PCB manufacturing PCB manufacturing
Home > Blog

CoWoS-S vs CoWoS-L: Interposer and Scaling Differences

September 25th, 2026

CoWoS-S vs CoWoS-L is primarily a comparison of interposer structures: S uses a continuous silicon interposer, while L combines an RDL-based interposer with embedded local silicon interconnects. Both can connect logic and HBM inside an advanced package.

L is not simply a larger version of S, and the letter does not determine a finished chip’s speed or price. The useful comparison is where dense routing is placed, how the package scales, and which design requirements each qualified implementation meets.

Conceptual comparison of a continuous CoWoS-S silicon interposer and a CoWoS-L RDL interposer with local silicon bridges

What Is the Main Difference Between CoWoS-S and CoWoS-L?

CoWoS-S places dense routing in a full silicon interposer; CoWoS-L uses local silicon bridges within a wider RDL interposer. The package substrate remains a separate structure beneath both.

Feature CoWoS-S CoWoS-L
Interposer structure Continuous silicon RDL structure with embedded LSI
Dense die-to-die routing Silicon interposer wiring Local silicon interconnect regions
Layout consideration Routing across the silicon interposer Bridge placement plus wider RDL routing
Memory integration Supports HBM configurations Supports HBM configurations
Connection to the PCB Through a separate package substrate Through a separate package substrate

This comparison describes the architecture. Exact routing rules, supported dies, dimensions, and qualification belong to the specific generation and product. TSMC’s official CoWoS technology overview is the primary reference for the structural distinction.

How Do Their Signal Paths Differ?

S routes die-to-die connections through the continuous silicon interposer; L uses local silicon regions for dense neighboring interfaces and RDL for the wider redistribution structure.

CoWoS-S schematic showing lateral routing across silicon and separate vertical connections

In an S layout, the silicon interposer is the common routing platform below the dies. In an L layout, a bridge must align with the interfaces it connects. That makes the die floorplan and local connection regions part of the architecture decision.

CoWoS-L schematic showing a local silicon bridge under adjacent die edges inside an RDL structure

Neither illustration defines all signal or power paths in a real product. A board designer should not infer package pin assignments, routing pitch, or electrical models from a simplified cutaway.

The engineering comparison is routing availability versus routing localization. S offers a continuous silicon platform on which to organize connections among the dies. L requires dense interfaces to line up with local silicon regions, while wider redistribution uses the surrounding RDL. L therefore makes bridge placement an explicit floorplanning constraint; S still has congestion, layer-count, and interposer-area constraints.

Which Architecture Supports Larger Interposer Areas?

CoWoS-L is TSMC’s local-silicon approach for extending the interposer platform beyond the area served by its published CoWoS-S offering.

The public overview lists S up to 3.3 reticles, approximately 2,700 mm², and identifies a 3.5-reticle L generation in production since 2024. These figures describe particular platform generations, not a fixed area ratio or a universal limit on every future product.

Compare dated, qualified configurations. Do not compare an S production specification with an L roadmap target and describe both as equally available. Interposer area also differs from package outline, usable die area, and system PCB dimensions.

Is CoWoS-L Faster or Cheaper Than CoWoS-S?

CoWoS-L is attractive when the design needs a larger integration area while retaining dense local links; CoWoS-S remains relevant when the die arrangement fits a qualified continuous-silicon platform. Neither architecture alone determines a finished device’s speed or price.

  • Bandwidth: compare the memory generation, interface width, transfer rate, and controller.
  • Electrical performance: compare the actual channels and supply network under matching conditions.
  • Cost: include dies, interposer, substrate, assembly, test, qualification, and production volume.
  • Availability: confirm the specific qualified configuration and supply schedule.

Replacing full-area silicon with local silicon changes the material structure, but it does not prove a lower finished-package price. Likewise, a larger interposer may accommodate a different die arrangement without making each connection faster.

A useful cost comparison is total build and test cost divided by the number of passing packages. Less full-area silicon is only one input. Bridge integration, substrate complexity, assembly losses, and the value of attached logic and HBM can outweigh that saving. Compare the same functional target and production assumptions, not unlike products.

For electrical comparison, use extracted channel resistance, capacitance, loss, and crosstalk at the intended data rate. A short but congested route is not automatically better than a longer route with a better reference environment. For thermal comparison, hold workload, cooling conditions, and temperature limits constant.

Design Situation Architecture to Evaluate
Existing die set fits a qualified silicon-interposer floorplan S, against its routing, power, and thermal limits
Integration area exceeds the published S offering L, with confirmed bridge map and qualified size
Dense connections concentrated at neighboring die edges L, if local bridge placement serves those interfaces
Released accelerator already chosen for a board Use its actual package documentation; the board fabricator does not choose its CoWoS variant

CoWoS-S vs CoWoS-L vs CoWoS-R: Where Does R Fit?

CoWoS-R uses an RDL interposer, S uses a silicon interposer, and L combines RDL with embedded local silicon interconnects.

Variant Interposer Approach
CoWoS-S Full silicon routing platform
CoWoS-R Polymer-and-copper RDL platform
CoWoS-L RDL platform with local silicon interconnects

CoWoS-S vs CoWoS-R compares silicon-based and RDL-based interposers. CoWoS-L vs CoWoS-R focuses on the local silicon interconnects added to L’s architecture. R should not be treated as merely another name for L.

Routing figures must stay attached to the relevant variant. A published RDL line-and-space value for R is not automatically the wiring pitch of an S interposer or an L silicon bridge.

CoWoS-L vs EMIB: Are They the Same?

CoWoS-L and Intel EMIB both use local silicon interconnect concepts, but the bridge integration is different: L embeds LSI in its interposer platform, while EMIB embeds a bridge in the package substrate.

That difference affects the surrounding interconnect architecture and design flow. The technologies are not interchangeable components or identical manufacturing processes. Intel also offers multiple EMIB generations, so a feature of one version should not be attributed to the entire family. See Intel’s advanced packaging overview for its current platform descriptions.

What Should a PCB Designer Compare?

A PCB designer should compare the released components’ footprints, power requirements, external interfaces, and mechanical limits—not select a board stackup from the CoWoS letter.

Distinct semiconductor, interposer, package substrate, and system PCB integration levels
  • Footprint: ball map, land pattern, escape routing, and keep-outs.
  • Power: supply rails, current demand, and decoupling requirements.
  • Signals: interface specifications, reference planes, and channel models.
  • Mechanics: package support, cooling attachments, and board constraints.
  • Assembly: component handling, soldering profile limits, and inspection requirements.

Our guide to advanced HDI PCBs covers board-level routing options. For the broader semiconductor context, our TSMC overview separates foundry technology from the finished electronic system.

FAQ About CoWoS-S vs CoWoS-L

Can the same PCB accept S and L packages interchangeably?

Only if the actual components are specified as compatible. Matching the CoWoS family does not establish a common ball map, dimensions, supply rails, signal assignment, or cooling requirement.

Does CoWoS-L always support more HBM than CoWoS-S?

No. A larger available platform can accommodate a more expansive die arrangement, but HBM count is also limited by the processor interfaces, memory configuration, routing, power, and cooling design.

Does CoWoS-S have to be replaced in every new design?

No. A qualified S implementation can remain suitable if it meets the required integration area and electrical, thermal, and manufacturing targets. A newer packaging option is not by itself a reason to redesign a working product.

Can CoWoS-S and CoWoS-L use the same memory generation?

Yes, the architecture name does not uniquely identify the HBM generation. Compatibility must be established for the specific logic, memory, and package implementation rather than inferred from S or L.

Is CoWoS-L the same technology as SoIC?

No. CoWoS-L connects dies through an interposer platform. SoIC is a die-stacking technology; a stacked die assembly can subsequently be integrated into a larger package. They address different integration levels.

Does either option require a fixed PCB layer count?

No. PCB layer count follows the external ball map, routing density, power distribution, channel requirements, and manufacturing rules. Interposer metal layers are not motherboard layers, so they cannot be converted into a PCB stackup count.

How Can We Help with the Surrounding PCB and PCBA?

At EBest Circuit, we provide PCB fabrication and PCBA services. For a design using an advanced package, we review the board files, stackup, and assembly scope against the actual component requirements. We do not equate these services with manufacturing CoWoS interposers.

Send your Gerber files, package information, quantities, and stackup requirements to sales@bestpcbs.com. Include the BOM and test requirements for assembly so we can discuss a board-level manufacturing solution.

You may also like

Tags: , , , ,
Posted in best pcb, PCB Design, PCB Technology | No Comments »

EMC Meaning in Electronics: EMI, Testing and PCB Design

September 23rd, 2026

The emc meaning in electronics is electromagnetic compatibility: the ability of equipment to work as intended in its electromagnetic environment without creating unacceptable interference for other equipment. In practical terms, an EMC-compatible product controls what it emits and continues operating when exposed to expected disturbances.

EMC is a system-level result, not a single component or PCB feature. Circuit architecture, PCB stackup, return paths, power integrity, cables, shielding, enclosure design, firmware states, and the test setup can all change the result. Engineers therefore need to address EMC during design and verify it against the standards that apply to the finished product.

EMC meaning shown with a PCB under near-field emissions testing

Key Takeaways

  • EMC stands for electromagnetic compatibility, the ability of electronic equipment to operate correctly without causing or suffering unacceptable electromagnetic interference.
  • The purpose of EMC is to let electronic devices coexist reliably by limiting emissions and maintaining immunity in their intended electromagnetic environment.
  • EMC testing checks both emissions and immunity under the standards, operating modes, cable arrangements, and performance criteria applicable to the finished product.
  • In the United States, EMC requirements depend on the product and its functions, so the applicable FCC rules and product standards must be confirmed for each device.
  • EMI risk can be reduced through continuous return paths, compact switching loops, correct decoupling, connector filtering, cable control, shielding, and appropriate enclosure design.
  • EMI and EMC are not interchangeable terms, because EMI is the unwanted electromagnetic disturbance while EMC is the broader goal of controlling emissions and maintaining immunity.

