Walk into any office and place a one-page letter on the photocopier. The copy looks fine to a casual eye. Now copy the copy. Then copy that. By the tenth generation, the page is grey, smudged, and barely legible. The continuous greyscale of the original blends a little more with each step. There is no point at which you can stop the rot, because every copy is "almost right" but never exactly right.
Now imagine a different machine. Instead of treating the page as continuous greyscale, it scans each pixel and decides "ink" or "no ink." It then prints fresh, sharp, full-density ink at every "ink" pixel and pure white at every "no ink" pixel. The output is not a noisy copy but a clean reconstruction. As long as the scan can correctly distinguish ink from background, the next generation is just as good as the original. You can copy ten million times without degradation. That is digital communications.
The implications are revolutionary.
- Regeneration without accumulation. Each repeater on a transcontinental fibre link looks at incoming bits, decides each as 0 or 1 with vanishingly small error, and re-launches a fresh, full-amplitude signal. Twenty thousand miles of cable can carry a stream that exits the far end as clean as it left the near end. An analog signal, by contrast, is the sum of every noise source it ever passed through.
- Error correction. Once the signal is bits, we can sprinkle in extra bits chosen so that the decoder can detect and undo a small number of errors. The penalty is a slight loss in throughput. The reward is a system that operates reliably at signal-to-noise ratios where an uncoded analog system would be useless. Voyager 1 reads back 160 bits per second from beyond Pluto using error-correcting codes that turn a soup of thermal noise into a numbered photograph.
- Encryption is natural. Once data is bits, you can XOR it with a key stream, run it through AES, sign it with HMAC. None of these operations have analog analogues. Every modern security primitive lives downstream of the bit boundary.
- Multiplexing is easy. Bits from many users can share a wire by interleaving in time. Analog circuits cannot easily share the same channel cleanly. Time-division multiplexing only became routine after telephony went digital in the 1960s with PCM.
- Signals can be processed in software. Once you have bits, the rest is arithmetic. Filters, equalizers, decoders, demodulators, all become code. Software-defined radios exploit this exhaustively.
The cost of going digital is bandwidth. To replace a 4 kHz telephone voice channel with PCM at 8 bits per sample, we use 64 kbps and need a channel that can pass that bit rate. We trade analog spectral efficiency for digital robustness. The trade has been worth making, every time, since the 1950s.
Photocopy chain analogy, applied. Picture a signal as ink on paper passing through a series of imperfect copy machines. Analog: each copy is slightly worse than the last, errors accumulate without limit, and at the end you cannot recover the original. Digital with regeneration: each copy is read by a strict "is this ink or not" decision, then redrawn cleanly. Errors do not accumulate. Digital with regeneration and error correction: even if a few pixels were misread, the redundancy lets the decoder restore the original. Three layers of protection, all impossible in analog.