Actually, in a really good amp 'clipping' should be inaudible ...
Huuuhh? Well, if you look at a decent cross-section of audio waveforms, from real CD tracks there is significant clipping in many cases - it can be disguised by having the peak bits not actually bumping the end stops, but it's a fake: it's easy to see that the waveform should have gone well above the peak bit, by comparing neighbouring wave peaks with the 'villain' - but the mastering applied a bit of cute chopping off and dithering at that point, so pretending that nothing actually 'bad' happened. I've seen clipping on all sorts of material - "high quality" jazz, classical, etc.
Yet, listening to these pieces at normal volumes I hear nothing untoward, there are no audible nasties at those "bad" bits. So, playing a piece with minor clipping it is not evident that something's wrong - therefore, playing non-clipped material so loudly that every now and again the amp clips also should not be audible - IF the amplifier circuitry is very well behaved ... and there's the rub of course ...
Huuuhh? Well, if you look at a decent cross-section of audio waveforms, from real CD tracks there is significant clipping in many cases - it can be disguised by having the peak bits not actually bumping the end stops, but it's a fake: it's easy to see that the waveform should have gone well above the peak bit, by comparing neighbouring wave peaks with the 'villain' - but the mastering applied a bit of cute chopping off and dithering at that point, so pretending that nothing actually 'bad' happened. I've seen clipping on all sorts of material - "high quality" jazz, classical, etc.
Yet, listening to these pieces at normal volumes I hear nothing untoward, there are no audible nasties at those "bad" bits. So, playing a piece with minor clipping it is not evident that something's wrong - therefore, playing non-clipped material so loudly that every now and again the amp clips also should not be audible - IF the amplifier circuitry is very well behaved ... and there's the rub of course ...
Clipping that was already applied to the signal before it was recorded onto the CD doesn't really count, Frank. It gets band-limited to 22 kHz, or whatever.
When clipping happens "live", in a power amp, it happens because the output's peak level control has been turned up to the point where it is attempting to squeeze the voltage across the amplifier itself (between power rail and output) into a too-small space above the output peak level and below the bottom of the ripple voltage waveform. It's just like violating the dropout voltage of a regulator. You get ripple-shaped chunks being gouged out of the output signal's waveform, which can leave jagged corners even if both the ripple and signal contain only rounded shapes. And if you go deep into clipping, the whole top of the output signal can get sheared off and you have square corners. (And similarly for the negative-polarity part.)
Because of the sharp edges it can create in the output waveform, clipping will possibly produce a spray of high-frequency stuff that has ben known to blow the ribbon tweeter fuses of my MG-3.6 speakers, and burn up tweeters in other types of speakers.
So it's actually true that a less-powerful amplifier is more likely to damage tweeters than a more powerful amplifier. e.g. Say that you usually use an amp that can provide 400 Watts per channel into 4 Ohms, and it can make the sound loud-enough without ever even getting close to clipping. But if that amp needs a minor repair and you temporarily have to use an amp that can only provide 150 Watts into 4 Ohms, it may not be loud-enough unless you turn it up to the point where it clips.
When clipping happens "live", in a power amp, it happens because the output's peak level control has been turned up to the point where it is attempting to squeeze the voltage across the amplifier itself (between power rail and output) into a too-small space above the output peak level and below the bottom of the ripple voltage waveform. It's just like violating the dropout voltage of a regulator. You get ripple-shaped chunks being gouged out of the output signal's waveform, which can leave jagged corners even if both the ripple and signal contain only rounded shapes. And if you go deep into clipping, the whole top of the output signal can get sheared off and you have square corners. (And similarly for the negative-polarity part.)
Because of the sharp edges it can create in the output waveform, clipping will possibly produce a spray of high-frequency stuff that has ben known to blow the ribbon tweeter fuses of my MG-3.6 speakers, and burn up tweeters in other types of speakers.
