Hello. My name is Gregg Scott, and I'm a senior applications engineer. And today, we're going to discuss measuring Class-D performance of Class-D amplifiers. Today, we're going to discuss several tests. And in each of those tests, we're going to discuss the following, test definitions, test setups, and the expected results, and also some pitfalls and traps that you may fall into on these tests.

So why is audio testing different? Audio testing uses methods that try to mimic our hearing. Our hearing is qualitative, where electrical performance tests are quantitative. There will always be that opinion part of the equation that cannot be evaluated or defined with data. But for audio amplifiers, tests were created to try to quantify the audio quality of an amplifier. So audio tests are unique to electronics testing.

Most tests use the same test setups. We have a DC power supply to power the amplifier IC. We have a signal generator that generates our signals. We have an LC filter for the Class-D amplifier. We have a speaker load, and then we have signal analyzer. Today, we'll use the Audio Precision, AP 500, as both the signal generator and the signal analyzer.

Shown here, is the Audio Precision user interface. In the far left column is our input and output configurations. The second column is our generator and analyzer control. And then, on the right side, is our measurement window.

The output power is a measure of the amplifier's ability to provide a voltage and current to a load. As we know, power is in watts and is the voltage times current. And audio amplifier is a voltage amplifier with high current capability. The voltage places the speaker cone in the correct position, and the current is what is needed to move the speaker to that position.

Areas that can get you into trouble with output power. Output power must be defined with a total harmonic distortion, THD, and load impedance defined, such as 25 watts into 4 ohms at 10% THD. Power is volts RMS times current RMS. And power is in watts. There is no such thing as P RMS, RMS power. And there is no such thing as peak power.

And maximum power is the output power of an amplifier with a square wave. Basically measures the power supply voltage for the amplifier and any losses in series with the load. Be careful with the test equipment. Make sure you have enough current in the power supply. Current limit supplies will skew your power downward.

As you can see across the x-axis, the amplifier can provide power from 0 watts to greater than 20 watts, as shown on this graph. That is why it is important to specify the THD level for the output power. Typically, the power is specified at 1% or 10% THD. In this graph, the amplifier has 21 watts at 1% and 25 watts at 10%. This was measured at 14.4 volts DC and a 4 ohm load.

We've been talking about total harmonic distortion. And what is harmonic distortion? Well harmonic distortion is the measurement of the non-linearity of the audio signal due to the audio amplifier. And an amplifier will always have some non-linearity. Total harmonic distortion is measured by injecting a very low THD sine wave into the input of the amplifier. This sine wave is measured at the output of the amplifier, and the test equipment will insert a notch filter to remove the sine wave, leaving only the THD and noise that is left to measure. The equation is the summation of the harmonics, plus noise, divided by the fundamental. And that will give you the THD+n number.

Well, we need to understand the measurement of THD+n. When using a THD meter, it provides only a number that is everything that is not the fundamental. This includes noise, signals that are not harmonic. And the harmonic structure is also not known. Shown below is a THD versus frequency curve. This is done at one volt RMS output at 14.4 volts into 4 ohms. As you can see, the THD is quite low, below 1% THD across a frequency band from 20 hertz to 20 kilohertz.

As discussed previously, measuring THD with a meter cannot determine the harmonic structure of the THD. The Fast Fourier Transform, or FFT, can provide the frequency components from the time domain signal. The non-linearity of each harmonics of the signal will be defined and quantified. This provides details of each harmonic, and also provides the details of the noise floor over frequency. It is known in audio that low order harmonics sound more natural than high order harmonics.

Shown here is an FFT plot of THD+n. As you can see, at 1 kilohertz, we have 2-volt RMS signal. And at 3 kilohertz, 5 kilohertz, 7 kilohertz, and 9 kilohertz, and so on, we have a harmonic structure shown.

The frequency response is the measure of magnitude and phase at the output as a function of frequency. A sine wave is swept from 10 hertz to over 20 kilohertz, to measure the amplitude and also to measure the output phase relative to the input signals phase. The frequency response is typically determined by the -3dB points at the lower frequency and the upper frequency. This is also known as the bandwidth of the amplifier. Also, this is commonly known as a Bode plot. The frequency response of a Class-D amplifier is controlled by external components. The low frequency cutoff is due to input coupling caps, and the high frequency cutoff is due to the LC filter.

When measuring frequency response, be aware of these pitfalls. Understand the frequency response limitation of your test setup, such as cables. Test equipment settings and filter settings are usually the culprit. Shown here is a typical frequency response graph of a Class-D amplifier. As you can see, the low frequency -3dB point is 10 hertz, and the high frequency -3dB point is beyond the test limit of the analyzer.

