Coaxial cable and digital encoded signals

[FOX]

I know that the first 100/ 200 Khz of the band of a coaxial cable carries distorsions, is not linear.The coaxial cable comes used to transmit digital encoded signals HDB3 or NRZ for example and if we examine their spectrum the lower frequencies have a better energy and here should have distorsions. I imagine that the information that we want to transmit has not contained in the lowest frequencies of the NRZ signal or HDB3 .If yes , where has the information concentrated, on which band of frequencies? And why ?

 

[flatulent]

equalizers

NRZ signals with truly random data, which these codings tend to produce, have their energy fall off with frequency. These codings are specifically used so that the bit timing can be more easily extracted at the destination end.

Price for a given performance is the driving function in signal and cable design. Many of these systems were stanardized around 1950 with the technology available at that time. The T1 system was designed to carry the same signals as used on already installed cables for analog FDM signals. It was designed to have repeaters at the inspection access areas (man holes).

In general, long cables are equalized. If there are no active repeaters (gain amplifiers), the equalization is done by passive RLC circuits with the inverse amplitude and phase response of the cable. If there are repeaters that contain ampliferes, the amplifiers are designed to have the gain to be the inverse of the cable before the signal is demodulated, retimed, regnerated, and sent out at full levels.

 

[FOX]

Thanks flatulent for fast reply ,
I understand the function of equalization , now. At this point the information primarily is contained in the frequencies to better energy ? Always , for other type of signal too . True?

 

[flatulent]

signal spectrum

In a truly random NRZ bit stream, the energy would be nearly continuously smeared across the spectrum from DC to infinity with a weighting function of sinx/x based on the bit width and rise/fall times. There would be a spike at DC.

One good source of more information on this is contained in books and articles on EMI radiation from PC boards. They will have graphs of signal spectral density vs bit width, bit rate, and rise/fall times.

 

[FOX]

Can u give me links to finds e-books and articles on this ?

 

[flatulent]

no

I am afraid not. But now that I have been thinking of this I remember that the sinx/x function in the frequency domain has a first null at the inverse of the bit period. This function is then multiplied by another sinx/x function in the frequency domain that has the first null at the inverse of the rise time.

If you want to draw a straight line approximation of the spectrum (vertical scale is amplitude in dB, horizontal scale is log frequency) start off horizontal until you come to the frequency equal to 1/(bit period). Draw a line going down at 20 dB/decade. Then at a frequency of 1/(rise time) bend the line a further 20 dB/decade down for a total of 40 dB/decade at all higher frequencies.