By Dan Roach
This month I propose to
finish off my diatribe from last time, about the importance of proper source
impedance when driving telephone lines, with a roundup of information about
analog lease circuits. Last time I was
describing an electrically long (i.e., >300 m) twisted pair, in terms of its
series inductance and resistance and shunt capacitance, as being similar to an
analog low-pass filter. The equivalent
circuit is shown in Figure 1A. I was
attempting to show how the response at the output of the line varied with R,
the impedance of the audio source. Why
is that?
Well, it’s because it
varies the Q of the series RLC circuit, of course, as we know from our AC
theory. In Figure 1B you can see how
the upper frequency response varies with R.
As the source impedance is reduced, the Q of the filter is increased,
and the attenuation of high frequencies is delayed until the inevitable plunge
to zero output at infinite frequency.
The traditional approach
to flattening the response is to equalize it with a parallel resonant RLC
circuit, arranged as in Figure 2A. If you look carefully, the bottom half of the
response curve (below resonance frequency) complements the response curve of
the uncorrected telephone line—add ‘em together and you should get unity! By adjusting the resonance frequency of the
equalizer to just above the frequency response desired, and varying the damping
resistor to adjust the loaded Q, one can get a fairly flat resultant response
from the program circuit. With this type
of equalizer, frequency response above resonance drops like a rock, reducing
out-of-band noise as an added benefit.
But since the equalizer is completely passive, it can’t boost
frequencies that have been lost in the line, it can only attenuate frequencies
that have less loss to balance the response.
A long line can have a lot of inherent high frequency loss, so that at
the output of one of these equalizers levels will need amplifying, sometimes a
lot! Which is why we’ve progressed from
the simple RLC equalizer shown, to more modern equalizers from folks like Tellabs and McCurdy Telecom that can provide gain, and
other features like phase equalization.
Because, you see, these old RLC circuits can kind of ruin the phase
response of a line. This subjectively
doesn’t sound too bad with a moderately-equalized circuit, but can show up as
an odd kind of hollow sound when extreme amounts of equalization have been
used.
At least we’ve chosen an
unloaded pair. Normally, telco adds loading coils every fraction of a mile, which
add series inductance, with the net effect that line attenuation is much
reduced in the voice-band (300-3kHz), but drops precipitously above that. Once the audio’s been through this kind of
mess, it’s impossible to smooth out to get better high end response. Instead of a smooth attenuation curve, you
end up with bumpy in-band response followed by a sharp drop-off.
One trick that
old-timers have been known to use can come in handy when you have a fairly
short loop and no budget for equalizing:
you can use a pair of repeat coils to drop the impedance of the source
from 600 to 150 ohms, which more closely matches the actual impedance of the
line. At the receive end, you pop in
another repeat coil to get from 150 back up to 600 ohms. The circuits need to be properly terminated
with 600 ohms at each end. The result is
flatter response and less attenuation than if you had left the circuit at 600 ohms
throughout.