When to Prototype, When to Simulate: Hard Lessons from RF Product Development
Simulation is fast and cheap — until it isn't. A practical framework for deciding when to spin a board and when to trust the model.
Every RF program faces the same fork: trust the simulation, or build hardware. The wrong call costs months. Here's the framework we use with clients.
Simulate First When the Physics Is Well-Bounded
Linear passive networks, well-characterized amplifiers in their linear region, and antenna patterns in free space all simulate beautifully. Tools like ADS, AWR, and HFSS will get you within a dB of measured behavior if your models are good. For these problems, prototyping first is wasted money.
Prototype Early When Reality Bites
Anything involving real-world clutter, multipath, hand effects, near-field interactions with enclosures, or non-linear behavior at compression — simulate to inform the design, then put hardware on the bench fast. We've seen 6 dB surprises from a plastic enclosure that "shouldn't matter."
The Hybrid Approach
Most successful programs interleave: simulate to bound the design space, prototype to validate the riskiest assumption, then iterate the model with measured data. The model becomes more trustworthy with each cycle, and by the time you're at engineering verification, simulation is genuinely predictive.
Budget for Two Spins
Single-spin success in RF is the exception. Plan and budget for at least two PCB iterations on any non-trivial RF design. Teams that pretend otherwise either pad their schedules silently or ship compromised products.