What Does EMC Mean in Electronics?

In electronics, EMC means that a device can coexist with other electrical and electronic equipment in its intended environment. The device must keep its electromagnetic emissions within the applicable limits and maintain acceptable operation when exposed to defined disturbances.

This definition has two complementary sides:

  • Emissions: electromagnetic energy leaving the product through radiation or conductors must remain below the applicable limits.
  • Immunity: the product must continue to meet specified performance criteria when exposed to disturbances such as RF fields, electrostatic discharge, transients, or conducted noise.

The phrase emc meaning in electronics is sometimes reduced to “no interference,” but that is incomplete. Every operating circuit produces electromagnetic energy. EMC engineering controls that energy and its coupling paths so nearby systems can operate together.

Electromagnetic compatibility balancing emissions and immunity

What Is the Difference Between EMC, EMI and EMS?

EMC is the ability to coexist, EMI is the disturbance that causes or may cause a problem, and EMS describes susceptibility to that disturbance. The terms are related, but they are not interchangeable.

Term Full Form Practical Meaning
EMC Electromagnetic compatibility The product limits emissions and tolerates specified disturbances in its intended environment.
EMI Electromagnetic interference Unwanted electromagnetic energy that can degrade or disrupt operation.
EMS Electromagnetic susceptibility The tendency of equipment to be affected by electromagnetic disturbances; immunity expresses resistance to them.

For example, a switching converter can be an EMI source, a cable can provide a coupling path, and a sensitive analog input can be the victim. An EMC design review looks at the complete source-path-victim relationship rather than treating the noise source alone.

What Does EMC Cover: Emissions, Immunity and Coupling?

EMC covers how disturbances are generated, how they travel, and how circuits respond. A useful diagnosis separates the source, coupling path, and victim because changing any one of the three can reduce the problem.

  • Conducted coupling: noise travels through power, ground, signal, or cable conductors.
  • Capacitive coupling: changing electric fields transfer energy between nearby conductors.
  • Inductive coupling: changing magnetic fields induce voltage in an adjacent current loop.
  • Radiated coupling: traces, cables, slots, or enclosures behave as transmitting or receiving structures.
EMC source coupling path and victim relationship in electronic systems

Fast edge rates often matter more than the nominal clock frequency. A low-frequency digital signal with sharp transitions can contain high-frequency energy, while a cable connected to the board can convert common-mode current into radiated emissions.

What Is EMC Testing?

The emc testing meaning is the controlled evaluation of a product’s emissions and immunity against specified limits, test levels, setups, and performance criteria. It is not one universal test; the correct program depends on the product category, intended environment, interfaces, power source, and target market.

If you are asking what is emc testing, the most useful distinction is between pre-compliance and formal compliance work. Pre-compliance measurements help find high-risk frequencies, cables, operating modes, and PCB areas before the final test. Formal testing follows the applicable standard using the required equipment, calibrated setup, distances, limits, and documentation.

Engineer performing a PCB EMC pre-compliance scan with a near-field probe

For a deeper test-oriented overview, see our guide to EMI and EMC testing in PCBs.

Which EMC Tests Are Common for Electronic Products?

Common EMC evaluations include conducted and radiated emissions plus several immunity tests, but not every product needs every test. The applicable product or product-family standard should define the final test plan.

Test What It Evaluates Typical Coupling Path
Conducted emissions Noise returned through power or connected conductors Power leads and cables
Radiated emissions Electromagnetic energy radiated by the product and its cables Enclosure openings, traces, cables, and common-mode currents
Radiated RF immunity Operation under an external radio-frequency field Enclosure, PCB, and cables acting as receiving structures
Conducted RF immunity Response to RF energy coupled onto connected cables Power and signal ports
Electrostatic discharge Response to direct or indirect ESD events Exposed metal, seams, displays, buttons, and connectors
EFT/burst and surge Response to fast repetitive transients or higher-energy events Power and external I/O wiring

Performance criteria also matter. A disturbance may permit no degradation, temporary self-recovery, or controlled operator intervention depending on the applicable standard. Teams should define acceptable behavior before the test instead of deciding after a failure.

Why Do PCBs Cause EMC Problems?

PCBs contribute to EMC problems when fast currents flow through large or discontinuous loops, when noise couples into sensitive nodes, or when common-mode energy reaches an efficient radiator such as a cable. The schematic may be correct while the physical current paths are not.

Frequent board-level causes include:

  • high-speed traces crossing a split or void in the reference plane;
  • large switching-current loops around converters, drivers, or power devices;
  • decoupling capacitors placed too far from IC power and ground pins;
  • long parallel routes that increase crosstalk;
  • filters placed far from the connector where interference enters or leaves;
  • poor shield or chassis termination that forces high-frequency current through signal ground;
  • layer changes without a nearby return-path transition;
  • uncontrolled cable common-mode current.

A PCB is only one part of the final electromagnetic structure. The enclosure, harness, display, power supply, grounding arrangement, connector shells, and installed accessories can turn a small board-level noise source into a compliance failure.

How Can PCB Layout Improve EMC Performance?

PCB layout improves EMC by reducing loop area, preserving continuous return paths, containing high-frequency fields, and preventing noise from reaching external cables. The most effective changes are usually architectural and placement-related, not late additions of random filter components.

PCB EMC layout practices showing return path, ground plane, and connector filtering
  • Route critical signals over a continuous reference plane and avoid plane splits beneath them.
  • Keep switching loops compact and place high-di/dt components close together.
  • Place decoupling capacitors at the relevant power pins with short connections to the reference plane.
  • Keep noisy power sections away from clocks, RF paths, analog inputs, and external connectors.
  • Place filtering and transient protection at the connector boundary.
  • Provide nearby stitching vias when signals change reference layers.
  • Control impedance, return paths, and pair geometry for fast interfaces.
  • Plan shield-can pads, via fences, chassis bonds, and keep-outs before routing is complete.

Our article on the circuit board ground plane explains why return-path continuity and layer references are central to both signal integrity and EMC.

Which Standards and Regulations Apply to EMC?

The applicable EMC requirements depend on the product, environment, interfaces, and market. There is no single “EMC certificate” or universal test list that covers every electronic device.

Common frameworks include the IEC 61000 series for EMC terminology and test methods, CISPR publications for many emissions requirements, FCC rules for relevant radio-frequency emissions in the United States, and the EU EMC Directive for equipment placed on the European market. Product-specific families can add different limits, ports, test levels, operating modes, and performance criteria.

Examples include IEC 61326-1 for certain measurement, control, and laboratory equipment and IEC 60601-1-2 for medical electrical equipment. These examples should not be treated as automatic selections. Confirm the product classification and current edition with the compliance laboratory or responsible regulatory specialist before freezing the test plan.

How Should You Prepare an EMC-Sensitive PCB for Manufacturing?

An EMC-sensitive board should be released with the stackup, reference planes, controlled-impedance requirements, grounding details, shield features, connector interfaces, and assembly notes clearly defined. Fabrication and assembly must preserve the electrical geometry on which the design depends.

Before sending the package, check that it includes:

  • Gerber, ODB++, or IPC-2581 data plus drill and route files;
  • a stackup drawing with materials, finished thickness, and copper requirements;
  • controlled-impedance definitions and coupon requirements where applicable;
  • a released BOM with filter, protection, and shielding components identified;
  • pick-and-place and assembly drawings with polarity and shield orientation;
  • notes for via filling, ground stitching, shield-can pads, and connector-shell grounding;
  • inspection and electrical-test requirements appropriate to the build.

An EMC test is still performed on the finished product or representative system, not on bare Gerber data alone. Manufacturing documentation reduces unintended variation, but it does not replace system-level compliance validation.

FAQ About EMC Meaning

What does EMC stand for in electronics?

EMC stands for electromagnetic compatibility. It describes whether electronic equipment can operate satisfactorily in its intended electromagnetic environment without causing unacceptable interference to other equipment.

Does EMC mean a device produces no EMI?

No. Every active electronic product produces some electromagnetic energy. EMC means emissions remain within the applicable limits and the product has the required immunity for its intended environment.

Is EMC the same as FCC or CE compliance?

No. EMC is the engineering and performance concept. FCC requirements, the EU EMC Directive, and product standards are regulatory or conformity frameworks that may define specific emissions, immunity, documentation, and assessment obligations.

Can a PCB manufacturer certify the finished product’s EMC?

A PCB manufacturer can build the specified stackup, impedance, grounding, shielding pads, and assembly features. Formal product compliance normally requires testing the representative finished equipment under the applicable standard and configuration.

Does a solid ground plane guarantee good EMC?

No. A continuous reference plane usually improves return-path control, but poor connector grounding, large switching loops, cable common-mode current, enclosure seams, or unsuitable filtering can still cause emissions or immunity failures.

Why can a prototype pass while a later version fails?

Changes to the stackup, component substitutions, cable routing, enclosure bonding, firmware modes, clock edges, power supply, or test configuration can change EMC behavior. Revision control should include both electrical design files and the representative test setup.

Is pre-compliance EMC testing worthwhile?

Yes. Near-field scanning and other pre-compliance checks can identify dominant frequencies, noisy PCB regions, cable effects, and operating modes before formal testing. They reduce risk, but they do not replace the required compliance test.

How Can EBest Circuit Support an EMC-Aware PCB Build?

EMC performance begins with system design, while consistent PCB fabrication and assembly help preserve the intended stackup, return paths, impedance geometry, grounding features, and shielding interfaces. At EBest Circuit, we support PCB fabrication and PCBA from prototype through volume production using the released manufacturing data and agreed inspection scope.

