Ferrite Bead and Shield Can Co-Design for Conducted and Radiated EMI on Switching Power Rails
Engineering guide to integrating ferrite bead pi-filters with shield cans on DC-DC converter PCBs to meet CISPR 25 Class 5 and IEC 61000-4-3 limits.
Executive Summary
Switching power supplies on automotive ECUs, ADAS modules, and infotainment head units are the dominant source of conducted and radiated EMI failures during CISPR 25 Class 5 and ISO 11452-2 qualification. Designers reach first for ferrite beads on the input and output rails of buck and boost converters, but ferrite beads above their self-resonant frequency become resistive shunts with collapsing impedance, and they do nothing for near-field magnetic coupling from the switching loop itself. This application note explains why a ferrite bead pi-filter must be co-designed with a board-level shield can over the switching node, defines the geometric and material rules that produce ≥60 dB shielding effectiveness through 6 GHz, and specifies the POCONS two-piece shield can, SMD pan nut, and spring contact families that resolve the failure mode.
Technical Specifications & Attenuation Data
A correctly executed shield-can-plus-filter strategy contains both the differential-mode conducted noise on the rail and the radiated H-field from the switching loop. The conducted path is dominated by the ferrite bead's impedance profile combined with input and output ceramic capacitors forming a C-L-C pi network; the radiated path is governed by the shield can's seam construction, ground-pad return inductance, and aperture geometry.
| Parameter | Specification | Standard | |-----------|--------------|----------| | Shielding effectiveness, 200 MHz–6 GHz | ≥60 dB | IEEE 299, MIL-STD-285 | | Shielding effectiveness, 6–8 GHz | ≥45 dB | IEEE 299 | | Shield can material | Nickel silver C7521, 0.20 mm wall | ASTM B122 | | Sheet resistance, plated nickel silver | ≤5 mΩ/sq | ASTM B193 | | Shield can fence pad ground stitch pitch | ≤2.5 mm (≤λ/20 at 6 GHz) | IPC-2221B | | Spring contact resistance, gold-plated BeCu | ≤25 mΩ initial, ≤50 mΩ after 10k cycles | IEC 60512-2-1 | | SMD pan nut pull strength | ≥45 N axial | IEC 60068-2-21 | | Conducted emissions limit, 150 kHz–108 MHz | CISPR 25 Class 5 (–9 to +63 dBµV) | CISPR 25 ed.5 | | Radiated emissions, 30 MHz–2.5 GHz | CISPR 25 Class 5 vehicle limits | CISPR 25 ed.5 | | Bulk current injection susceptibility | 200 mA, 1 MHz–400 MHz | ISO 11452-4 | | Reflow compatibility, peak | 260 °C, 30 s above 217 °C | J-STD-020 |
The 60 dB target is not arbitrary. CISPR 25 Class 5 radiated limits in the FM and DAB bands sit 30–45 dB below typical unshielded buck converter emissions at the 5th–20th switching harmonic. A 30 dB margin on top of that floor is required for production variation, antenna factor uncertainty, and aging of seam contact. Anything less than 55 dB measured shielding effectiveness on the bench will not survive a chamber retest after thermal cycling.
Common Design Pitfalls
- Treating the ferrite bead as a low-pass filter. Root cause: bead datasheets advertise impedance at a single test frequency (commonly 100 MHz), but the impedance curve drops above the SRF as the parasitic shunt capacitance takes over. Observable consequence: radiated peaks at the 30th–60th switching harmonic exceed CISPR 25 Class 5 by 8–15 dB even with a "correctly sized" bead. Mitigation: select beads with SRF at least 3× above the highest harmonic of concern, place ≥4.7 µF X7R MLCCs on both sides of the bead, and never rely on the bead above 200 MHz — that is the shield can's job.
- Insufficient ground pad copper area under the shield can fence. Root cause: a narrow ground trace creates a high-inductance return path, and the shield can fence becomes an antenna driven by the very currents it is meant to contain. Observable consequence: cavity resonance at f = c / (2L) where L is the longest internal dimension, producing 10–20 dB peaks at predictable frequencies (e.g., a 25 mm can resonates near 6 GHz). Mitigation: ground fence pad must be ≥1.5 mm wide, continuous, stitched to an internal ground plane with vias every ≤2.5 mm, with no plane splits crossing under the fence.
- Switching node routed under the shield can perimeter. Root cause: the SW node is the loudest H-field source on the board; routing it within 2 mm of the fence couples directly into the seam. Observable consequence: shielding effectiveness collapses from 60 dB to 30 dB in the 100–500 MHz band. Mitigation: keep SW node and inductor body inside a keepout that is ≥3 mm from any fence wall, and route the SW return on the layer immediately below with a solid plane.
- Single-piece shield can without removable lid for rework or tuning. Root cause: post-EMC tuning (snubbers, additional ceramics, layout cuts) is impossible if the can must be desoldered. Observable consequence: schedule slip of 2–4 weeks per debug cycle in the chamber. Mitigation: specify a two-piece shield can with a frame soldered to the PCB and a removable lid retained by spring fingers or pan nuts, allowing iterative debug without thermal damage to the BGA underneath.
