Low Power Audio Amp Experiments
This web page contains some experiments on simple, low power, speaker output audio amplifiers. Presented are ideas, some measurements and examples of audio amplifiers which will likely sound better than the IC audio chips commonly seen in many receiver projects. This web page is a follow-on to this one and is a completely new area of experimentation for me. Audio amps were built using both split and positive power supplies. In all cases the complimentary power followers were driven by an op-amp. I tried building some power amps using discrete transistor differential amplifier stages with current sources as the driver, but the noise performance and simplicity of the NE5532 or NE5534 op-amp was superior.
Split or Bi-polar Power Supply Audio Amplifiers
In order to build up some split power supply amplifiers, a basic power supply was constructed and the schematic is shown in Figure 1 above. I found it was essential to regulate the voltage or hum would appear on the output. Choose a standard value fuse that is rated somewhere just above the maximum current you measure. I used 2 different AC output power transformers which were in the 18-24 volt, 375 mA to 1 amp range. The LEDS are strongly suggested. They inform you when there is power applied and their relative brightness will also often fall when higher current is being drawn on one side or the other. This alerted me to an accidental solder bridge to ground on more than 1 occasion.
The split power supply is shown in the photograph to the left. The retro Bakelite fuse holder is from an old tube audio amplifier. If you are wondering why the copper clad board is so large, the power supply is part of a future project. Some builders would use even greater value filter capacitors than those shown. Heat sinks on the voltage regulators are required for supplying DC to higher power amplifiers.
Rod Elliot Headphone Amp
The first amp I built is shown in the Figure 2 schematic above. This amplifier was designed by Rod Elliot and is used with his permission. Rod's ESP web site is a virtual treasury of audio design information. If you are into understanding audio design, visiting his web site is strongly recommended. Rod sells printed circuit boards for all of his circuits if you prefer this building method . Note I have made some minor modifications to some part values. although primarily designed as a low distortion, high power, headphone amplifier, it drives an 8 ohm speaker very well. I was able to drive this amplifier as high as 0.68 watts average power with a pure sine wave output during analysis. Power measurement is discussed in the next section. Note that on this web page, I quote the entire stage quiescent current. Since the op-amp and the 2 (or more) power followers are a "package", it is a lot easier to just measure the current at the power supply lead(s) of the stage than unsolder and lift up a transistor lead. In this case, with no input single, the stage current was about 9.5 mA. I did check and about half the current is going to the op-amp with the other half to the transistor pair. This is a wonderful sounding amplifier and Rod has an entire web page devoted to it, so I will not comment further.
There are a variety of suitable transistor pairs for audio power amplifiers depending on the power output you are choosing. I stock just a few; BD139-140, TIP 41C-42C, NTE 128-129. The higher beta 2N3904-3906 or 2N4401-4403 pairs worked well in the low power, single power supply amplifiers shown on this web page. I also performed some higher output power experiments which required the TIP and BD transistors and these are not shown.
Above photograph. A breadboard of the Rod Elliot headphone amplifier. This early version had a temporary output capacitor. When first testing a new circuit that has a direct speaker output, it might be a good idea to temporarily use an output capacitor until you measure your voltages and current and feel your transistor temperatures. This will save your speaker if you made a big mistake and/or blow up the transistors when you first power it up.
Amplified Diode Biased Audio Amp
Above in Figure 3 is a split power supply audio amp using an "amplified diode" to control the bias. The bias transistor was wedged between one of the output transistors and a piece of copper clad board to allow thermal tracking. The 10K bias control resistor was a trimmer type suspended over the copper clad board in most of my bread boards. Usually, you just need to set and forget about this resistor after initial set up. I adjusted the bias by watching in my oscilloscope with a low level, 1 KHz sine wave connected to the input. I measured the various voltages and stage current at quiescent and have indicated these values in red for learning purposes. The bias current range was 7.2 to 154 mA when turning the 10K trimmer pot from 1 extreme to the other. The maximal clean output average power of this amp was 0.78 watts. I used press on heat sinks for the NTE128-129 pair and they ran quite warm to touch. These TO39 type packaged transistors are somewhat difficult to heat sink compared to the TIP/BD transistor packages where you can just bolt on a heat sink of any size that is required. Please remember that the metal tab on the TIP and BD transistors is connected to the collector terminal.
In the above photograph is my first bread board of the Figure 3 amplifier. This particular version had TIP transistors, a 4 ohm speaker and an 470 uF output capacitor. Note the full size 10K bias control potentiometer on the left hand side. This was purely a experimenter's bread board, but it sounded amazing when listening to music through it.
Harmonic Distortion and Measuring Output Power
In Figure 4 is the formula used to calculate the average power of the
circuits on this web page. For example if you measure 6 volts peak to peak on
the oscilloscope, (3 volts peak voltage) and
your resistive load is 8 ohms, the average power is 560 milliwatts. At any point in
AC waveform there is power and it may be reported using a variety of ways.
Was it clean? distorted? a peak value? an RMS value? - often it is unclear.
