In a previous article, we looked at the rigorous design of a two-stage transistor amplifier. Here, we are going to put together the building blocks of an IC-based audio amplifier. This is simpler in some ways, but harder in others as the very high gain available can lead to nasty feedback effects. We could finish up with a high-powered audio oscillator instead of the amplifier.

Choice of Amplifier Device

There are many to choose from, and we will look at a few of them. But first, we need to understand some requirements or specifications to make a wise choice. These requirements apply equally to small op-amp amplifiers and the next expensive Hi-Fi you may want to buy.

  • Power output
  • Frequency response
  • Input impedance
  • Output impedance
  • Noise performance
  • Distortion
  • Power supply requirements

The main question in choosing an amplifier is how powerful you need it to be? Will it be a headphone amplifier, a PA system, or a guitar amplifier?

Here are three common and widely used amplifiers with different power ranges:

SpecificationLM386LM1875LM3886
Supply voltage12V60V80V
Output power0.5W25W60W
THD0.2%<0.6%0.03%
Noise i/p VNot spec3uV2uV
Gain bandwidth prod300kHz5.5MHz8MHz

Characteristics of a Good Audio Amplifier

Power Output

Often, when designing an amplifier like this, you might (if you are like me) be constrained by time, availability of parts, and cost. The most expensive part of this amplifier by far is the power supply mains transformer, and scratching through your junk box might be your starting point in deciding which amplifier chip to use. Too high a PSU voltage, and you can’t use the LM1875; too small a VA rating, and you won’t be able to make use of the LM3886’s power. Whether the transformer has suitable balanced voltage taps will also govern if it will be suitable for a split power rail or not. A split power rail will enable you to dispense with a large output coupling capacitor and even run the amplifier as a DC amplifier if needed.

So the first design criterion is power output. Don’t get carried away by silly comparisons. A 20W amplifier is not very much louder than a 10W amplifier; only 3dB, and that’s about the smallest change your ear will detect. A 10W amplifier is not 10 times louder than a 1W amplifier; only twice as loud. The more important aspect is the dynamic range.

Let’s say you had a CD player with a dynamic range of 60dB. This means that if the smallest soft sound you wanted to reproduce were at a level of 1mW, the amplifier would need to reproduce the loudest one at a thousand Watts!

Input and Output Impedances

The output impedance is not actually that much of a concern. All of these amplifiers are designed to load into 8Ω and often much less, provided you don’t exceed the total device dissipation. If you really want to know what it is, apply a sine wave to the input at about ¼ power level and load the output with suitable size resistors until you find the value that drops the loaded value to half the unloaded value—that resistor is equal to the output impedance.

The input impedance is more important. If it is too high, you may run into noise or radio interference, and if it is too low, you will load the source. Most amplifiers are best designed to match a Hi-Fi standard of 47kΩ, but this is not the best if you design for the lowest noise. Professional mixers and PA systems are designed for low impedance (600Ω) microphones, giving them better noise performance.

Frequency Response and Power Bandwidth

An audio amplifier must reproduce signals from 20Hz to 20kHz (although very few people except children can actually linearly hear this range). So frequency response is usually taken to mean the amplifier must deliver its rated power output across this range with no more than a 3dB drop in level.

On the other hand, power bandwidth tells us what frequency response the amplifier can achieve given a desired gain or amplification. The LM386, for example, with a gain-bandwidth product of 300kHz, can only give a gain of 30 at 10kHz or 10 at 30kHz. The other two amplifiers are vastly superior, and the LM3886 with a value of 8MHz should give a gain of 8000/20 at 20kHz or 400 times.

But why would you want to do this? If you applied huge amounts of negative feedback, you would reduce the THD to ridiculously low levels—0.03% as in the data. The fact is, no one can hear this, not even 1%, and what you will achieve is abysmal TID (Transient Intermodulation Distortion) and phase shift distortion, something amplifier developers and Hi-Fi sales brochures are very careful to ignore.

Distortion

THD or Total Harmonic Distortion is not such a bad thing, and anything less than 0.5% is fine. In fact, your own ear is producing more than 0.5% THD inside your head anyway. Clipping distortion is a bad thing when the amplifier clips (the signal has reached the supply rail levels), the signal is going square and is thus full of harmonics. Most speaker systems are rated at their power level to expect 75% of the energy to be below 250Hz.

