Discrete BJT, Low Power Audio Power Amp Experiments
Introduction
Presented are some experiments undertaken to learn more about low power audio amplifiers. All circuits were work bench developed using mostly popcorn-grade transistors. The overarching design goal was to get a pure sine wave output using a differential pair as the voltage amp and having negative feedback from the output to the input base terminal of one the differential amplifier transistors. I failed to achieve this goal.
There is a paucity of instructional material concerning single supply, low power audio amplifiers. Modern conventions include using relatively high DC voltages, spit power supplies and sophisticated differential voltage amplifiers with current sources and active loads (using current mirror type circuits).These are constructs of fabulous audio power amplifiers, however tend to be way over the head of the average lay-person hobbyist. The typical radio experimenter is likely going to build a receiver using a 12-15 volt single power supply. More often than not, the builder will spend a great deal of time and effort on the RF circuits but just use an IC audio power amp such as an LM386. I also did this this for many years, but evolved to using the better audio amps depicted in EMRFD. The study of audio power amplifier design began in October 2008. I knew next to nothing about differential amps, level shifters or how to properly bias power follower stages. To learn about these circuits, some experiments were performed and a few of them have been published on this web site.
This web page shows evolving experiments with the hopeful outcome of designing a decent audio amplifier for an upcoming HAM radio receiver. People with advanced knowledge or a degree in electronic engineering may find these experiments too basic and non-quantitative however, I believe that the more qualitative aspects of an experimenter's journey can also be useful. The whole purpose of this web site is to share ideas; warts and all!
These experiments began with 2 weeks of frustrating attempts to build an "op-amp". Please refer to Figure 1. This op-amp had a differential transistor at the input. The first transistor used a standard voltage divider bias and the second transistor received its bias from the DC feedback loop. This is common practice in IC op-amps.
Shown to the left is the basic problem encountered with
my early audio power amplifiers such as Figure 1. The distortion seen is the result of the bias offset; that is, each transistor
of the input differential pair had a different DC voltage on its base. Radically
different bias voltages are generally undesirable and pretty much wreck any chance of delivering a
pure sine wave to the speaker. Figure 1 is not a good design and should only be
replicated for study purposes. All sorts of methods were used to try to remedy the
problem, but none of them worked well.
It was hypothesized that starting with
more basic designs might be fruitful. The
op-amp experiments were put on hold and the careful study of
differential amplifiers was undertaken. I sent one of my designs to Wes, W7ZOI via email. He
replied via a text file containing some tips that were very helpful.
Click here for the file
(pdf format).
To eliminate bias
offset problems, audio power designs involving differential
amplifiers were avoided for awhile. However, it is not possible to learn about a transistor topology by
avoiding it! It was decided that if the negative feedback from the output to the input transistor stage went to
the voltage amp emitter instead of its
base, a constant bias could be maintained and perhaps some advancement in
knowledge might occur. To test whether applying negative feedback to the emitter
of the first voltage amplifier stage would work, Figure 2 was built and tested. It worked
very well. Output power is low; only 35 mW, but this was okay. When you push the
input voltage (to get distortion) into this particular amp, each half of the AC
waveform distorts identically. I had never seen this in my audio amps before.
Usually one half distorts first and as you keep increasing the input
signal, at some point the other half also distorts. The feedback loop has a
RC high pass network to improve stability and to provide audio low pass filtering. This circuit gave me a confidence boost which
the struggles with circuits like Figure 1 had eroded.
Later on, I learned that
bringing DC feedback to the emitter of a voltage amp in a power amplifier was or
is a
common technique. Electronics can be so humbling!
In the above Figure 3, the series diodes have been replaced by a active diode or DC level shifter transistor. This was done to facilitate experiments with different power amp topologies in the basic design as it evolved. The bias can be now easily changed by adjusting the level shifter voltage divider.
Here in Figure 4 is an advanced, higher power version of Figure 3. Darlington emitter followers increased the power gain, but even more output power was desired.
Please refer to Figure 5. In order to build differential amplifiers, transistor matching for beta (Hfe) is necessary. Figure 5a shows the simple transistor matcher used. 8 transistors were measured and 2 were selected as depicted because they had the same DC collector voltage . Any reasonable resistor values can be used in Figure 5a. A simple differential-style amplifier was built to study the amplifier's DC voltages; this is shown in Figure 5b. This particular voltage divider bias circuit forces the 2 transistors bias values to be equal.