For projects involving industrial interfaces, power conversion, fast digital links, or shielded connectors, our industrial network PCB design checks provide a useful release reference. Send your Gerber files, stackup, BOM, assembly data, quantities, and EMC-related manufacturing notes to sales@bestpcbs.com for engineering review and a quotation.

You may also like

Tags: , , , ,
Posted in best pcb, PCB Design, PCB Technology | No Comments »

Flexible Printed Circuit: Materials, Types and Connections

September 22nd, 2026

A flexible printed circuit (FPC) is an electrical circuit built on a thin, bendable insulating film, usually polyimide, with patterned copper conductors. It can connect boards, carry components, or follow a three-dimensional enclosure where a rigid PCB will not fit.

Flexibility does not mean unlimited bending. A circuit folded during installation and one moving inside a hinge need different constructions. The materials, bend zone and connector ends must be designed together.

Amber flexible printed circuit with copper routing, a smooth bend and gold contact fingers

How Does a Flexible Printed Circuit Work?

An FPC carries power and signals through copper tracks just as a rigid PCB does; its thin film substrate lets those tracks follow a controlled bend. The copper is patterned into separate conductors, while insulating layers prevent contact between adjacent circuits.

A flexible printed circuit board can include pads, vias and mounted components, not just parallel wires. In a camera module, for example, one custom-shaped flex can route signals from the sensor board around a mechanical obstruction to the main board. Component locations remain supported while the connecting section bends.

What Are the Main Types of Flex Circuits?

The main constructions are single-sided, double-sided and multilayer flex; rigid-flex combines flexible interconnect sections with integrated rigid circuit sections.

Construction Copper structure Typical reason to choose it
Single-sided flex One conductive layer Simple routing with a thin bend region
Double-sided flex Two conductive layers, usually joined by plated holes More routing paths or a reference layer
Multilayer flex Three or more conductive layers Dense routing where a thicker stack is acceptable
Rigid-flex Rigid circuit sections integrated with flex layers Rigid component areas connected without separate cable connectors

Adding layers increases routing space but also changes bending stiffness. A multilayer design that fits a stationary enclosure is not automatically suitable for repeated movement.

Which Flexible Printed Circuit Board Material Should You Choose?

Polyimide with copper foil is a common choice for soldered FPC assemblies; the copper type, adhesive system and protective coverlay must then match the temperature and bending requirements.

  • Base film: polyimide provides electrical insulation and tolerates the thermal processing used for many assembled flex circuits. Polyester is an alternative for suitable lower-temperature constructions, not an automatic replacement in a reflow-soldered design.
  • Copper: rolled-annealed copper is commonly evaluated for repeated flexing. Electrodeposited copper is also used, but foil grade and fatigue performance matter more than the abbreviation alone.
  • Laminate: adhesive-based constructions bond copper to film with an adhesive layer. Adhesiveless laminates remove that separate bonding layer from the copper-to-polyimide interface, helping reduce stack thickness.
  • Coverlay: a protective film and adhesive cover the tracks, with openings at solder pads and contacts.

For a concrete coverlay example, DuPont Pyralux LF7001 combines 13 µm adhesive with 13 µm polyimide, while LF0110 lists 25 µm for each. These are individual coverlay constructions, not finished FPC thicknesses.

Exploded single-sided flex stack showing coverlay, coverlay adhesive, patterned copper and polyimide base

How Is an FPC Different from an FFC or a Rigid PCB?

An FPC has a custom circuit pattern, an FFC typically has parallel flat conductors, and a rigid PCB uses a substrate intended to stay rigid.

Feature FPC FFC Rigid PCB
Routing Custom tracks, branches and pads Usually parallel conductors Custom tracks, planes and vias
Form Custom outline with designed bend zones Usually a flat ribbon Fixed board shape
Component mounting Possible on supported areas Usually used as an interconnect cable Standard component platform

An FFC can be the simpler option for a straight connection between compatible sockets. Choose FPC when the circuit needs branching, an unusual outline, components or controlled routing. A flex board with a bonded stiffener is still not the same as an electrically integrated rigid-flex board.

How Do Static and Dynamic Bending Change the Design?

Static, or flex-to-install, circuits are bent into position and normally remain there; dynamic circuits must survive repeated movement without conductor fatigue.

For installation-only routing, define the formed shape, bend radius and assembly sequence. For dynamic routing, also specify the travel, cycle requirement, speed and operating environment. A hinge that passes one assembly bend has not demonstrated its service life.

  • Keep components, solder joints and plated holes outside the working bend zone.
  • Route conductors across the bend line rather than along it, and avoid abrupt changes in width.
  • Use smooth curves instead of creases; keep the moving section clear of sharp enclosure edges.
  • Evaluate the thinnest practical stack and suitable copper foil before adding layers or shielding.
Comparison of a fixed installation bend and a rolling flex loop for repeated motion

How Is the Minimum Bend Radius Determined?

The minimum bend radius comes from the complete flex stack and required bend life, not from the polyimide thickness alone. Copper, adhesive and coverlay all contribute to the thickness and strain of the bend region.

For example, our rigid-flex DFM guide gives 6× composite thickness for single- and double-sided flex sections and 12× for sections with three or more copper layers. Under that guide’s construction assumptions, a 0.10 mm section at 6× gives a 0.60 mm radius. These are design guidelines, not a guarantee for every FPC or dynamic application.

Record the approved radius on the drawing and validate the actual construction at the required motion and temperature. Changing the copper thickness, shielding or coverlay after approval can invalidate the earlier bend assessment.

How Do You Match a Flexible Printed Circuit Connector?

Match the connector’s pitch, position count, contact side and specified FPC mating thickness before finalizing the tail drawing. Matching the number of contacts alone is not enough.

For example, Hirose FH12 is an FPC/FFC connector family offering 0.5 mm and 1 mm pitches. The exact part drawing determines the contact orientation and acceptable tail dimensions; the family name is not a complete interface specification.

  • Contact geometry: confirm finger width, pitch, exposed length and edge-to-contact dimensions.
  • Mating thickness: include the circuit, bonded stiffener and adhesive at the insertion area.
  • Contact side: verify whether the socket contacts the upper or lower face of the inserted tail.
  • Retention: allow access to the latch and keep cable pull or bending loads away from the connection.

A local PCB stiffener can support the tail and establish its mating thickness. Its edge should not force bending directly beside exposed contacts or solder joints.

FPC tail, local stiffener and open ZIF connector with total mating thickness indicated

Flexible Printed Circuit Manufacturing Process

The flexible printed circuit manufacturing process patterns copper on a flexible laminate, forms any required interlayer connections, adds insulation and reinforcement, and tests the finished circuit.

  1. Prepare the laminate: select the specified copper-clad film and prepare the panel for imaging.
  2. Form the circuit: image and etch the copper tracks; drill and plate interconnections where the construction requires them.
  3. Protect the conductors: align and laminate coverlay with openings for pads and contacts.
  4. Complete exposed surfaces: apply the specified pad or contact finish and bond local stiffeners.
  5. Profile and test: cut the outline, inspect dimensions and check electrical continuity and isolation.

The exact sequence changes with layer count, via structure and finish. The etched FPC process explains copper pattern formation in more detail. Coverlay registration and adhesive flow deserve particular attention because a partly covered contact may not mate or solder correctly.

What Changes When Components Are Assembled on Flex?

Flex assembly needs support beneath the circuit during printing, placement and soldering so the thin panel stays flat and the joints remain unloaded.

  • Carrier or fixture: support the panel without obstructing pads or distorting the intended outline.
  • Moisture control: follow the laminate and assembly process requirements before thermal exposure; do not apply one universal baking recipe.
  • Reflow profile: qualify the temperature profile against the solder paste, components and complete flex material system.
  • Inspection: check solder joints and pad alignment, then use electrical or functional testing appropriate to the assembly.

After soldering, handle the assembly by supported areas. Pulling a loose tail to lift the board can transfer force into small pads or joints even when the soldering itself was acceptable.

Where Are Flexible Printed Circuits Most Useful?

FPCs are useful where a circuit must fit a narrow three-dimensional space, connect moving sections or reduce separate wire connections.

  • Cameras and displays: connect modules arranged on different planes.
  • Wearable devices: follow compact enclosure shapes and connect small component islands.
  • Medical instruments: route signals through constrained assemblies with application-specific material and reliability requirements.
  • Battery monitoring: distribute sensing connections across a cell arrangement without treating thin sensing traces as the main power bus.
  • Moving mechanisms: use a qualified dynamic flex section where the motion profile suits the construction.

For a flat assembly with ample space and no movement, a rigid PCB may be easier and less expensive. FPC is valuable when its geometry solves a real packaging or interconnection problem.

FAQ About Flexible Printed Circuits

Can an FPC be stretched?

Not like an elastic band. Conventional polyimide-and-copper flex is designed to bend, not stretch freely. Stretchable electronics require different materials or conductor geometries.

Can a torn flex circuit be repaired?

Some accessible traces can be repaired in controlled rework, but a repair changes local stiffness and may not survive repeated bending. Replacement is usually more appropriate for damaged moving sections or contacts that must fit a socket precisely.

Can flex carry high-speed signals?

Yes, with a designed transmission-line structure. Specify the impedance target, reference conductors, dielectric thickness and connector transition; bending and any change from solid to hatched shielding should be included in the assessment.

Is every amber flex circuit made from the same material?

No. Color does not identify the laminate grade, copper type or adhesive. The material specification and stackup are the reliable references.

What makes a custom FPC expensive?

More layers, complex outlines, poor panel utilization, fine features, multiple stiffeners and special materials can add cost. Compare quotations using the same construction and test requirements rather than board area alone.

