High-Speed Sensor Hub Bring-Up for Robotics

High-speed sensor hub bring-up for robotics Executive summary This case study (CS01) documents a rigorous bring-up and verification approach for a robotics “sensor hub” PCB that aggregates multiple sensors (e.g., cameras, IMUs, time-sync peripherals) and forwards synchronized data to a compute module over one or more high-speed interfaces. The work focuses on early risk reduction across signal integrity (SI), power integrity (PI), and electromagnetic interference (EMI), while explicitly addressing robotics-specific reliability drivers—vibration and connector strain—and supply-chain longevity through planned lifecycle alternates. The case study is structured around four pillars that repeatedly appear in high-speed board failures: (1) insufficient timing margin due to impedance discontinuities and skew, (2) degraded return-path continuity (plane splits, poorly stitched layer transitions), (3) instability from PDN impedance peaks, and (4) emissions that are discovered only at the end of a program rather than in pre-compliance. The approach combines controlled-impedance stack-up targets, partitioned power domains with measurable PDN targets, and a staged bring-up sequence (power → clocks → data links → sensors) with instrumented checklists and logs for traceability. 1 Meta title (en-US): High-speed Sensor Hub Bring-up for Robotics: SI/PI, EMI Pre-compliance, and Vibration- Robust PCB Practices Meta description (en-US): Analytical case study of robotics sensor-hub PCB bring-up: controlled- impedance stack-up, partitioned power, return-path continuity, EMI pre-compliance, and measurable SI/PI outcomes.

Engineering approach Stack-up options and controlled-impedance intent The stack-up strategy is driven by two constraints: preserve return-path continuity and achieve stable, reproducible impedance for critical links. Where feasible, high-speed pairs are routed on outer layers with an adjacent ground plane, consistent with manufacturer guidance that prioritizes adjacent reference planes for high-speed differential routing and cautions against routing choices that complicate control of the electromagnetic environment. 12 Table: stack-up options for a robotics sensor hub (comparative design trade study) Stack-up Typical Risks / failure modes to Strengths When to choose option routing use manage Limited isolation; fewer 4-layer Short high- “reference choices”; more Minimal lane (Signal / speed escape Lowest cost, simple congestion near count, short GND / PWR / + modest fabrication connectors; higher risk of traces, tight Signal) lane count plane splits/voids under BOM budget routes 6-layer Two solid reference Must manage plane Mixed high- (Signal / Balanced: planes enable segmentation carefully; speed + GND / Signal / high-speed + cleaner return via transitions need moderate PWR / GND / dense IO paths and better intentional return-path complexity Signal) plane pairing stitching sensor hub 8-layer More routing Multi-camera (Signal / Higher lane channels; improved Higher cost; more vias; hubs, heavy GND / Signal / count, better isolation between needs robust via strategy compute IO, GND / PWR / domain domains; easier to (anti-pads, back-drill if aggressive EMI Signal / GND / isolation keep reference needed) targets Signal) continuity The design intent behind selecting 6–8 layers is to make return paths local and predictable (minimize loop area), a consistent theme in high-speed practice and manufacturer guidance on plane voids and stitching. 13 Impedance targets and high-speed geometry rules Impedance targets are anchored to manufacturer-recommended values for representative high-speed interfaces used in embedded vision and sensor hubs. For example, one embedded platform layout guide specifies 90 Ω differential ±15% and 50 Ω single-ended ±15% for MIPI D‑PHY CSI‑2/DSI nets, along with strict skew guidance (intra-pair skew ~1 ps order) and a clear warning about reference planes: if power is used as a reference, add ~10 nF stitching capacitors between PWR and GND on both sides of the connection to provide a return path. 14 Analog Devices’ guidance for CSI‑2 D‑PHY recommends impedance-controlled 100 Ω differential routing and recognizes lane operation up to ~1.25 GHz (2.5 Gbps). 2 Table: representative impedance and timing targets for “sensor hub” interfaces | Interface class | Nominal target | Typical tolerance (design spec) | Margin-sensitive constraint | Primary source basis | |—|—:|—:|—| | MIPI CSI‑2 / D‑PHY | 90–100 Ω differential | ±15% (common starting point) | Intra-pair skew ~ps order; minimize vias; maintain reference continuity | Platform layout guidance for CSI‑2/DSI and D‑PHY routing practices 15 | | High-speed USB-class diff pairs (if present) | Differential pairs require strict geometry control | Governed by interface compliance, but routing discipline is similar | Avoid probe/test points on high-speed diff, isolate from clocks/switchers, control via stubs | TI high-speed differential routing rules and via guidance 16 | | LVDS / serialized video link classes (e.g., FPD-Link) | Controlled pair routing to reduce emissions and prevent errors | Interface-specific | TX placement near connector reduces skew; focus on “error free, low emission” constraints | FPD-Link PCB/interconnect guidelines 17 | Partitioned power domains with measurable PDN targets A robotics sensor hub is typically mixed-signal: sensors and references can be analog-sensitive, while compute-side interfaces and serializers are digitally noisy. This motivates partitioned power domains and well-defined grounding/decoupling points.

