Measurements, Testing, and Calibration Power and Signal Measurement Informational

What is the correct setup for measuring two-tone intermodulation distortion and IP3?

Two-tone IP3 (third-order intercept point) measurement characterizes the intermodulation distortion of an amplifier, mixer, or receiver by applying two equal-amplitude CW tones and measuring the resulting third-order intermodulation products. Setup: (1) Two signal generators: SG1 at frequency f1, SG2 at frequency f2, both at the same power level P_tone. Frequency spacing: f2 - f1 = 1-10 MHz (typical; wider spacing for wideband devices, narrower for narrowband). Both frequencies must be within the device passband. (2) Combining: use a resistive combiner (Wilkinson power divider) to combine the two tones. Important: the combiner must have high isolation between the SG ports (> 20 dB) to prevent the SGs from intermodulating with each other. A directional coupler or isolator on each SG output further prevents SG-to-SG coupling. (3) Connect: combined output → cable → DUT → cable → spectrum analyzer. (4) Power calibration: the per-tone power at the DUT input = P_SG - combiner_loss - cable_loss. A Wilkinson combiner has 3 dB loss per path (by design) plus 0.2-0.5 dB insertion loss. So P_tone_DUT ≈ P_SG - 3.5 dB. Verify with a power meter. (5) Measure: on the SA, observe the two fundamental tones at f1 and f2, and the IM3 products at 2f1-f2 and 2f2-f1. Measure the power of the fundamentals (P_fund) and the IM3 products (P_IM3). (6) Calculate IP3: IIP3 = P_tone_DUT + (P_fund - P_IM3)/2 (dBm). OIP3 = P_fund_out + (P_fund_out - P_IM3_out)/2. Example: P_tone at DUT input = 0 dBm. P_fund at SA = +15 dBm (15 dB gain). P_IM3 at SA = -25 dBm. Delta = 15 - (-25) = 40 dB. IIP3 = 0 + 40/2 = +20 dBm. OIP3 = +15 + 40/2 = +35 dBm.

Category: Measurements, Testing, and Calibration

Updated: April 2026

Product Tie-In: Power Meters, Spectrum Analyzers, Signal Generators

IP3 Measurement Techniques

IP3 is the most important linearity specification for components in multi-signal environments. Accurate IP3 measurement requires careful attention to test setup to avoid measurement artifacts that can make the DUT appear better or worse than it actually is.

Common Measurement Pitfalls

(1) SG-to-SG intermodulation: if the two signal generators couple through the combiner (insufficient isolation), they generate IM3 products at the combiner output before the signal reaches the DUT. These products add to the DUT IM3, making the DUT appear more nonlinear. Prevention: use a Wilkinson combiner with > 20 dB isolation + 6 dB attenuator pads on each SG output (adds 12 dB isolation). Verify: disconnect the DUT and look at the combiner output on the SA. The IM3 products should be < -80 dBc (below the fundamental level by at least 80 dB). If visible: add more isolation (attenuators or isolators). (2) Spectrum analyzer compression: as with P1dB measurement, the SA must not compress. Ensure the fundamental tones at the SA input are well below the SA 1 dB compression point (typically +10 to +20 dBm for most SAs). Use an external attenuator before the SA to reduce the input level. (3) SA dynamic range: the IM3 products are typically 30-60 dB below the fundamentals. The SA must have sufficient dynamic range to measure them. Use the minimum resolution bandwidth (RBW) that provides adequate signal-to-noise for the IM3 products. Narrower RBW = lower noise floor = better dynamic range (but slower measurement). (4) Tone power level: IP3 should be measured at power levels well below P1dB (in the small-signal region where the 3:1 dB slope of IM3 vs input power holds). A common practice: measure at P_in = IP1dB - 20 dB. If P1dB = +10 dBm: measure IP3 at P_in = -10 dBm. At higher power levels (near compression): the IM3 products deviate from the ideal 3:1 slope, and the calculated IP3 depends on the measurement power level.