Shown above is the copper clad board used for the power amplifier experiments which used a "differential-style" voltage amplifier at the input.
A single-ended Input and output voltage amplifier was used. Feedback is to the emitter of this voltage amplifier. This is clearly not a differential amplifier. A simple emitter follower was used to drive the output transistors. The shunt 470 pF capacitor was necessary to prevent the first voltage amplifier from oscillating. The output transistors were a compound or Sziklai pair. Quiescent current draw for this amplifier was fairly high as shown in Figure 6.
Please refer to Figure 7. The output compound pair were augmented with an additional parallel transistor as shown. This worked and increased the output power to 160 mW. All of the amplifiers on this page were tested with an audio source and 8 ohm speaker load. This amplifier sounded great, but was quite a current hog. I believe they were biased more towards Class A operation. The quiescent current had to be high for both Figure 6 and 7 (as shown) to eliminate crossover distortion in the finals.
Shown above is a photograph of Figure 7 with an 8 ohm resistive load.
In Figure 8, the power followers are a now in a standard Darlington pair configuration.
Shown above in Figure 9 is another power follower experiment. An additional transistor is placed in parallel with each final transistor. The result is a boost in output power to a very usable level for popcorn radio receivers. This circuit does have its drawbacks, namely the negative feedback is going through each of the 2 voltage amplifiers.
Please refer to Figure 10. The negative feedback is moved to the base terminal of one of the voltage amps. Considerable time was spent adjusting resistor values to try to obtain a non-distorted. sine wave output. Later it was realized that this power amplifier did not have any negative feedback from output to input. This is because the 82K resistor is AC grounded via the 1 uF capacitor. It was difficult to get a pure sine wave, although I came very close. Note the bias and emitter voltage differences.
Shown above is a photograph of the Figure 10 output. There is some distortion on the top and bottom peaks of the sine wave. Still, this amplifier sounded very good through a speaker.
Please refer to Figure 11. This amplifier was an attempt at chasing a perfect sine wave output. We all chase our windmills; this seems to be mine. Q1-Q3 are Fairchild MPSA18, low noise BJTs. although their beta is lower than the 2N3904, they are very nice for audio use. Two were randomly grabbed from my parts bin and their collector voltages matched perfectly in the Figure 5 transistor matcher. This alone is a reason to use these transistors in differential-style amplifiers. Metal film type resistors were chosen for the collector and emitter degeneration resistors. This was an attempt to lower noise and better match the 2 transistors. Current mirror transistors were wired to provide constant current sources for the voltage amp plus the DC level shifter/power follower circuit. Constant current sources are so cool; although the AC signal is swinging, the DC current remains rock steady. You can optionally replace Current source 1 with a 665 ohm 1% (somewhat better choice) or just a 680 ohm 5% resistor and Current source 2 with a 1K5 ohm resistor wired per Figure 10. If a standard 13.8 volt power supply is used, the average clean power output will increase to 175 mW or so.
Further attempts to design and build a circuit that used a true
differential voltage amplifier were attempted over 3 nights. I was able to improve the voltage gain by 30-50%, however, all of my designs resulted in a distorted output. It seems that my design goal had proven elusive
once again. The best discrete
BJT amplifier I could design (Figure 11) does not have a differential voltage amplifier,
nor has negative feedback from output to input!
I am uncertain whether this is a bad thing, but the failure to reach my goal
was a little disheartening.
Shown above is a photograph of the output of Figure 11. This is an awesome sine wave. This amp sounds just lovely and quiet.
Epilogue - March 14, 2009
I hesitated to present these experiments as they represent the failure to obtain my goal of building a differential amp with output to input feedback via 1 of the transistor base terminals. This was an exercise in humility; a testimony to how little I know about designing audio amplifiers. As a lay-person it is great not to be constrained by rules and conventions, however, the skill and experience of more knowledgeable builders is so highly desirable. Upon reflection, some good things were yielded. These include a transistor matcher, new ways to increase power gain were found, a reasonable, low power, non-feedback amplifier was built and some valuable experience with audio circuits was gained. It is hoped that one day, my goal will be achieved. Feedback and help is greatly wanted. Thank you.