How Can EBest Circuit Support Your Flex Circuit Project?

At EBest Circuit, we support flexible PCB and rigid-flex projects from prototype fabrication through production and assembly. Our PCB and PCBA services date back to 2006, and we can review the circuit, component sourcing and assembly requirements together.

For your flexible circuit project, send the Gerber or ODB++ data, stackup, bend drawing, connector part number and quantity to sales@bestpcbs.com. If assembly is required, include the BOM and placement data. We will review the material construction, reinforcement and manufacturing requirements before confirming a quotation.

You may also like

Tags: , , , ,
Posted in best pcb, FPC, PCB Design | No Comments »

CoWoS-L Explained: RDL and LSI for Larger AI Packages

September 21st, 2026

CoWoS-L combines a redistribution-layer (RDL) interposer with local silicon interconnects (LSI). The RDL spans the larger routing platform; embedded silicon bridges provide dense links at selected die interfaces. This division helps integrate more logic and high-bandwidth memory without using one continuous silicon interposer across the full area.

Conceptual CoWoS-L package with embedded local silicon interconnects

What Is CoWoS-L Packaging?

CoWoS-L packaging is TSMC’s CoWoS architecture that embeds local silicon interconnects within an RDL-based interposer for large multi-die computing packages.

A CoWoS-L cross section contains logic dies and HBM at the top, an interposer incorporating RDL and local silicon beneath them, and a separate package substrate below. RDL means redistribution layer: patterned conductors redistribute connections across the platform. LSI means local silicon interconnect: dense silicon-based routing placed where neighboring die interfaces need it.

Calling this only an “organic interposer” misses the embedded silicon and molded integration structure. It is not silicon-free, and its package substrate is not the system PCB. For the foundry context, see our introduction to TSMC’s manufacturing technologies.

How Do RDL and LSI Work Together in CoWoS-L?

LSI handles dense local die-to-die connections, while RDL distributes connections over the wider interposer footprint.

Local die-to-die signal path within the silicon bridge above wider RDL routing
Structure Location Interconnect role
LSI Selected neighboring die interfaces High-density local links using fine silicon-based wiring
RDL Across the wider interposer Broader redistribution and integration with the embedded structures
Package substrate Below the interposer Connections toward the finished component’s board interface

TSMC describes CoWoS-L LSI with multiple layers of submicron copper wiring and connections including SoC-to-SoC, SoC-to-chiplet, and SoC-to-HBM. The bridge occupies the region serving those interfaces; it is not simply a small silicon support underneath an arbitrary part of the package.

Consider two logic dies whose high-density interfaces face each other. An LSI region can connect those edges while the wider RDL carries other connections. Rotating one die may move its interface away from the intended landing region, forcing changes to bridge position and routing. Die orientation and LSI placement must therefore be co-designed.

The bridge does not generate bandwidth by itself. Link width, signaling rate, electrical characteristics, and the transmitting and receiving circuits still determine usable throughput.

Why Can CoWoS-L Support Larger Packages?

CoWoS-L extends the interposer footprint through a wider RDL platform while concentrating fine silicon routing at local interfaces, instead of enlarging one continuous silicon interposer everywhere.

TSMC’s public overview identifies a 3.5-reticle CoWoS-L generation that entered production in 2024. This is a dated technology generation, not a permanent maximum. Reticle multiples describe interposer scale relative to an exposure field; they do not specify the component’s outer dimensions, BGA pitch, or HBM count.

  • More placement area: the platform can accommodate additional or larger logic and memory components when supported by the package design.
  • Localized fine wiring: high-density silicon regions follow the die interfaces rather than covering the entire footprint.
  • Separate size limits: interposer area, package substrate outline, and cooling assembly envelope remain different dimensions.

Larger still means more manufacturing coordination. Mold, copper, silicon, and the organic substrate respond differently to temperature. Warpage, interconnect stress, routing yield, and cooling all require qualification. Less full-area silicon does not prove that every CoWoS-L product is cheaper than a CoWoS-S alternative.

How Is a CoWoS-L Package Manufactured?

The CoWoS-L process flow integrates local silicon interconnects into a molded RDL platform, attaches the top dies, and completes the assembly on a package substrate.

  1. Plan the floorplan: align SoC, chiplet, and HBM interface locations with the required LSI regions and reserve power-routing space.
  2. Integrate the embedded structures: incorporate the local silicon elements, and applicable embedded passive components, into the reconstituted interposer structure.
  3. Form the RDL connections: create the redistribution wiring that connects the embedded regions and wider package interfaces.
  4. Attach the top dies: TSMC describes a chip-last approach in which the interposer platform is prepared before top-chip assembly.
  5. Complete substrate integration and verification: assemble the package substrate and thermal structure and verify links, power behavior, and reliability.
CoWoS-L functional stages from floorplan through assembly and verification

Critical controls include embedded-element position, RDL registration, surface planarity, and joint formation. A displaced bridge or a local height error can compromise the fine die interface even if the overall package outline is correct.

This sequence describes functional stages, not a proprietary process recipe. Exact mold materials, bonding temperatures, tolerances, and inspection acceptance criteria must come from the qualified package process.

How Does CoWoS-L Support Power Delivery?

CoWoS-L supports power delivery through its interconnect network and the integration of stand-alone embedded deep trench capacitors (eDTCs) beneath the SoC.

Embedded eDTC beneath the SoC within the RDL interposer above a separate package substrate

TSMC identifies this eDTC integration in its official CoWoS technology overview. Placing capacitance near the load can shorten the local current loop and reduce the inductive penalty of supplying rapid current changes. The benefit depends on the actual connection geometry, capacitor characteristics, and complete power network.

  • Local transient support: nearby capacitance supplies part of a fast current demand before more distant supply paths respond.
  • Power-path coordination: package conductors, board planes, capacitors, and voltage regulators must meet the device’s supply limits together.
  • Impedance control: capacitance and interconnect inductance can create resonances, so adding capacitance is not automatically an improvement at every frequency.

For an illustrative target, a permitted 30 mV voltage change during a 100 A current step gives Z = ΔV/ΔI = 0.3 mΩ. These are example inputs, not a CoWoS-L rating. A real design needs its own tolerance and frequency-dependent package, board, and regulator models.

Where Is CoWoS-L Packaging Used?

CoWoS-L is used for large AI and HPC packages that need dense local connections among logic dies, chiplets, and HBM on a larger interposer platform.

  • Multi-die AI accelerators: local bridges connect the compute interfaces, while the wider floorplan accommodates logic and memory integration.
  • HBM-based computing packages: SoC-to-HBM connections combine local routing density with space for the required memory arrangement.
  • Chiplet-based HPC designs: dense connections between selected chiplets support integration without making every part of the routing platform silicon.

TSMC’s disclosed SoC-to-SoC, SoC-to-chiplet, and SoC-to-HBM connection forms support these application categories. They are not a claim that every chiplet processor or AI accelerator uses CoWoS-L. Naming a particular product requires a disclosed packaging variant, not an inference from its HBM count or performance.

What PCB Assembly Checks Matter for CoWoS-L Packages?

PCB assembly checks must follow the specific finished component or module’s land pattern, handling limits, reflow instructions, warpage requirements, and thermal-mechanical design.

  • Board interface: confirm whether the delivered item is a directly soldered component or a module with its own board/connector interface.
  • Land pattern and escape routing: use the released ball map and pad geometry; LSI dimensions do not become PCB trace dimensions.
  • Moisture handling and reflow: use the device’s specified storage, exposure, and thermal-profile limits, not a generic “CoWoS-L temperature.”
  • Coplanarity and support: review the component’s allowed deformation, board support, heatsink loads, and attachment keep-outs.
  • Inspection and testing: agree on accessible joint-inspection methods and electrical or functional tests appropriate to the assembly. Visual inspection alone cannot verify hidden BGA joints.

Our advanced HDI PCB guide covers board-level routing options. At EBest Circuit, we review PCB fabrication and PCBA requirements against the actual design and component documentation—not the interposer’s marketing dimensions. Send your Gerber files, stackup, BOM, quantities, and component assembly specifications to sales@bestpcbs.com. This board-level support is separate from TSMC’s CoWoS-L package manufacturing.

FAQs About CoWoS-L

Does CoWoS-L eliminate silicon?

No. It retains silicon in local interconnect regions. The wider RDL platform changes where silicon is used, not whether silicon is present.

Is LSI another processor?

No. In CoWoS-L, LSI is a local silicon interconnect structure that routes signals between die interfaces; it is not an additional computing die.

Is one LSI bridge enough for every package?

Not necessarily. The number and location of bridges follow the interfaces being connected. Multiple logic dies and HBM stacks can require several local regions.

Does an embedded eDTC replace PCB decoupling?

No. Embedded capacitance supports part of the local supply network. Board capacitors and the voltage regulator must still satisfy the component’s wider power-delivery requirements.

Is 3.5 reticles the maximum CoWoS-L size?

No. It identifies a published production generation rather than a permanent ceiling. Obtain the qualified dimensions and availability for the intended design instead of treating a roadmap target as a released package.

You may also like

Tags: , , , ,
Posted in best pcb, PCB Design, PCB Technology | No Comments »

CoWoS-S Packaging: Silicon Interposers for AI and HPC

September 21st, 2026

CoWoS-S packaging places logic dies and high-bandwidth memory (HBM) side by side on a silicon interposer. Fine metal wiring connects the dies across that interposer, while through-silicon vias connect it vertically to a separate package substrate. This arrangement supplies the dense, short memory connections needed by AI accelerators and high-performance computing systems.

Conceptual CoWoS-S package with logic and HBM above a continuous silicon interposer

What Is CoWoS-S Packaging?

