PNA / SA bench
Two-tone envelope, IMD3 / IP3, P1dB, and THD. Power is per tone at the DUT plane unless noted.
One setup feeds all three sections below. The per-tone level and impedance give the envelope, the frequencies give the product spectrum, and the measured IM3 gives the intercepts. Changing the reference plane converts the tone level by the gain.
Two equal CW tones add in voltage at envelope peaks. Peak envelope power is +6.02 dB vs one tone — the DUT sees twice the CW VOPP you would calculate from a single tone at that dBm. That is why a two-tone IMD sweep compresses sooner than a CW P1dB sweep at the same per-tone power.
Odd-order products sit on a uniform grid of spacing Δ: order 2k+1 lands at f1 − kΔ and f2 + kΔ. Type frequencies with a unit or SI prefix (1G, 100 MHz); bare numbers use the default unit. Edit f2 or Δ and the other follows.
Optional. Whichever floor is highest is the one that limits the IM3 measurement; the analyzer's own third-order products usually dominate, which is why input attenuation changes the answer. The tone level here is at the analyzer, after whatever loss sits between it and the DUT.
| Order | Product | Frequency | Offset from f1 | In DUT band |
|---|
The envelope beats at Δ, so bias and video paths need several times Δ of bandwidth. If the upper and lower IM3 products differ, sweep Δ and watch the asymmetry: that is a memory effect, not a measurement error.
dBc is relative to one output tone; absolute IM3 dBm is measured at the DUT output. With the tone level at the DUT input, IIP3 = Pin,tone + Δ/2; at the output, OIP3 = Pout,tone + Δ/2. OIP3 = IIP3 + G. Use the small-signal cubic region, below compression. Measurement reference.
OP1dB = IP1dB + G0 − 1 dB. If you enter a measured Pout at the same Pin, compression is (Pin + G0) − Pout. On a PNA, G0 is small-signal |S21| after power cal at the DUT plane.
Optional. Enter the fundamental frequency to place the harmonics, and a device corner frequency to estimate how much the DUT's own rolloff hides. The correction assumes the nonlinearity comes before the band limit and the device is not slewing; a feedback amplifier moves the other way, because loop gain falls with frequency too.
The fundamental places every harmonic at n·f0. The analyzer maximum flags harmonics you physically cannot reach, and the passband flags harmonics the device was never meant to pass.
The corner and pole count are a model of the response shape, and they convert a measured harmonic into an estimated intrinsic one. They cannot replace the passband: relative attenuation saturates at 20p·log₁₀(n), so one pole can never account for more than 6 dB on H2 however far above the corner you drive.
| n | Frequency | Measured dBc | Band-limit | Intrinsic dBc | Notes |
|---|
THD = √(Σ 10Hn/10) with Hn in dBc. A single −40 dBc harmonic is 1% THD. Harmonic dBm = Pfund + dBc. Receiver harmonics on a VNA can fake DUT THD — check with a notch or a spectrum analyzer when it matters.
Order-of-magnitude only
| Item | Thumb | Why it bites you |
|---|---|---|
| Two equal tones | PEP = Ptone + 6 dB | Envelope Vpp is 2× the CW VOPP at that per-tone dBm |
| IMD3 vs P1dB | OIP3 ≈ OP1dB + 10 dB | Cubic class-A estimate; real parts wander several dB |
| Tone step | +1 dB tones → +3 dB IM3 | 3rd-order: ΔIM3 = 3 ΔPin. If it is not ~3:1, you are not in the cubic region |
| Diff two-port | VOPPdiff = 2 × SE | Same per-port dBm, twice the swing across the pair |
| THD vs H2 | −40 dBc ≈ 1% | THD is RSS of all harmonics, not just H2 |
Static page. Calculations stay in your browser.