Noise and Signal Recovery

Every real signal picks up some unwanted noise on its way from transmitter to receiver — from electrical interference, from the thermal jostling of electrons inside components, from a dozen other sources. What happens next is where digital signals earn their keep: a noisy digital signal can often be repaired almost perfectly, while a noisy analogue signal is damaged for good.

What noise is, and where it comes from

Noise is a small, unwanted, essentially random fluctuation added on top of a genuine signal. It can arise from several sources: electrical interference from nearby equipment or power lines, thermal effects inside the components of the circuit itself (random motion of charge carriers, which becomes more significant as temperature rises), or interference picked up as a signal travels through a transmission medium such as a cable or the atmosphere. Whatever its origin, the practical effect is the same — the signal that arrives at the receiver is not quite identical to the signal that was sent; some small, unpredictable extra variation has been added to it.

Every real communication or storage system has to contend with noise to some degree, and how well it copes with that noise — how much of the original signal can be recovered afterwards — is one of the most important properties of the system.

Recovering a noisy digital signal

Digital signals have a crucial advantage here. Because a digital signal is only ever supposed to take one of a small number of discrete levels (in the simplest and most common case, just "high" and "low"), a receiving circuit doesn't need to know the exact voltage that was sent — it only needs to decide which of the two levels the received (possibly noisy) voltage is closer to. If a "high" signal picks up a little noise on its way to the receiver, it might arrive slightly below or above its ideal value, but as long as the noise hasn't been large enough to push it past the halfway decision threshold toward "low", the receiver still reads it correctly as "high".

This means a receiving circuit can effectively regenerate the signal: it reads the noisy incoming voltage, decides which discrete level it was intended to be, and outputs a fresh, clean version of exactly that level — discarding the noise entirely. The original data is recovered essentially perfectly, provided the noise added along the way was smaller than the margin between the two decision levels. This is the formal version of the robustness argument first introduced in the "Analogue and Digital Signals" lesson.

Why analogue signals can't be repaired this way

An analogue signal offers no such safety net. Because an analogue signal's genuine, meaningful information is its exact continuous value at each instant, there is no separate "intended level" to round back to — any value at all is potentially a legitimate part of the original signal. Once noise has been added to an analogue signal, a receiving circuit has no way to distinguish how much of the received value is genuine signal and how much is added noise; the two are inseparably mixed together. That corrupted information is permanently lost — it cannot be recovered by any amount of clever processing at the receiving end, because the processing has no way of knowing what the true original value was.

Weighing up digital sampling: advantages and disadvantages

Converting a signal to digital form is not automatically the "correct" choice in every situation — like most engineering decisions, it involves a genuine trade-off, and this balanced weighing-up is a common extended-response exam theme.

• Advantages. Much stronger noise immunity and the ability to recover the original data almost perfectly (as covered above); digital signals are also easier to process (using standard digital logic and computing), easier to store and copy without any loss of quality between copies (no "generation loss", unlike repeatedly copying an analogue recording), and easier to error-check, since extra bits can be added specifically to detect or even correct transmission errors.
• Disadvantages. Digitising a signal introduces quantisation error, an unavoidable small approximation that has no equivalent in a perfect (idealised) analogue representation; the process requires additional circuitry — an analogue-to-digital converter (ADC), and very often a digital-to-analogue converter (DAC) to convert the signal back to analogue form for its final use, such as driving a loudspeaker — adding cost and complexity; and, because more data is needed to represent the same original signal once digitised, a digital system typically requires a higher bandwidth to transmit the same information in the same time, unless the data is compressed.

In practice, the huge gains in noise immunity, processing flexibility and copying fidelity almost always outweigh these costs for modern systems, which is why digital representations are now used almost universally for audio, video, telecommunications and data storage — but the disadvantages are real engineering constraints, not merely theoretical footnotes, and a complete answer should acknowledge both sides.

Mark as complete