Because the transmitter and receiver often have different electrical requirements, sometimes you have to AC couple the signal, which means you put a series capacitor in the circuit that blocks the DC component of the transmitted signal.
Now, remember that DC is just the average value of the signal. Imagine you are using the simplest tranmission scheme, which is Pulse-Amplitude modulation, Non-return-to-zero (PAM-NRZ). This means the voltage is high for 1, low for 0. Also, say that in your transmission system the level for 1 is 2V higher than for 0, or as it is often described, the signal alternates between 1 and -1 V (relative to the DC level).
Some examples:
If you have an equal number of 1s and 0s, the average value is 0.5*1 + 0.5*(-1) = 0. Which is what you expect.
If you have 9 1s on average for each 0, your average value you be 0.9*1 + 0.1*(-1) = 0.8.
If you have all 1s, the average value is just 1.
As you can see, this is no good because the DC level at the receiver is data dependent and this causes all kinds of problems.
There are various ways to fix this. For example:
- DC couple your TX and RX
- Use a "DC balanced" line code. Basically encode the data such that 1s and 0s are approximately zero (even if they aren't in your real data). Examples of a DC balanced line code are Manchester encoding or 8b-10b encoding.
- Use a DC restoration circuit. Basically a slow filter which holds the baseline as the data changes.