Manufacturer mixed-signal guidance emphasizes that noise between analog and digital ground planes affects noise margin and should be controlled; one cautionary criterion is that the voltage between the two ground domains should not exceed ~300 mV in the described context, and it recommends appropriate decoupling and careful domain grounding strategies. 18 Decoupling guidance also emphasizes that high-frequency noise is best reduced with low-inductance, close‑placed ceramic capacitors and that decoupling capacitors must connect directly to a low-impedance ground plane (short traces/vias) to be effective. 19 To convert “good decoupling practice” into a measurable engineering goal, this case study uses the target impedance framework for the PDN. A PDN design textbook defines target impedance as the maximum allowable PDN impedance such that rail noise does not exceed the noise specification, with the simple relation: ΔV Z = noise target I max-transient and provides an example: ±50 mV noise with 1 A worst-case transient implies ~50 mΩ target impedance.

The same source also notes that typical digital PDN noise allowances are often on the order of ~5% of supply, while more sensitive blocks (e.g., ADCs/PLLs) may require <1% PDN noise and bandwidth relevant up to multi‑GHz. 21 For modern high-speed systems, Keysight notes it is “not uncommon” for microprocessor-based systems to require PDN target impedance of <10 mΩ from DC up to a few GHz, which sets a realistic expectation for the measurement challenge even if a sensor hub’s own targets are less aggressive. 22 Power domain partitioning diagram (conceptual) [Robot Input Power] –> [EMI/Surge/Reverse Protection] –> [Primary DC/DC] | +–> [Digital Core Rails] –> (SoC/FPGA/Serializer) | | | +–> High-speed I/O rails (e.g., D-PHY power) | +–> [Sensor/Analog Rails] –> (IMU/ADC/Reference/Clock) | +–> Quiet reference/ clock island This diagram is implemented with the explicit rule that each domain has its own local high-frequency decoupling loop to a low-impedance reference plane, and domain coupling points (if any) are intentional and reviewable. 23 Return-path continuity and stitch strategy Return-path continuity is treated as a first-class constraint, not a “layout nice-to-have.” The design principle is to keep the return current close to the forward path (minimize loop area), because uncontrolled return paths increase both SI risk and emissions risk. 24 Two practical rules from manufacturer guidance are used: • If a signal must cross a plane split, place stitching capacitors across the split near the crossing to provide a return path, minimize loop area, and reduce the impedance discontinuity introduced by the split; one guideline recommends stitching capacitors ≤ 1 µF placed as close as possible to the crossing. 25 • When using a power plane as the reference plane for MIPI D‑PHY routes, add ~10 nF stitching capacitors between PWR and GND near the transition so the high-frequency return current has a local path. 14 Signal return-path illustration (layer transition risk) The figure below is representative of the failure mode: a differential pair transitions layers, but the return current lacks an adjacent reference continuity (missing nearby ground vias), forcing return current to detour and increasing loop area. 26 Bring-up sequence and gating criteria Bring-up is structured as a gated sequence that prevents “debugging everything at once.” It is aligned to common board bring-up guidance emphasizing early-stage visibility and controlled boot conditions (e.g., ensuring basic debug/UART paths exist). 27 On the sensor-side, MIPI CSI-2/D‑PHY initialization is treated as a staged procedure: initialize the PHY, program the host controller register set, confirm stop-state readiness, configure the sensor, and confirm clock reception—an example sequence is explicitly described in an NXP application note. 28 Bring-up flowchart (mermaid) flowchart TD A[Start: board arrives] –> B[Visual inspection + assembly sanity checks] B –> C[Continuity/shorts check on power rails] C –> D[Power-up with current-limited supply] D –> E{All rails within spec and sequenced?} E — No –> E1[Stop: isolate fault, rework, retest] E — Yes –> F[Clock bring-up: oscillators/PLLs/ref clocks] F –> G{Clocks stable (freq/ppm/jitter budget acceptable)?} G — No –> G1[Stop: debug clock tree, grounding, PDN noise] G — Yes –> H[Digital access: JTAG/UART/boot visibility] H –> I[High-speed link bring-up: PHY lock + basic traffic] I –> J{Link stable at target rate?} J — No –> J1[SI debug: TDR/eye/probing plan, via/return-path review] J — Yes –> K[Sensors: configure, stream, timestamp/trigger verify] K –> L[EMI pre-compliance scan + near-field hotspots] L –> M{Meets pre-compliance margin target?} M — No –> M1[Mitigate: return-path fixes, filtering, shielding, reroute next rev] M — Yes –> N[Exit: release bring-up report + rev plan] Deliverables and verification plan PCB manufacturing package The manufacturing package is designed to minimize ambiguity between design intent and fabrication output, and to enable measurable verification of high-speed constraints.

