ADL5602 RF gain block: matching, bias, and linearity testing

ADL5602 suits RF/IF gain stages from 50 MHz to 4 GHz, with a single 5 V supply, nominal fixed 20 dB gain, and internally matched 50 Ω ports. Internal matching simplifies application circuitry but excludes PCB traces, connectors, bias inductors, and external loads. Assess small-signal transmission, output compression, intermodulation, noise, and thermal conditions separately to determine design suitability.

Use input dynamic range to locate the gain stage first

Record target frequencies, signal bandwidth, average/peak input, preceding loss, allowed downstream level, and load matching. 20 dB means 100 times power gain and 10 times voltage ratio at equal impedance; it does not hold for arbitrary inputs. Datasheet P1dB, OIP3, and noise figure carry test conditions; one frequency's number cannot represent the entire band.

For example, a −20 dBm single-tone input gives approximately 0 dBm ideal small-signal output. Raising input to 0 dBm gives 20 dBm by linear extrapolation, at which point perform a compression sweep instead of continuing fixed-gain predictions. This example allocates instrument ranges, not an approved continuous output point. For modulated signals, include peak-to-average ratio and adjacent-channel requirements, checking whether instantaneous peaks reach nonlinear operation.

RFOUT carries both RF output and DC bias

In the basic connection, 5 V reaches RFOUT through a bias inductor, with input/output capacitors blocking DC. Do not treat RFOUT as DC-free or connect it directly to instruments that prohibit DC. The manufacturer's example uses 0.1 µF coupling capacitors and multistage supply bypassing; substitutions also require self-resonance and target-frequency impedance checks, beyond nominal capacitance.

Multiplying 5 V by the datasheet's typical 89 mA quiescent current gives approximately 0.445 W, indicating the DC input magnitude. Actual dissipation also depends on RF input/output and signal-dependent bias. Record actual supply current and component temperature; this static estimate is not the thermal input for all conditions.

Include pads and connectors in the 50 Ω channel

Calculate RF trace width from the actual stackup, transition smoothly between connector pads and traces, and minimize test-point stubs. Avoid long parallel input/output routes; maintain continuous local ground with vias connecting ground layers. Ground splits or openings beneath traces change return paths and can cause reflections even when widths match.

Soldering of the exposed pad and ground pins affects both thermal and high-frequency behavior. If temperature rises gradually and gain changes over time, check soldering, supply, and cooling first. When reusing an evaluation board, retain the external topology and review dielectric structure, connectors, and component packages. A top-layer screenshot alone cannot establish matching internal ground references.

Measure compression and two-tone intermodulation separately

For P1dB testing, increase input at a fixed frequency and compare measured output against small-signal extrapolation. Two-tone testing uses equal-amplitude nearby frequencies, recording fundamentals and third-order products. Ensure source isolation before the combiner and avoid spectrum-analyzer input compression. Change a known output attenuation to check whether instrument-generated intermodulation affects results.

Example: each fundamental is 5 dBm and one third-order product is −65 dBm, a 70 dB separation. Linear extrapolation gives OIP3≈5＋70/2＝40 dBm. This is an extrapolated intercept, not an output capability of 40 dBm. Measure where intermodulation exceeds the noise floor without obvious compression; record tone spacing, resolution bandwidth, input attenuation, and calibration planes.

Troubleshooting normal gain with excessive distortion

Check instrument through connections and known attenuation before connecting the board. If reducing analyzer input substantially lowers IM3, investigate instrument nonlinearity first. If supply voltage falls with increasing input, check supply capacity and bias-inductor drop. For abrupt changes in one band, check coupling components, connectors, and reflections. Change one factor at a time and retain original-revision reproduction results so incidental changes are not mistaken for fixes.

Confirm preceding loss and bandwidth before measuring noise figure

A receive chain needs a noise budget as well as gain. Cables, switches, and attenuators before the amplifier reduce its input signal and change system noise figure. A board gain-stage measurement is therefore not automatically the complete product's antenna-port value. Record calibration reference planes, noise-source state, preceding loss, and instrument compensation so results map to the actual chain.

If only observing output noise on a spectrum analyzer, resolution bandwidth, detector mode, and input termination also affect results. Hold these settings constant when comparing layouts. Replace the application signal with a 50 Ω termination first, then enable nearby digital and supply circuits individually. Discrete spectral lines appearing with particular activity usually require coupling-path investigation rather than attributing all noise to the component itself.

Substitution review should compare gain, return loss, compression, intermodulation, noise, and current at matching frequencies, plus pins, package, and bias. A lower noise figure may accompany different linearity headroom; do not select solely by one superior promotional specification. Complete input/output records are needed to decide between one stage, distributed multistage gain, or added filtering and attenuation between stages.

Control interstage isolation during cascade verification

With actual preceding and following stages, ports are no longer always ideal 50 Ω. Establish a baseline with attenuators or specified loads, then reconnect actual devices and compare gain, return loss, and intermodulation. If switching one stage causes oscillation or spurs, check reflections, isolation, and supply coupling instead of only increasing interstage distance.

Include filter insertion loss in the cascade dynamic-range budget. Before a gain stage it changes the noise budget; after it, the gain stage may encounter large out-of-band signals first. Choose placement from actual interference and linearity headroom. Record preceding/following stage states and terminations, retaining standalone evaluation-board and complete-product results separately.

For production verification, first select points representative of operating-band and power boundaries, then fix test-chain attenuation, cable losses, and calibration methods so operators produce comparable output-power results.

Can S-parameters directly establish modulated-signal quality?

S-parameters describe reflection and transmission under small-signal linear conditions; they do not replace modulation-error, adjacent-channel leakage, or compression tests. Conversely, one passing modulation test does not establish stability across the band. Deliver frequency, power, bias, board revision, and instrument settings. Recalculate dynamic range and thermal budgets before widening bandwidth or replacing components.

Define the output-power reference plane in deliverables

Different cable lengths can produce different instrument readings for the same board. State whether power is referenced to the component output, board-edge connector, or instrument, and which cable-loss curve provides compensation. If the fixture contains attenuation or filtering, retain its model and calibration files so fixture loss is not mistaken for gain changes.

For production sampling, choose points representative of operating frequencies and power, defining expanded tests for anomalies. Small-signal gain sampling and full linearity verification have different coverage; state these separately instead of labeling quick screening as full-band, full-power confirmation.