TI Training home
Power Supply Design training resources
Power Supply Design Seminar 2018 technical resources
Power solutions for Class-D audio amplifiers Impact of Power Supply on Audio Performance
Power Supply Design training resources
4.2 Impact of Power Supply on Audio Performance
So now let's go on to part two of the presentation, which is the impact of the power supply on audio performance.
The performance of any audio amplifier can be measured by taking into account several parameters. For example, the ones indicated in this slide, like transient distortion, frequency response, output impedance, and power supply rejection ratio, impact directly what is the most important parameter, the total harmonic distortion, or THD.
Here, the transient distortion shows the ability of the output of the amplifier to follow the input signal. The THD, or total harmonic distortion, is defined as the ratio of the contribution of all harmonic components to the value of the fundamental frequency.
A very high quality system shows a total harmonic distortion value of less than 0.025%, as shown here in the second bullet. While 0.1% is recognized as a maximum value for a typical power amplifier nowadays.
In the top right side waveform, we can see how power, current, and voltage look like on the output of the amplifier. Since the load is considered as a resistor for simplicity, load current will be in phase with voltage. So both are sinusoidal waveforms. The instantaneous power will follow sinusoid, like the green curve on the right.
If a power supply has a non-zero impedance, there will be a drop on it with the amplitude proportional to the model of the power supply impedance, and the frequency equal to twice the audio frequency.
A test on a real class D amplifier has been performed by supplying it with a non-zero output impedance PSU in addition to an external 250 hertz signal injected into the power VDD of the amplifier according to the block diagram. You can see here in the bottom left picture.
Here we have a closed loop class D amplifier, and E of S represents a 250 hertz signal. The result is shown in the graph on the bottom right corner. As we can see here, in addition to the obvious one kilohertz audio signal, we also have modulation due to the 250 hertz added and subtracted with twice the power supply ripple, which happens in 2 kilohertz and later frequencies. The other spikes are also explained on the graph.
The reason we tested the class D amplifier noting the two times power supply unit ripple and an external noise source was to understand how the output impedance of the power supply unit affects the whole system's audio quality.
For this reason, we tested a TPA3251 EVM supplied at 36 volts, and we compared the performance of this class D audio amplifier by supplying it with a high quality bench supply.
A comparison has been performed by supplying the EVM directly connected to the power supply unit, and then a second time with the 2 ohm resistor as shown here in the top right picture. The 2 ohm resistor simulates the equivalent of an output impedance of a power supply that we want to develop later.
The purpose of this test was to quantify a closed loop class D amplifier's ability to reject VDD power supply ripple. These bullets list the test conditions.
The audio evaluation module, or EVM, is equipped with high values of electrolytic capacitance and power of VDD. It didn't make sense to use these large capacitors and test the system with very, very low impedance, therefore, we removed those capacitors and added only two 120 microfarad capacitors to avoid excessive ripple voltage from the switching frequency.
According to the table on the bottom left, we can see that the power supply unit's ripple with a 0 ohm resistor is very small. The ripple measures below 200 millivolts.
The total harmonic distortion's well below the EVM specs. By adding a 2 ohm resistor, the power supply unit ripple increases a good deal, and at 400 hertz, we can see here in the table, that it's 1890 millivolts, which is about 2 volts, and is a very large signal.
Even with this high ripple, thanks to the fact that the EVM works in a closed loop, the worst case total harmonic distortion at 5 kilohertz stays below 0.016%, which is an excellent value. The data's reported on the graph in the bottom right corner. The curve colors in the bottom right are associated and matched with the colors on the left table for easier analysis.
Another cause of audio quality degradation comes from clipping. So, for example, when the duty cycle of a class D amplifier approaches 0% or 100%, it can no longer regulate the audio signal, and strong clipping occurs, as can be seen in the figure on the left where green and blue signals are just the outputs of each bridge, and the red dotted line is the power of VDD.
Clipping an audio signal creates high frequency harmonics, as shown in the graph on the right. And since the clip waveforms are close to a square wave, the amplitude of each harmonic keeps high peak and high RMS values even at the frequency outside the audio band. Furthermore, at low frequency, the distortion can be dangerous for woofers because of its mechanical and thermal limits and due to the limited swing amplitude of a coil.
At light music volume, it's possible to reduce the VDD of a class D amplifier and reduce switching losses as already discussed in the previous slides. This is called Class G mode. It's important to recover from Class G to full VDD voltages soon as a clipping situation occurs.
So let's talk about the output impedance of a power supply. In these two graphs we have an example of a DC to DC stage of an offline power supply. On the left graph, we have three curves. The yellow one shows the open loop gain, the green one shows the output impedance and open loop conditions, and the blue line shows the output impedance and closed loop conditions.
According to the formula on the top right corner, the open loop impedance is divided by 1 plus the open loop gain. This open loop impedance is greatly decreased thanks to the fact that we have a high gain at the denominator.
This is valid for voltage mode modulation, as in this case as well as in current mode modulation. In the graph on the right, we actually made a comparison between the voltage mode, which is the yellow line and current mode output impedance, which is the blue line.
At the crossover frequency, the closed loop and the open loop impedances match, as we can see here on the left graph at the red line. Since we verified that even with 2 ohm series impedance, the total harmonic distortion of an audio amplifier still shows very good performance, we can accept that our power supply can have 2 ohms of output impedance at the crossover frequency FCO.
At frequency lower than 1 kilohertz, the output impedance decreases greatly, both in voltage, and current mode. Therefore, we can expect an even better performance regarding total harmonic distortion in this frequency range. We'll see later that according to the minimum output capacitance requirement to satisfy 2 ohm impedance a very small output capacitor can be selected.
It must be noted that this small output capacitance has to comply with the high demand and transient response for a DC to DC converter. For this reason, we have to calculate the overshoot and undershoot of a peak current mode isolated converter.
The related formulas are sorted in the table here. The limits are defined by the absolute maximum value of the power supply voltage, PVDD, of the audio amplifier and our main PVDD absolute max, which is 38 volts, along with the limit from PVDD clipping.
In practice, it's mandatory that during heavy load transients for power supply, its output voltage never goes beyond those two values to avoid failure clipping. Our ESR, equivalent series resistance, represents the series resistance of the output capacitor C out, while ESL, or the equivalent series inductance, is neglected because of its small contribution to the transient response performance.
The crossover frequency has been selected at 5 kilohertz because this is a typical bandwidth of an isolated DC to DC converter employing an optocoupler to close the loop. The optocoupler introduces a pull in the transfer function in the range of 10 kilohertz to 40 kilohertz. Since this frequency range is not always perfectly known, whether from production or due to aging, it's better to close the feedback loop a little bit earlier, and we typically choose a frequency between 3 and 7 kilohertz, such as 5 kilohertz.
February 8, 2018
This training series describes how to properly design a power supply for a high-power Class-D amplifier based on the output impedance requirements as well as on typical requirements like average and peak power demands. In this video, we discuss the details of audio performance and how it is affected by the power supply.
This course is also a part of the following series