Minimum deliverables (fab + assembly + test): – ODB++ dataset (preferred when the manufacturer supports it) because it is an integrated product model describing “what” must be manufactured, covering fabrication, assembly, and test. 29 – IPC‑2581 dataset (preferred alternative or parallel deliverable) as a standardized exchange between PCB designer and manufacturing/assembly facilities. 30 – A netlist suitable for bare-board electrical testing using an IPC‑D‑356 family format, which is explicitly defined as a data format for transmitting bare-substrate electrical test information and supporting test point assignment/positioning. 31 – Explicit controlled-impedance requirements (net classes → impedance targets → tolerance → coupon requirements), anchored to the controlled-impedance design guide standard. 32 Bring-up checklist and log artifacts Bring-up artifacts are treated as “first-order deliverables” because they compress debug time and enable repeatability across revisions.

Table: bring-up checklist (condensed, test gates + evidence) Stage Gate condition Evidence captured Typical tools All rails in tolerance; no Rail voltages, inrush/current Bench supply, DMM, Power unexpected current draw vs time, thermal scan scope, IR camera Reference clocks stable; no Frequency measurement, Scope + phase noise/ Clocks gross jitter/phase jitter snapshots, clock enable jitter options (as instability sequencing available) Debug ports reachable; UART logs, JTAG scan chain Debug probe + UART Digital access boot visibility present report dongle 27 TDR traces; via stub High-speed Channel meets impedance confirmation; impedance TDR/TDT tools 33 physical layer and discontinuity limits coupon results High-speed Link stable at [target rate]; BER counters/logs, retrain Protocol analyzer/ functional error metrics stable counts FPGA counters Sensors stream Frame drop count, time-sync Application + logic Sensor layer deterministic frames; logs analyzer timestamps/trigger align This checklist is paired with a bring-up log template (date, board rev, firmware rev, measurement setup, pass/fail, deviation notes) to maintain traceability for the iteration timeline.

SI/PI risk notes and measurement plan The measurement plan is driven by two central ideas: (1) measurements validate models and reduce risk in high-speed design, and (2) TDR/VNA provide direct visibility into where the interconnect departs from the target behavior. 34 Table: key SI/PI risks and how they are verified Risk Why it matters Primary measurement Acceptance framing Return-path Discontinuities within Increases loop TDR discontinuity discontinuity at layer allowed envelope; no high- area, reflections, localization; near-field transitions or plane field hotspots at crossings EMI risk scan splits 13 Risk Why it matters Primary measurement Acceptance framing Stub length controlled (or Adds insertion TDR + structural Via stubs / resonant back-drilled) per routing loss; eye collapse review; back-drill structures guideline; eye margin at higher rates verification acceptable 35 Timing margin PDN impedance via PDN impedance below PDN impedance loss; potential 2‑port VNA technique; target impedance; rail noise peaks and rail noise jitter coupling transient rail noise within % spec 36 Reduces noise Keep domain-to-domain Differential ground Mixed-signal ground margin; potential noise within stated noise measurement at noise damage if constraints; verify critical points excessive decoupling topology 37 Table: test equipment and measurement methods (recommended baseline) Domain Instrument Method Why it is chosen TDR supports impedance/ Characteristic impedance, uniformity checks and finding SI TDR/TDT discontinuity location, discontinuities such as gaps in pair skew return paths and launches 38 2‑port VNA techniques enable VNA with 2-port PDN impedance profiling measurement well below 1 Ω PI low-Z fixtures into mΩ regime (down to ~mΩ), addressing PDN regimes of interest 22 Pre-compliance can use non- compliant tools if sufficient Spectrum analyzer Hotspot identification, EMI (debug) margin is applied; near-field + near-field probes mitigation feedback loops probing accelerates root cause ID CISPR 16-1-1 defines EMI receiver / Receiver characteristics EMI measurement apparatus spectrum analyzer aligned to CISPR 16-1-1; (alignment to characteristics; CISPR 32 defines meeting CISPR emissions context per standards) emissions requirements for MME criteria CISPR 32 Logic analyzer / Trigger/timestamp Confirms link + application-level Sensors protocol capture correlation; frame drop timing behavior (as applicable) profiling Bring-up iteration timeline The bring-up plan is intentionally iterative because high-speed issues frequently require one or more physical revisions to remove SI/PI “pathologies” that are not fully predictable from CAD alone.

