Spectrum Analyzer PCB: RF Design and Manufacturing
Why the Board Sets the Noise Floor
In a spectrum analyzer, the printed circuit board is not a passive carrier for the design. At gigahertz frequencies the geometry of a trace is part of the circuit: its width sets the impedance, its length sets the phase, its surface roughness sets the loss, and the ground plane beneath it decides how much of the signal leaks into the neighbouring channel. Two instruments built from the same schematic and the same components can differ in noise floor and phase noise purely because one board was laid out with more attention than the other.
That is the reason a spectrum analyzer board is treated as a radio-frequency design first and a digital design second, even though a modern instrument contains a substantial amount of digital processing.
The Signal Chain as It Appears on the Board
The receive path starts at the input connector and passes through protection and attenuation before reaching the front-end amplifier. A low-noise amplifier sets the noise figure for everything downstream, which is why its ground, its supply decoupling, and the short, straight trace leading into it are laid out before any other part of the board.
The amplified signal then enters the mixer, where it is combined with a local oscillator. The oscillator itself usually comes from a phase-locked loop with a voltage-controlled oscillator or a synthesizer, and its reference has to be kept away from anything that switches or radiates, because reference contamination appears directly as phase noise.
Filtering follows, whether with surface-acoustic-wave devices or lumped elements, and the signal is then digitised by a high-speed converter and processed by a field-programmable gate array or a digital signal processor. The analog and digital halves of that chain have very different requirements, and most of the layout work is in keeping them from interfering with each other.
Material Selection
Standard FR-4 is adequate below roughly two gigahertz, where its loss and its variation in dielectric constant are tolerable. Above that, the dielectric loss dominates and the material choice becomes a performance decision rather than a cost one.
PTFE-based laminates and filled hydrocarbon materials such as Rogers 4350B are the common answers, with dielectric constants around 3.5 and loss factors an order of magnitude below FR-4. The stability of the dielectric constant across temperature matters as much as its absolute value, because a drifting constant shifts the impedance of every trace on the board and with it the calibration of the instrument. Where a design mixes materials, the stackup has to keep the low-loss material in the RF layers and confine standard FR-4 to the digital and power layers.
Stackup for a Radio Board
Four layers suit a simple RF module, six layers suit a mid-range instrument, and eight to twelve layers appear in laboratory-grade equipment that combines a wideband RF chain with substantial digital processing. The organising principle is the same at every count: every RF trace needs a solid ground plane directly beneath it, and the plane should be continuous from the connector to the mixer without a split anywhere along the path.
Microstrip is used on the outer layers where components must be mounted, and stripline appears on inner layers where shielding from both sides is worth the extra dielectric loss. Symmetry matters in the stack because it keeps the board flat through lamination and reflow, and via transitions are used sparingly, with back drilling where a stub would resonate inside the band of interest.
Layout and Shielding
Fifty-ohm impedance held to within five percent is the baseline requirement, and it applies to the launch at the connector, the traces between stages, and the fan-out to each component. Any change in width along the way is a discontinuity, so the transitions are tapered or compensated rather than left to chance.
Radio paths are kept short and direct, and the components along them are arranged in a line rather than scattered, because a short path has less loss and fewer opportunities to couple into something else. Amplifier stages are separated from each other with ground vias and, where the isolation requirement is severe, with metal shields soldered to the ground plane.
The analog and digital grounds are kept apart in layout and joined at a single defined point, so that the return currents of the fast digital section do not flow under the RF front end. Supply rails are decoupled at each stage with capacitors placed as close to the pin as the layout allows.
What the Fabricator Has to Hold
Fine-line etching on a low-loss material is harder than on FR-4 because the copper surface is smoother and the etchant behaves differently, so the fabricator must control the etch to keep the impedance inside tolerance. Impedance test coupons travel with the panel and are measured with time-domain reflectometry, and the results are supplied with the board rather than assumed.
Surface finish is chosen for the assembly that follows: ENIG is flat and robust for fine-pitch parts and shielded cans, and immersion silver offers lower loss where the assembly schedule allows. Optical inspection, X-ray inspection for hidden joints, and electrical test complete the sequence, and on RF work the microsection is worth looking at because the dielectric thickness it reveals explains the impedance the coupon measured.
Assembly Is Half the Problem
Radio assemblies are less forgiving than ordinary digital boards. Components are small, tolerances are tight, and the laminate conducts heat differently from FR-4, so the reflow profile has to be developed for the material rather than inherited. Placement accuracy matters for the fine-pitch converter and the amplifier, and shield cans must be soldered with continuous contact to the ground plane or the isolation they were meant to provide is lost.
After assembly comes alignment and calibration. Connector launches, filter responses, and gain stages are measured and trimmed, and the results are only as stable as the board beneath them. A board that drifts with temperature will not hold its calibration regardless of how carefully it was tuned, which is why material stability and mechanical rigidity matter as much as electrical performance.
Where the Cost Comes From
Low-loss laminates cost several times what FR-4 costs, and they are bought in smaller volumes with longer lead times. Fine-line processing on those materials adds yield risk, and impedance testing, X-ray inspection, and shield assembly add work. Small quantities of a complex board carry a high unit price because setup dominates, and the same board in production volume falls steeply once the process is stable.
The savings that matter usually come from getting the design right the first time. A stackup agreed with a fabricator that builds RF boards regularly, an impedance target that the material can hold, and a layout that keeps the RF and digital sections apart are worth more than any negotiation on a quotation. A partner able to handle PCB manufacturing and the PCB assembly that follows, and to supply an impedance report with the panels, removes most of the risk from that handover.
FAQ
Can a spectrum analyzer board be built on FR-4? Below about two gigahertz, yes. Above that the dielectric loss and the variation in dielectric constant make a low-loss laminate necessary.
What impedance tolerance is normal? Plus or minus five percent for RF boards, which requires controlled etching and impedance coupons measured during production.
Why is the ground plane so critical? It provides the return path that defines the impedance of every trace above it, and any split in it creates a discontinuity and a radiation path.
Is back drilling needed? Where via stubs would resonate within the frequency band of interest, yes. It removes the unused barrel that would otherwise cause a notch in the response.
How is the assembly different? The profile must be developed for the laminate, placement is tighter, and shield cans have to be soldered with continuous ground contact.
Conclusion
A spectrum analyzer board is a radio-frequency design in which the laminate, the stackup, the trace geometry, and the shielding all participate in the measurement. Choosing the material from the frequency range, keeping the return path continuous, holding impedance inside five percent, and separating the analog and digital sections are the decisions that determine whether the instrument meets its noise floor. Everything else is execution. For related topics, see our notes on PCB design and layout and PCBA testing.
