Audio Amplifier June 2021

The making of a high quality audio amplifier was my long-term wish. It also gave me a motivation during study at university and encouraged me to think deeper about the circuits we had studied. When I got enough knowledge about circuits and simulation, I started making my own, beginning with a single transistor in voltage follower (Class A) topology then studying other amplifier circuits and in the end constructing my own topology from well designed amplifiers.

My enthusiasm for audio encouraged me to write my master thesis and even scientific paper about a low distortion (-100dB THD) audio amplifier. And even though I have made a lot of research, I do not consider my audio journey has come to an end and in the future when I gather more knowledge I will be making even better amplifier.

The beginnings

My journey started by making a simple voltage follower topology as in Figure 1. I thoroughly analyzed the circuit and its shortcoming such as Class A operation which resulted in huge power consumption. The advantage of its simplicity is guaranteed stability since only transistor transfer characteristic is involved in signal amplification. In the process I gathered knowledge of all the basic transistor limitations: exponential I_emitter(U_base) characteristic, nonlinear small signal amplification, U_CE_saturation... which all contribute to non ideal current amplification.

Figure 1: Simple amplifier realisation (LEFT) and schematic (RIGHT)

Then I started researching other topologies on the internet. Most of them were made from a huge number of transistors, made for huge powers (over 100W) and did not consider the transistor advancement in gain capability. Therefore designs did not offer very low distortion and consumed a lot of power.

But I have been able to find a simple topology from Linsley Hood which I also analyzed and executed shown in Figure 2. The topology consists of 4 transistors: PNP with feedback for input, NPN for voltage translation for output stage and dual NPN output stage. The idea from Linsley has very low THD (under -60dB) and can be configured into Class AB with setting output stage bias with R_bias4 without compromising the distortion.

Figure 2: Linsley Hood amplifier realisation (LEFT) and schematic (RIGHT)

Proposed design

My design goal was to construct a 10W amplifier with -100dB total harmonic distortion with simple, efficient and compact topology. The design also needed to be constructed without AC coupling capacitors, because they take space, distort signals and are adding additional unnecessary cost. And lastly, all the components to produce and amplify signal from USB audio protocol need to fit into an elegant enclosure with VU meters. A block diagram of system is shown in Figure 3.

Figure 3: Block diagram of proposed design

In Figure 4 a schematic of proposed amplifier is shown by combining a dual NPN output stage with differential input pair design from general purpose operational amplifier LM2904. I was able to simulate distortion levels around -100dB which means the topology meets the specifications. After many simulations the design was redrawn in Altium designer and a whole circuit was drawn. I used my knowledge gathered from Laboratory Power Supply project to make custom voltage/current regulators for all power rails. Standard voltage regulators such as 78L05 were used for simple signal rails, because they required fixed voltage. A logarithmic converter was also made by following a design note from TI LOG101. For DA conversion PCM1794 with less than -100dB distortion alongside TI OPA1678 low distortion opamp for signal conditioning was used. After schematic was finished, the circuit was drawn with considering the case so minimal afterwork is needed when mounting the circuit in and out the case. One quality of life improvement is the usage of TO-220F power transistors that have plastic casing and do not need additional isolation to prevent short circuiting the TAB connection. Also layout design consisted mostly of proper trace sizing, symmetry and correct signal connections to prevent signal offsets. The only digital part of the circuit is digital side of DA converter where continuous GND plane was poured and input signals were isolated from USB-to-I2S card. The circuit with different parts is shown in Figure 5.

Figure 4: Schematic of proposed amplifier

Figure 5: Circuit layout of proposed amplifier with different parts

Testing

After promising simulations and circuit design I ordered parts and assembled the PCB. There were some mistakes made in custom power supply regulation but everything else worked without a hitch. Our local instrumentational distributer Amiteh was kind enough and lent me an audio spectrum analyzer Keysight U8903B which helped me characterize the audio amplifier. Audio amplifier was able to amplify signal with up to -100dB THD up to a 1kHz range, when feedback loop was optimized as shown in Figure 6. Another optimization was done by choosing the best quiescent current without load for output stage for lowest distortion which is around 20-40mA as measured in Figure 7. A performance comparison to other amplifiers is shown in Figure 8 where we can see poor performance of Linsley Hood amplifier of only -50dB THD, because of use of low quality bypass capacitors. In-circuit distortion of PCM1794 conversion was also measured and is around -92dB together with preamplification stage as shown in Figure 9. Lastly a cheap DIY method for measuring THD of amplifier was used by generating sine signal with PC via USB-to-I2S card, converting it with PCM1794 ADC and measuring the output of proposed amplifier with Tascam DR-22WL recorder through voltage divider. Comparison with Keysight measurement and setup is shown in Figure 10 and 11. DIY method is of course not as good as Keysight instrument, but for distortion levels over -90dB it is an appropriate.

Figure 6: Distortion at different feedback loop compensations

Figure 7: DIstortion at different quiescent currents of output stage

Figure 8: Distortion comparison of different audio amplifiers

Figure 9: Distortion of PCM1794 ADC and preamplifier

Figure 10: Distortion comparison between Keysight and DIY measurement method

Figure 11: Measurement setup comparison Keysight (LEFT) and DIY method (RIGTH)

Enclosure design and fabrication

As mentioned in the beginning a corresponding enclosure for the circuit was drawn in Fusion 360 which is simple and powerful program free to use for hobby and students. The program also offers fabrication layout generation which I utilized in order to mill the side panels of the amplifier with CNC machine 6040. A lot of work was also taken by making taps in the longitudinal sides, but despite the effort, the final product looks much more clean that way.

Figure 12: Fusion 360 design and manufacturing workflow

Battery protection

The amplifier has been proudly used and enjoyed since the construction. First it was intended to be used with AC transformer and rectified directly on the circuit, but the idea was dropped, because AC noise and 50Hz rectification caused 10dB worse SNR compared to battery powered circuit with 76dB. The amplifier was powered on and off by connecting and disconnecting high power connector. One time I forgot to disconnect the connector and the batteries drained, but in the process positive battery drained first because of higher consumption on positive rail where ADC converter and isolator are connected. A problem occurred on the proposed amplifier side where middle point was tried to be sustained. Because there was no positive voltage, electrolytic capacitors on the amplifier side started to charge with negative voltage. So magic smoke started to slowly spray out of the positive caps during the night. Luckily no other damage or fire happened and I was immediately prompted to construct battery protection circuit which will protect battery as well as amplifier circuit. On of the professors on the university suggested using TL431 in a comparator regime and this lead to the construction of an extremely simple and low power protection circuit shown in Figure 13. The circuit also used a button on the amplifier to power on, off the amplifier so no more fiddling with the connector. The protection circuit also had a latch which prevented oscillation of switches when low voltage occurred and was reset when power button was depressed for longer time.

Figure 13: Simple battery protection circuit with latch

Conclusion

The project was a great success. Design of an amplifier of -100dB distortion is more than initially aimed. In the process a lot of different skills were used which is exactly what I look for in my projects. The downside is only consumed time that is harder to find by the day. Future goal is to improve the SNR to the same value than distortion (100dB) by using better topology such as the one operational amplifier OPA1678 uses and maybe even improving the THD figure. Since I am working on integrated circuits, an integrated version of the ADC and amplifier would be also nice to develop and such that has active biasing regardless the ambient temperature.