EM validation only. Do not change the schematic, the layout, the routing, the stackup, or the net properties. Do not ask me questions.

Use the EM simulator on the routed layout of this project. Run one differential-pair simulation for each of these pairs. Use the actual routed copper of this project and the stackup of this project:

1. ETH_TX_P / ETH_TX_N. Target: 100 ohm differential.
2. ETH_RX_P / ETH_RX_N. Target: 100 ohm differential.
3. USB_DP / USB_DM. Target: 90 ohm differential.

Use the same solver settings for every pair:
- Solver preset: balanced.
- Overrides: max_timesteps 200000, end_criteria_db -40, pml_cells 8.
- Frequency range: 0 to 1 GHz.
- Use the differential_pair setting (p_net and n_net), so that SDD11, SDD21, and Zdiff are reported.

For each pair, report:
- Zdiff in ohm (the value and the method: TDR or frequency domain).
- SDD11 and SDD21 in dB at 100 MHz and at 480 MHz.
- SCD21 (mode conversion) in dB at 100 MHz and at 480 MHz, if the simulator reports it.
- The routed length of the P net and of the N net in mm, and the P minus N mismatch in mm.
- The number of vias on each net of the pair.
- Solver status: did the energy reach -40 dB before max_timesteps (yes or no), the number of timesteps used, and the simulation run time.
- Passivity: is every |S| value at or below 1 at all frequencies (yes or no).

If a run does not converge, or if a result is not passive, mark that pair as INVALID and give the reason. Do not retune the geometry. Do not report an impedance value for an INVALID pair as a result.

Finish with one table that has one row for each pair, then stop.