Cirrus Logic AN257 User Manual

AN257

Application Note

MEASURING THE PERFORMANCE OF POWER SUPPLY

REJECTION FEEDBACK AS USED IN CS44800/44600

DIGITAL CLASS-D PWM AMPLIFIERS

by Randy Boudreaux and Sean Davis

1. Introduction

Most currently available digital Class-D amplifiers are termed “open-loop” systems since no mechanisms have been developed to provide real-time feedback to reject unwanted tones (ripple) generated by the power supply, or the noise coupled onto the power rail due to ripple currents flowing through large decoupling capacitors. Since the high sides of the power MOSFETs are connected directly to the power supply rail, all tones and associated harmonics generated on the power supply rail are coupled onto the output of the audio amplifier channel.

Costly detection and feedback circuitry must be added to the power supply to account for output voltage variations due to load current fluctuations. In addition, the output voltage rail must be designed to remain stable as the input AC voltage deviates from the nominal 120 V. Any drift from the nominal design point will be directly reflected in the output DC voltage rail causing audible volume changes during playback. To overcome these inherent limitations of conventional “openloop” Class-D amplifiers, tightly regulated power supplies are commonly employed.

Utilizing the integrated PSR Feedback, it is shown that a digital Class-D PWM amplifier implemented with the CS44800/44600 PWM Controller is now an effective power supply “closed-loop” system by employing real-time feedback of the voltage rail. Real-time voltage feedback now enables this digital Class-D PWM amplifier to operate with any type of power supply, with any measure of output regulation. To demonstrate the rejection/compensation for the above-mentioned deficiencies, a procedure is described below that uses the CDB44800 development board to emulate a system which contains a large amount of tones and harmonics on the power supply rail.

Power supply costs can now be reduced and other power supply technologies used. “Open-loop” digital Class-D PWM amplifiers prohibit the use of traditional low-cost, dependable, low-EMI-radiated-noise, unregulated, linear power supplies. Published application notes for current ClassD amplifier systems on the market recommend the use of switch mode power supplies with very low output voltage ripple and tight regulation for both load and line variations, in conjunction with large, expensive, low-ESR electrolytic capacitors. The CS44800/44600 PSR circuitry eliminates these dependencies and allows for inexpensive power supply alternatives to be used.

www.cirrus.com

FEB ‘05

AN257REV1

AN257

2. PSR Feedback Measurement Procedure

2.1 Power Supply Rejection Performance Test

The test platform used for the performance measurement plots below consists of the CDB44800 (half-bridge channels), with Vpower = +40 V, and a resistive load of 6 Ω. Channel A of the input audio stream was set to channel 3 on the development board and channel B was set to channel

1) Use an unregulated power supply if available. The PSR rejection performance can be seen using any type of power supply, however the actual amount of rejection due to PSR feedback will be more visible with a lightly regulated power supply than with a highly regulated power supply. PSR calibration should be done before any of the channels are enabled so an accurate representation of the nominal voltage rail can be captured. See the CS44800/600 datasheet for the PSR calibration routine.

2) Set up the board under test for 2-channel operation with 6priate script file such that amplified audio is playing from the board and PSR is calibrated. Verify that PSR Feedback is disabled (CS44800/44600 bit 5 in register 34h set to 0b). The two channels being used should be in separate power packages, such as channel 3 and channel 8 on the CDB44800, to minimize the effects of switching noise.

3) Using a digital audio source (such as an Audio Precision), set channel A to be a wave and set channel B to be a 0-dB, 60-Hz sine wave. Channel B is being used to emulate the harmonics generated by an inexpensive, poorly regulated power supply. The 60-Hz switching output of channel B will generate a corresponding ripple voltage on the rail, which will couple into the channel under test, in this case

channel A.

Ω resistive loads and execute the appro-

-1-dBFS, 1-kHz sine

4) With only the channel A PWM output turned on, use an analog analyzer (such as an Audio Precision) and take an FFT of amplitude versus frequency for channel A. The 1-kHz tone should be present with an amplitude of

d B

-1 dBFS. Figure 1 shows the results with only channel A enabled as a baseline.

-10

-20

-30

-40

-50

-60

-70

-80

-90

-100

-110

-120

-130

-140

-150 20 20k50 100 200 500 1k 2k 5k 10k

ColorSweep Trace Line St y l e Thick Data Axi s Comment

1 1 Red S ol i d 1 Fft.Ch.1 Am pl Left Channel B = disabl ed, P SR feedback di sabled

Figure 1. FFT Amplitude vs. Frequency,

Channel A = 1 kHz,

-1 dBFS, Channel B = disabled, PSR feedback disabled

2 AN257REV1

AN257

5) Turn on the channel B PWM output and take an FFT of amplitude versus frequency for channel A. The 1-kHz tone should be present with an amplitude of 60-Hz tone and associated harmonics. Figure 2 shows the FFT of channel A with both channel A and channel B PWM outputs enabled. The original 1-kHz tone is shown at Hz tone from channel B shown at

-50 dBFS. The full scale, 60 Hz tone being played back on

channel B’s MOSFET devices causes an associated 60-Hz ripple current on the power voltage rail. This ripple current, along with the capacitor’s equivalent series resistance (ESR), causes the discrete tones on the power supply rail. Notice the 2nd, 3rd, 4th, 5th, etc. harmonics at 120 Hz, 180 Hz, 240 Hz, 300 Hz, etc. due to the system non-linearities. Because all of these tones are being modulated onto channel A’s audio output by the power MOSFETs switching at a 384-kHz rate, these discrete tones will also be modulated onto the 1-kHz tone being played back (see tones grouped around 1 kHz). These modulated tones appear as symmetrical, equidistant tones on each side of the 1-kHz tone. The amplitude and frequency of each modulated tone is easily calculated using standard FM modulation formulas.

-10

-20

-30

-40

-50

-60

-70

-80

-90

-100

-110

-120

-130

-140

-150 20 20k50 100 200 500 1k 2k 5k 10k

-1 dBFS (with modulated side tones) along with a

-1-dBFS, with the coupled 60-

ColorSweep Trace Line St y l e Thick Data Axi s Comment

2 1 Blue Solid -1 Fft.Ch. 1 A m pl Left Channel B = 60 Hz , 0 dB F S , PSR feedbac k di s abl ed

Figure 2. FFT Amplitude vs. Frequency,

Channel A = 1 kHz,

-1 dBFS, Channel B = 60 Hz, 0 dBFS, PSR feedback disabled

6) Enable PSR Feedback (CS44800/44600 bit 5 in register 34h set to 1b). Take an FFT of amplitude versus frequency on the output of channel A. The 1-kHz tone should be present with an amplitude of

-1 dBFS, however the 60-Hz tone and the modulated side tones will be greatly diminished in ampli-

tude. Figure 3 shows the results of having both channels on and PSR feedback enabled.

AN257REV1 3

+ 4 hidden pages