Tone Spacing Effects

The tone spacing (delta_f = f2 - f1) can affect the measured IP3: (1) Memory effects: amplifiers with asymmetric bias networks or thermal time constants show different IM3 levels at different tone spacings. This is characterized by the "sweet spot" (a tone spacing where IM3 is minimized due to cancellation of different distortion mechanisms). (2) Narrow tone spacing (< 100 kHz): the IM3 products are very close to the fundamentals. SA resolution bandwidth must be narrow enough to separate them (RBW < delta_f / 3). Measurement is slow. (3) Wide tone spacing (> 10 MHz): the IM3 products may fall outside the device passband (if the device has a bandpass filter). Also: the gain flatness across 2× the tone spacing must be considered (if gain varies significantly between f1, f2, 2f1-f2, 2f2-f1: the simple IP3 formula is less accurate). (4) Standard tone spacings: for general-purpose testing: 1 MHz is common. For satellite transponder testing: 14 MHz (two carriers in a 36 MHz transponder). For cellular testing: per 3GPP specifications (specific test tones and power levels).

IP3 vs Input Power

In theory: for a weakly nonlinear device, IM3 power increases at 3 dB/dB of input power, while the fundamental increases at 1 dB/dB. The IP3 (extrapolated intercept point) remains constant regardless of measurement power level. In practice: (1) At very low input power: IM3 is below the noise floor of the SA and cannot be measured. (2) At moderate input power (10-30 dB below P1dB): the 3:1 slope holds and IP3 is approximately constant. This is the ideal measurement region. (3) At high input power (within 5 dB of P1dB): the IM3 products deviate from the 3:1 slope (they increase faster or saturate). The calculated IP3 from a single measurement at this power level is not representative. Best practice: measure IM3 at multiple input power levels (e.g., -20, -15, -10, -5, 0 dBm). Plot IM3 and fundamental power vs input power. Verify the 3:1 and 1:1 slopes. Extract IP3 from the linear (small-signal) region of the plot.

IP3 Measurement Equations

IIP3 = P_in + (P_fund - P_IM3)/2 dBm
OIP3 = P_out + (P_out - P_IM3)/2 dBm
IM3: 2f₁-f₂ and 2f₂-f₁
P_IM3 slope: 3 dB/dB of input (ideal)
OIP3 = IIP3 + Gain (dB)

Common Questions

Frequently Asked Questions

What is the difference between IIP3 and OIP3?

IIP3 (input IP3): the intercept point referred to the input of the device. IIP3 = P_in + delta/2, where delta is the difference between the fundamental and IM3 at the output. OIP3 (output IP3): the intercept point referred to the output. OIP3 = P_out + delta/2. Relationship: OIP3 = IIP3 + gain (dB). For a 20 dB gain amplifier with IIP3 = +10 dBm: OIP3 = +10 + 20 = +30 dBm. Convention: for receivers and LNAs, IIP3 is preferred (it indicates how strong an input signal the device can handle). For transmitters and power amplifiers, OIP3 is preferred (it indicates the linearity of the output signal).

How do I calculate cascade IP3?

For cascaded stages: the overall IP3 is limited by the individual stage IP3 values: 1/IIP3_total = 1/IIP3_1 + G1/IIP3_2 + G1×G2/IIP3_3 + ... (all values in linear power, not dB). In dB: use: 10^(-IIP3_total/10) = 10^(-IIP3_1/10) + 10^((G1-IIP3_2)/10) + 10^((G1+G2-IIP3_3)/10). The stage with the worst IIP3 relative to the signal level reaching it dominates. Usually: the mixer or the IF amplifier (which sees the highest signal levels due to pre-amplification) limits the cascade IP3. Design rule: the mixer IIP3 should be at least 15 dB above the maximum signal level at the mixer input.

What about IP2 (second-order intercept)?

IP2 characterizes the second-order nonlinearity that produces output at f1±f2 (sum and difference frequencies). IP2 is particularly important for: (1) Direct-conversion (zero-IF) receivers: the difference product at f1-f2 falls at baseband (DC to a few MHz), directly interfering with the desired signal. (2) Wideband receivers: second-order products from two strong signals can fall on a weak desired signal. IP2 measurement: similar to IP3 but measure the product at f1+f2 or f2-f1 instead of 2f1-f2. IIP2 = 2×P_in - P_IM2 + gain (note: the slope for IM2 is 2:1, not 3:1). Typical values: IIP2 = +40 to +65 dBm for receiver front-ends (much higher than IIP3 because even-order distortion is largely canceled in balanced/differential circuits).

Keep Reading