Tutorial: Transistor Amplifier Clipping

[Flint]

I hope to get close to explaining the most basic ideas around amplifier clipping.

When a signal contains lots of bass frequency information along with treble information, the treble information "rides" on the wave of the bass info. So, any amplitude in the treble will extend above and below the amplitude of the bass info.

The first image below are the waveforms of two sine waves, one at 50Hz and one a 220Hz:

 

[Flint]

If I were to combine them in a way that no clipping occurs, they would look like the second image:


If you follow the waveform closely, you will see the amplitude of the 50Hz wave at about 50% on the scale and the 220Hz waveform is "riding" on top of the 50Hz wave and increasing the amplitude to 100%.

 

[Flint]

If I were to combine the two waveforms and keep the 50Hz wave at 100%, it would look like the third image, below:

Notice that the 200Hz signal is now clipped off while the 50Hz signal is not.

 

[Flint]

This is part of the reason why tweeters often fail when amplifiers are underpowered for the signals they are being asked to drive. A loud bass signal can cause the treble to clip while not actually consuming all the power for itself.

Simply by removing the bass from the signal before it reaches the amplifer will reduce this potential issue pretty easily. Back when turntables were the norm, nearly all receivers and integrated amplifiers had subsonic filters that would filter out everything below 20Hz, or so, in order to reduce this potential problem. It was common for turntables to have LOTS of information below 20Hz that would cause the amplifier to put out energy at frequencies the speakers could reproduce anyway. Since the amps were driving that deep bass, they were more likely to clip the higher frequencies which distorted the desired musical information and could potentially damage the speakers.

In a HT, filtering the bass out of any speaker that cannot reproduce it will reduce the potential for this issue arising. Hence my general recommendation to set all speakers to small and adjust the crossover on the receiver at or slightly above the limits of the speakers. This not only ensures that the bass in the signal that cannot be reproduced by the speakers is sent to the subwoofer so that you will hear it, it also protects the speakers and can reduce the likelihood of clipping distortion in the midrange and treble.

In case you have never heard clipping, I have created 5 music files you can download and listen to.

Right click on the link below and SAVE the file, do not open it from the browser. Download it to your hard drive, scan it for viruses and then extract the MP3s in the file. Once extracted you will have five mp3 files -
http://www.soundenvironments.org/Data_Files/clipped.zip

First, listen to the file named "unclipped.mp3" to hear the original recording, unclipped.

Then, I suggest you listen to "clipped3dB.mp3". As an analogy, imagine a 100watt amp playing the unclipped file at the absolute maximum loudness without any clipping. Then, imagine the loudness remains the same, but the 100watt amp is replaced with a 50 watt amp. The clipped3dB file is what the music would sound like.

Next, play the "clipped6dB.mp3" file. This is like moving from the 100watt amp to a 25watt amp while keeping the loudness the same.

Then play the "clipped10dB.mp3" file. This is similar to moving from a 100watt amp to a 10watt amp.

Finally, play the "clipped20dB.mp3" file. This is an extreme example of moving from a 100watt amp to a 1watt amp.

Keep in mind, I am modelling the nature of transitor amplifiers to clip when the signal exceeds their voltage capabilities. Amps with soft clipping circuits and tube amplifiers do not react in this way.

 

[Flint]

If the amplifer's maximum output is close to or more than the maximum input capabilities of your tweeters, then a small about of clipping can cause them to fail more quickly. However, if the power output of your amp is considerably less than the capabilities of the tweeters, no amount of clipping can cause them to fail.

The gotcha is that you have no idea what your tweeters can truly handle. While the speaker system as a whole might be able to handle 200 watts (or whatever), most high quality tweeters only handle about 20 watts to 50 watts continuously into their operating range (cheap tweeters handle much less). Since tweeters are generally more efficient than the woofers they are teamed with, they have pads that reduce the incoming signal to match their output to the woofer's. However, when an amp is clipping, there is much more energy concentrated in the highest frequencies, so even with the pads, more energy is getting to the tweeters than with a non-clipped signal with the same average level (like a larger amp's signal). So, a 100 watt amp is likely to blow a tweeter when clipped, whereas, a 20 watt amp will not likely ever blow a quality tweeter.