Table: bring-up iteration plan (timeline template) Iteration Primary goal Typical activities Exit criteria Current-limited power-up; Prove power/clock/ clock validation; TDR on Rev A (first debug access; Stable boot + debug; no coupons and representative prototype) establish SI/PI critical SI discontinuities channels; first sensor link at baselines reduced rate Re-route patch wires; swap Isolate single Rev A.1 (rework termination options; add Repeatable high-speed dominant failure spin, if needed) stitching/decoupling where training/streaming mode designed Final routing changes: via Stable operation at Rev B (first Hit [target rate] and stub control, return-path [target rate]; EMI margin “integration- EMI pre-compliance stitching, domain isolation; [Y] dB; documented ready”) margin pre-compliance scan measurement report Rev C Lock manufacturing Second-source footprints/ Released manufacturing (manufacturing- package and qualification; finalize dataset + obsolescence ready) alternates IPC-2581/ODB++ handoff plan The iteration count [N] and timing are project-dependent; the table is included as an engineering-control artifact rather than a schedule guarantee.

Outcomes with quantitative placeholders and realistic benchmark ranges Because this is a publishable case study template, outcomes are given as placeholders to be filled with measured results; each placeholder is paired with realistic ranges and how those ranges are grounded in the referenced literature.

Table: quantitative outcome placeholders and what “typical” looks like Measurement Metric Placeholder Typical realistic range and why method For CSI‑2 D‑PHY, one manufacturer guide cites operation up to ~2.5 Gbps (1.25 GHz), while Sustained Link counters + [target another platform guidance frames ~1 GT/s per high-speed protocol rate] lane as a design point; a realistic robotics target lane rate validation often falls within ~1–2.5 Gbps/lane depending on sensor format and lane count. 40 Measurement Metric Placeholder Typical realistic range and why method Expect at least one deliberate “return-path Return-path continuity” intervention in early spins (stitching Layout review + integrity fixes [N] vias/caps near layer transitions or plane splits), TDR + near-field applied because vendors explicitly document it as a scan mitigation when splits are unavoidable. 41 Target impedance is defined as ΔVnoise / Imax- transient; a worked example gives 50 mΩ for 50 mV / 1 A. For fast digital systems, “typical” 2‑port VNA PDN PDN target allocations are often ~5% rail noise, with more impedance + [X] mΩ impedance sensitive blocks driven toward <1% and transient rail bandwidth extending into multi‑GHz; Keysight noise notes <10 mΩ targets are not uncommon for microprocessor-class PDNs. 42 A PDN text provides a rough sensitivity example Clock/jitter on the order of ~1 ps jitter per mV of PDN noise Period jitter/ impact from [Y] ps p‑p in a cited scenario, illustrating why PDN control phase noise + rail PDN noise can dominate timing margin in high-speed noise correlation systems. 21 Pre-compliance is intended to uncover problems early; vendors note non-compliant equipment can be used if sufficient margin is applied, and pre-compliance workflows include near-field Spectrum scan + EMI pre- probing + spectrum analysis to identify near-field probes; compliance [Y] dB dominant coupling paths. Exact margin targets compare to CISPR margin are program-specific, but the case study framework enforces that the margin must be explicitly specified and justified in the verification plan.

