Satellite payload operators have been migrating to higher-order modulations for years. DVB-S2's 16APSK and 32APSK, the higher-tier modes in DVB-S2X, and similar formats in proprietary HTS uplink standards all carry more bits per symbol than QPSK. They also impose significantly tighter constraints on the PA linearity that the uplink terminal must provide.
This article examines the relationship between output intercept point (OIP3), the modulation order in use, and the required back-off at the PA output. It covers how to size OIP3 for a given link configuration, where GaN-on-SiC changes the available design space relative to GaAs, and what the OIP3 specification on a PA datasheet actually tells you versus what you need to compute yourself. Falcomm design targets where mentioned are design goals, not production specifications.
Why OIP3 Becomes the Binding Constraint in High-Throughput Links
In a satellite uplink carrying QPSK, the PA can run relatively close to P1dB. QPSK has low peak-to-average power ratio (PAPR), typically 0-2 dB, and the modulation quality requirement (EVM or SNR) is relaxed enough that moderate PA compression is acceptable with digital pre-distortion. The binding PA constraint in a QPSK link is usually output power, not linearity.
In a 32APSK link with a spectral efficiency of 4.5-5 bits/symbol, the situation is different. The constellation requires the PA to resolve amplitude differences of roughly 1 dB between the inner and outer ring. AM-AM compression at moderate output back-off directly distorts those amplitude separations, producing EVM that degrades bit error rate. PAPR for a filtered 32APSK signal is typically 3-5 dB, and the instantaneous envelope peaks must be reproduced linearly to avoid clipping-induced spectral regrowth into adjacent channels.
The out-of-band spectral regrowth is where OIP3 enters. Third-order intermodulation products from a broadband signal fall at offsets equal to the channel bandwidth, producing interference into adjacent satellite channels that the operator must meet as a spectral mask requirement. The IMD product level relative to the carrier is directly related to OIP3: IMD3 = 2 * (OIP3 - P_out). For a required carrier-to-IMD3 ratio of 30 dBc at a given output power, OIP3 must exceed P_out by 15 dB.
Sizing OIP3: From Link Budget to PA Specification
The link budget for a high-throughput satellite uplink specifies the required carrier-to-noise-and-distortion ratio at the satellite receiver. The distortion contribution from the uplink PA appears as an additive interference source. The total interference budget at the satellite includes thermal noise from the uplink path, interference from adjacent uplink terminals, and the IM products from the transmitting terminal's PA.
For a typical Ka-band HTS uplink with a 30 dBc carrier-to-interference requirement attributable to PA IM products, and a terminal PA output power of 34 dBm, the required OIP3 is 34 + 15 = 49 dBm. This calculation uses the two-tone approximation, which slightly underestimates the required OIP3 for a wideband modulated signal. A more conservative sizing adds 1-3 dB to the two-tone result to account for broadband IMD behavior.
The OIP3 requirement then constrains the allowed output back-off relative to P_sat. A PA with OIP3 = 49 dBm and P_sat = 38 dBm operates at 11 dB below OIP3. The corresponding P_out at 30 dBc IM3 suppression is approximately OIP3 - 15 dB = 34 dBm, which is 4 dB below P_sat. This 4 dB back-off figure is the system operating margin: the PA can deliver 34 dBm output with adequate linearity, and P_sat of 38 dBm is the ceiling.
If the required OIP3 is 49 dBm and the available PA has OIP3 = 45 dBm, the operating point must shift to maintain the 30 dBc IM3 requirement. P_out would need to fall to 30 dBm, an 8 dB reduction, which either reduces the link budget (lower EIRP) or requires a higher-gain antenna to maintain the required E/N0 at the satellite. Neither is free.
OIP3 vs Output Power: The GaN Advantage in Practice
The design space for PA selection in a satellite terminal is defined by OIP3 at a given output power, with efficiency as the third axis. A PA that delivers OIP3 = 49 dBm at P_out = 34 dBm with high efficiency is more valuable than one that requires higher P_sat (and therefore higher DC power) to achieve the same OIP3 at the same output level.
GaN-on-SiC devices show a favorable OIP3 vs P_out relationship at Ka-band compared to GaAs, arising from two factors. First, the higher supply voltage allows the device to deliver a given P_out at a lower fraction of its P_sat. A GaN device with P_sat of 40 dBm delivering 34 dBm output is operating 6 dB below saturation. A GaAs device with P_sat of 36 dBm delivering 34 dBm output is operating 2 dB below saturation. At lower back-off, the GaAs device is further into compression and its OIP3 at that operating point is lower than its small-signal two-tone measurement would suggest.
Second, the AM-AM curve for GaN devices operating at moderate back-off is generally flatter than for GaAs at the same relative operating point. The third-order Taylor coefficient of the transfer function, which determines OIP3 in the small-signal model, is related to how rapidly the gain begins to compress near P1dB. A flatter AM-AM curve means the gain is still close to small-signal gain at 3-4 dB back-off, and the IMD products at that back-off are lower than the simple OIP3 extrapolation would predict.
We are not saying GaN is more linear per unit gate periphery than GaAs across all operating conditions. The claim is more specific: at the combination of output power level and back-off ratio required by high-order modulation satellite links, GaN-on-SiC devices operating at their optimized bias point consistently show higher OIP3 at a given output power level than comparable GaAs designs, while also dissipating less power to achieve that output level.
What the Datasheet OIP3 Actually Tells You
OIP3 on a datasheet is measured with a two-tone CW test using tones separated by a fixed offset (typically 1-100 MHz depending on the application band). The measured result is the extrapolated third-order intercept from the fundamental and IM3 power levels at a specific input power. Several things can make this number optimistic for a real satellite uplink scenario.
First, two-tone OIP3 is measured at a single frequency pair, typically at band center. OIP3 varies across the frequency allocation, often by 2-4 dB from center to band edge at Ka-band. A terminal operating at band edge may see lower OIP3 than the datasheet center-frequency figure.
Second, two-tone OIP3 assumes both tones are at equal power. In a multi-carrier or pre-filtered OFDM uplink, the spectral loading is distributed across the channel bandwidth. The effective OIP3 under wideband signal loading differs from the two-tone result because the intermodulation products from multiple spectral components add coherently at some frequencies and incoherently at others.
Third, OIP3 is bias-dependent. It can be improved by operating the device at higher quiescent current (deeper class A bias), which also reduces efficiency. A datasheet that quotes OIP3 at a bias point specifically optimized for linearity may not reflect performance at the bias point you would select for efficiency. Always confirm that the OIP3 and efficiency figures on the same datasheet are measured at the same bias condition, not at two different operating points selected to make each individual spec look better.
Applying This to the FA-2400 Series Design
For the FA-2400 series, OIP3 sizing is driven by the 32APSK and DVB-S2X uplink requirements for LEO terminal applications. Our design target places OIP3 at sufficient margin above the P_out operating point to satisfy the spectral mask requirements without requiring active DPD at the terminal, which reduces terminal complexity. Whether a given terminal design uses DPD is the terminal designer's choice; our goal is to make DPD optional rather than required for basic spectral compliance.
This is a design target, not a certified measurement. It is based on our simulation work using our GaN-on-SiC device models and the published OIP3 performance ranges for comparable device geometries in the literature. We will characterize and publish actual OIP3 data as our evaluation process advances. Until then, the target gives a sense of the design intent and the linearity-versus-efficiency tradeoff we are making in the FA-2400 architecture.