Spectral properties of EFM encoding

Question

Data is written to CD/DVD using an encoding called eight-to-fourteen modulation, where each eight bits of data are encoded by fourteen bits on disk. The encoding was designed to give the encoded signal some useful properties, such as limiting consecutive identical bits (for clock recovery).

One property that I'm struggling to find motivation for is the low power density at low frequencies. Can someone explain why this property is useful?

the spectrum (power density function) of the encoded sequence vanishes at the low-frequency end

(source: Wikipedia)

EFM combines high information density ... with low power at the low-frequency end of the modulation bit stream spectrum.

(source: abstract of related paper)

Answer 1

As already pointed out by Marcus in another answer, you'd want to get rid of the DC component of a signal so you can use AC-coupled stages on your signal processing chain. This would be possible by using a simpler encoding.

The additional advantage of having low power density function (PDF) at low frequencies is that this makes the AC-coupling easier, since you can use "smaller caps" (i.e. caps with smaller capacitance) and consequently amplifiers with smaller input impedance (to avoid attenuation, the input impedance should be much higher than the impedance seen "looking back" at the preceding stage, and this latter includes the impedance of the coupling cap).

A signal with no DC level, but with high PDF at low frequencies would require bigger coupling caps to avoid signal integrity issues. To be more specific: lower capacity caps would lead to a lower cut-off frequency that is higher, hence leading to more signal distortion because the frequency response of the system would attenuate more in the frequency range where the energy content of the signal is more significant.

Answer 2

The most easy to come up with explanation is that you practically never want to have a DC component – simply because

1. you can pass that through a capacitor, e.g. to couple it to a different bias,
2. having "data-changeable" DC means that in one moment, your semiconductors might be working at a different operating point than at the next moment, de-linearizing response.