Number of High-speed sensor bring-up commonly requires hardware iteration because via stubs, plane voids, and Bring-up log + spins to [N] connector transitions are frequent SI inflection revision delta stable points; TI emphasizes vias degrade SI and must analysis streaming be engineered, not ignored. 44 Tools and standards used Core textbooks and high-speed references – Signal and Power Integrity—Simplified 45 for foundational SI concepts: uniform transmission lines, return-path primacy, and “measure to validate.” 46 – Principles of Power Integrity for PDN Design—Simplified 47 for PDN target-impedance definition, PDN noise budgeting intuition, and PDN-to-jitter coupling examples. 48 – Electromagnetic Compatibility Engineering 49 as a comprehensive EMC text that explicitly includes precompliance measurements and PCB layout/stack-up topics. 50 – High-Speed Digital Design 51 as a canonical high-speed design reference emphasizing return current, stack examples, and connector EMI mechanisms (cited here via the author’s official synopsis and topic outline). 52 Standards – IEC 53 IEC 60068-2-6: sinusoidal vibration testing method for assessing mechanical robustness and performance degradation under vibration. 54 – CISPR 32: emissions requirements framework for multimedia equipment, including objectives for spectrum protection and reproducibility. 55 – CISPR 16-1-1: defining characteristics/performance of measurement equipment (EMI receivers/spectrum analyzers) for radio disturbance measurement. 56 – IPC 57 IPC-2141: controlled-impedance design guide (used as the controlling reference for impedance specifications and procurement language). 32 – IPC-D-356: bare-substrate electrical test data format for netlist/test processing workflows. 31 – IEC 62402: structured requirements and guidance for obsolescence management and design-time strategies to minimize obsolescence risk. 11 – ECIA EIA-364-28F: vibration test procedure for electrical connectors and sockets, used to qualify connector performance under vibration severities. 58 Manufacturer application notes and official guidance – Texas Instruments 59 high-speed layout and return-path guidance, including plane split stitching capacitors and via discontinuity mitigation. 60 – Analog Devices 61 mixed-signal grounding/decoupling best practices and CSI‑2 D‑PHY routing guidance. – NXP Semiconductors 63 sensor-side bring-up sequencing for CSI‑2/D‑PHY and data-rate calculation guidance. 64 – Renesas Electronics 65 quantitative PCB power-line targets for MIPI-related domains (e.g., resistance/ inductance guidance near the device). 66 – Keysight Technologies 67 TDR measurement capabilities (impedance, discontinuities, differential characterization) and 2‑port VNA low-impedance PDN measurement guidance. 68 – Tektronix 69 EMI pre-compliance instrumentation expectations and the principle that non-compliant gear can be used when sufficient margin is applied. 9 – IPC-2581 Consortium 70 dataset exchange framing for manufacturing handoff. 30 1 4 24 34 46 51 59 https://ptgmedia.pearsoncmg.com/images/9780132349796/samplepages/ 9780132349796.pdf https://ptgmedia.pearsoncmg.com/images/9780132349796/samplepages/9780132349796.pdf 2 40 67 https://www.analog.com/media/en/technical-documentation/user-guides/public-gmsl2- hardware-design-and-validation-guide.pdf https://www.analog.com/media/en/technical-documentation/user-guides/public-gmsl2-hardware-design-and-validation- guide.pdf 3 44 61 https://www.ti.com/lit/pdf/ssztcm4 https://www.ti.com/lit/pdf/ssztcm4 5 12 16 26 35 https://www.ti.com/lit/pdf/slla414 https://www.ti.com/lit/pdf/slla414 6 54 https://webstore.iec.ch/en/publication/544 https://webstore.iec.ch/en/publication/544 7 https://www.sciencedirect.com/science/article/abs/pii/S0026271418309478 