This review page is supported in part by the sponsors whose ad banners are displayed below

Intermodulation Distortion
If you like to listen to simple music as played by an unaccompanied sackbutt, harmonic distortion figures might be perfectly relevant to you but with a lot of music you will find that intermodulation distortion becomes the elephant on the dance floor.

Harmonic distortion is closely related to intermodulation distortion in that both result from the same nonlinear distortion of the gain device; but IM distortion reflects what happens when there is more than one tone involved. And that describes most music. With a single tone, lower order harmonics such as 2nd and 3rd are not as discernible and in real life, most instruments (including vocal) contain a fairly rich set of these harmonics anyway. They are considered musical.

But when two tones are passed through a nonlinear device, the amplitude of each of the tones is altered or modulated by the other tone. The result is a series of 'sidebands', i.e. additional tones occurring at the sum and difference of the original frequencies. These additional tones are not generally musically related.

Worse, real music consists of very many tones passing through the nonlinear gain device and each of these interacts with each of the others. The result will be very complex and very unmusical.

Figure 7 shows a distortion waveform resulting from two tones passing through a gain stage with both 2nd and 3rd order nonlinearities having 1% coefficients. The two tones have equal amplitude and they are one octave apart. The signal peaks are about 1.8 volt, and the distortion peaks are about .09 volts or 5%, and the ratio of RMS averaged distortion divided by the RMS signal is about 4%.

This distortion doesn't look so bad but it is obviously higher and more complex than single-tone distortion. Let's see what happens when there are lots of frequencies involved. In Figure 8 we see a waveform consisting of 7 non-harmonically related tones of equal amplitude from 100Hz to 2800Hz. If we run this signal through the same gain stage and subtract the original signal, we get the distortion seen in Fig 8:

Not very pretty, is it? Now the distortion is getting really complex with lots of harmonics, and the peaks are up around .9 volts. That's 11 times the .08 volt figure of the single tone and the ratio of the RMS distortion to the RMS input signal is about 8%.

My point? IM distortion is the elephant on the dance floor.
Much of the time IM distortion simply forms a complex “noise floor” which masks musical detail. At lower levels, it takes the life out of the music and makes it uninteresting, even irritating. It isn't as noticeable with very simple music but it stands out with orchestral material as if the instruments were covered by a veil.

At high distortion levels, the sound simply turns to mud and we turn it down.

Or off.

Negative Feedback
In 1927 Harold Black invented the negative feedback amplifier in which the output signal of an analog gain circuit is compared with the input signal so as to improve the performance. There are many ways to achieve this effect, all involving negatively amplifying the difference between the input and output so as to minimize this difference.

A simple version can be made from a 3 pin device like the parts from Figure 1, in a circuit that looks like this:

Here we see a single part (Mosfet) in a network of four resistors forming an inverting amplifier with some gain. R3 and R4 set up the 'open loop' gain (the gain you get without negative feedback) and R1 and R2 set up a negative feedback loop.

If we remove R2 so that there is no feedback, and assume that the Mosfet is a high-gain part, then we would see an 'open loop' gain of approximately the ratio R3 / R4. It is not difficult to make R3 / R4 a fairly high figure, giving an open loop gain of maybe 10 times (20dB) or even 100 times (40 dB) larger than the input voltage. A typical example of 100 times would be R3 at a value of 1Kohm and R4 at 10 ohms.

If you put R2 back in the circuit, you will find the gain reduced. If the ratio of R2 / R1 is much less than the open loop gain and if the value of R2 is much higher than R3, then the gain of the whole stage becomes quite close to R2 / R1. Examples for this would be R2 at 100Kohm and R1 at 31.6Kohm, for an output gain of approximately 3.16 times (10dB).

Simply put, the difference between the two gain figures is considered the amount of feedback. If the open loop gain is 40dB and the gain with feedback is 10dB, then the amount of feedback is 30dB. In actual audio circuits, the amount of feedback can range between 0dB (none at all) to as much as 100dB (100,000 times). Sometimes you don't need that much but a typical integrated circuit op-amp comes with that much open loop so there it is.

Negative feedback is good at reducing all forms of distortion, linear and nonlinear. As a concept, it's pretty straightforward: You create one or more gain stages in series in order to get enough gain to equal the final gain figure you want, plus the amount of feedback you think you want to use.

As the feedback figure exceeds 20dB or so, you find that all the measurements will improve by the amount of additional feedback. If the open loop distortion of the amplifier is 5%, then 60dB of feedback should make it about .005%. It's relatively easy to construct additional stages or to milk existing stages for more open loop gain, so why not 80dB for .0005%?

Sounds like something for nothing, doesn't it?

Not quite. I think it's a bit more like a credit card – convenient if used wisely but carrying interest payments and penalties when it's not.