So it's actually true that a less-powerful amplifier is more likely to damage tweeters than a more powerful amplifier. e.g. Say that you usually use an amp that can provide 400 Watts per channel into 4 Ohms, and it can make the sound loud-enough without ever even getting close to clipping. But if that amp needs a minor repair and you temporarily have to use an amp that can only provide 150 Watts into 4 Ohms, it may not be loud-enough unless you turn it up to the point where it clips.
Last edited:
Tom, that's exactly what can be put on a CD, it's quite easy to create perfect square waves, if 1 and -1 are the maximum excursions, then on the track you could get, say: 0, .2, .5, .7, 1, 1, 1, 1, 1, -1, -1, -1, -1, etc - if the player output passes this through cleanly that's what the amp sees ...
The point is obviously not to be silly about it ... a touch now and again, normally won't cause a problem. But, the real issue is whether that "distortion", as a momentary 'glitch', is audible ...
Triangle waves were popular for testing for clipping and slew rate limits using only an oscilloscope for observation. I suspect Dick you just may have better gear than that. 
With a digital recording there is a maximum signal level often 2 volts rms. Typical professional amplifier gain is 26 dB. Many consumer amplifiers have higher gain.
As very few use a preamp to raise the gain, the maximum output voltage would be 2 x 20 x 1.4 = 56 volts peak. Add about 6 more volts to keep the output stage happy and clipping should not be a problem.
Now as 2.83 volts is the efficiency reference level a 62 volt rail amplifier should be able to drive a loudspeaker to 26 dB above that reference number.
So an inefficient loudspeaker rated at 82 dB/W would produce 108 dB at 1 M. At a distance of 4 M from a pair of loudspeakers that would produce about 99 dB peaks. For very conservatively recorded classical music there could be as much as 30 dB of head room.
With a bit of room gain due to reverberation that would be a bit shy of reproducing front row in a real hall. It would do back row. Now a normal recording would have 20 dB or less so you could reproduce front row. However in virtually all concert halls front row is not as well balanced as further back!
Now if you wish to produce rock music at concert volumes you will require peak levels of 115 dB. So you won't find too many 98 dB/W loudspeakers and at those levels you will loose another 3 dB due to voice coil heating.
Now if you really do live away from any neighbors then you can look at double stacked loudspeakers and another amplifier to reach that level.
The system I just brought up will do 112 dB for 60,000+ seats from 50 to 10,000 Hz.
With a digital recording there is a maximum signal level often 2 volts rms. Typical professional amplifier gain is 26 dB. Many consumer amplifiers have higher gain.
As very few use a preamp to raise the gain, the maximum output voltage would be 2 x 20 x 1.4 = 56 volts peak. Add about 6 more volts to keep the output stage happy and clipping should not be a problem.
Now as 2.83 volts is the efficiency reference level a 62 volt rail amplifier should be able to drive a loudspeaker to 26 dB above that reference number.
So an inefficient loudspeaker rated at 82 dB/W would produce 108 dB at 1 M. At a distance of 4 M from a pair of loudspeakers that would produce about 99 dB peaks. For very conservatively recorded classical music there could be as much as 30 dB of head room.
With a bit of room gain due to reverberation that would be a bit shy of reproducing front row in a real hall. It would do back row. Now a normal recording would have 20 dB or less so you could reproduce front row. However in virtually all concert halls front row is not as well balanced as further back!
Now if you wish to produce rock music at concert volumes you will require peak levels of 115 dB. So you won't find too many 98 dB/W loudspeakers and at those levels you will loose another 3 dB due to voice coil heating.
Now if you really do live away from any neighbors then you can look at double stacked loudspeakers and another amplifier to reach that level.
The system I just brought up will do 112 dB for 60,000+ seats from 50 to 10,000 Hz.
Other than ignoring Nyquist and reconstruction. It may be worthwhile for your understanding to do a couple actual experiments and see what waveforms (and spectra) result from non-physical data before expounding on it.
That's fair comment. As you say, it's little more complex than I implied, even in the theoretical sense - I should have considered a little more carefully what Tom said, and replied with more clarity about what I was trying to convey.
Implementations of DACs also vary from the theoretical, in how they actually behave in real life when fed these sorts of waveforms, further complicating matters.
