www.ti.com
CLASS-D RC LOW-PASS FILTER
R
FILT
R
L
R
FILT
C
FILT
V
L=VIN
V
OUT
R
ANA
C
ANA
R
ANA
C
ANA
C
FILT
To APA
GND
AP AnalyzerInput
RCLow-PassFilters
Load
V
OUT
w
V
IN
wO
R
ANA + RFILT
R
ANA
1+j
(
)
=
(5)
f =
c
2 x f
max
(6)
SLOS516B – OCTOBER 2007 – REVISED MARCH 2007
An RC filter is used to reduce the square-wave output when the analyzer inputs cannot process the pulse-width
modulated class-D output waveform. This filter has little effect on the measurement accuracy because the cutoff
frequency is set above the audio band. The high frequency of the square wave has negligible impact on
measurement accuracy because it is well above the audible frequency range, and the speaker cone cannot
respond at such a fast rate. The RC filter is not required when an LC low-pass filter is used, such as with the
class-D APAs that employ the traditional modulation scheme (TPA032D0x, TPA005Dxx).
The component values of the RC filter are selected using the equivalent output circuit as shown in Figure 36. RL is the load impedance that the APA is driving for the test. The analyzer input impedance specifications should be
available and substituted for RANA and CANA. The filter components, RFILT and CFILT, can then be derived for the
system. The filter should be grounded to the APA near the output ground pins or at the power supply ground pin
to minimize ground loops.
Figure 36. Measurement Low-Pass Filter Derivation Circuit-Class-D APAs
The transfer function for this circuit is shown in
Equation 5 where
ω
O = REQCEQ, REQ = RFILT || RANA and
CEQ = (CFILT + CANA). The filter frequency should be set above fMAX, the highest frequency of the measurement
bandwidth, to avoid attenuating the audio signal. Equation 6 provides this cutoff frequency, fC. The value of RFILT must be chosen large enough to minimize current that is shunted from the load, yet small enough to minimize the
attenuation of the analyzer-input voltage through the voltage divider formed by RFILT and RANA. A general rule is
that RFILT should be small (~100 ) for most measurements. This reduces the measurement error to less than
1% for RANA ≥ 10 k.
An exception occurs with the efficiency measurements, where RFILT must be increased by a factor of ten to
reduce the current shunted through the filter. CFILT must be decreased by a factor of ten to maintain the same
cutoff frequency. See
Table 3 for the recommended filter component values.
Once fC is determined and RFILT is selected, the filter capacitance is calculated. When the calculated value is not
available, it is better to choose a smaller capacitance value to keep fC above the minimum desired value
Copyright 2007, Texas Instruments Incorporated
27