Signal
A signal is a physical quantity, most often a voltage or a current, that varies in a controlled way so that the variation carries information. The voltage on a microphone wire, the pulses on a network cable and the light flickering in an optical fibre are all signals: the medium is only a vehicle, and what matters is the pattern imposed on it. Electronics is, in large part, the craft of creating, moving, reshaping and detecting such patterns without losing the message.
Analog and digital signals differ in how the pattern encodes the message, and that difference explains why transistors are used so differently in the two worlds. An analog signal varies continuously and is proportional to what it represents: a microphone voltage rises and falls in step with air pressure, so every tiny wobble is meaningful and every tiny error is a distortion. A digital signal agrees in advance that only a few distinct levels count, usually two, so a receiver only has to decide high or low. Because the decision throws away small deviations, noise can be wiped out at every stage, which is why a copy of a copy of a digital file is perfect while a copy of an analog tape is not. A transistor driven hard between off and on is a digital device; one held in the middle of its range to amplify gently is an analog one, and the same silicon can do both.
Frequency, amplitude and waveform are the three vocabulary words for describing how a signal moves. Amplitude is how large the swing is, frequency is how many cycles occur per second (measured in hertz), and waveform is the overall shape, whether a sine wave, a square wave or something irregular like speech. Fourier showed that any practical waveform can be built by adding sine waves of different frequencies, so a signal's bandwidth, the range of frequencies it contains, tells how fast a circuit must be to handle it. A sharp-edged digital pulse contains very high frequencies, which is why fast processors are plagued by the same wire effects as radio gear.
Noise is the unavoidable partner of every signal, and the signal-to-noise ratio is the number that says whether the message survives. Noise is any unwanted random variation: thermal agitation of charge carriers in every resistor, the granularity of discrete electrons in a current, interference picked up from neighbouring wires. The signal-to-noise ratio compares signal power to noise power and is commonly quoted in decibels. Claude Shannon's 1948 paper A Mathematical Theory of Communication turned this into a hard limit: the capacity of a channel, in bits per second, is bandwidth multiplied by the base-2 logarithm of one plus the signal-to-noise ratio. No cleverness can push reliable information through a channel faster than that bound, though good coding can approach it, which is the principle behind modern wireless links and storage devices.
The word signal carries two senses that are easy to blur: the physical carrier and the abstract information. In the physical sense, a signal is the actual voltage on a particular wire at a particular moment, something an oscilloscope can display. In the abstract sense, the signal is the message itself, which can be moved from a copper wire to a radio wave to a magnetic domain to a flash memory cell and still be the same signal. Shannon's achievement was to make the abstract sense measurable, in bits, independent of the hardware. A transistor sits exactly at the join of the two senses: it is a physical device whose output voltage is controlled by an input voltage, so it can regenerate a degraded pulse, restoring the abstract message that the physical carrier has partly lost.
A common misconception is that a digital signal is not really electrical or is somehow immune to physics. Inside a chip a digital one is simply a voltage near the supply level and a zero is a voltage near ground, and the transitions between them are as analog as anything else, with finite slopes, overshoot and noise. Digital design works because logic gates restore levels at every step, giving noise margins, not because the underlying signals are exempt from noise. When margins shrink, as with lower supply voltages or faster clocks, bit errors appear. Modern relevance abounds: high-speed links between processors and memory use multilevel signals and equalisation precisely because the analog nature of the wire reasserts itself, and AI accelerators spend a large share of their power moving signals, not computing with them.
Read signal as information riding on a physical variation: the carrier is merely voltage or current, the message is the pattern, and noise sets how much pattern can be trusted.