You run the simulation.
The impedance looks right.
The insertion loss is acceptable.
The layout passes review.
Then the first PCB arrives.
And the measured result is not what you expected.
This happens more often than many engineers would like to admit.
The immediate reaction is often:
“Something must be wrong with the simulation.”
Sometimes it is.
But quite often, the bigger problem is that the simulation and the manufactured PCB are not actually describing the same physical structure.
That difference is where many high-frequency PCB problems begin.
The PCB in Your CAD Tool Is Not the PCB You Receive
In a design tool, a transmission line might look like this:
Trace width: 0.18 mm
Copper thickness: 35 μm
Dielectric: 0.20 mm
Target impedance: 50 Ω
Everything looks precise.
But manufacturing introduces real physical dimensions.
The finished trace may have a different profile after etching.
The dielectric thickness may differ from the nominal value.
Copper thickness can change through plating.
Registration can affect the relationship between layers.
The actual laminate may have electrical properties that differ from the simplified value used during simulation.
None of this means the manufacturer has produced a “bad PCB.”
It means that the real board has physical variables that the original model may not have fully represented.
A 50 Ω Trace Is Not Just a Width
One of the easiest mistakes is to think:
“I need a 50 Ω trace, so I just need the correct trace width.”
Not quite.
The impedance depends on the complete transmission-line structure.
That includes:
- Trace width
- Copper thickness
- Dielectric thickness
- Dielectric properties
- Reference-plane position
- Trace geometry
- Surface characteristics
- Manufacturing tolerances
Change one of these variables and the electrical result can change.
This is why a trace width calculated for one stackup cannot simply be copied to another board construction.
The Stackup Can Quietly Change Your Result
Imagine that your simulation assumes a signal layer is 0.15 mm above the reference plane.
During fabrication, the actual finished dielectric spacing is different.
The trace itself has not changed in the CAD file.
But its electromagnetic environment has changed.
That means the impedance can change too.
This is particularly important when the design has limited electrical margin.
For high-frequency boards, the stackup should therefore be treated as an electrical design parameter, not simply a mechanical layer list.
Vias Are Another Common Surprise
A signal trace can look perfect from the top view.
Then it changes layers.
Suddenly the signal encounters:
- A via
- A pad
- An antipad
- A reference-plane transition
- Possibly a via stub
At lower frequencies, these details may be relatively forgiving.
At higher frequencies, they can become part of the transmission path.
This is why a design can have excellent straight-line impedance and still show an unexpected discontinuity around a layer transition.
The lesson is simple:
Don't simulate only the trace. Simulate the important transitions too.
Copper Is Not an Ideal Rectangle
CAD software makes copper look beautifully simple.
Real copper is not.
The fabrication process can affect:
- Trace width
- Sidewall shape
- Copper thickness
- Surface roughness
- Plating distribution
For ordinary signals, these variations may have limited impact.
For demanding RF and high-speed applications, they can become part of the loss and impedance budget.
This is one reason material selection alone does not guarantee high-frequency performance.
A low-loss laminate combined with poorly controlled geometry is still a problematic transmission structure.
Then There Is the Material Problem
A material datasheet may provide a Dk value.
But engineers should be careful about treating that number as a universal constant.
The effective dielectric behavior used in an actual PCB structure can depend on:
- Frequency
- Test method
- Resin content
- Glass construction
- Lamination
- Direction
- Manufacturing construction
This is particularly important when a simulation relies on a single simplified dielectric value.
The more demanding the design, the more important it becomes to understand what material data was actually used in the model.
The Prototype Should Close the Loop
A prototype should not simply answer:
“Does the circuit turn on?”
For a high-frequency design, the prototype can answer a much more useful question:
“How closely does the manufactured physical structure match the model?”
Depending on the application, useful validation can include:
- TDR
- VNA measurements
- Insertion loss
- Return loss
- Impedance testing
- Cross-section analysis
- Critical-dimension inspection
The goal is not to collect as many test reports as possible.
The goal is to find out where the simulation and hardware begin to disagree.
The Manufacturer Should See the Important Electrical Requirements
A common mistake is to send a manufacturer only the fabrication files and assume the files explain everything.
For a demanding high-frequency PCB, it can be useful to clearly communicate:
Target impedance
Critical signal layers
Reference planes
Stackup
Material
Copper requirements
Impedance tolerance
Via requirements
Special testing
This gives the manufacturing team context.
For example, a trace that appears visually unimportant may actually be part of a critical 50 Ω RF path.
A manufacturer cannot reliably prioritize what is electrically critical if that information is never communicated.
The Real Problem Is Usually the Gap Between Teams
Many high-frequency PCB failures are not caused by one dramatic mistake.
They happen because several small assumptions are never aligned.
The designer assumes one dielectric thickness.
The manufacturer builds another within its process capability.
The simulation assumes one copper geometry.
The finished board has another.
The RF engineer expects a particular connector transition.
The mechanical design changes the launch geometry.
Each individual decision may appear reasonable.
Together, they can produce a board that behaves differently from the original model.
A Better High-Frequency PCB Workflow
Instead of:
Design
↓
Simulation
↓
Gerber
↓
Manufacturing
↓
Problem
try:
Electrical requirements
↓
Preliminary stackup
↓
Manufacturer review
↓
Simulation
↓
Layout
↓
DFM / impedance review
↓
Prototype
↓
TDR / VNA / electrical validation
↓
Production
The important change is not adding more steps.
It is moving manufacturing feedback earlier in the process.
That can prevent an expensive redesign after the layout is already finished.
What I Would Check Before Ordering the First High-Frequency PCB
Before sending the design to fabrication, I would ask five questions:
1. Is the stackup actually achievable by the selected manufacturer?
2. Are the dielectric and copper dimensions used in simulation realistic for the finished board?
3. Which traces and transitions are electrically critical?
4. How will controlled impedance be verified?
5. What measurements will be used to compare the prototype with the simulation?
If those questions do not have clear answers, the design may not be ready for production yet.
The Takeaway
A high-frequency PCB does not fail simply because the engineer chose the “wrong” material or calculated the “wrong” trace width.
Sometimes the bigger issue is the gap between the model and the manufactured structure.
Simulation describes an electrical model.
Manufacturing creates a physical object.
High-frequency PCB engineering is about making those two things agree as closely as practical.
That is why stackup definition, manufacturing tolerances, via transitions, material data, impedance control, and prototype measurement all matter.
If you are working on an RF, microwave, radar, telecom, wireless, or high-speed digital project, it is worth reviewing the complete high-frequency PCB manufacturing approach before releasing the final fabrication package.
For a broader reference covering high-frequency PCB materials, design considerations, and manufacturing capabilities:
High-Frequency PCB — Highleap Electronics
The best time to discover that your simulated PCB and your physical PCB are different is before you manufacture the first batch.
Top comments (0)