1997 Nov 17
20
Philips Semiconductors
Preliminary specification
ISO/MPEG Audio Source Decoder
SAA2502
7.4.7
D
YNAMIC RANGE COMPRESSION
The baseband audio output resulting from MPEG
decoding has a high dynamic range (theoretically
>200 dB, practically up to 120 dB for the 22-bit output
mode).This feature is very attractive from the high quality
audio standpoint of view, but such high dynamic range is
undesirable when there is a relatively high level of
background noise (e.g. for car radio). For those
applications the SAA2502 offers the possibility of built in
dynamic range compression:
Internal dynamic range compression is offered. Thus
any standard MPEG encoded bit stream may be
compressed i.e. no added compression information is
required.
The dynamic range compression algorithm is fully
parameterised. All major characteristics are
programmable through the control interface:
– Level of compression
– Maximum compression
– Compression offset
– Compression release rate (compression attack rate
has to be fixed).
The dynamic range compression algorithm is based on a
(in time varying) amplification factor, which is equally
applied to all audio output samples. The value of the
amplification factor is calculated on basis of the current
audio output power level for each (sub)frame of 384 output
samples. The applied power to amplification curve is
shown in Fig.11. All characteristics of the curve are
programmable:
Compression slope minimum = 0, maximum = 0.996
Maximum amplification minimum = 0 dB,
maximum = 23.81 dB
Offset minimum = 0 dB, maximum = 47.81 dB.
Offset values close to 0 dB may result in clipped output
signals. This is especially true for signals with a high
amplitude-to-power ratio (an extreme example of such a
signal is a maximum amplitude unit impulse).
The occurrence of this effect can be avoided by selecting
an offset value close to or greater than 15 dB.
In the context of dynamic range compression definition,
the 0 dB power reference level is defined as a sine wave
shaped output signal with maximum amplitude in just one
(right or left) channel.
The calculation will result in an new amplification factor
every 384 samples (i.e. from 8 ms at 48 kHz to 24 ms at
16 kHz sample rate). Subsequent amplification factors
may vary considerably.
An example showing two large step type discontinuation is
shown in Fig.12. It is undesirable to apply large increasing
amplification steps immediately. Consequently increasing
the amplification factor is limited to the ‘release rate’ which
is also programmable:
Minimum release rate =
(1.46 dB/s at 48 kHz; 0.488 dB/s at 16 kHz)
Maximum release rate =
(46.87 dB/s at 48 kHz; 15.625 dB/s at 16 kHz).
Decreasing amplification factors, must be applied almost
immediately to avoid overflow when the audio power
increases rapidly; thus attack rate is non-programmable
and fast.
384 samples
384 samples
Fig.11 Dynamic range compression characteristic.
handbook, halfpage
MGE478
maximum
amplification
compression
slope
amplification
(dB)
0 dB
power (dB)
offset
Fig.12 Amplification change rates.
handbook, halfpage
MGE479
audio
signal
power
amplifi-
cation
release
rate
attack
rate
time