CoWoS-S is TSMC’s silicon-interposer version of Chip-on-Wafer-on-Substrate packaging: the “S” identifies the silicon interposer that carries the die-to-die wiring.

Within the CoWoS semiconductor packaging family, its distinguishing feature is a continuous silicon interposer beneath the top dies. The logic and HBM are separate components connected on this shared routing platform, rather than one monolithic chip. HBM itself contains vertically stacked memory dies, but the side-by-side arrangement of logic and memory on the interposer is commonly described as 2.5D integration.

TSMC develops the package technology; the system board sits at a different manufacturing level. Our introduction to TSMC’s manufacturing technologies explains that broader context.

What Is Inside a CoWoS-S Package?

A CoWoS-S package contains logic dies and HBM above a silicon interposer, fine die-attachment connections at their interfaces, and a package substrate beneath the interposer.

Part Position Function
Logic die Above the interposer Processes data and controls memory access
HBM stack Beside the logic die Provides wide-interface, high-bandwidth memory
Microbumps Between top dies and interposer Connect die pads to interposer wiring
Silicon interposer Below logic and HBM Routes dense connections between dies
Through-silicon vias (TSVs) Through the interposer thickness Carry connections to its underside
Package substrate Below the interposer Redistributes connections toward board-facing terminals

The CoWoS S silicon interposer is primarily an interconnect platform, not an additional processor. It can also incorporate passive functions such as integrated capacitance. The silicon interposer, organic package substrate, and system PCB are therefore three distinct structures—not interchangeable names for the same board.

Underfill and the thermal assembly complete important mechanical and heat-transfer functions. Their materials depend on the qualified package design; a conceptual cross section does not specify an actual lid, thermal-interface material, or assembly thickness.

How Does CoWoS-S Connect Logic Dies and HBM?

Logic-to-HBM signals travel from a die’s pads through microbumps, laterally along the interposer’s metal wiring, and through another set of microbumps into the memory interface.

Lateral logic-to-HBM routing and vertical TSV connections to the package substrate

This lateral memory path is different from the vertical path through interposer TSVs toward the package substrate. An HBM data signal does not have to travel down to the system PCB and back up to the neighboring memory stack. Keeping many connections within the package supports a wide memory interface without routing that interface across the board.

  • Connection density: fine interposer wiring accommodates many parallel signal connections in a small area.
  • Shorter paths: adjacent die placement reduces the distance compared with off-package memory routing, although actual delay and loss depend on the layout.
  • Matched interfaces: logic memory controllers, HBM generation, pad maps, and package routing must work together. An interposer alone does not set bandwidth.

CoWoS S bump pitch refers to center-to-center spacing at a specified bump interface. It must not be confused with TSV pitch, metal line spacing, or board-level BGA pitch. For example, a hypothetical 40 µm pitch with 20 µm-wide pads leaves a nominal 20 µm edge gap; those illustrative dimensions are not a CoWoS-S specification.

Microbump, TSV, and BGA pitches refer to different package interfaces

How Is a CoWoS-S Package Manufactured?

The CoWoS-S process flow prepares the silicon interposer, attaches the logic and memory dies to it, and integrates the resulting assembly with a package substrate.

  1. Design the interconnect platform: coordinate die locations, HBM interfaces, routing layers, TSVs, and power connections.
  2. Fabricate the interposer: form fine metal interconnects and the required through-silicon connections. Large designs may use lithographic stitching across exposure fields.
  3. Prepare the backside connections: thinning and backside processing provide access to the interposer’s vertical interconnects.
  4. Attach the top dies: align the logic and HBM interfaces with the interposer connections and form the die joints, with appropriate mechanical reinforcement.
  5. Complete and verify the package: integrate the package substrate and thermal structure, then check electrical operation and package reliability.

These are public functional stages, not a recipe for TSMC’s proprietary production line. Bonding temperatures, process ordering, tolerances, and inspection limits require the applicable qualified process documentation.

Reliability depends on more than electrical continuity. Silicon, copper, underfill, and substrate materials expand differently during temperature changes. TSMC’s 2013 CoWoS reliability work examined underfill and lid choices, including AlSiC versus copper, in relation to interconnect fatigue. That dated study illustrates why the complete assembly must be qualified; it does not prescribe a lid material for every modern package.

What Limits CoWoS-S Interposer Scaling?

CoWoS-S interposer scaling is constrained by stitched routing, defect exposure, wafer utilization, assembly yield, and mechanical control—not by an absolute one-reticle limit.

TSMC has publicly described reticle stitching to extend silicon interposer area beyond a single exposure field. Its 2021 fifth-generation CoWoS-S publication reported an approximately 2,500 mm² interposer supporting multiple logic dies and eight HBM stacks, with five layers of submicron copper wiring and second-generation integrated capacitors. This is a specific published configuration, not a mandatory HBM count or permanent size ceiling.

  • Stitching: interconnects crossing exposure boundaries require controlled pattern alignment and continuity.
  • Yield: increasing area exposes more routing and structures to possible defects; the economics also depend on die and assembly yields.
  • Mechanical integration: larger assemblies require coordinated substrate, underfill, lid, and cooling design.
  • Usable floorplan: logic sizes, HBM placement, power regions, and keep-outs determine how much area is actually available.

Interposer area is not the package outline: the substrate and thermal assembly can extend beyond it. Nor can an area in square millimeters be converted into a package width without knowing the shape.

For larger integration footprints, CoWoS-L uses a wider RDL platform with local silicon interconnects rather than extending one continuous silicon interposer. The TSMC CoWoS overview describes these architecture options; their full comparison belongs in a separate S-versus-L discussion.

Where Is CoWoS-S Packaging Used?

CoWoS-S packaging is used in high-end computing products that need dense connections between processing dies and high-bandwidth memory, including AI acceleration and HPC.

  • AI accelerators: memory bandwidth helps feed compute units with model weights and intermediate data. The benefit depends on the workload, not simply the presence of HBM.
  • HPC processors and accelerators: simulations and other data-intensive calculations can require substantial bandwidth between compute and memory.
  • Multi-die computing designs: a shared interposer provides dense routing between separately fabricated logic and memory components.

TSMC’s 2022 annual report identifies CoWoS-S with high-end HPC and AI and describes HBM3-related qualification. Together with the published eight-HBM fifth-generation example, this provides concrete application context without assuming that every GPU, accelerator, or HBM product uses CoWoS-S. A named chip’s packaging variant still requires its own disclosure.

How Does a CoWoS-S Package Connect to a PCB?

A finished CoWoS-S component connects to the PCB through its package substrate and board-facing terminals, commonly a BGA interface—not through the interposer’s microbumps.

Four distinct levels: dies, interposer, package substrate, and system PCB

The component’s released land pattern, ball map, power requirements, and assembly guidance determine the board design. The PCB routes external interfaces and supplies power; it does not reproduce the fine logic-to-HBM wiring inside the package.

  • Layout: select escape routing and via structures from the actual BGA geometry.
  • Power: design planes, regulator connections, and board decoupling for the component’s current and voltage limits.
  • Assembly: follow the device-specific handling, reflow, warpage, and inspection requirements.
  • Mechanical support: coordinate heatsink attachment, board support, and package keep-outs.

Our guide to advanced HDI PCBs explains board-level routing structures. At EBest Circuit, we support PCB fabrication and PCBA projects around qualified components; we do not claim to manufacture TSMC’s interposers. Send your Gerber files, stackup, BOM, quantities, and package assembly requirements to sales@bestpcbs.com for a board-level review.

FAQs About CoWoS-S

Is CoWoS-S the same as SoIC?

No. CoWoS-S provides interposer-based integration, while SoIC addresses a different level of die stacking and bonding. These technologies can be combined in a larger integration scheme.

Are HBM TSVs and interposer TSVs the same structures?

No. HBM TSVs connect dies within the memory stack; interposer TSVs pass through the separate silicon interposer. They occupy different parts of the package.

Does every CoWoS-S package contain eight HBM stacks?

No. Eight stacks describe one published configuration. The actual count depends on the processor interfaces, memory requirements, floorplan, and qualified package design.

Does reticle stitching mean joining separate pieces of silicon?

No. It joins lithographically patterned regions across exposure fields on the interposer. It does not mean gluing individual silicon tiles together.

Can HBM be replaced like a DIMM?

No. HBM is integrated into the package assembly rather than installed in a board-level memory socket. Replacement is not equivalent to changing a server DIMM.

You may also like

Tags: , , , ,
Posted in best pcb, PCB Design, PCB Technology | No Comments »

M8 vs M9 CCL: Key Differences for AI Server PCBs

September 18th, 2026

M8 vs M9 CCL is becoming a core material question for AI server PCB design. As 800G networks move toward 1.6T architectures and high-speed SerDes channels move toward 224G-class signaling, M9-class CCL is increasingly discussed as the next step beyond mature M8-class materials.

However, the real engineering question is not whether M9 has a lower nominal Df. The question is whether the complete PCB channel needs enough additional loss margin to justify the higher material cost, tighter supply chain, qualification work, and more demanding multilayer fabrication process.

M8 vs M9 CCL comparison for AI server PCBs showing high-speed multilayer constructions

Key Takeaways

  • M8 and M9 are performance-class descriptions, not universal IPC laminate grades. The exact resin, glass fabric, copper foil, and electrical properties depend on the supplier and material part number.
  • M8-class CCL is already a mature ultra-low-loss option for many 800G switches, AI servers, and high-speed networking boards.
  • M9-class CCL targets even lower transmission loss for next-generation 1.6T networking, longer SerDes channels, and systems with tighter channel-loss budgets.
  • The transition from M8 to M9 is not simply a lower-Df resin upgrade. It usually involves the complete material system: resin, glass fabric, copper profile, lamination behavior, and PCB process control.
  • 224G-class signaling does not automatically require M9. Channel length, via count, connector loss, copper roughness, routing topology, and insertion-loss budget all matter.
  • A 52-layer PCB does not automatically need M9 on every signal layer. Material choice should follow the electrical requirement of each critical channel.
  • M8 remains a practical choice when simulation and validation show sufficient insertion-loss, impedance, and reliability margin.
  • For RFQs, “M9 material” is not enough. Buyers should specify the exact laminate, Dk/Df conditions, copper foil, glass style, stackup, impedance, insertion-loss target, and qualification requirements.

