Method and apparatus for regenerating a modified duobinary signal

DETAILED DESCRIPTION OF THE INVENTION It is well known that signals transmitted over a cable pair will suffer considerable degradation with respect to amplitude and delay distortion.
The function of a regenerative repeater is to accept the degraded signal and to produce, at the output, a replica of the original signal as it was transmitted either from the originating station or the preceding repeater.
A block diagram of the preferred embodiment of a regenerative repeater for modified duobinary signals is shown in FIG.
1.
For a description of the modified duobinary technique, reference should be made to U.
S.
Pat.
No.
3,457,510 and to U.
S.
Pat.
No.
3,993,953, both of which are incorporated herein by reference.
Not as a limitation, but as a means of particularizing the description, the following discussion will only consider a repeater which is operating at a bit rate of 3.
152 Mb/s and providing a channel capacity equivalent to 48 voice-grade channels.
Because of the encoding in accordance with the modified duobinary rules using 100% duty cycle pulses, the output power spectrum density has a null at DC and a frequency numerically equal to 1/2 the bit rate, i.
e.
, 1.
576 MHz.
Thus, the power spectral density nearly is exactly the same (except for the constant scaling factor) as that for a 24-channel PCM system which employs a bipolar digital signal with 50% duty cycle pulses at a rate of 1.
544 Mb/s.
Also, as with the bipolar system, the power spectrum peaks at about 772 kHz.
The spectral densities for bipolar and modified duobinary pulse trains are derived and compared below.
The general expression for the spectral density of a random pulse train was derived by W.
R.
Bennett, "Statistics of Regenerative Digital Transmission," Bell System Technical Journal, November 1958, pp.
1501- 1543.
The expression for spectral density is: ##EQU1## where G(f) = Fourier transform of pulse shape 1/T = speed in bits/second, where T is the duration of the pulse time interval in seconds.
R(k) = autocovariance of the pulse sequence