 

[Flint]

Here's the deal with square waves and tweeter voice coils. The power that goes into a driver's voice coil is converted into magnetic flux which in turn causes the voice coil to move in or out, depending on the positive or negative voltage on the coil. Most of the energy is converted to heat but about 5% to 40% is converted into physical motion which translates into acoustic sound. The motion of the voice coil dissipates power in heat and physical motion. So, to ensure the delicate voice coil in a tweeter doesn't over heat and fail (usually by melting the coil wire or the enamel coating - causing a short), the heat needs to be wicked away. Some tweeters have a oily fluid with iron suspended in it filling the voice coil gap which can help wick away the heat, but it has draw backs as well. Others rely heavily on simple airflow to pull the heat away. Ultimately, the power handling is affected by the phsyical motion of the voice as it needs to be cooled.

With a regular, rounded wave form, the coil is always in motion. This is a good thing. However, when the driving signal is a square wave, the coil will jump outward at the absolute maximum velocity it can accomplish and stop with the voltage holding it in place. Even the the coil isn't moving at all, there is still current flowing through it to hold it out of its resting place until the voltage changes. Then the voltage will swing the other way and the coil will jump back at the greatest velocity it can manage and sit still until the voltage swings back again. This is a problem on two levels. One is that the voltage is not being converted into physical motion. The other is that the methods for draining the heat away from the voice coil are no long functional as the coil is not moving around and pushing air around or mixing the ferro-fluid around. The heat is staying in the coil. As such, a square wave is more likely to destroy a voice coil than any analog wave with soft edges and contant motion.

I just described the reason most standard dynamic speakers fail due driving too much power into them. However, a clipped signal and sqaure waves are more dangerous to tweeters because by their very nature, sqaure waves are comprised mostly of high frequency information. When the crossovers filter out the bass from the signal and pass just the treble to the tweeter, that will include nearly all the energy that makes a square wave square. And a clipped amp is basically putting out various levels of square waves. Also, the sharper the edge is between the round analog wave and the flat top that makes it a square wave, the more intense the high frequency power. This is why "soft-clipping" circuits are used in a large percentage of higher quality amps. That simple circuit can reduce the amount of high frequency energy in the clipped signal significantly and protect your tweeters in situations where the amp is clipping constantly.

 

[Flint]

Here is a spectral analysis of a 100Hz sqaure wave:

Notice the intensity of the harmonics above 2kHz that will be sent to the tweeters along with all the other information in the musical signal.

 

[Flint]

Some amplifiers have clipping indicators on them that are pretty good. However, most of the time the only way to really know is when you can hear it. After all, the dynamic range of the CD is so great, a good, uncompressed recording will likely have lots of peaks that get clipped by the amp most of the time anyway. A short clip on a single dynamic peak is nothing to worry about - after all, Soundhound made it quite clear that in recording music, producers and engineers rely on clipping as a method for "fattening" the sound.

The thing to look out for is when clipping is occuring for durations of several seconds.

 

[Flint]

Here''s another example of what I am trying to convey about clipping adding lots of energy to the high frequencies.

I created a test tone signal with four sine wavs at 41Hz, 440Hz, 2kHz and 12kHz. I then measured the FFT Spectrum and got this chart:

 

[Flint]

I then did the same clipping levels as the MP3 files at 3dB, 6dB and 10dB and measured the FFT Spectrum Graph:

Notice how much more high frequency energy there is at a mere 3dB of clipping, which most of you could not hear is the sample file.

 

[Flint]

With 6dB of clipping, the high frequency noise is increased be about 12dB on average.

 

[Flint]

With 6dB of clipping, the high frequency noise is increased be about 12dB on average.

So, as you can see, clipping will greatly increase the amount of energy the tweeter has to dissipate.