LNA Input Matching on RO4350B: Noise Figure Versus Return Loss

The source impedance that minimizes an LNA's noise figure is not generally the impedance that maximizes gain or gives the best board-level input return loss. On a RO4350B board, choose the matching network against a receiver noise budget, gain, stability and the required input match. A deep S11 notch by itself is not evidence of a low-noise receiver.

This guide extends the amplifier PCB overview with an LNA-specific decision: how to compare noise matching and power matching without confusing their reference planes. It does not prescribe a component network for an unspecified device.

Three targets, three different questions

Analog Devices' noise-matching treatment explains optimization using noise parameters and source impedance. Use a device's noise data and S-parameters at the same frequency, bias, temperature and reference impedance. S-parameters alone do not supply Γopt, minimum noise factor Fmin, or equivalent noise resistance Rn.

The input coefficient of a bilateral device depends on the output termination: Γin = S11 + S12S21ΓL/(1 − S22ΓL). Consequently, blindly using conjugate(S11) as the source target assumes a particular output condition. Keep the matching network and device models at consistent planes before comparing gain and noise.

A small, reproducible noise-parameter example

Consider an entirely hypothetical small-signal device at 2.4 GHz. Assume NFmin = 1.00 dB, Rn = 10 Ω, Z0 = 50 Ω, and Γopt = 0.4 + j0. These are invented teaching inputs, not measurements or specifications of a named LNA. No physical bias point or temperature dependence is modeled; the example uses the conventional 290 K source-noise reference.

Use the standard noise-parameter equation F = Fmin + 4(Rn/Z0)|Γs − Γopt|² / [(1 − |Γs|²)|1 + Γopt|²]. Here F and Fmin are linear noise factors; NF = 10 log10(F) in dB. Convert 1.00 dB to Fmin = 10^0.1 before adding the mismatch term. All three Γs values below are real and are referenced to 50 Ω at the device input looking back toward the source.

For the middle row, the added noise factor is 0.8 × 0.2² / (0.96 × 1.4²) = 0.017007. Adding this to 1.258925 and converting to dB gives 1.0583 dB. The source impedance is Zs = 50(1 + Γs)/(1 − Γs), so Γs = 0.2 corresponds to 75 Ω at this plane.

Moving from Γs = 0 to Γs = 0.4 lowers the calculated device noise figure by about 0.22 dB in this example. It does not tell us the board's return loss or gain: no S-parameters or matching-network model were supplied. In particular, −20 log10|Γs| is not the assembled board's input return loss. This distinction prevents a source-pull result from being mislabeled as an S11 measurement.

Include loss before selecting the network

A more elaborate match may improve the impedance presented to the device but add component or interconnect loss. In a separate matched cascade example, a passive input network at 290 K with loss L has noise factor L and gain 1/L. Friis' relation gives Ftotal = L × FLNA; in dB, its insertion loss adds directly to the LNA noise figure. Thus a 0.30 dB input loss before a 1.00 dB LNA produces 1.30 dB total noise figure under those assumptions.

That cascade example must not be added mechanically to the earlier mismatched source-pull rows. For the actual circuit, simulate the lossy network and noise parameters together, including mismatches and physical temperature. Compare performance at the receiver input plane, not just the ideal device plane.

What RO4350B changes—and what it cannot determine

The Rogers material datasheet provides laminate properties, not an assembled LNA noise figure. Line geometry, copper, pads and dielectric construction affect the network presented to the input pin. Include these structures in an EM model and combine it with component models that retain Q, parasitics and self-resonance.

Keep the sensitive input path short enough to meet its loss budget, provide the intended ground return, and assess coupling from clocks, digital nets and bias circuits. Use the thermal-conductivity guide to distinguish laminate properties from the complete heat path. Temperature changes can affect device operating point and noise; laminate choice alone cannot remove those effects.

Before adding an external match, check the part architecture. ADI AN-2622 notes that many LNAs are internally matched. Start from the selected device's recommended application circuit and determine whether external tuning is intended. RO4350B is not automatically the best material for every low-frequency receiver or millimeter-wave module.

Validate the tradeoff at the correct planes

1. Set noise figure, return loss, gain, linearity and stability requirements over the operating band, bias and temperature ranges.
2. Sweep candidate networks with device noise data, S-parameters, component tolerances and the actual board geometry. Check out-of-band stability as well as in-band matching.
3. Measure S-parameters and noise figure using defined fixtures, output loading and calibration planes. Record losses ahead of the DUT and apply appropriate corrections with their uncertainty.
4. Confirm results on the assembled receiver, including filtering, protection, enclosure and supply behavior.

Keysight's noise-figure measurement documentation describes noise-parameter measurements and measurement setup considerations. A VNA return-loss trace and a noise-figure result answer different questions; keep both in the validation report.

Prepare the amplifier PCB inquiry

Follow the device-reference-to-board workflow. Provide device package and footprint, RF band, proposed stackup, finished copper and mask details, impedance requirements, board dimensions, quantities, dissipated power and thermal structure. Add coupon requirements and clarify whether assembly or RF testing is included.

Use the RFQ preparation page, then request an amplifier PCB quotation from PCBSync. Specify the deliverables required from the fabricator; a board quotation does not itself certify receiver noise figure, matching or stability.