EMC SIPI Technical Discussion Group Meeting

On December 11th, we successfully organized our first EMC+SIPI Technical Discussion Group session. This is an exclusive, small community where members share hands-on experience and lessons learned in a trusted environment.

The purpose of the group is to exchange ideas, enhance practical design expertise, and foster meaningful professional connections within a social and collaborative setting.

Meeting notes from the session are shared below. If you are a Senior or Principal Engineer and are interested in joining this group, please feel free to reach out to us.

Date: 11/12/2025

Venue: Telonic Instruments, Wokingham

Attendees:

Min Zhang

Rod Macpherson

Art Shchatsko

Karthik Guruchandran

Arun Khilnani 

Phyu Phyu Khine

Steve W

Doug discussing the new R&S equipment they represent 

Notes:

Min – Case Studies on EMC Issues

Case 1 – 144 MHz Failure

Case 2 – Wi-Fi Module on Motherboard

Case 3 - SPI line between Microcontroller and Flash

Across all three case studies, transmission-line impedance matching did not appear to be the dominant factor. While board-level shielding or stripline routing of clock and communication signals provided some benefit near the fundamental clock frequency, it had little impact on higher-order harmonics, which were responsible for the majority of the EMC failures.

The evidence indicates that grounding quality and clock signal drive strength are the primary contributors to the emissions, rather than transmission-line mismatch. Excessive drive strength (fast edge rates) excite common-mode currents and harmonics that are not effectively mitigated by impedance control alone.

Reducing edge rate (e.g. through FBs) and implementing low-impedance grounding paths were consistently more effective than transmission-line matching in reducing radiated emissions.

Min consulted Dan Beeker, who suggested the power distribution network (PDN) was poorly designed. But it is very hard to prove. 

Case studies on DC-DC converter 200-300 MHz Radiation

Power architecture:

24 V → 12 V → 3.3 V SMPS chain.

A sensor was powered from the 12 V rail.

Broadband noise observed around 277 MHz.

Connecting the sensor’s metal can directly to the ground plane increased noise, likely due to a new common-mode noise path.

Suggested actions:

Perform near-field probing to locate the ~200 MHz noise source.

Externally excite the unpowered board and look for SWR dips.

For this frequency range, the broadband response is most likely due to resonance caused by the layout rather than the component itself. Min needs to take all suggestions into account and perform additional measurements.

Karthik SIMBEOR for PCB simulations

Karthik was doing his presentation on SIMBEOR

Used for high-speed digital signals greater than 15 Gbps on PCBs.

HFSS and CST are not very useful for this specific application. Keysight ADS may be a more suitable alternative.

A case study of a pair of signal vias with surrounding ground stitching vias shows that the impedance is not constant up to 70 GHz.This approach can be used to optimise the number of ground stitching vias and the number of PCB layers.

The entire PCB can be imported into Simbeor for simulation.

Simbeor simulations provide a good level of accuracy in approximately 7 minutes, compared to many hours for HFSS, and Simbeor is also faster to set up. 

Prepreg exhibits worse RF performance than the core material because compression during lamination changes the epoxy content within the stack-up.

MMPX connectors are guaranteed up to 50 GHz; the manufacturer provides guidance on how to connect to them. They are purely surface-mount connectors.

HFSS costs approximately £40k. Simbeor costs approximately £12k. Altium and other PCB tools can estimate propagation delay, but these estimates do not account for via stitching and related effects.

Art Shchatsko – Medical Device EMC Design (32-bit MCU)

Art presenting his PCB design 

This case study described the EMC design of a PCB for a large, metal-enclosed Class I medical device used for cleaning surgical instruments. Although the system had many external cables, it contained no high-frequency circuitry and was not safety-critical.

A connector-centric EMC strategy was used. All I/O pins (signals, power, and 0 V) were protected directly at the connector with ferrites, X2Y capacitors, and bi-polar TVS diodes. After filtering, connectors were intentionally placed off the ground plane, with the ground plane cut away beneath them to prevent noise coupling into the PCB.

The power input used a dedicated filter chain, starting with high-voltage (330 V) MLCCs in parallel with TVS devices, followed by a two-stage LC filter.

A 6-layer PCB stack-up was defined early and optimised for low impedance using via-in-pad:

L1: Components and some signals

L3: Power sub-plane

L5: Signals

L6: 0 V and decoupling

To handle external transients, 16 spring contacts around the rear PCB perimeter provided a low-impedance connection to the metal chassis, efficiently shunting EFT energy away from the electronics.

The design passed without requiring any PCB re-spins.

Key Takeaways

The PCB stack-up drives the entire EMC performance and must be defined early.

Early collaboration between electrical, mechanical, and software teams is critical.

Bench testing with realistic cables, loopbacks, and dummy loads should be done early.

Early EMC testing significantly reduces project risk and rework.

Arun – Issues in DC Output Power Electronics

DC grids are increasingly used for power distribution.

CISPR 16 specifies measurements down to 9 kHz, but there are no established standards for DC grids.

The concept of impedance is not well defined for pure DC, particularly in the context of noise measurements.

A conventional LISN cannot be used on a DC grid, as it introduces resonance and results in a non-constant impedance at very low frequencies.

Automotive systems and data centres currently use DC grids; servers typically operate at around 300 V DC.

Arun is researching methods to measure noise on DC grids and identifying the equipment required to perform accurate and repeatable measurements.

Dinner