What Do M8 and M9 Mean in CCL Materials?

M8 and M9 are commonly used as performance-class descriptions for high-speed, low-loss CCL systems, rather than as universal IPC material designations.

Two materials described as M9 can therefore have different:

  • Resin chemistries
  • Glass fabrics
  • Copper foil profiles
  • Resin contents
  • Dk and Df values
  • Test methods
  • Processing characteristics

This distinction matters during procurement. A PCB drawing that simply states “M9 material” still leaves too many variables open for accurate stackup design, loss modeling, purchasing, and fabrication.

The correct starting point is the CCL manufacturer and exact material part number, followed by its tested electrical properties, copper construction, glass style, and approved alternatives.

Exploded M8-class and M9-class CCL layer structures with copper foil resin and glass fabric

M8 vs M9 CCL: What Are the Main Differences?

M8-class CCL is already an ultra-low-loss material system for high-speed computing and networking, while M9-class development pushes dielectric loss, conductor loss, dimensional stability, and multilayer process control further for longer and faster channels.

Because M8 and M9 are not universal IPC grades, the most useful comparison combines the general performance direction with published material examples. Panasonic MEGTRON 8 provides a documented M8-class reference, while Doosan DS-7409DYQ illustrates the direction of next-generation material development for 1600G networking and AI accelerators.

Comparison M8-Class CCL M9-Class / Next-Generation CCL
Typical platform target 800G switches, current AI servers, high-speed networking 1.6T switches, next-generation AI accelerators, longer ultra-high-speed channels
Published material example Panasonic MEGTRON 8 R-579Y(U) / R-579Y(N) Doosan DS-7409DYQ
Published application positioning High-speed networking; supports 800GbE Super ultra-low loss for 1600G and AI accelerators
Dk example 3.08 / 3.13 @ 14 GHz 2.50 @ 10 GHz
Df example 0.0012 / 0.0016 @ 14 GHz 0.0006 @ 10 GHz
Tg, DMA 220°C 220°C
Td, 5% weight loss 370°C 380°C
T288 >120 min 120 min
X/Y-axis CTE 17–20 ppm/°C 6 ppm/°C
Z-axis CTE below Tg, α1 50 ppm/°C 25 ppm/°C
Z-axis CTE above Tg, α2 270 ppm/°C 150 ppm/°C
Thermal conductivity Depends on exact construction 0.4–0.5 W/m·K
Water absorption 0.06% 0.06%
Published copper example H-VLP3, 1 oz HVLP3, 1 oz
Published peel strength 0.7 kN/m 0.5 kgf/cm (~0.49 kN/m)
Resin / dielectric direction Mature ultra-low-loss system Further optimized extremely low-loss system
Glass direction Low-Dk / ultra-low-Df glass options Greater emphasis on very-low-loss and dimensionally stable glass construction
Copper requirement Low-profile copper already important Copper roughness becomes more critical as dielectric loss falls
Manufacturing window Relatively mature for experienced high-speed PCB fabs Tighter lamination, registration, copper-interface, thickness, and reliability control
Supply maturity More mature and broadly qualified Fewer qualified material and stackup combinations
Cost direction High Generally higher
Best selection basis Meets channel loss with sufficient engineering margin Consider when M8 no longer provides enough channel-loss or qualification margin

These values show the direction of material development, but they should not be treated as a direct M8-to-M9 percentage comparison. The Panasonic and Doosan data use different frequencies, material constructions, and potentially different measurement conditions. For an actual PCB design, engineers should compare qualified laminate part numbers using the same Dk/Df test method, frequency, glass style, resin content, and copper profile.

M8 vs M9 CCL performance comparison for AI server PCB materials

Why Are AI Server PCBs Moving from M8 Toward M9?

The main reason is that the available channel-loss budget is becoming tighter as data rates increase and signal channels become more difficult.

Several trends are happening at the same time:

  • 112G-class channels are moving toward 224G-class signaling.
  • 800G network architectures are moving toward 1.6T.
  • AI server boards are using more complex multilayer stackups.
  • Long SerDes channels may pass through more vias and connectors.
  • Phase consistency and impedance control are becoming more demanding.

Higher data rate does not automatically mean M9 is mandatory. A short 224G-class channel with limited via transitions and carefully controlled copper roughness may have a very different loss budget from a long midplane or switch-board channel.

The decision therefore depends on the entire transmission path, including routing length, connector loss, via structure, copper profile, stackup, and any retimers in the architecture.

AI server PCB transition from 112G to 224G and 800G to 1.6T with tighter loss budget

How Do M8 and M9 Differ in Dk, Df and Transmission Loss?

Df is usually the first specification engineers compare because it directly influences dielectric loss, but Dk and the test conditions behind both values also matter.

Public material data show the general direction clearly. MEGTRON 8 publishes Dk values around 3.08/3.13 and Df values around 0.0012/0.0016 at 14 GHz, while DS-7409DYQ publishes Dk 2.50 and Df 0.0006 at 10 GHz.

Df mainly affects dielectric loss. As trace length and frequency increase, a lower Df can reduce the amount of signal energy lost in the dielectric.

Dk affects impedance, propagation velocity, and trace geometry. A lower Dk can help some high-speed designs, but the lowest nominal Dk is not automatically the best choice. Stability across frequency, temperature, glass construction, and production lots can be just as important.

For real channel modeling, engineers should use material data measured under conditions that match the qualified stackup as closely as possible rather than comparing datasheet numbers from unrelated test methods.

Why Is M9 a Material-System Upgrade Rather Than Just a Lower-Df Resin?

M9-class development is better understood as a system-level material optimization. Reducing resin Df alone does not guarantee sufficiently low PCB channel loss.

Resin System

The resin must reduce dielectric loss while still supporting multilayer PCB manufacturing. Engineers also need to consider resin flow, thermal reliability, adhesion, dimensional behavior, and CAF resistance.

A resin that looks excellent electrically but creates an unstable lamination process does not solve the complete PCB problem.

Glass Fabric

Glass fabric contributes to effective Dk, Df, dimensional stability, and fiber-weave behavior. Advanced low-loss glass constructions can help reduce dielectric loss and limit local variations that contribute to skew and phase mismatch.

Glass style also affects pressed dielectric thickness and manufacturing behavior, so it should be defined at stackup level rather than treated as an invisible material detail.

Copper Foil

Once dielectric loss is reduced, conductor loss becomes a larger part of the total channel budget. Smoother copper becomes increasingly valuable because surface roughness increases conductor loss at high frequency.

This is why many M9-class designs place greater emphasis on very-low-profile copper in addition to lower-loss resin and glass. Our existing HVLP Copper Foil for AI Server PCBs guide explains this conductor-loss mechanism in more detail.

M9 material system upgrade showing low-loss resin advanced glass fabric smooth copper foil and multilayer PCB

How Much Does Copper Roughness Matter in M8 and M9 PCBs?

Copper roughness can determine how much of the theoretical low-loss advantage of a laminate remains after PCB fabrication.

At high frequency, current concentrates near the conductor surface because of the skin effect. If that surface is rough, the effective current path becomes more complex, increasing conductor loss.

This means a lower-Df resin can still deliver disappointing channel performance if the signal layer uses an unsuitable copper surface.

For M8 and M9 projects, engineers should evaluate:

  • Copper foil type
  • Roughness data such as Rz or Rq
  • Treatment side
  • Foil thickness
  • Inner-layer surface treatment
  • Final insertion-loss performance

The incoming copper foil grade is only one part of the story. PCB processing can also change the effective copper-dielectric interface, so overly aggressive inner-layer treatment can reduce the benefit of starting with smoother foil.

M9 does not automatically mean HVLP5. Published next-generation low-loss material data can still include HVLP3 copper constructions, which is why the M9 label alone should never be translated into a mandatory copper grade.

Smooth copper versus rough copper showing lower and higher conductor loss in high-speed PCB traces

When Is M8 CCL Still Enough for an AI Server PCB?

M8 remains a strong choice when the complete channel meets the required electrical targets with sufficient engineering margin.

Typical situations include:

  • High-speed channels are relatively short.
  • Via transitions are limited.
  • Connector loss is manageable.
  • Simulation shows acceptable insertion loss.
  • Existing M8 materials are already customer-qualified.
  • Production experience with the stackup is mature.
  • Supply availability and lead time are important.
  • Only some signal layers carry the most demanding SerDes channels.

An M8 stackup may therefore remain entirely appropriate even in an advanced AI server platform.

If an M8-based design already meets insertion-loss, impedance, and reliability requirements with sufficient margin, moving to M9 may increase cost without producing a measurable system-level benefit.

This is especially important for boards where only a small portion of the routing approaches the channel-loss limit.

When Should Engineers Consider M9 CCL?

M9 becomes more relevant when the M8 design begins to consume too much of the available channel-loss margin.

Typical triggers include:

  • Long 224G-class SerDes channels
  • 1.6T switching architectures
  • High-loss midplane or backplane routes
  • Multiple via transitions
  • Very dense high-speed routing
  • M8 simulation results close to the insertion-loss limit
  • Customer-specified or platform-qualified M9 materials
  • Architectures trying to preserve margin without adding additional retimers

The decision should still be based on the actual channel model. A short trace on a high-layer-count PCB may have less need for M9 than a long route on a lower-layer board.