Noise and weighting curves-- noise voltage can be tricky to measure accurately. The main problem when measuring noise is to have a test setup that is quieter than the device under test. The next concern is to only measure noise that is in the frequencies that we care about. So we use weighting curves that are used to filter the noise. In the US, we typically use A-weighting, which uses the Fletcher-Munson curves that represent the equal loudness contour of human hearing. Another way to measure noise is to use the noise FFT. The spectrum shown here is from 20 hertz to 20 kilohertz. And the noise is very low-- minus 100 dBV to minus 120 dBV.

Gain of an amplifier-- this is the voltage gain of the amplifier, and it is output voltage divided by the input voltage. It is often expressed in dB. This is thought to be a simple measurement, so it can be incorrectly performed with poor results. The gain should be measured in the middle of the frequency response where the gain is known to be stable. And the gain structure of the system needs to be known. There can be sneaky resistive dividers in the system that can reduce their gain. And also, the equipment limitations need to be understood. And they can change the gain measurement also.

Measuring DC offset-- an amplifier that drives a speaker should have 0 volts DC across the output terminals. This is easily measured with a DC volt meter. No signal should be present, and a typical load should be present, such as a speaker or load resistor. The amplifier should be allowed to thermally settle for a proper measurement. The voltage will not be stable. DC voltmeters measure average RMS of the DC voltage, and this includes noise and voltage drift.

We also need to understand why this is important. Well, as we stated earlier, the speaker position is due to the voltage. And a DC voltage would move the speaker to an undesirable position and use the current to hold the speaker there. The speaker would also heat up. And if the DC is large enough, the heat could damage the speaker or the voice coil.

Measuring crosstalk-- this is the ability of one channel of an audio amplifier to influence the output of another channel in the system. This is important to know how well the channels are isolated from each other. And crosstalk is measured in dB. And the lower the number, the better. This is measured by placing a sine wave on one channel and measuring the signal on another channel.

This is a typical crosstalk versus frequency curve. Channel 1 is driven, and the gain is shown to be approximately 21 dB and is influencing channel 2. The crosstalk value on channel 2 is from minus 105 dB at 20 hertz to minus 78 dB at 20 kilohertz.

Defining signal to noise ratio-- the definition of a signal to noise ratio is exactly that, the ratio of a signal to the noise voltage. And the noise voltage can be found in the datasheet. And the signal voltage is the variable.

Well, some use 1 volt RMS as a signal level. That's typical of preamplifiers. And some use the maximum unclipped signal level. And that's typical of amplifiers. And then, some use random values to make their spec look good. So signal to noise ratio should always have a signal level or voltage defined. We measure it by using the maximum unclipped signal level for our Class-D amplifiers and A-weighted noise voltage.

Efficiency and power dissipation-- efficiency is the measurement of how well an audio amplifier transfers the power from a power supply to the speaker load. Power dissipation is the amount of power that is lost and must be dissipated by the amplifier. Efficiency is measured by measuring the input power and the output power, and then dividing the output power by the input power. And power dissipation is the difference between the input power and the output power.

The efficiency test setup-- we have a power supply that provides the voltage to power our amplifier IC. This power supply will have a current draw from the amplifier, and we also have to measure the voltage at the power supply pin. We multiply the current and the voltage at the power supply pin together, to get the input power. And we measure the output power across the load, but we have to remove the losses from the filter from the equation, so we can measure the efficiency of the amplifier IC.

Efficiency pitfalls-- we need to know what part of the circuit is being measured. Are we measuring the whole system? Or are we measuring only the Class-D integrated circuit? We need to know where hidden losses are located-- the power supply wires, the speaker wires, the PCB traces, et cetera. And we also need to understand the ripple on the DC power supply, so that we can accurately measure the input power. Shown here, are typical Class-D efficiency and power dissipation curves. On the left-hand side is the efficiency curve, and on the right-hand side is the dissipation curves.

Power Supply Rejection Ratio, or a PSRR-- the Power Supply Rejection Ratio is the ability of an amplifier to maintain its output voltage as its DC power supply voltage is varied. So PSRR is equal to the change in your power supply voltage divided by the change in the Vout. This parameter quantifies the amplifier's ability to reject power supply noise. And the test setup is quite complicated, as shown in the next slide.

The PSRR test setup has an additional power supply amplifier inserted into the power supply rail. The signal generator sends a signal to the power supply amplifier, and the power supply amplifier provides DC voltage, plus the signal on top of the DC voltage. And this powers the amplifier IC. And of course, we still have the LC filter and the speaker load. And the signal analyzer measures the signal across the speaker load.

PSRR pitfalls-- one pitfall is that the test signal on the power supply is distorted. This is typically due to either the power supply amplifier cannot drive the load, or the power supply amplifier can't drive a high enough bulk capacitor that is on your amplifier's peak DC supply.