Electromagnetic Simulation of Your Routed Board, From the Flux Chat

It is Thursday, and Jonas has to order the first boards of a 2.4 GHz sensor on Monday. The layout is done: an edge-mount SMA connector, a grounded coplanar feed of about 20 mm, a matching network, and a chip antenna at the board edge. The design checks are clean. One question is still open: is the feed from the connector to the matching network really 50 Ω, now that it has ground pours on both sides, stitching vias, and a 1.5 mm substrate?

Before, this question took a day. Jonas exported the board, opened a field solver, rebuilt the stackup, drew the ports by hand, set up the mesh, and waited. Then he copied the results into a slide for the design review. Most of the time, he did not do it, and the first prototype was the test.

Now Jonas asks Flux: "Simulate RF_FEED from the SMA connector J1 to the antenna matching network with 50 Ω ports, from 0.5 to 4 GHz. Show S11, S21, and the input impedance at 2.45 GHz." Flux reads the routed copper and the stackup of the board, puts the ports on the feed, and runs the simulation. A few minutes later, the answer is in the chat: the charts, the values at 2.45 GHz, and the board, unchanged. Jonas sends the link to his lead and orders the boards.

A full-wave simulation of the board you routed

The Flux agent runs a full-wave electromagnetic simulation of the real board: the copper that you routed, the layers and dielectrics of your stackup, the ground pours next to the trace, and the return plane under it. It does not use an ideal model of the line. It simulates the geometry that goes to the factory.

You do not set up the solver. Flux finds the net, puts a port at each end of the trace, picks the mesh for the stackup, and sizes the simulation box. It checks each result for physical sense, for example that a passive feed does not give out more power than it gets. Then it shows the charts and the full report in the chat: the effective settings, where each port sits, and the values at the frequencies you named.

A real board: the feed of a 2.4 GHz antenna

We ran Jonas's question on a real 2.4 GHz RF test board in Flux. The board has an edge-mount SMA connector, a grounded coplanar feed on the top layer with a ground pour on each side and stitching vias along it, a solid ground plane on the bottom layer, a matching network, and a chip antenna. Flux simulated the feed from the SMA pad to the first part of the matching network. The result:

• S11 at 2.45 GHz: −33.7 dB. Less than 0.05 % of the power reflects back to the connector.
• S21 at 2.45 GHz: −0.1 dB. The signal goes through the feed with almost no loss.
• Input impedance at 2.45 GHz: 51.9 + j0.9 Ω. The feed is on its 50 Ω target.

The impedance chart shows the whole band. The feed stays between 44 and 54 Ω from 0.5 to 4 GHz, so the match holds across the Bluetooth, Wi-Fi, and Zigbee channels, and well past them.

Then Jonas asked for a field map at 2.45 GHz. Flux showed the electric field 500 µm above the top copper and in a vertical cut along the feed. The field stays on the feed and in the gaps to the ground pours next to it, from the connector to the matching network. This is how a good coplanar feed keeps its energy.

What you can ask Flux to simulate

• Reflection and loss. S11 and S21 of a routed net between ports that you choose, over the frequency range that you need.
• Impedance. The input impedance of a feed over frequency, and the TDR impedance along a trace.
• Crosstalk. How much of a fast signal couples into the nets next to it.
• Differential pairs. SDD11, SDD21, and the differential impedance of a USB, Ethernet, or LVDS pair.
• Time domain. The step response of a net, and an eye diagram at the bit rate that you choose.
• Field maps. The electric and magnetic fields around a net, above the board and in a cut through it. You see where the energy goes.

One simulation, every role on the team

Jonas is one example. Here is how other people on a hardware team use it.

The RF engineer: two feed widths, one thread

Aisha designs the radio section of a smart lock. She asks Flux for S11 and the input impedance of the antenna feed from 2 to 3 GHz. Then she makes the feed trace a little narrower, asks Flux to run it again, and compares the two charts in the same thread. She keeps the width that is closest to 50 Ω.

