A signal that appears only with both blocks active needs a different question
A prototype produces an unexpected RF tone when two circuit blocks operate together. Either block alone looks acceptable. A partition between them may appear to be the obvious next step, but the observation does not yet identify a field leaking from one block into the other. The new frequency could arise through nonlinear interaction in the product, a changed operating condition, or distortion in the measurement chain.
Before commissioning a multi-cavity shielding frame, establish what creates the tone and where the relevant signals meet. This investigation can turn an uncertain metal change into a targeted decision: interrupt a demonstrated coupling path, change the circuit conditions that generate the product, or correct a misleading measurement. The aim is not to diagnose intermodulation from one spectrum screenshot.
Build an operating comparison that preserves the rest of the product
Compare the permitted states with block A active, block B active, and both active. Include an appropriate inactive baseline when the architecture allows it. Use the same observation point and product assembly. Keep the shield position, cables and receiver settings comparable so that the state change remains interpretable.
Disabling a block can also alter the shared power load, clock distribution, receiver gain or software activity. Record those consequences at the time of the comparison. A regulator entering a different mode when one radio is disabled can produce a new noise pattern without the two RF signals mixing. Likewise, a receiver that changes its gain when a transmitter starts is no longer being observed under the same condition.
Where possible, use an approved control that separates the suspected RF activity from those other changes. The system owner must determine whether such a state is valid. Do not disconnect a required load or force an undocumented mode merely to produce an apparently cleaner comparison.
Use frequency relationships to form a testable hypothesis
When multiple frequencies reach a nonlinear element, additional frequency components can result. Candidate intermodulation products include sums and differences of the input frequencies and their integer multiples. A tone near one of those combinations is a clue, not proof of its origin or the location of the nonlinearity.
List the actual active source frequencies, including relevant clocks or local oscillators where known. Compare the unexpected tone against a small set of plausible relationships. Do not search indefinitely through high integer combinations until an accidental match appears. Frequency accuracy, source drift and the measurement resolution limit how strongly an apparent match supports a hypothesis.
If the product permits a controlled change to one source frequency, predict how the candidate product should move before running the comparison. Keep the other source and the observation method unchanged as far as practical. A reproducible movement matching the prediction strengthens the frequency relationship; it still does not establish whether the mixing occurred on the PCB or inside the analyzer.
Exclude a measurement-chain product before redesigning the shield
A spectrum analyzer contains nonlinear stages too. Strong signals can generate internal distortion, especially when sensitivity settings expose its front end to large levels. The test arrangement therefore has to be assessed as a possible source of the unexpected tone.
Use the instrument manufacturer's procedure to compare suitable attenuation and gain conditions while keeping the product operating consistently. Interpret the readings at a common reference plane, with external losses and displayed correction settings accounted for. Simply inserting a component and seeing every displayed signal fall does not separate product distortion from instrument distortion.
Track the main signals, the suspect tone and the visible noise floor together. If the tone becomes unresolvable after attenuation, the result may only show inadequate measurement sensitivity. A credible check needs enough range to observe the relevant signals without overloading the receiving chain. No universal attenuator setting applies to every analyzer, probe or product.
Include external preamplifiers and active probes in this review. Passing the analyzer's internal overload check does not demonstrate that an upstream accessory remained linear. Preserve the physical coupling arrangement during instrument comparisons; moving a near-field probe changes the signal presented to the entire chain.
Find where the necessary signals can meet
After the measurement survives those checks, draw a focused path from each candidate source to the suspected nonlinear region. The relevant meeting point could be an active input, a shared circuit connection or another interface requiring investigation. Do not assign a nonlinear property to a shield contact merely because it is nearby.
Separate two questions: where is the additional frequency generated, and how does that component reach the affected circuit or observation point? A partition may reduce a source reaching a susceptible input. It may also reduce an already generated signal escaping from a region. Those are different interventions and can require different boundaries.
Follow connected routes as well as space between components. Shared supply or reference impedances can couple circuit activity through paths that a vertical wall does not interrupt. A field probe showing a strong signal above a component identifies a useful observation location; it does not independently prove the complete route into the victim.
Compare a partition against the specific hypothesis
A useful partition trial changes the proposed field path while preserving the operating state and other important interfaces. Its result should be assessed against the predicted source and victim behavior, not only against the height of one displayed peak. Temporary conductive material can change grounding, loading, clearances and cable placement at the same time.
If a trial lowers the candidate source at the suspected input and the unwanted product falls consistently, the case for that boundary becomes stronger. If only the observed radiation changes while the functional problem remains, the trial may have altered the escape path instead. If supply-related behavior changes at the same time, the comparison needs further separation.
Use a shield drawing review once the electrical hypothesis identifies which regions must be separated and which connections cross the proposed wall. A manufacturable wall needs a defined relationship to the PCB and lid, but those geometric decisions follow the diagnosed task. They cannot establish that task by themselves.
Two hypothetical outcomes lead to different next steps
Consider a fictional board with two independently selectable RF sources. An unexpected tone appears only when both run. A permitted frequency adjustment moves the tone as predicted, but a properly interpreted analyzer attenuation comparison shows that the receiving chain contributes the dominant displayed product. The next step is to obtain a valid measurement, not to approve new shield tooling from the original trace.
In a second hypothetical case, the measurement remains stable across suitable instrument conditions. A source-frequency change supports a particular mixing relationship, and controlled observations associate the problem with a susceptible input region. A temporary boundary changes that input exposure and the functional symptom together without changing the circuit mode. This supports a more detailed boundary trial, while leaving the exact nonlinear mechanism for circuit-level confirmation.
Neither example provides a universal signature or a predicted attenuation result. They show why the same initial observation, two blocks active and a new tone present, can justify very different engineering work.
Choose the correction at the point supported by the evidence
Possible corrections include reducing unwanted coupling into a nonlinear stage, changing filtering or operating conditions through the circuit owner, or revising the physical shield boundary. The choice depends on the demonstrated chain. A larger enclosure does not inherently prevent two signals already sharing an electrical route from interacting.
When a partition remains justified, evaluate a representative shield sample with the intended PCB contacts, lid and operating assembly. Repeat the relevant state combinations and functional checks. A favorable local spectrum is useful evidence for that configuration; it is not a substitute for the product's broader RF or EMC acceptance work.