This is not the case with clipping; there could now be 75% in the expensive tweeter’s frequency range, and they are now severely overdriven and sure to die soon. Crossover distortion is not something to expect in this kind of amplifier. Here is an article on TID distortion and phase accurate amplifiers to help explain more.

Noise Performance

Any noise in the power supply will appear in the amplifier output, so it goes without saying that the PSU should either be well regulated or have very substantial decoupling capacitors to eliminate ripple. Large decoupling caps will also prevent low frequency oscillating (or ‘motor-boating’) in the amplifier. Amplifiers are also vulnerable to strong RF signals picked up on any input wires and so it is important to tailor the amplifiers frequency response to only what you can hear. The remaining noise is a function of what is happening at the input impedance.

Noise occurs in all electronic circuitry, the main types are:

  • Thermal noise
  • Shot noise
  • Flicker (1/f) noise
  • Burst noise
  • Avalanche noise

The only one we are concerned with, as it is the only one we can control and is the worst offender, is thermal noise. Thermal noise is a linear function of an equivalent resistor representing the amplifier’s input impedance. It increases ten times for each 100 times increase in resistance, so the input impedance should be kept as low as possible. There is an excellent article on noise here.

Some Audio Amplifiers

Sample Project

We are going to design an amplifier based on the LM1875 amplifier.

LM1875 Design

Actually this is fairly simple, you can just copy the circuit off the National or Texas design sheet here

LM1875 from TI datasheet, split PSU version

I changed a few things and used the following design which works very well and very stably.

I added C1 to force an upper-frequency cut-off at about 30kHz. C2 and R6 are added to provide a constant output impedance at frequencies above the range to prevent oscillation (R6 only becomes visible to the output at about 150kHz). C4 and R4 control the lower frequency cut-off.

If a dual supply is used, C7 can be omitted and the output connected straight to the speaker. The gain is set by the ratio of R3/R1, i.e., 22. R1 and C8 set the low frequency cut off when R1=C8, i.e., at about 4Hz if you wanted to make this a bit higher 1/C=1k*2*π*F, i.e., for 20Hz about 8uF. The volume is usually a log pot at the input; the 47k log will be fine. D1 and D2 are just reverse connection protection diodes. The supply voltage should be 25V and must never exceed 30V.

LM1875 Pinout

LM1875 Pinouts

Feedback and Grounding Issues

But before we attempt to make an amplifier with breadboard or Vero-board or a PCB layout, we will look at some important reminders.


Diagram 1
Diagram 1 shows a rather badly drawn simple amplifier schematic. As with all schematics, no attempt to show wiring layout has been made and there is no clue as to the terrible secret lurking behind the simple facade.

Diagram 2: Hidden Impedances
Diagram 2 shows the danger. The red blocks are the ever-present complex impedances in the wiring and PCB tracks. These are both resistive and inductive. The resistive part is constant with rising frequency, but the inductive part is not and soon takes over in magnitude as the frequency rises. Any current drawn through these impedances will cause a voltage to appear across them.

Imagine every time the speaker needs to draw a bite of current; it will cause a volt drop across Z1, effectively in series with the rest of the amplifier. Likewise, as the amplifier demands a current, it will cause a volt drop across both Z2 and Z3. Finally, any attempt by C to charge will do the same through Z4 and Z5 and insert this volt difference to the input of the amplifier, which will now oscillate.

Diagram 3: Proper Grounding
Diagram 3 shows the way to overcome all these problems. Everything has its ground connection brought to one point, and this is usually the point where the main smoothing/reservoir capacitor goes to the ground. Note that we are not talking about noise or mains earthing here; just oscillating and feedback. In many cases, the reservoir capacitor is going to be quite large—at least 1000uF. In the same way, to prevent a volt drop from being inserted into the +ve supply line, the +ve end of the cap should be as close to the amplifier +ve pin as possible.

These are very rewarding amplifiers to build! If you obey the grounding rules, they are guaranteed to work the first time and provide a very clean listening performance. The LM36 makes a great headphone, intercom, or radio amplifier, and the LM1875 and LM3886 make great Hi-Fi or guitar amplifiers. Other uses are bi-amping, surround sound, and PA uses.