To be clear, I measured the peak voltage on a pure, undistorted sine wave into an 8 ohm resistor. Stated power values are the mean sine wave power calculated with the formula shown. See this somewhat controversial link for details. You may not agree with my methodology, however, it allows you to compare the circuits on this web page. If you really must know the peak power, multiply the stated average sine wave power by 2. I will leave the power measurement and calculation debate up to scholars; as a lay-person, I need something simple.
The bench voltage measurement was as follows: The amplifier was connected to a 1 KHz pure sine wave generator and the 10K volume control pot was advanced just until any sign of distortion of the amplifier output sine wave appeared. Voltage measurement was taken at the point just before distortion occurred.
It is difficult to photograph a sine wave without a tripod. Motion, the angle, light reflection and
jpeg graphic compression all
wreck the perfect sine wave. In the Figure 4 graphic above is a typical
1 KHz output
waveform from my power amplifiers (squeaky clean) at the amplifiers maximum
average power level.
To the right is a photograph of my AF signal generator. This is an old, tube device but the output sine wave is beautiful. I did not perform spectrum analysis with a computer audio sound card program and will leave this up to audiophiles. These audio amps sound great; especially when compared to the IC audio power amps that many of us tend to use in our receivers.
In the above photograph is A, an 8 x 1 ohm resistor load and B, an 8 ohm load made from parallel 1/2 watt 10 and 39 ohm resistors. In dummy load A, I used 5 two watt metal film resistors plus 3 half watt resistors. In the future, I will obtain 3 more 2 watt resistors and replace the 1/2 watt resistors for a 16 watt rating. For a quick resistive load, B is the way to go for most of the circuits on this web page.. You can make a 4 ohm load from parallel 4.7 and 27 ohm resistors. In truth, a single resistor or any combination of resistors adding up to the desired load R value will work.
In Figure 5 above are some scope waveforms ranging from mildly distorted to full-on dirty.
Power Amplifier Concerns
Although the amplifiers on this page are 0.15-0.8 watts or so, they can consume relatively large current compared to the usual voltage amplifier circuits we build. Some potentially helpful tips to help keep away ground loops, oscillations and thermal run away are suggested as follows:
- Connect your negative speaker terminal directly to the AF power amp (do not use a common ground for the negative speaker terminal).
- Use big power supply line bypass capacitors (no 10 uF caps here)
- Keep your audio amplifier copper clad board separate from the rest of your circuit boards and star ground it to your main power supply ground point.
- Use heat sinks on your final transistors and voltage regulator(s) when you go for bigger power
- Watch your layout - keep the output away from the input etc.
- Watch your emitter resistor power ratings in "higher wattage" amplifiers. Burning resistors stink.
Single Power Supply Audio Power Amps
Since most 12 volt power supplies are actually closer to 14 volts; these experiments were performed with a typical radio bench DC power supply at 13.69 volts.. Figure 6, 7a and 7b represent evolving experiments aimed at obtaining greater output power.
Shown above in Figure 6 is the fundamental design using one op-amp and 2
power followers. It is shown in an AC output power measurement configuration.
The bias current range was about 5 to 100 mA when turning the 10K trimmer bias
control pot from one extreme to the other. Maximal sine wave average output
power was only 141 mW. Nevertheless, it might be loud enough for some receiver applications.
I connected this amplifier to a VCC of 15 volts. The maximal sine wave average output power was then 220 mW. In all of the single supply audio amps presented , increasing the VCC increased the maximal power output. Driving these amplifiers beyond a pure sine wave output power resulted in predictable harmonic distortion plus the re-emergence of crossover distortion in the output. This was an incredible learning; how could there be crossover distortion re-emerging in a amp that was properly biased to begin with? Increasing the bias current to the maximum level did not remove this crossover distortion. After emailing this question to Rick, KK7B and Wes, W7ZOI, and reading their replies, my best guess was that at some power level, the 5532 op-amp can not provide enough current to properly drive the complementary symmetry pair. The AC current in the output transistors may be limited by the base drive of the op-amp and they were no longer forward biased at the crossover point.
Above photograph. This is the Figure 6 amp driven past the point where the sine wave is pure. Note the crossover distortion blips on the sine wave. The base drive current for the power follower pair all comes from the op-amp. At this point there is likely not enough base drive to keep the base emitter junctions forward biased.
In the above photograph I blacked out the room and photographed the same
scope waveform as above while shaking the camera from side to side. This adds some
horizontal spreading of the signal and provided more information about what was
happening as compared to a single, clear oscilloscope trace.
I should mention that this crossover distortion blip occurred in all of the Figure 2 to 6 amps when they were driven past the point where a pure sine wave was seen. It is clear that maximal available power from a simple audio amplifier like this (one NE5532 op-amp plus 2 power followers) is constrained and thus its application is limited. Greater output power is possible using a split supply per Figures 2 and 3, however, a typical radio project has a 12 volt, single power supply. These basic amplifiers with a single power supply, may be very appropriate for projects such as a compact radio receivers or a code practice oscillator project, but not for applications where you require louder audio.
Shown above in Figure 7a is an easy method to get more output power from the
Figure 6 amplifier; add another set of complimentary pair current amplifiers.
