How does gain flatness affect modulated signals?

A wideband modulated signal like 256QAM OFDM occupies many subcarriers across the channel bandwidth. If the amplifier gain varies across this bandwidth, different subcarriers receive different amplification. The subcarriers at the gain peaks are amplified more than those at the dips, distorting the constellation diagram and degrading EVM. For 256QAM with a required EVM of minus 32 dB, the gain flatness across the channel bandwidth must be better than 0.5 dB peak-to-peak. For 1024QAM: better than 0.3 dB. This is why wideband digital communications systems specify gain flatness as tightly as they specify noise figure and linearity. Gain slope (a systematic tilt from low to high frequency) is equally harmful and can be corrected with a simple linear equalizer, while gain ripple (periodic variation) requires more complex equalization.

How is gain flatness measured?

Connect the device under test between a vector network analyzer's ports and measure S21 across the operating bandwidth. The gain flatness is the difference between the maximum and minimum S21 values: flatness = max(S21) minus min(S21) in dB. Some specifications split this into gain slope (the linear component of the variation from low to high frequency) and gain ripple (the periodic or non-linear component). The measurement resolution bandwidth must be narrow enough to resolve any ripple: if the ripple period is 50 MHz, the VNA frequency step must be less than 10 MHz to capture the peaks and valleys accurately. Temperature and input power level both affect gain flatness; specifications should state the temperature range and power level at which flatness is measured.

How can gain flatness be improved?

Several techniques address different types of gain variation. Gain slope is corrected with a passive equalizer: a simple resistive network that attenuates more at the higher-gain frequencies, flattening the response at the cost of overall gain reduction. Gain ripple from impedance mismatches between cascaded stages is reduced by improving the match (using attenuator pads between stages, at the cost of noise figure and gain). Digital predistortion systems in modern transmitters include frequency-dependent correction tables that compensate for the entire chain's amplitude and phase variation. Temperature-induced gain drift is corrected with digital gain control tables indexed by temperature sensor readings.

RF Design

Gain Flatness

A 5G base station transmitter amplifies a 100 MHz channel with 256QAM OFDM. The PA datasheet says 30 dB gain with ±0.5 dB flatness. In reality, the gain is 30.5 dB at the channel center and 29.5 dB at the edges: a 1 dB peak-to-peak variation. The subcarriers at the edges receive 1 dB less amplification than those at the center, distorting the constellation. The EVM degrades from −38 dB (meeting the 256QAM spec) to −30 dB (failing it). A passive slope equalizer, a simple 3-element RC network, tilts the response by +1 dB across the band, restoring flatness to ±0.15 dB. The EVM returns to −38 dB. Gain flatness is not a secondary spec; it directly determines modulated signal quality.

Category: RF Design

Spec: Peak-to-peak dB across BW

Impact: EVM, constellation quality

Gain Flatness Requirements by Modulation

Modulation	Required EVM	Max Gain Flatness	Channel BW	Application
QPSK	−15 dB	±1.5 dB	5 to 20 MHz	Satellite, IoT
16QAM	−20 dB	±1.0 dB	20 MHz	LTE
64QAM	−25 dB	±0.75 dB	20 to 100 MHz	LTE-A, WiFi 5
256QAM	−32 dB	±0.5 dB	100 MHz	5G NR, WiFi 6
1024QAM	−35 dB	±0.3 dB	160 MHz	WiFi 6E, 5G-Adv
4096QAM	−38 dB	±0.2 dB	320 MHz	WiFi 7

Gain flatness definition:
GF = max(S₂₁) − min(S₂₁) dB (across specified BW)

EVM degradation from gain ripple:
EVM_ripple ≈ 20·log(10^GF/20 − 1) dB
1 dB flatness: EVM ≈ −25.3 dB contribution
0.5 dB flatness: EVM ≈ −31.5 dB contribution

Cascaded system flatness (worst case):
GF_total = √(GF₁² + GF₂² + ... + GF_N²) (RSS)

Common Questions

Frequently Asked Questions

Impact on modulated signals?

Unequal subcarrier amplification distorts the constellation. 256QAM needs ±0.5 dB across channel BW. 1024QAM: ±0.3 dB. Gain slope (tilt) is fixable with a passive equalizer. Gain ripple needs complex digital correction in the DPD system.

How is it measured?

VNA S₂₁ sweep across BW. Flatness = max − min. VNA step must resolve ripple (50 MHz ripple needs <10 MHz steps). Specify at stated temperature and power level. Separate slope from ripple for equalization planning.

How to improve flatness?

Slope: passive RC equalizer (simple, costs gain). Ripple from mismatch: attenuator pads between stages (costs NF/gain). Digital: DPD with frequency-dependent correction tables. Temperature: gain tables indexed by temp sensor.

Related Terms

← Frequency Multiplier Group Delay →

Signal Quality

Gain Flatness Impact Calculator

Enter modulation order, channel bandwidth, and measured gain flatness. Compute the EVM contribution from gain ripple and determine if equalization is needed to meet the standard's requirements.

Check EVM Impact