|Elliott Sound Products||Why Do Tweeters Blow When Amplifiers Distort?|
Copyright © 2001 - Rod Elliott (ESP)
Page Updated June 2015
A vexing question, regularly asked and rarely answered properly - Why Do Tweeters Blow When Amplifiers Distort? The answers are actually quite simple, but the common misconception is that the distortion creates harmonics, and the additional harmonic content destroys the tweeter.
Not really - woofers and midrange drivers can also blow from a distorted amp, and this is unlikely to have anything to do with harmonics. Certainly, there are additional harmonics generated, and they will be at relatively high levels, but rarely high enough to cause more than relatively minor stress to the tweeters (in particular).
It is also worth looking at the article Amplifier Clipping, as this provides some additional information that is not covered below. Although it usually does not affect tweeters, the effects on low frequency drivers can be very harmful.
Because of the nature of music and the over-use of compression, it is not only tweeters that are at risk. Woofers can also be damaged, due simply to excessive continuous power. In the end, it doesn't matter if the signal to the speaker is clipped or not - if you feed excessive power into a loudspeaker, it will fail.
Note that this article has been substantially changed (twice now). The conclusions are the same as the original, but new graphics have been added to show the waveform and an FFT of the tweeter signal under three different drive conditions. Hopefully the new information will make the article clearer, and the new simulations show the fundamental and harmonics so you can see the relative effects of clipping induced compression and harmonic generation. While the latter is certainly a contributor to the total power, its influence is far less than you may have expected.
A great part of the mystery is uncovered when we look at two aspects of music - the average versus peak power, and the energy distribution of typical music material.
It is commonly accepted (and quite valid) that music has a peak to average ratio of about 10-20dB. This means that if the signal is being amplified by a typical 100W amplifier, the amp's power rating limits the absolute maximum power to 100W (give or take a little). Since this is the peak, the average must be somewhat lower, and we will assume 10dB for the sake of convenience. Average power is therefore 10W or less at the onset of clipping.
This is not dynamic range per sé, but it is most certainly a part of the overall dynamic range of the music signal. The term 'dynamic range' usually refers to the very quietest up to the very loudest passages in a given piece of music. In some cases, there is no variation whatsoever - it starts loud, is loud in the middle, and (just to be different) finishes ... loud. The peak to average level may also be compressed, but it is difficult to reduce it to less than 10dB without it becoming flat and lifeless. If done incorrectly, it can simply become a jumbled mess with no intelligibility whatsoever (and no, I'm not going to take this to its logical conclusion and denounce various styles that may be classified as music to only a select few )
Most speakers are rated for a continuous power and an instantaneous power - the voice coil and to a lesser degree the suspension can withstand short bursts at much higher powers without damage. This does not imply that such power will be reproduced cleanly, and it will almost certainly be with a large increase in speaker distortion. The peak power rating defines the maximum transient power the loudspeaker can handle without suffering electrical or mechanical (stress induced) damage.
Nearly all tweeters are rated to 'system power', and this will usually be quoted relative to a specific crossover frequency. A hypothetical tweeter may be rated at 100W system power when crossed over at 3,000Hz. The power that it can withstand is not 100W! Not at any frequency or for any duration. The actual bandwidth-limited long-term average power for most tweeters is around 10W, but many can't even handle that much without some distress.
Figure 1 - Power Distribution Chart
The above power distribution table is approximate (as must be the case), and applies for 'typical' music - whatever that may be. If we look at the case for a crossover frequency of 3kHz, we can see that 85% of the power is in the low frequency spectrum, and only 15% in the high frequencies above 3kHz. It is not difficult to deduce from this that the peak power to the tweeter will be in the order of 15W at full power from the amplifier, with the average hovering around 1.5 watts.
This is the way the system was designed to be used, and as long as the power amp does not clip, all is well (well, almost - read on).
When an amplifier is overdriven, the sound becomes distorted. This manifests itself in many ways, but the two we are interested in are the generation of harmonics, and the reduction of dynamic range - both the true dynamic range and the peak to average ratio. Let's assume that the amp is overdriven by a mere 3dB, so the average level is now 20W, and the peaks are clipping. With many systems (or listeners), this will be virtually inaudible. Careful listening will uncover the fact that there is distortion present, and there is a definite reduction of intelligibility.
