Why Digital, and Why Binary
Start here. Send one value down a noisy wire as a level and as a bit, and only one of them survives.
Skip to the animationA digital signal carries one of a small number of discrete values instead of a continuous voltage, which means noise can be discarded rather than accumulated — and two values is the choice with the widest safety margin and the simplest hardware.
The analogue problem
In an analogue system, the quantity of interest is the voltage: 2.4 V might mean 24 °C. Between any two voltages there is another one, so a single wire theoretically carries unlimited precision. That sounds like a strictly better deal than two levels, and on a perfect wire it would be.
Real wires pick up interference from every motor, radio and neighbouring track. If 2.4 V leaves and 2.29 V arrives, the received value is a perfectly legal analogue value — indistinguishable from a genuine 2.29 V. No circuit downstream can detect the error, let alone repair it.
Worse, the errors accumulate. Each stage adds its own, so a signal through ten stages is measurably wrong and through a thousand is meaningless. This is why a tenth-generation tape copy is unlistenable.
The digital answer: bands, with a gap
Stop using the exact voltage. Declare that anything above a threshold means 1 and anything below another means 0, and deliberately leave the range between them unused.
| Band | Typical 5 V range | Meaning |
|---|---|---|
| Logic 1 | 2.0 V – 5.0 V | 1, whatever the exact voltage |
| Forbidden | 0.8 V – 2.0 V | no valid signal ever sits here |
| Logic 0 | 0.0 V – 0.8 V | 0, whatever the exact voltage |
Send 3.3 V, receive 2.9 V after the same interference — still comfortably logic 1. The information was never in the precise voltage, so corrupting the voltage did not corrupt the information. That unused middle range is the noise margin, and it is the entire product.
Regeneration is the property that matters
A gate does not pass its input along. It decides which band the input is in, and then drives its own output to a full, clean level. The noise is discarded rather than carried forward.
That is why a signal can pass through billions of transistors and arrive exactly as correct as it started. Errors cannot accumulate, because every stage throws the accumulated error away. Analogue circuits have no equivalent, and this single asymmetry is why digital won.
- Noise margin
- How much interference a signal can absorb and still be read correctly.
NM_high = V_OH − V_IHandNM_low = V_IL − V_OL: the gap between what a driver guarantees to output and what a receiver promises to accept. - V_OH / V_OL
- The output levels a gate guarantees to drive — deliberately further from the middle than the input thresholds require.
- V_IH / V_IL
- The input levels a gate guarantees to interpret correctly. The deliberate gap between these and the output levels is what creates the margin.
Why exactly two levels
More levels would carry more data per wire — four levels is two bits, ten is about 3.3. But squeezing more levels into the same 0–5 V narrows every gap, and margin is the product being bought.
- Widest margins. Two levels leaves the largest possible gap, which is what tolerates temperature, ageing and interference.
- Matches the hardware. A transistor is a switch — naturally either conducting or not. No comparator is needed to tell two states apart.
- Boolean algebra already exists. A complete, well-understood mathematics for two-valued logic was available a century before anyone needed it.
- Cheap and reproducible. Only two states must be manufactured reliably, across billions of devices.
Multi-level signalling is not fantasy — flash memory stores several bits per cell exactly this way, and pays for it in error-correction machinery and reduced write endurance.
A bit is a band, and meaning is a convention
One wire distinguishes two values: one bit. Put n wires side by side and you can distinguish 2ⁿ patterns — 8 wires give 256.
The hardware attaches no meaning to a pattern. Deciding that 10000001 is 129, or −127, or the letter Á, or an instruction, is a convention chosen by whoever designed the system. Choosing those conventions is the whole of the next module.
Advantages and disadvantages
Advantages
- Noise is discarded at every gate instead of accumulating.
- Perfect copying and storage — a bit is either right or obviously wrong.
- Circuits can be designed with algebra and verified before being built.
- Two states match what a transistor naturally does, so devices are cheap.
Disadvantages
- The real world is analogue, so every input needs an ADC and every output a DAC.
- Quantisation throws information away — resolution is capped by the number of bits.
- Representing a smooth quantity takes many wires or many clock cycles.
- Switching between levels radiates interference and burns power on every transition.
Watch it work
Check yourself
question 1 / 4
One question at a time. Pick an answer to see why it is right or wrong, then move on — there is no score to keep and nothing is saved.