The trigger for M9 should be the channel-loss budget and platform-qualification requirement, not the PCB layer count alone.

Does a 52-Layer or Higher-Layer PCB Automatically Require M9?

No. A 52-layer or higher-layer PCB does not automatically require M9 material on every layer.

A high-layer-count AI server PCB may contain:

  • Critical high-speed SerDes signal layers
  • Lower-speed control signals
  • Clock or management interfaces
  • Power planes
  • Ground planes
  • Auxiliary signal layers

Only some of these layers may be sensitive enough to justify the most aggressive low-loss construction.

In some designs, engineers may consider a hybrid material strategy in which the most demanding signal regions use a higher-performance material system while other layers follow a different qualified construction. Whether this is practical depends on lamination compatibility, reliability, stackup design, and the fabricator’s process capability.

Layer count therefore tells you how difficult the board may be to manufacture, but it does not by itself define the required loss class.

What PCB Manufacturing Challenges Increase with M9-Class CCL?

M9-class materials can narrow the PCB manufacturing window because the electrical advantage must be maintained through lamination, drilling, plating, and inner-layer processing.

Important manufacturing challenges include:

  • Multilayer lamination control
  • Resin-flow consistency
  • Pressed dielectric thickness
  • Layer-to-layer registration
  • Copper adhesion
  • Inner-layer surface treatment
  • Mechanical drilling quality
  • Plated-through-hole reliability
  • Controlled impedance
  • Insertion-loss verification
  • Material-lot control

Extremely smooth copper needs enough adhesion without introducing excessive surface roughness during inner-layer treatment. This creates a direct trade-off between electrical performance and process robustness.

Advanced glass constructions can also change drilling behavior. Where a qualified M9 construction uses harder low-loss glass, tool wear, hole-wall quality, and drilling parameters may require tighter control.

Before quoting an M9 PCB, the fabricator should review the exact laminate system, stackup, copper foil, glass style, finished thickness, via structure, impedance requirements, and customer qualification rules.

How Should Engineers Choose Between M8 and M9 for a Real PCB Stackup?

The most reliable selection method is to start with the channel requirement and work backward to the material system.

  1. Define the signaling rate and channel topology. Identify connectors, vias, trace lengths, and any retimers.
  2. Build the insertion-loss budget. Determine how much loss the PCB portion of the channel can consume.
  3. Identify the longest and most critical channels. Do not optimize every route based on the worst case unless necessary.
  4. Simulate an M8 stackup using realistic copper roughness. Nominal Df alone is not enough.
  5. Compare the remaining margin with an M9 alternative. Evaluate whether the improvement is electrically meaningful.
  6. Confirm the stackup with the PCB fabricator. Check actual cores, prepregs, glass styles, copper foils, thickness tolerances, and material availability.

The fabricator should be involved before the stackup is frozen. A theoretically ideal dielectric thickness or glass style may not be the construction the factory can source and process consistently.

This step becomes more important as layer count increases because small dielectric-thickness and registration changes can affect impedance, board thickness, and production yield across many lamination interfaces.

What Should Buyers Include in an M8 or M9 PCB RFQ?

An RFQ should define the material system and electrical target clearly enough that the PCB manufacturer does not have to guess what “M8” or “M9” means.

RFQ Item Why It Matters
CCL manufacturer and exact part number Avoids ambiguous M8/M9 descriptions
Approved alternative materials Supports supply continuity
Dk/Df and test frequency/method Makes electrical data comparable
Copper foil type and roughness Controls conductor-loss assumptions
Glass style Affects loss, skew, thickness, and processing
Resin content Influences pressed dielectric behavior
Layer count Defines stackup complexity
Finished board thickness Sets lamination target
Impedance table Defines trace geometry requirements
Insertion-loss target Establishes electrical acceptance criteria
Coupon / S-parameter requirement Defines production verification
Customer-approved material list Controls qualification compliance

For a high-speed AI server project, the RFQ should also state the via structure and identify the most critical high-speed layers whenever possible.

Writing only “M9 material, 52 layers” is not enough for a reliable technical quotation because two suppliers may interpret that requirement using very different laminate, copper, and glass constructions.

M8 and M9 PCB RFQ checklist including exact material Dk Df copper foil glass style stackup impedance and insertion loss

FAQ About M8 and M9 CCL

1. Is M9 CCL always better than M8 CCL?
No. M9 CCL generally targets lower transmission loss, while M8 CCL remains suitable when it meets the channel budget with sufficient margin.

2. Can M8 CCL support 224G-class signals?
Potentially, yes. The result depends on trace length, copper roughness, via count, connectors, routing topology, and the complete insertion-loss budget rather than the M8 label alone.

3. Does every M9 CCL use Q-glass?
No. The search phrase M9 CCL Q glass describes one material direction, but the exact glass fabric must be checked in the qualified laminate construction.

4. Does M9 CCL require HVLP5 copper foil?
No. The required copper profile depends on the qualified material system, channel-loss target, signal layer, and customer approval.

5. Does a 52-layer PCB need M9 on every layer?
No. Material selection should follow the electrical role of each layer and the loss requirements of the critical channels rather than total layer count alone.

6. What information is needed to quote an M9 PCB?
Provide the exact laminate, approved alternatives, stackup, copper foil, glass style, finished thickness, impedance requirements, insertion-loss target, via structure, quantity, and qualification requirements.

Selecting between M8 and M9 requires more than comparing nominal Dk and Df values. Channel length, copper profile, glass construction, stackup, via structure, manufacturing capability, and customer qualification requirements should be evaluated as one system.

For an AI server or high-speed network PCB review, send your Gerber files, stackup, target material, impedance table, insertion-loss requirements, finished board thickness, via structure, and expected quantity to sales@bestpcbs.com.

You may also like

Tags: , , , ,
Posted in AI PCB, best pcb, PCB Design, PCB Technology | No Comments »

PCB Cards: Types, Card Edge Design, Assembly & Manufacturing Guide

September 18th, 2026

The term PCB cards appears frequently in electronics, but it does not describe one standardized type of printed circuit board. An engineer may use the term for a plug-in communication card, a bare circuit card, an assembled controller board, or even an NFC business card.

For PCB manufacturers, that ambiguity matters. Before fabrication or assembly begins, the actual card function, mechanical interface, component state, stackup, connector requirements, and testing scope must be defined. This guide explains the common meanings of PCB cards and the design and manufacturing details that make card-type PCBs different.

PCB cards including a card-edge board, assembled controller card, NFC PCB card, and diagnostic test card

Key Takeaways

  • A PCB card is an informal term for a printed circuit board used as an electronic card or module. Depending on context, it may mean a bare PCB, an assembled circuit card, or a plug-in board.
  • PCB cards include controller cards, PCIe-style cards, communication cards, card-edge modules, diagnostic cards, NFC cards, and PCB business cards.
  • PCB card, circuit card, PCBA, and circuit card assembly are related terms but should not automatically be treated as identical.
  • Plug-in PCB cards often use gold fingers along the board edge to mate directly with a socket or card-edge connector.
  • Gold-finger design depends on connector pitch, PCB thickness, plating, bevel, mechanical tolerance, insertion cycles, and the manufacturer’s fabrication process.
  • A PCB being called a “card” does not define its laminate. Standard FR-4, high-Tg FR-4, and low-loss materials can all be used depending on the application.
  • Manufacturing files should clearly specify the board outline, connector interface, gold fingers, stackup, impedance, assembly data, and testing requirements.

What Is a PCB Card?

A PCB card is a printed circuit board used as a functional electronic card or module. Depending on the industry and project documentation, it can refer to either a bare printed circuit board or a populated assembly.

The word “card” is especially common when the board is designed to:

  • Plug into another PCB or backplane
  • Fit into a defined slot or chassis
  • Act as a replaceable functional module
  • Use edge contacts as an electrical interface
  • Carry a specific control, communication, storage, or test function

Typical examples include network interface cards, industrial I/O cards, motor-control cards, PCIe expansion cards, communication modules, and diagnostic cards.

The important point is that “PCB card” does not automatically tell a manufacturer whether components are assembled. That information must come from the BOM, assembly drawing, pick-and-place data, and purchasing specification.

What Are the Main Types of PCB Cards?

PCB cards can be grouped by their mechanical interface, assembly state, and intended function.

PCB Card Type Typical Example Main Characteristic
Bare circuit card Unassembled controller PCB PCB fabrication only
Assembled circuit card Industrial control module Components already mounted
Plug-in PCB card PCIe or communication card Inserts into a socket or backplane
Card-edge PCB Memory or interface module PCB edge acts as connector
Test card Diagnostic or interface board Used for test and verification
NFC PCB card Smart identification card PCB antenna and NFC IC
PCB business card Promotional electronic card PCB used as functional branding item

The same card can fit more than one category. For example, a communication board can be both an assembled circuit card and a card-edge PCB.

For manufacturing, the functional label is less important than the actual requirements: stackup, board dimensions, copper thickness, connector geometry, surface finish, components, and test criteria.

Main types of PCB cards including bare circuit card assembled card card-edge module test card and NFC PCB card

PCB Card vs PCB vs PCBA vs Circuit Card Assembly

These terms overlap, but they describe different things in manufacturing documentation.

Term Most Common Meaning
PCB Printed circuit board; may refer to a bare board or the general technology
PCB card Informal term for a board used as an electronic card/module
Circuit card Often another name for a PCB or functional electronic card
PCBA Printed circuit board assembly with components installed
CCA Circuit card assembly; populated functional card
Plug-in card PCB module designed to insert into another system
Card-edge PCB PCB with plated edge contacts that mate directly with a connector

A bare PCB card has copper circuitry, holes, pads, solder mask, and surface finish but no electronic components. Once components are assembled, it becomes a PCBA or circuit card assembly.

