DEVICE DESCRIPTION

A PCM Codec–Filter is used for digitizing and reconstructing

the human voice. These devices are used primarily for

the telephone network to facilitate voice switching and transmission.

Once the voice is digitized, it may be switched by

digital switching methods or transmitted long distance (T1,

microwave, satellites, etc.) without degradation. The name

codec is an acronym from ‘‘COder’’ for the analog–to–digital

converter (ADC) used to digitize voice, and ‘‘DECoder’’ for

the digital–to–analog converter (DAC) used for reconstructing

voice. A codec is a single device that does both the ADC

and DAC conversions.

To digitize intelligible voice requires a signal–to–distortion

ratio of about 30 dB over a dynamic range of about 40 dB.

This may be accomplished with a linear 13–bit ADC and

DAC, but will far exceed the required signal–to–distortion

ratio at larger amplitudes than 40 dB below the peak amplitude.

This excess performance is at the expense of data per

sample. Two methods of data reduction are implemented by

compressing the 13–bit linear scheme to companded

pseudo–logarithmic 8–bit schemes. The two companding

schemes are: Mu–255 Law, primarily in North America and

Japan; and A–Law, primarily used in Europe. These companding

schemes are accepted world wide. These companding

schemes follow a segmented or ‘‘piecewise–linear’’ curve

formatted as sign bit, three chord bits, and four step bits. For

a given chord, all sixteen of the steps have the same voltage

weighting. As the voltage of the analog input increases, the

four step bits increment and carry to the three chord bits

which increment. When the chord bits increment, the step

bits double their voltage weighting. This results in an effective

resolution of six bits (sign + chord + four step bits) across

a 42 dB dynamic range (seven chords above 0, by 6 dB per

chord).

In a sampling environment, Nyquist theory says that to

properly sample a continuous signal, it must be sampled at a

frequency higher than twice the signal’s highest frequency

component. Voice contains spectral energy above 3 kHz, but

its absence is not detrimental to intelligibility. To reduce the

digital data rate, which is proportional to the sampling rate, a

sample rate of 8 kHz was adopted, consistent with a bandwidth

of 3 kHz. This sampling requires a low–pass filter to

limit the high frequency energy above 3 kHz from distorting

the in–band signal. The telephone line is also subject to

50/60 Hz power line coupling, which must be attenuated

from the signal by a high–pass filter before the analog–to–

digital converter.

The digital–to–analog conversion process reconstructs a

staircase version of the desired in–band signal, which has

spectral images of the in–band signal modulated about the

sample frequency and its harmonics. These spectral images

are called aliasing components, which need to be attenuated

to obtain the desired signal. The low–pass filter used to attenuate

these aliasing components is typically called a reconstruction

or smoothing filter.

The MC145484 PCM Codec–Filter has the codec, both

presampling and reconstruction filters, and a precision voltage

reference on–chip.