Carrier Recovery and Phase Noise in Digital Receivers

Carrier recovery is the receiver function that estimates and removes carrier frequency and phase error from a modulated signal. In a coherent digital receiver, the constellation is meaningful only after the receiver has aligned its local oscillator, digital numerically controlled oscillator or carrier-tracking loop with the incoming waveform.

Phase noise is the short-term instability of that phase reference. It can blur constellation points, increase error vector magnitude, weaken demodulation, stress forward error correction and trigger fallback to lower modulation modes. A receiver can have enough received power and still fail because its carrier estimate is too noisy or too slow.

Principle

The transferable rule is:

A digital receiver must remove carrier frequency and phase error fast enough to keep constellation decisions stable, while rejecting oscillator noise and interference that should not become tracking motion.

This principle is separate from symbol timing. Symbol timing decides when to sample. Carrier recovery decides how to rotate the received complex samples so that the constellation axes are correct. Practical receivers need both.

A carrier-recovery review should state:

1. carrier frequency tolerance and expected Doppler or oscillator offset;
2. acquisition range and acquisition time;
3. residual frequency offset after lock;
4. residual phase error, phase-noise contribution and loop bandwidth;
5. modulation order and allowed phase-error budget;
6. EVM, FEC, BER, packet-error and lock-state evidence;
7. behavior during fading, burst acquisition, handover, temperature drift and reference-clock faults.

Without these checks, a high-order modulation mode may be selected from SNR alone even though phase stability is not good enough for its constellation spacing.

Carrier Error As Constellation Rotation

After downconversion, many receivers represent the signal as a complex baseband sequence:

where:

• r(t) is the received complex baseband signal;
• s(t) is the ideal transmitted symbol waveform after channel effects;
• n(t) is noise and interference;
• \phi(t) is residual carrier phase error.

A simple phase-error model is:

where:

• \Delta f is residual carrier frequency offset;
• \phi_0 is static phase offset;
• \phi_n(t) is time-varying phase noise or tracking error.

The first term is important. Even a small residual frequency offset produces phase error that grows with time. If the receiver does not track it, the constellation rotates continuously.

The phase increment per symbol is:

where T_s is the symbol period. A small per-symbol increment can still become a large packet-level rotation if the packet or frame is long.

Worked Example: Residual Frequency Offset And Phase Budget

A receiver uses a 64-QAM mode with:

The symbol period is:

Before carrier tracking, the oscillator offset is:

The per-symbol phase rotation is:

That looks small per symbol. Over a 2\ \text{ms} burst, however:

Without carrier tracking, the constellation would rotate three full turns during the burst. The receiver must acquire and track the carrier, not merely tolerate the offset.

After acquisition, assume the residual frequency offset is reduced to:

For a 200\ \mu\text{s} decision interval:

Assume the release budget allows:

Measured residual phase-noise jitter is:

and the static phase bias estimate is:

Use a conservative phase budget:

Substitute:

The phase margin is:

Engineering Comment

The carrier-recovery screen passes, but the margin is not large. The remaining 1.24^\circ must cover measurement uncertainty, oscillator aging, temperature, traffic patterns and implementation changes. If the same receiver is moved to a higher modulation order, longer packet, noisier reference or more mobile channel, the phase budget should be recalculated.

Phase Noise And EVM

For small random phase error, a useful screening relation is:

where \sigma_\phi is in radians. If RMS phase error is:

then the phase-noise contribution to RMS EVM is approximately:

Engineering Comment

This does not replace a full EVM measurement because EVM also includes thermal noise, distortion, IQ imbalance, quantization, equalization error, timing error and channel-estimation error. It is still a useful diagnostic. If power-based SNR is acceptable but measured EVM is poor, phase noise and carrier tracking should be investigated with the same seriousness as noise figure or interference.

Loop Bandwidth Tradeoff

Carrier recovery is a feedback problem. A narrow loop can reject phase noise but may leave residual frequency offset, Doppler or oscillator drift. A wide loop can track frequency changes quickly but may convert noise, interference or decision errors into phase motion.

The correct loop behavior depends on the waveform:

• burst receivers need acquisition fast enough for the preamble or training sequence;
• mobile receivers need tracking that can follow Doppler and fading dynamics;
• fixed microwave links usually prefer stable tracking and low residual phase noise;
• OFDM receivers need frequency offset control because subcarrier orthogonality is sensitive to residual offset;
• high-order QAM needs lower residual phase error than lower-order modulation.

The loop setting should therefore be a system decision, not a hidden firmware constant. It affects throughput, error rate, fallback behavior and operations diagnostics.

Validation Evidence

Carrier recovery should be validated with both physical-layer and service evidence:

Evidence	Why it matters
acquisition range and time	proves the receiver can lock from expected offsets
residual frequency offset	shows whether packet-level rotation remains bounded
residual phase error or phase-noise estimate	checks constellation stability
EVM by modulation mode	exposes decision-quality loss
BER, FEC and packet-error trends	connects phase stability to service behavior
lock, unlock and cycle-slip events	reveals recovery and false-lock risk
temperature, supply and reference-clock tests	exposes drift and oscillator sensitivity
spectrum and interferer tests	separates carrier tracking from RF impairment

The strongest diagnosis comes from agreement between evidence streams. If received power is stable while EVM, FEC corrections and carrier-lock alarms degrade together, the problem is more likely to be carrier recovery, phase noise or oscillator reference quality than simple path loss.

Common Failure Modes

Common carrier-recovery failures include:

1. oscillator frequency offset outside acquisition range;
2. loop bandwidth too narrow for Doppler, drift or burst acquisition;
3. loop bandwidth too wide, allowing noise or interference to modulate the carrier estimate;
4. reference-clock phase noise that dominates the constellation;
5. cycle slips during fades or low-SNR intervals;
6. poor preamble, pilot or decision-directed tracking under the selected mode;
7. firmware or FPGA latency changes that destabilize the loop;
8. accepting an aggressive modulation and coding mode from SNR alone.

These failures often appear as unexplained EVM growth, burst FEC corrections, intermittent loss of lock, constellation rotation, or mode fallback that seems more severe than the link budget predicts.

Practical Acceptance Rule

A practical acceptance rule is:

1. estimate worst-case initial carrier offset and acquisition time;
2. measure residual frequency offset after lock;
3. budget residual phase drift, RMS phase noise and static phase bias;
4. compare the result with a modulation-specific phase allowance;
5. verify EVM, BER, FEC and packet behavior under realistic load and environment;
6. document alarms and fallback thresholds for operation.

The result should say whether the selected modulation can tolerate the carrier-recovery error with margin. A digital receiver is not validated by SNR alone. It is validated when energy margin, carrier phase margin, symbol timing margin and service evidence agree.