The speakers - both tweeters and woofers, are now being asked to absorb twice the power that would be normally obtainable, and the power is more constant - the signal is compressed by the power amp. Add to this the additional harmonics generated by the clipping waveform, and the tweeter may actually be getting up to 3 times the continuous power that was available before clipping. Peak power remains the same, since it is limited by the amplifier's power supply voltage.
Now, let's overdrive the amp by 10dB. The amp is delivering in excess of 100W, since it is reproducing square waves much of the time. The woofer will be subjected to perhaps a continuous 100W of power, and around 15W continuous will be available to the poor tweeter. Of this, probably less than 1% will be converted into sound (1% represents an efficiency of about 92dB/W/m). Ferrofluid helps, but virtually no hi-fi tweeter can withstand that sort of continuous power for any duration.
The tweeter was never designed for that! Just look at a 10W wirewound resistor for example. It is big and chunky, and made from a ceramic material that is designed to handle a lot of heat. Run one at 10W to find out just how much heat you will get. There is very little airflow around the tweeter voice coil, and the heat has nowhere to go. The result is that the voice coil will quickly overheat, and the adhesive that bonds the coil to its former, the former itself, and even the enamel insulation on the coil will be damaged. The result (naturally) is a dead tweeter.
As for the woofer - unless it is designed to take 100W or more continuous sinewave power, it will also overheat and eventually die. It takes a lot longer, because there is airflow around the voice coil, and the coil is bigger and has greater thermal inertia, but die it must if the abuse is maintained.
The following diagrams illustrate the above. The waveforms below are the result of simulation, but 'real life' will show exactly the same things as described. The waveform used for the simulation was made up from the following signals ...
|Frequency||Peak Amplitude (V)||Relative Amplitude (dB)|
This waveform is not an attempt to reproduce any musical instrument or section of music - it is simply a batch of frequencies that make up a suitably interesting waveform. Of great importance is the ability to use this signal to demonstrate how the relative levels at various frequencies are affected by clipping. The peak to average ratio of the signal used is around 8.56dB. As noted above, this is (typically) about what we can expect with music, and because no frequency is below 160Hz it is realistic for a mid+high system used with a separate subwoofer. Unfortunately, it is simply not possible to cover all possibilities, but I believe that the test waveform used is sufficiently realistic to provide a useful result.
The signal (both normal and clipped) was passed through a passive 12dB/octave Linkwitz-Riley crossover network set at 3kHz. This is very typical of the results seen with a conventional passive speaker network. Lower order passive filters and/or lower crossover frequencies will make the situation worse.
The simulated 'amplifier' used for the following charts and graphs has a supply voltage of ±30V, so has a nominal rating of 56W into 8 ohms. This was selected for convenience as much as anything else, but the results are no different with real amplifiers - regardless of power. Once the amp clips, more power is available to both the woofer and tweeter, and surprisingly little of the extra is the result of harmonics.
The first set of graphs was with the amp's input set so that the system was right to the maximum voltage swing, but with no clipping at all. Total amp output was measured at 11.2V RMS, so the peak to average (RMS) voltage ratio is 8.56dB. The gain was then increased by 6dB for the 'slight clipping' example, and a further 6dB (12dB total) for 'moderate clipping'.
Figure 2A - Normal Signal, Without Clipping
This shows the waveform at a level just below clipping. The red trace is the full range signal, green is the signal to the tweeter, and blue is the woofer signal. Notice that the tweeter signal is continuous - there are no breaks or discontinuities. This is important as we shall see from the next waveforms.
Figure 2B - Full Range Spectrum, Without Clipping
The above shows the spectrum of the composite waveform, with the amplifier voltage set at just below the onset of clipping. This is the reference, and the spectrum simply shows each frequency at its preset level. The RMS voltage of the tweeter signal is 3.7V, compared to 11.2V (RMS) for the total signal before the crossover.