Implementations of DACs also vary from the theoretical, in how they actually behave in real life when fed these sorts of waveforms, further complicating matters.
Yes, as far as I know the pre-anti-aliasing has a cutoff frequency of just above 20kHz to keep the ADC happy. So any harmonics above that can be present but will be (much) attenuated. In fact, your square wave will look like any square wave that went through a bandlimit filter. As it should be; this stuff is an intergral part of the overall CD chain design.
So that's the kind of rounded-corner square wave your amp will see.
jan
Last edited:
Any harmonics that get past the anti alias filter will of course alias to a lower frequency.
Absolute nonsense, if you do not understand the basics, try it before you write, look at CD player analog output.
Pavel, I've already acknowledged the sloppiness on my part - no need to be an echo ...
Of course the analogue version of the digital waveform won't be a 'perfect' square wave, there will a 'rounding' of the corners in some sense. Precisely how that rounding will manifest is then dependent on a whole lot of implementation variables.
And, the clipping of an amp's rails also won't be a perfect slice off - stray capacitance and inductance will guarantee that ...
Of course the analogue version of the digital waveform won't be a 'perfect' square wave, there will a 'rounding' of the corners in some sense. Precisely how that rounding will manifest is then dependent on a whole lot of implementation variables.
And, the clipping of an amp's rails also won't be a perfect slice off - stray capacitance and inductance will guarantee that ...
Again, there is nothing stopping someone placing a severely a-musical signal on a CD track - how any ADC operates is irrelevant, the 'damage' can easily be done in a digital editor, if so desired. I in fact have a test CD with those very sort of tracks ...
So, it all then depends on how the reconstruction is done, as you say. There will the theoretically correct behaviour - and then what actually occurs for a particular implementation, and example, of DAC and output buffer. It's already been noted that some DACs misbehave even if they just see a maximum amplitude signal as part of a non-clipped signal, let alone where this is combined with 'pure' clipping as well. The real world doesn't always work as well as it should, especially when $$$, or 'funny' ideas intrude ...
So, it all then depends on how the reconstruction is done, as you say. There will the theoretically correct behaviour - and then what actually occurs for a particular implementation, and example, of DAC and output buffer. It's already been noted that some DACs misbehave even if they just see a maximum amplitude signal as part of a non-clipped signal, let alone where this is combined with 'pure' clipping as well. The real world doesn't always work as well as it should, especially when $$$, or 'funny' ideas intrude ...
I'm not saying you are wrong, but could you say where this has been previously "noted"? And could you make clearer what the heck you are talking about? What is "pure clipping", and how does it differ from whatever the other thing you are talking about is (presumably "impure clipping")? I have to say this sounds like a lot of hand-waving that lacks any meaning, or at best can mean whatever someone wants it to mean.
Would have to play a bit with Google for this, because it's been awhile since I ran across mentions of this: in objective measuring of CD players and DACs one of the tests is specifically to measure the output when a maximum level sine wave is played, where there are known peaks of +1, and -1 recorded, because in some cases spurious artifacts occur, a significant drop in distortion performance is noted.
Further, a perfectly valid, non-compressed, 'undistorted' digital signal can be generated such that the reconstructed analogue waveform contains a peak value beyond what the digital maximum represents - there's a term for this 'construction' - it's quite easy to draw an example, if one thinks for a minute. And, again, some DAC circuitry does not reconstruct this correctly, the analogue clips, even though the digital representation is correct, or 'pure'.
Further, a perfectly valid, non-compressed, 'undistorted' digital signal can be generated such that the reconstructed analogue waveform contains a peak value beyond what the digital maximum represents - there's a term for this 'construction' - it's quite easy to draw an example, if one thinks for a minute. And, again, some DAC circuitry does not reconstruct this correctly, the analogue clips, even though the digital representation is correct, or 'pure'.
To reduce the chance of any more twaddle type reactions, I did a little hunting, via Google - I would suggest that those that "don't get" it read this, Issues with 0dBFS+ Levels On Digital Audio Playback Systems | Audioholics, and spend some time digesting the concept ...
- Status
- Not open for further replies.