The high-speed designer: the camera clock and its neighbors

Marco routes the camera interface of an ESP32-S3 camera module. The 24 MHz pixel clock runs 42 mm from the camera connector to the ESP32-S3, between the data lines D4 and D5. He asks Flux: "Check that the ground plane is unbroken under PCLK, D4, and D5. Then give me the impedance of PCLK and its crosstalk into D4 and D5, from 10 MHz to 2 GHz."

We ran this on a real camera board in Flux. First, Flux checked the copper under the three traces and their six pads: the ground plane under them is unbroken. Then it ran three simulations on the routed board, and all three are valid. The result:

• PCLK impedance: 50.6 Ω along the line. The clock line is on its 50 Ω target.
• Crosstalk at 24 MHz: −47 dB into D4 and −48 dB into D5. At the clock frequency, the neighbors get less than 0.5 % of the voltage.
• Crosstalk near 900 MHz: −21 dB. The coupling grows with frequency, so the fast edges of the clock couple more than the clock itself.

Now Marco knows two things before he orders the board. The routing of the clock line is good. And if he sees noise on D4 or D5, the edge rate of the clock is the first thing to change, not the routing.

The hardware designer: a USB 2.0 pair from the connector to the bridge

Priya designs a USB sensor hub. The USB 2.0 pair runs about 27 mm on the top layer, from the USB-C connector to the USB bridge chip, with no vias and a solid ground plane under it. The layout tool sized the pair for 90 Ω. She asks Flux: "Simulate USB_D+ and USB_D- as a differential pair from the connector to U1, from 1 MHz to 3 GHz. Show SDD11 and SDD21."

We ran this on a real USB board in Flux. The run is valid. The result:

• At 240 MHz, the fundamental of 480 Mbps: SDD11 −23.7 dB and SDD21 −0.2 dB. The pair delivers the signal to the bridge chip with almost no loss.
• Up to 1 GHz: SDD21 stays above −0.4 dB. The fast edges get through too.

The chart also shows the pair above the USB 2.0 band: SDD11 rises to about −12 dB near 1.5 GHz. Priya has the full picture of her pair before she orders the board.

The mechanical engineer: a smaller board, the same feed

Lena makes the enclosure of a wearable sensor 4 mm shorter, and the board outline changes with it. The router moves the antenna feed around the new outline. Before she signs off on the enclosure, she asks Flux to simulate the new feed and compare it with the old one. The electrical team sees the result in the project, without a meeting.

The mechanical engineer: a USB pair around a mounting boss

Tobias moves a mounting boss of a handheld meter closer to the USB-C connector, so the USB pair must route around the new hole. He asks Flux for the differential impedance and SDD21 of the new route, and an eye diagram at 480 Mbps. He decides with the result of the real route in front of him.

The hardware lead: proof for the design review

Priya leads the hardware team. Before each build, she asks Flux to simulate every controlled-impedance net on the board and to write one summary file with the charts. In the design review, the team looks at the results of the real routed board, not at a calculator value.

Prompts to try

• "Simulate the antenna feed from the SMA connector with 50 Ω ports, and show S11 and the impedance at 2.45 GHz."
• "Show me a field map of the RF feed at 2.45 GHz, above the board and in a cut along the trace."
• "What is the TDR impedance along the clock trace from the oscillator to the MCU?"
• "Show the differential impedance and SDD21 of the USB pair."
• "Make an eye diagram of the data line at 1 Gbps."
• "How much does the SPI clock couple into the analog input next to it?"

How it works

Open a project with a routed board and ask your question in the chat. Name the net, the ports, and the frequencies, or let Flux pick them from the design. Flux runs the simulation in the cloud, so your computer stays free, and you can keep working while it runs. Every result stays in the thread, and Flux can save the report as a project file, so the next person sees what was simulated and what came out.

Try it

Open a routed board in Flux and ask for the impedance of the feed that you worry about most. Then ask for the next one.