Figure 2C - Tweeter Spectrum, Without Clipping
Figure 2C shows the spectrum of the signal presented to the tweeter. As you can see, it reflects the respective levels, with the low frequencies suitably attenuated by the crossover network.
Figure 3A - Slight Clipping
The clipping is visible, and the peak amplitude has increased. The overall spectrum shows that the background energy level (i.e. any frequency that was not part of the input) has risen from around 10mV to an average of about 100mV below 20kHz. This is relatively insignificant, but also note that the peak levels of the input signals are also greater than the previous case.
Of particular concern are those frequencies below the crossover point (3kHz) - 400Hz, 1kHz and 2kHz peaks are noticeably higher than in the previous graph - a clear sign of impending danger. Overall level to the tweeter has increased from 3.7V to 6.4V.
Figure 3B - Tweeter Spectrum, Slight Clipping
The spectrum overall shows that the average level of the signal and harmonics has increased as described above. At this level, there is some degradation of sound quality, but it could easily be missed during casual listening.
The tweeter voltage is now 6.4V RMS, and the total signal measures 19.3V. This means that the total signal voltage has increased by 4.7dB, and the tweeter signal level has increased by 4.76dB.
Figure 4A - Moderate Clipping
Increasing the gain further (another 6dB), the waveform is now visibly distorted, and if you look at the green trace (the tweeter) you can see that there are very noticeable 'dropouts'. When the overall signal clips, sections of high frequency signal are simply lost, and replaced by harmonics of the clipped waveform. These harmonics do not replace (musically speaking) the frequencies that were removed by clipping.
The relative amplitudes of the tweeter signal and the full range signal show us that while the tweeter voltage has increased to 8.9V, the overall is now 24.7V. This is an increase of 7.6dB for the tweeter, and 6.9dB for the overall signal (compared to the unclipped waveform).
Figure 4B - Tweeter Spectrum, Moderate Clipping
The overall level is now much greater than before - it has risen quite dramatically over the entire spectrum. Not only are the wanted frequencies at a higher level, but so are the harmonics of the lower (clipped) frequencies. However, the background 'noise' of the harmonics is still more than 10dB below the peaks, so the contribution from harmonics is not great.
What we do see is the increase of signal level to the tweeter, with the moderate clipping example showing a 7.6dB increase. Remember that this translates to an average power increase of over 5 times to the tweeter. If the tweeter would normally be expected to handle a peak power of 15W and an average power of perhaps 5W (based on Figure 1 and a 100W amplifier), a 7dB increase will take that average to 25W! The tweeter will not survive.
The above examples represent a total gain increase of 12dB from the non-clipping condition to the moderate clipping states. As the gain is increased further, the 'power compression' effect described becomes worse. At no time does the additional energy of the harmonics created by clipping exceed a level of more than around -7dB with respect to the fundamental frequencies - unless one goes absolutely berserk and forces the amplifier into total square-wave clipping. That will occur with 20dB of excess gain (i.e. 20dB above the clipping level, based on the waveform described above).
With the volume advanced by 12dB above clipping, the average power is well above danger level. The distortion will be highly audible with most music, but is not at all uncommon. Parties are probably the worst offenders, although I have heard many high power (10kW or more) PA systems driven to at least this level of distortion.
Bear in mind that most "ordinary" (non-audiophile) people will not be aware that there is anything wrong at this point! During 'festivities' (for want of a better term, liberal quantities of alcohol enriched fluids will ensure that ears remain unresponsive to the assault. Given the number of times I have heard about peoples' speakers "blowing up" during (or after - allegedly) a party, we can safely conclude that the requirement for more noise is far greater than any need for quality.
Few tweeters will last long with sustained power at a level easily achieved with an overdriven 100W amplifier - they are simply not designed for it. Midrange speakers and woofers will probably be pushed beyond their limits as well. Remember too, that not only is the "in band" signal increased, but so is the out of band signal. The tweeter must also handle many times the normal power at all frequencies below the crossover frequency. This alone can cause damage, without the heating effects of such a high sustained power level.
