Field Guide · algorithm

Also known as: early-late gate, early/late gate, ELG

Early-late gate timing recovery is a symbol-synchronization technique that finds the best instant to sample each symbol by exploiting the symmetry of the matched-filter pulse: it takes one sample slightly early and one slightly late relative to the current estimate, and if the two differ in magnitude the sampler is off-centre and must be nudged.1 It is one of the classic clock-recovery methods that lets a receiver lock its symbol clock to a transmitter with no shared timing reference.

on-time (peak) early late |early| = |late| ⇒ locked early late off-centre ⇒ error
On a symmetric matched-filter pulse the early and late samples are equal only when the middle sample sits on the peak; any imbalance is a signed timing error that steers the clock.

How it works

After matched filtering (e.g. a root-raised-cosine receive filter), each symbol appears as a pulse whose peak marks the correct sampling instant. Because that pulse is symmetric about its peak, a sample taken τ before the peak and one taken τ after it have equal magnitude — when you are sampling on the peak. The loop exploits this:

  • Form a timing error ≈ |y(early)| − |y(late)| (an energy/magnitude comparison, so it works regardless of the symbol’s polarity).
  • If early > late, the true peak is earlier than assumed → advance the clock; if late > early, retard it. At the peak the two are equal and the error is zero.
  • Feed this error through a loop filter to a numerically-controlled clock or interpolator, forming a closed timing loop that converges and then tracks slow drift.

Early-late gate typically wants roughly two samples per symbol (or an interpolator to synthesize the early/late points) and a reasonably symmetric pulse for the equal-magnitude assumption to hold.

Contrast with Gardner and Mueller-Muller

Early-late gate is the intuitive ancestor of the detectors used most today:

  • Gardner also uses samples around the symbol but forms its error from the midpoint sample between two symbols times their difference; it needs 2 samples/symbol, is non-data-aided, and is notably carrier-phase independent, which is why it dominates modern PSK/QAM modems.
  • Mueller-Muller is a decision-directed, one-sample-per-symbol detector — cheaper in samples but it needs reliable symbol decisions and is more sensitive to residual carrier offset.

Early-late gate is simple and robust but its explicit early/late samples cost extra computation or interpolation compared with Gardner, and like all these detectors it degrades under low SNR where the pulse shape is buried in noise.

Relevance to SDR

Symbol-timing recovery is mandatory in every digital receiver: without it the sampler drifts off the symbol peaks and error rates collapse. Early-late gate and its variants appear in modem chipsets, satellite and telemetry receivers, and SDR toolkits. Trunked-radio decoders recover a 4800 sym/s clock for C4FM/π-4-DQPSK modes this way. GopherTrunk performs symbol-timing recovery in its demodulators using Gardner/Mueller-Muller-style detectors; early-late gate is the foundational scheme those refine.

Sources

  1. Early-late gate — Wikipedia, on the early/late magnitude-comparison timing-error detector; see also symbol-synchronization loop background. ↩

See also