https://www.sciencedirect.com/science/article/abs/pii/S0026271418309478 8 https://ece.uwaterloo.ca/~oramahi/IEEE-TEMC-Electromagnetic-Nov2004.pdf https://ece.uwaterloo.ca/~oramahi/IEEE-TEMC-Electromagnetic-Nov2004.pdf 9 43 45 https://download.tek.com/document/37A_60141_1_HR_Letter.pdf https://download.tek.com/document/37A_60141_1_HR_Letter.pdf 10 55 https://webstore.iec.ch/en/publication/22046 https://webstore.iec.ch/en/publication/22046 11 https://www.csagroup.org/store/product/iec_059531/?srsltid=AfmBOorxIgroQleKg5gl8Su- iERRNAc1PAjiaXnc7VdgW1DJEGuJzfZd https://www.csagroup.org/store/product/iec_059531/?srsltid=AfmBOorxIgroQleKg5gl8Su-iERRNAc1PAjiaXnc7VdgW1DJEGuJzfZd 13 25 41 53 60 https://www.ti.com/lit/pdf/spraar7 https://www.ti.com/lit/pdf/spraar7 14 15 https://community.nxp.com/pwmxy87654/attachments/pwmxy87654/qoriq-grl/9543/2/ High_speed_interfaces_layout__layout_design_guide.pdf https://community.nxp.com/pwmxy87654/attachments/pwmxy87654/qoriq-grl/9543/2/ High_speed_interfaces_layout__layout_design_guide.pdf 17 65 https://www.ti.com/lit/pdf/snla002 https://www.ti.com/lit/pdf/snla002 18 37 69 https://www.analog.com/media/en/training-seminars/tutorials/MT-031.pdf https://www.analog.com/media/en/training-seminars/tutorials/MT-031.pdf 19 23 62 63 https://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf https://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf 20 21 42 48 57 https://ptgmedia.pearsoncmg.com/images/9780132735551/samplepages/ 9780132735551_CH01.pdf https://ptgmedia.pearsoncmg.com/images/9780132735551/samplepages/9780132735551_CH01.pdf 22 36 https://www.keysight.com/zz/en/assets/7018-08474/application-notes/5989-5935.pdf https://www.keysight.com/zz/en/assets/7018-08474/application-notes/5989-5935.pdf 27 https://docs.altera.com/r/docs/683192/18.1/an-763-intel-arria-10-soc-device-design-guidelines/board- bring-up-considerations https://docs.altera.com/r/docs/683192/18.1/an-763-intel-arria-10-soc-device-design-guidelines/board-bring-up-considerations 28 64 https://www.nxp.com/docs/en/application-note/AN5305.pdf https://www.nxp.com/docs/en/application-note/AN5305.pdf 29 49 https://odbplusplus.com/wp-content/uploads/sites/2/2020/03/odb_spec_user.pdf https://odbplusplus.com/wp-content/uploads/sites/2/2020/03/odb_spec_user.pdf 30 https://www.ipc2581.com/ https://www.ipc2581.com/ 31 https://shop.electronics.org/ipc-d-356/ipc-d-356-standard-only https://shop.electronics.org/ipc-d-356/ipc-d-356-standard-only 32 https://shop.electronics.org/ipc-2141/ipc-2141-standard-only https://shop.electronics.org/ipc-2141/ipc-2141-standard-only 33 38 68 70 https://www.keysight.com/us/en/assets/7018-01461/application-notes/5989-5763.pdf https://www.keysight.com/us/en/assets/7018-01461/application-notes/5989-5763.pdf 39 56 https://webstore.iec.ch/en/publication/60774 https://webstore.iec.ch/en/publication/60774 47 66 https://www.renesas.com/en/document/apn/pcb-design-guidelines-mipi-csi-mipi-dsi-usb20-and-pci- express-gen2?srsltid=AfmBOorJ0pV4jhQ97UI-2-XZZuQiVRYaamRnRGQD3BGvEtS86VOcVt_D https://www.renesas.com/en/document/apn/pcb-design-guidelines-mipi-csi-mipi-dsi-usb20-and-pci-express-gen2? srsltid=AfmBOorJ0pV4jhQ97UI-2-XZZuQiVRYaamRnRGQD3BGvEtS86VOcVt_D 50 https://www.wiley-vch.de/en/areas-interest/engineering/electromagnetic-compatibility- engineering-978-0-470-18930-6 https://www.wiley-vch.de/en/areas-interest/engineering/electromagnetic-compatibility-engineering-978-0-470-18930-6 52 https://www.sigcon.com/books/bookHSDD.html https://www.sigcon.com/books/bookHSDD.html 58 https://store.accuristech.com/standards/ecia-eia-364-28f? product_id=2585175&srsltid=AfmBOoqqDb2KwLTIX5fQGSN0neTTV50nIQfX-NXpI2qh-JCnLrl0fbw1 https://store.accuristech.com/standards/ecia-eia-364-28f? product_id=2585175&srsltid=AfmBOoqqDb2KwLTIX5fQGSN0neTTV50nIQfX-NXpI2qh-JCnLrl0fbw