Is the last example extreme? Not at all - I have actually understated the reality. When any system is driven so hard that the distortion is obvious to even untrained ears, it is reproducing close to square waves much of the time. A 100W amplifier driven to full power with a square wave will produce 200W, and when driven into heavy clipping with a music signal can easily achieve well over 100W average continuous power.
Conclusion ? ---- Exit tweeters (and/or midrange drivers and/or woofers).
The term 'power compression' is used here to describe the way that the power level is increased as an amplifier clips. Although it is covered above, the original drawings I used did show the effect rather well, although I was never very happy with the drawings themselves. The new images are a capture from the radio that I have connected to my PC, and hopefully show the effects better than before.
Figure 5 - Original Full Bandwidth Signal - No Clipping
Figure 5 shows the captured signal, and the red line indicates the clipping level. The signal shown was captured over about 20 seconds, and as you can see, the signal never reaches the red line. The average can be determined roughly by eye, and it's around 0.4 on the scale (blue line). Since the chart shows voltage, that's roughly ⅙ power, or 16W for a 100W amp.
Figure 6 - Full Bandwidth Signal - Heavy Clipping
Advancing the volume control by only 6dB (double the voltage gain) pushes nearly all peaks beyond the red line (representing the amp's maximum power), so the signal clips heavily. The average power is now around 0.7 on the scale - this is the half power point (3dB below maximum power), so the average is now 50W for the same 100W amplifier. This means that the power delivered to the woofer is doubled, and power to the tweeter is slightly more than doubled (because of the additional harmonics).
Figure 7 - Full Bandwidth Signal - Severe Clipping
A further 6dB gain increase is very bad news indeed. The average voltage is at least 0.9 on the scale, and that means that the average power is a minimum of 80W. Because the signal is clipped so badly, there will be times when the output power will be well over 100W. To understand this, remember that a squarewave has double the power of a sinewave, so a 100W (sinewave) amplifier can deliver 200W of a squarewave signal.
With this much clipping, the tweeter will be subjected to far more continuous power than it was ever designed for. As in the previous example, it will not survive.
These last three diagrams are another way of showing exactly the same thing as described with the example power waveforms, but over a longer period of time. The result is the same in either case (since they describe the same thing). All loudspeaker drivers in your system are at risk, and the tweeter (being the most fragile) is usually the first to go.
As for claims that a bigger amplifier won't clip, so won't blow your speakers - be very careful! Remember that I have covered clipping at 10dB (or a little more) above the amplifier power, and although highly undesirable, it does happen. 10dB above 100W is 1kW! Will your speakers survive being driven with peak power of 1kW? The answer (of course) is no!
All the above is valid when an amp clips, but what about speaker drivers that blow for no apparent reason, and with no clipping at all? Well, the people who create the final mix for CD (and those who actually transfer the final mastered tracks to CD format) have been using compressors for a long time - almost as long as music has been transferred to mass media in fact. What is different now, is that many of them seem to use their compressors with increasing vigour. The result is music that is compressed so heavily that the 10dB peak to average ratio we used in earlier examples no longer applies. Each individual recorded track is compressed, the final mix is compressed, and it may be compressed yet again before it is transferred to CD. For reasons that I cannot fathom, it seems that every new CD that is released has to be 'hotter' (louder) than the last. Guess what people - there is a limit, and we reached it long ago !
Towards this end (and to make matters worse), you will find some CDs that are pre-clipped, with the peaks of the occasional loud signal neatly cut off at the top and bottom. With a peak to average ratio of perhaps 6dB in the worst cases, we need to examine the effects on the loudspeaker drivers.
Using the same 100W amp as above, it was already established that the average power would be around 10W at the onset of clipping. Although this is not a trivial amount of heat, most speakers will be able to dissipate it without too much difficulty. As the peak to average ratio is reduced by compression, the effect is exactly the same as clipping the amp, except we don't get the harsh distortion.
