HF which would not be present at all in an unclipped waveform.
• Differential compression where the clipping limits the LF level, but permits the HF to rise by an amount comparable with what would be possible in the absence of clipping. Thus damage occurs due to the HF which would be present in the waveform even if its shape was unaltered by clipping

Then he said:

​

On the basis of the analysis carried out, and the examples used, the amount of HF/LF differential clipping which occurred did not seem to be very great. ...​

​

But still concludes:

​

In general, therefore, it seems quite plausible to assume that damage may sometimes occur due to any one of the following mechanisms, or various combination of them:​

• HF creation by clipping of the LF.
• Rise in the HF level of the original signal which – although limited by clipping by an amount similar to that which occurs at LF – rises to a high level.
• Leakage of LF power.
• Non-themal (sic) effects due to an increase in the peak displacement or accelerations.

To summarize, the author of the article had some hypotheses, but was unable to prove them, but concluded that they were still "seem quite plausible". How insightful is that?

Still, this article didn't address the question: Will an amplifier that does not clip (i.e. essentially with unlimited power) be "safer" (to a loudspeaker with the typical relative power handling capacities across the frequency range) than one that clips?

 

[Deleted member 46664]

Still, this article didn't address the question: Will an amplifier that does not clip (i.e. essentially with unlimited power) be "safer" (to a loudspeaker with the typical relative power handling capacities across the frequency range) than one that clips?

Yes a higher powered amplifier with less or no clipping is safer ... because of the normal decay in power inherent in music.
In well produced recordings, you may hit 100 watt peaks from deep bass or snare drums but at the same time a symbol or a flute that sounds equally loud is only asking for 2 or 3 watts. This decay is natural in all music.

As long as the frequency based fall off in output levels is maintained you can get to higher power levels in the bass region (hence higher SPL) without hitting the tweeter's power limits above the crossover frequency.

Tweeters generally have far lower power handling than woofers ... For example: 100 watt bookshelf speakers often use 10 watt tweeters.

Clipping tends to compress the energy above the crossover frequency giving it proportionally more energy than it would normally produce. Thus in a severe clipping situation it is possible for a low powered amplifier to exceed the power handling of a tweeter. This in turn causes excessive heat in the voice coil and will burn through the wire if sustained.

With a speaker rated for 10 to 100 watts ... You will be far more likely to blow a tweeter with a 20 watt amp than with a 100 watt amp.