Analog tells you it is dying while it dies. Digital does not. It is flawless, and then it is nothing — and the transmission before the last one sounded exactly like the first.
Every operator new to digital voice says some version of that sentence, usually with a note of complaint in it. The repeater is armchair copy, quiet and clean, better audio than anything analog ever gave them — and then, mid–word, silence. No fade. No warning. Nothing to react to.
It feels like a fault. It is not. That is the design, working exactly as intended.
This page is about why, because nobody tells a newcomer, and because the behavior is not a flaw to be tolerated — it is the direct price of the thing they came to digital voice for in the first place.
FM degrades continuously. That single word is the whole difference.
Drive away from an analog repeater and the signal walks down a staircase you can hear every step of. Full quieting. Then a whisper of hiss under the voice. Then more hiss. Then the voice riding on top of noise. Then picket-fencing as the flutter sets in. Then a voice buried in a wash of static that you can still, somehow, almost copy — because the human ear is a magnificent error-corrector, and analog hands it something to work with all the way down.
And here is the part that matters most, the part you never notice until it is taken away:
Analog tells you where you stand. The signal reports its own health, continuously, in the only currency you have — the sound in your ear. You always know how much margin is left, because you can hear it being spent.
You have been reading that report your entire operating life. You have almost certainly never thought about it once.
A digital signal is not a sound being carried. It is a stream of bits, and a bit is either recovered correctly or it is not. There is no such thing as a slightly noisy bit.
Real radio paths damage bits constantly — multipath, flutter, weak signal, interference. So every digital voice mode wraps its bits in forward error correction: extra, redundant bits sent along with the payload, enough that the receiver can detect damage and repair it without asking for a retransmission. There is no time to ask. It is a live conversation.
Error correction does its job completely. Repair a damaged frame and what comes out the other side is not “mostly right” — it is exactly right, bit for bit identical to what was sent. Perfect audio. The listener hears no evidence that anything was ever wrong, because in the recovered stream, nothing is wrong.
So the damage is hidden. Not reduced — hidden. Every dB of margin you spend on the way out to the fringe is silently absorbed by the error correction, and your audio stays flawless while it happens.
Until it doesn’t. Error correction has a budget: a finite number of errors per frame it can repair. Push past that and it cannot fix the frame, cannot partially fix the frame, and has nothing sensible to hand the vocoder. So you get a burst of digital garble — the R2-D2 noise — and then, as more frames fail, silence.
That threshold is the cliff. It is the edge of the error-correction budget, and you walk off it without ever having been told you were near it.
The dot is your last good transmission. It sounds exactly like your first.
Look at the two curves again and ask a different question: not “where does each one end?” but “how much of each curve is actually usable?”
The analog curve starts degrading immediately. Long before it reaches the bottom of the chart it has passed through scratchy, then annoying, then genuinely hard work to copy. That whole lower stretch is technically a signal and practically a headache.
The digital curve does not degrade at all until it stops. Every bit of the path where digital still works, digital works perfectly.
Analog spends its margin on noise. Digital spends its margin on staying perfect. Same margin. Different currency. Digital takes what analog would have squandered on hiss and hands it back to you as clean audio over a wider useful range.
You trade the warning for the range. Below the cliff, analog would have been miserable anyway — and in the stretch where they overlap, digital simply sounds better. That is the bargain, and for most operating, most of the time, it is a good one.
What you gave up is the report. And the report, it turns out, was doing more work than anyone realized.
Everything practical about the digital cliff follows from one fact:
You cannot hear yourself getting weak. Your last transmission before the cliff sounds identical to your first. There is no cue, no scratchiness, no moment where a careful operator thinks “I’d better find a better spot.” You are full quieting, and then you are not there.
This has consequences that catch people out over and over:
An operator who has internalized this stops treating the drop as a malfunction and starts treating it as a boundary — a fixed feature of the landscape, like a canyon wall. You do not argue with it. You learn where it is.
The cliff is universal to digital voice, but its shape is not identical across the six modes. Three things set how abrupt the edge is:
| What varies | Why it changes the cliff |
|---|---|
| The error-correction scheme | A bigger FEC budget means the mode absorbs more damage before failing — the flat part of the curve runs further out, and the drop, when it comes, is later and harder. |
| The vocoder | Different codecs fail differently. Some degrade into recognizable garble that a listener can still partly follow for a moment; others go straight to nonsense. This is why the drop sounds different from mode to mode. |
| The channel width | A narrower channel carries less signal energy for a given power. NXDN’s 6.25 kHz channel is remarkable engineering, but narrow channels have less to work with at the fringe. |
Notice what that middle row is. The vocoder row. The same row that explains why transcoding is hard, why AMBE dongles exist, and why the open-source community built Codec 2 — it also explains why the cliff sounds different on D‑STAR than it does on DMR.
That is the hinge, and it is worth saying plainly: the vocoder row of the Rosetta Stone is doing more work than any other single fact in digital voice. Learn what your mode’s codec is and half the mysteries — bridging, dongles, transcoding, and the shape of your own cliff — stop being mysteries.
Digital voice did not take away the fade. It took away the warning, and gave you back the range.
Analog told you where you stood, and charged you hiss for the privilege. Digital gives you flawless audio for as long as it gives you anything at all, and then it stops, without comment.
Once you know that, the drop stops feeling like a broken radio and starts feeling like what it is: the edge of the map. It was always there. Analog just talked you toward it, and digital lets you walk.