With a 6dB peak to average ratio, the poor loudspeaker now has to cope with an average power of 50W as the peaks reach clipping level. Unless you are dealing with serious professional drivers that are designed for continuous high power, 50W becomes a lot of heat to dispose of. Should you be using a 200W amp, that increases to 100W - continuous! If you think 100W is trivial, try holding a 100W resistor that is dissipating its rated power. Actually, don't! You will burn yourself quite badly. Remember that over 99% of all power that goes into a domestic loudspeaker is dissipated as heat, with less than 1% being converted into sound.
As you can imagine, it is entirely possible to destroy drivers without ever letting the amp clip. Speaker systems rated for 200-500W almost always refer to peak power, and the average (or continuous) power rating is in reality as little as 10% of that claimed. This is perfectly alright for most people in a domestic environment, where very high SPL isn't needed. If you like to listen really loud, then you will almost certainly have problems with drivers failing. Perhaps the loudspeaker manufacturers haven't quite got the message that a great deal of music is compressed much harder than they anticipated.
The above applies to tweeters, midrange drivers and woofers equally - all can be damaged by excessive sustained power. Regardless of claims, there are actually few loudspeakers that can withstand 100W of continuous average power for any length of time. To be able to do so requires large diameter voicecoils (preferably with aluminium formers to aid heat removal), vented pole-pieces, finned magnet assemblies and somewhere for the heat to go - preferably not just into the cabinet where it can't escape. This is not a trivial exercise, and professional drivers that can withstand that sort of abuse are very expensive indeed.
Finally, the CD (and other digital formats) allow the recording engineer/ producer to have as much bass as they want, plus some more for good measure. This can extend to very low frequencies - below audibility in fact. Tuned (vented) loudspeaker enclosures (usually) provide better bass extension than sealed types, but if there is any material recorded (and amplified) that is below the box tuning frequency, this will cause massive woofer cone excursion because the rear of the cone is unloaded below the vent tuning frequency. You won't hear the deep bass, but the cones will move very energetically indeed, and can easily reach physical excursion limits. Loudspeaker damage is guaranteed if this is done at high power levels, and can include buckled and/or detached voicecoils, broken lead-in wires and torn suspensions.
A persistent myth in the audio industry is that clipping damages tweeters, so you should use a bigger amp to ensure more headroom so the amp won't clip. This claim is simply bollocks! Take the 100W amp described above, and replace with an amp big enough to prevent clipping ... even with the additional 12dB input signal as shown in Figure 7. Since a 100W amp was just below clipping with an average output of 16W, if we add 12dB that takes the peak amp power to 1.6kW (near enough) and the average power will be 254W.
Do you imagine for an instant that this amp won't blow the tweeters (and everything else) if the input level is increased by 12dB (until it's just below clipping)? Everything will fail, and usually fairly quickly if the speaker was designed for a 'nominal' 100W input. It is simply nonsense to imagine that the loudspeaker drivers in a 100W speaker can survive an average power of over 250W and peak power of up to 1.6kW.
If a user often turns their amp up to beyond clipping levels, they will do the same with a bigger amp. They might even turn it up more, because it won't have the distortion component which increases apparent loudness until the average power is a great deal higher. Such users will never hear signs of speaker distress if they can't even hear gross clipping. Speaker failure is a certainty, even if their 1.6kW amp only ever clips a few transients. They can expect the tweeter to fail, and the woofer to catch on fire.
So, while it's perfectly alright to allow perhaps 3dB or so of headroom for the power amps, that relies on that fact that it is reserved as headroom! If you use the extra power then there's no headroom any more, and all the effects explained will still happen, but at even higher power levels than described above.
Headroom implies that there is extra power for the occasional transient. If you use the extra power, you no longer have any 'headroom' !
|Copyright Notice. This article, including but not limited to all text and diagrams, is the intellectual property of Rod Elliott, and is Copyright © 2001-2006. Reproduction or re-publication by any means whatsoever, whether electronic, mechanical or electro- mechanical, is strictly prohibited under International Copyright laws. The author (Rod Elliott) grants the reader the right to use this information for personal use only, and further allows that one (1) copy may be made for reference. Commercial use is prohibited without express written authorisation from Rod Elliott.|