I found 33-39 ohms to be a good emitter resistor value during my experiments.
Many hi-fi amp builders will use greater emitter resistor values, however, a
design goal was to get more output power from our 12 volt supply. Series emitter
resistors are used to improve linearity and operating-point stability. I kept
the final power follower pair emitter resistor values at 1 ohm to get maximal
output power. An output 10 ohm + 0.1 uF low pass filter was used to help prevent
oscillations in view of the low emitter resistor values on the finals.
The top 10K bias resistor was lowered to 6K8 to facilitate "more linear" setting of the output transistor bias with the 10K trimmer potentiometer. It did not help much. Setting the bias is very delicate procedure and you must turn the screw driver very slowly. In my bread board, the optimal stage bias current was 22.3 mA but anything around 20 mA should be fine. If you do not have an oscilloscope, after ensuring that there is no input signal, connect an ammeter in series with the positive power supply lead. Turn the bias potentiometer with with the screwdriver until you get close to 22 mA. If you only have a voltmeter, the rule of thumb of 1.1 seems to work... Measure the voltage across the final 2N3904-2N3906 bases and ensure the difference is at least 1.1 volts while adjusting the 10K trimmer pot. Personally I do my biasing with an oscilloscope at at least 2 different frequencies on the signal generator, however the for mentioned methods will work okay. This is a popcorn stage and a popcorn web site after all!
For the lowest potential noise, consider using metal film type resistors in your audio amps and "polysomething" capacitors wherever AC signal is coupled to another component or ground, excluding the output capacitor.
Further experiments to increase output power were frustrating. Finally a compound or Sziklai pair was trialed and increased the average power to over 400 mW as shown in Figure 7b . I used a small piece of copper clad board on the finals for a heat sink, although they really didn't get that warm. Ideally, the amplified diode should also be glued onto one of the heat sink boards for thermal tracking. This amplifier is now in a chassis as a bench reference audio amp for receiver testing.
In the above photograph is the Figure 7b prototype. I am using a new miniature potentiometer for my experiments that I bought from Digi-Key. The base has 2 leads which can be soldered right on the copper clad board for easy anchoring and removal after testing.
Another view of the bread board on which the Figure 6, 7a and 7b experiments were conducted.
KK7B Headroom Boosting Emitter Capacitors
We first learned about using large value emitter caps in audio amp complimentary
EMRFD. Experimentation revealed
capacitors do 2 things: 1. Can increase
the amplifier sine wave headroom and 2. Add some low pass filtering. I learned
from Rick, KK7B, that he designed his
EMRFD amp to
achieve low output power, low distortion and lower DC current drain. He desired
a clean output audio amplifier for his R2 series of receivers without needing a
lot of quiescent current or heat sinks on the 2N3904-2N3906 pair. The caps were
added to make the amplifier think it had much lower emitter resistors at AC than
the 22 ohm resistors he used in the EMRFD projects. When Rick made measurements and
simulations of the amplifier, it was very stable, had low distortion and
provided a very nice clean sound at all signal levels, from very weak signals in
a few milliwatts of noise, to music driving the speaker. The result is
outstanding and Rick's design was the catalyst for my own interest in audio
I performed experiments with these capacitors and found that they increased my amplifier power and head room in some cases, and that the boost is indirectly proportional to the emitter resistor value. With 1 ohm output transistor emitter resistors the boost is generally not that significant. With 4.7 ohm or greater emitter resistors, they can make a big difference and you might consider trialing them for more power and headroom as appropriate. They can also add a nice, warm sound to your audio amp. Refer to EMRFD for numerous examples of this technique.
Popcorn Audio Amplifier
What follows is a popcorn or "poor man's" audio power amp using the
2N3904-2N3906 pair. To meet true popcorn criteria, all of the capacitors used in my breadboard were
electrolytic and you can substitute different values from your own junk box.
It would be better to use "polysomething" capacitors for the NE5532 pin 5 and
6 signal capacitors if you have them. I normally use a 1 uF to 4.7 uF poly-type capacitor
in series with the 4K7 resistor on
pin 6. The 270 pF feedback capacitor could be
omitted or substituted with a higher or lower value to suit whatever high
frequency roll off you desire.
The transistor glue-on heat sinks seen in the 1 bread board photograph are completely unnecessary. This BJT pair were used in other higher power experiments as well. The 22 uF capacitor between the transistor bases is essential from my experimentation. Without this capacitor, the amplifier headroom decreases and crossover distortion occurs. You can use the other half of the 5532 for a preamplifier or use a NE5534 instead. A 741 op-amp would be a horrible substitution. The NE5532 performance is breathtaking considering its low cost. 181 mW is surprisingly loud. All resistors are quarter watt rated. What a fun little amplifier! The schematic is Figure 8.
The bread board of the Figure 8 popcorn amplifier using transistors without heat sinks.
I tried putting a current source on the Figure 8 bias and it made no difference to the amplifier characteristics according to my simple oscilloscope, listening and DC analysis.
One of the full wave rectifier, voltage regulator and filter bread boards used in these experiments. I went as high as 24 VCC on some single supply amps I tested and was getting over 5 watts average output power