- Aperture sizing that ignores the highest harmonic. Root cause: ventilation holes or component-access windows sized for visual inspection rather than λ/20 of the worst harmonic. Observable consequence: directional radiation lobe through the aperture, +12 dB above the cavity-only emission. Mitigation: maximum aperture dimension ≤ λ/20 at the highest frequency requiring 60 dB attenuation; for 6 GHz that is 2.5 mm. Use multiple small apertures rather than one large slot.
PCB Footprint & Soldering Profile Guidelines
The shield can fence pad is the single most critical layout element. Specify a 1.5 mm wide solder pad with 0.15 mm soldermask defined opening on the inner edge and non-soldermask defined on the outer edge, giving the fence wall a controlled wetting line. Courtyard clearance to adjacent components must be ≥1.0 mm to allow rework station access; for designs using SMD pan nuts as lid retention, place the nut footprint on a 0.10 mm thicker copper pad to absorb the 45 N axial preload without lifting.
Stencil design follows IPC-7525B: use a 0.12 mm laser-cut stainless stencil with a paste aperture ratio of 90% relative to the pad to compensate for the long thermal mass of the fence wall. Aperture-to-pad area ratio should sit between 0.66 and 0.80; a 1:1 aperture starves the joint and produces voids visible as dark streaks under X-ray inspection.
Reflow profile per J-STD-020 and IPC J-STD-001H: preheat ramp 1.0–2.5 °C/s from 25 °C to 150 °C, soak 60–120 s between 150 °C and 200 °C, ramp to peak at 1.0–3.0 °C/s, peak temperature 245–260 °C, time above liquidus (217 °C for SAC305) 30–90 s, cooling ramp ≤6 °C/s. Two-piece shield can frames are qualified for two reflow passes; the lid is installed mechanically post-reflow and post-test, never reflowed.
Spring contacts and pogo pins specified for board-to-shield grounding follow IPC-7711/7721 rework guidance: do not hand-solder above 350 °C tip temperature for more than 3 s on the contact pad, and reflow the contacts in the same pass as the shield can fence to prevent oxidation of the gold plating.
Recommended POCONS Components
Custom Two-Piece Shield Cans — POCONS USA tools two-piece nickel-silver shield cans to customer DXF in 4–6 weeks, with frame heights from 1.5 mm to 12 mm and footprints up to 80 × 80 mm. The two-piece architecture solves Pitfall 4: the frame solders in the main reflow pass, the lid snaps in for chamber test, and the lid lifts for rework without thermal stress on the BGA underneath. Specify when the design must clear CISPR 25 Class 5 and requires post-build EMC tuning. Browse the family at /products/shield-cans/.
SMD Pan Nuts — Surface-mount pan nuts provide a threaded mechanical retention point on the PCB for lids, brackets, or test fixtures rated to 45 N axial preload. Use them on shield can lids that must survive automotive vibration profiles per ISO 16750-3 without spring-finger fatigue. Standard thread sizes M1.6, M2, and M2.5 in nickel-plated brass. See /products/smd-pan-nuts/.
Spring Contacts and Pogo Pins — Gold-plated beryllium copper spring contacts deliver ≤25 mΩ initial contact resistance and survive 10,000 mating cycles per IEC 60512-5-2, making them the correct choice for shield-to-chassis bonding, board-to-board RF grounds, and ATE fixtures verifying shielded module compliance. Standard pitches from 1.27 mm to 4.0 mm. See /products/spring-contacts/.
For boards where the bead-plus-shield strategy must hit CISPR 25 Class 5 on the first chamber visit, send the placement file and the harmonic budget to applications@poconsusa.com for a fence-pad and aperture review before tape-out.
Application note produced by POCONS USA engineering team. Contact applications@poconsusa.com for design review.
Frequently Asked Questions
Why does a ferrite bead alone fail to suppress switching noise above 100 MHz on a buck converter?
Ferrite beads exhibit a self-resonant frequency typically between 80 MHz and 300 MHz where the bead's parasitic capacitance dominates and impedance collapses. Above SRF, a 600 Ω @ 100 MHz bead may present <20 Ω, so radiated harmonics from a 2.2 MHz switcher (44th harmonic = 96.8 MHz) couple unattenuated. A shield can over the switching node is required to contain near-field magnetic coupling that the bead cannot address.
What attenuation does a two-piece shield can deliver from 200 MHz to 6 GHz?
POCONS two-piece nickel-silver shield cans deliver ≥60 dB shielding effectiveness from 200 MHz to 6 GHz when seam pitch is held below λ/20 of the highest harmonic of concern, with a fence pad ground stitch pitch ≤2.5 mm and 0.20 mm wall thickness. Performance degrades to ~40 dB above 8 GHz due to seam radiation.
What is the lead time and MOQ for custom two-piece shield cans with non-standard footprints?
POCONS USA produces custom two-piece shield cans at MOQ 1,000 units with 4–6 week lead time for tooled parts and 7–10 days for prototypes from existing tooling families. Send DXF or STEP of the keepout area plus the target attenuation band to applications@poconsusa.com for a no-charge DFM review.