Impairments to voice channel transmission

Impairments to voice channel transmission

Impairments are the changes that a voice-frequency channel (nominally 300–3400 Hz) makes to a signal so that the received signal is not an exact copy of the transmitted one. They reduce speech quality and data accuracy. Main impairments: attenuation and attenuation distortion, phase (delay) distortion, noise, echo, crosstalk, frequency offset and jitter.

Attenuation distortion

• The loss of a channel is not the same at all frequencies in the band; different frequency components of the signal are attenuated by different amounts.
• Caused by cable capacitance (more loss at high frequency), transformers and filters (more loss near band edges).
• Specified as loss at each frequency relative to the loss at 1004 Hz (reference), e.g. −1 to +3 dB from 500–2500 Hz.

 Relative
 loss(dB)
   +  ┐                          ┌
      │ \                      / │
   0  │   ─────────────────────  │
      └───┴────────┴──────────┴──┴── f
         300      1004      3400 Hz

• Effect: speech sounds muffled; data pulses distorted. Corrected by equalisers and loading coils.

Phase (delay) distortion

• Different frequencies travel with different delays because the phase shift is not linear with frequency. Measured as envelope delay distortion – group delay $\tau = d\phi/d\omega$ relative to the minimum delay (usually near 1700–1800 Hz).
• Delay rises sharply near band edges due to filters.
• Little effect on speech (ear is insensitive to phase), but serious for data: causes intersymbol interference. Corrected by delay equalisers.

Noise

Unwanted signals added to the channel:

• Thermal noise ($kTB$, white) – sets the noise floor.
• Intermodulation noise – from non-linear devices (sum/difference frequencies).
• Crosstalk – coupling from adjacent pairs (NEXT, FEXT).
• Impulse noise – spikes from switching, lightning; causes burst errors in data.
• Quantisation noise in PCM systems.

Noise is measured with weighting (dBrnC, psophometric dBmp) and limits the signal-to-noise ratio of the channel (a voice channel needs SNR of about 30 dB or more).