In procurement documents, engineers should therefore avoid relying only on the words “PCB card” or “circuit card.” Specify whether the requirement covers bare-board fabrication, component assembly, programming, functional testing, or a complete tested module.

For a broader terminology comparison, see our guide to Circuit Card vs Circuit Board.

PCB card versus PCB versus PCBA versus CCA terminology comparison

How Does a Plug-In PCB Card Work?

A plug-in PCB card is designed as a removable electronic module that connects electrically and mechanically to another board, backplane, or system socket.

A typical connection path is:

PCB card → edge contacts or connector → motherboard/backplane → system power and signals

The card may carry:

  • Processor or FPGA circuitry
  • Network interface
  • Storage interface
  • Industrial I/O
  • Motor control
  • Power conversion
  • Data acquisition
  • Test electronics

Unlike a permanently wired PCB, a plug-in card must also satisfy mechanical requirements such as insertion alignment, connector retention, board thickness, chassis position, and repeated mating.

High-speed cards add another layer of complexity because the connector interface becomes part of the signal channel. Differential-pair geometry, breakout routing, via stubs, reference-plane continuity, and connector insertion loss may all affect performance.

This is why plug-in PCB cards should be treated as both an electrical design and a mechanical interface.

Plug-in PCB card showing gold fingers mating with a card-edge connector for power and signals

What Is a PCB Card-Edge Connector?

A PCB card-edge connector uses plated contacts along the edge of the PCB itself as one half of the connector interface.

The PCB slides into a matching socket, and spring contacts inside the connector press against the plated pads—commonly called gold fingers.

A card-edge interface may include:

  • Single-sided or double-sided contacts
  • Different contact pitches
  • Power and signal contacts
  • Ground contacts
  • Staggered contact lengths
  • Keying slots
  • Beveled insertion edges
  • Controlled PCB thickness

This construction removes the need for a separate board-mounted connector on the card itself, which can save space and reduce component count.

However, the PCB edge becomes a precision mechanical feature. Board thickness, routing profile, bevel, plating, pad position, and connector tolerance all need to match the mating connector drawing.

For high-speed interfaces, the card edge also becomes part of the controlled-impedance signal path.

What Design Rules Matter for PCB Card Gold Fingers?

Gold fingers must provide low contact resistance, corrosion resistance, and sufficient wear resistance for the expected mating cycles.

The main design items include:

  • Contact pitch
  • Finger width and length
  • PCB thickness
  • Nickel and gold plating requirements
  • Solder-mask clearance
  • Copper-to-edge spacing
  • Bevel angle and depth
  • Keying position
  • Connector insertion depth
  • Required mating-cycle life

Hard gold is commonly preferred for repeated-mating card-edge contacts because it offers better wear resistance than standard solderable surface finishes. The exact gold thickness should follow the connector life, customer specification, and fabrication capability rather than a universal value.

The leading edge may also require a bevel to reduce insertion force and prevent the connector contacts from being damaged. Bevel geometry is determined by the card thickness and mating connector; common designs may use an angled edge in roughly the 30°–45° range, but the connector drawing should control the final specification.

Gold fingers should also be kept free of solder mask, silkscreen, solder paste, surface contamination, and routing damage. For high-current card contacts, power fingers may need wider copper geometry than ordinary signal contacts.

PCB card gold finger design showing hard gold pitch bevel edge and board thickness

What PCB Materials and Stackups Are Used for Card-Type Boards?

“PCB card” describes the board’s role, not its laminate type. The correct material depends on electrical performance, mechanical strength, thermal environment, and product qualification requirements.

Application Common Material Direction
General controller card Standard FR-4
Industrial card High-Tg FR-4
High-speed communication card Low-loss / high-speed laminate
High-current power card Heavier copper or thicker copper planes
Flexible interface card FPC
Compact 3D module Rigid-flex PCB

A simple industrial I/O card may use a conventional multilayer FR-4 construction, while a PCIe or high-speed networking card may require lower Dk/Df materials, controlled impedance, tighter dielectric tolerances, and insertion-loss verification.

Mechanical requirements also matter. Card-edge connectors are often designed for a specific finished board thickness, so the PCB stackup must meet both electrical and connector-fit requirements.

For high-speed cards, stackup decisions should also consider differential impedance, reference-plane continuity, trace-to-plane spacing, copper roughness, glass weave, via structure, and insertion-loss budget.

The material should therefore be selected from the interface and channel requirements rather than from the word “card.”

What Are PCB Test Cards Used For?

A PCB test card is a board designed to support electrical testing, diagnostics, qualification, or connection between a device under test and test equipment.

Depending on the project, the term may describe:

  • Diagnostic card
  • Interface card
  • Fixture interface PCB
  • Production test board
  • Burn-in card
  • Signal breakout card
  • Calibration board

For example, a test card can connect production equipment to a DUT through pogo pins, edge connectors, sockets, or cable interfaces.

Some test cards carry active circuits for signal conditioning or measurement, while others mainly route signals between test equipment and the product.

Because PCB test card is not one standardized construction, an RFQ should define the actual application. Useful information includes DUT interface, test voltage/current, signal frequency, connector type, insertion cycle requirements, expected test volume, controlled-impedance requirements, and functional test procedure.

This prevents a manufacturing supplier from treating a demanding high-cycle test card like an ordinary low-volume PCB.

What Are NFC PCB Cards and PCB Business Cards?

NFC PCB cards are thin printed circuit boards that integrate an NFC antenna, NFC IC, and sometimes LEDs, sensors, QR codes, or other electronics.

Most NFC systems operate at 13.56 MHz. A typical NFC PCB card may include:

  • PCB loop antenna
  • NFC IC
  • Matching or tuning components
  • Memory
  • LED indicator
  • QR code or printed information

PCB business cards often use dimensions close to the ISO ID-1 credit-card format, approximately 85.60 × 53.98 mm, although custom sizes are also common.

A PCB business card does not have to include NFC. It may instead use a QR code, LED circuit, USB interface, small development circuit, measurement reference, functional tool, or decorative copper artwork.

For an NFC version, antenna geometry is critical. Trace width, spacing, number of turns, board thickness, copper environment, nearby ground planes, and metal objects can change antenna inductance and resonance.

The electrical tuning should therefore be verified on the final PCB construction rather than assumed from the CAD geometry alone.

NFC PCB card and PCB business card showing antenna coil NFC IC and 13.56 MHz operation

How Are PCB Cards Manufactured and Assembled?

PCB card manufacturing follows the standard PCB production flow, with additional attention to edge geometry and connector interfaces where required.

A typical process is:

Gerber / ODB++ review → PCB fabrication → profile routing → surface finish → gold-finger processing → electrical test → SMT → THT → inspection → functional test

For a card-edge PCB, fabrication may require extra controls for:

  • Gold-finger plating
  • Edge bevel
  • Connector key slots
  • Board thickness
  • Finger position
  • Profile tolerance
  • Contact-edge quality

During PCBA, the production flow can include solder paste printing, SPI, SMT placement, reflow soldering, AOI, THT insertion, wave or selective soldering, X-ray where required, programming, and functional testing.

The card connector area should be protected during assembly so that solder, flux, scratches, or handling contamination do not degrade the contact surface. Dimensional inspection is also more important than on a PCB that never mates with a precision slot.

PCB card manufacturing and testing flow from Gerber design to fabrication gold fingers SMT AOI and functional test

How Should PCB Cards Be Inspected and Tested?

The test plan should match the PCB card’s function rather than automatically applying every available inspection method.

For bare PCB cards, common checks include:

  • Electrical continuity/isolation test
  • Board dimensions
  • Finished thickness
  • Hole and slot dimensions
  • Gold-finger geometry
  • Surface finish
  • Controlled impedance where required

For assembled cards, additional inspection can include SPI, AOI, X-ray, ICT, flying-probe test, functional test, and firmware programming.

Card-edge modules may also require contact continuity, connector fit, insertion/removal inspection, finger plating verification, and mechanical gauge checks.

High-speed cards may require TDR, impedance coupon testing, insertion loss, S-parameters, eye-diagram testing, or system-level validation. Not every PCB card needs these tests; the acceptance plan should follow the interface, signal rate, reliability target, and customer specification.

What Files Are Needed to Quote a Custom PCB Card?

A complete RFQ should define both the PCB and the card interface.

File / Requirement Purpose
Gerber or ODB++ Defines PCB fabrication data
Stackup Defines layer and dielectric structure
Board outline Controls card dimensions
Mechanical drawing Defines slots, cutouts and tolerances
Connector drawing Confirms mating interface
Gold-finger specification Defines plating and bevel requirements
BOM Defines assembly components
Pick-and-place file Provides SMT coordinates
Assembly drawing Confirms component orientation
Impedance table Defines controlled-impedance requirements
Test specification Defines acceptance and functional testing
Quantity Supports material and process planning

If a card-edge connector is used, also specify finished PCB thickness, finger pitch, hard-gold requirement, bevel requirement, insertion-cycle requirement, and keying dimensions.

For high-speed PCB cards, include the target stackup, differential impedances, critical interfaces, and any insertion-loss requirement.

The more clearly the interface is defined, the less risk there is that a board can pass electrical fabrication checks but fail to fit or function correctly in the final system.

FAQ About PCB Cards

1. What is a PCB card?
The search query what is PCB card usually refers to a printed circuit board used as a functional electronic card or module. It may be bare, assembled, or designed as a plug-in module.

2. Is a PCB card the same as a circuit card?
Often, but not always. Both terms can describe a printed circuit board, while the exact meaning depends on the industry and project documentation.

3. Is a PCB card the same as a PCBA?
Not necessarily. A PCBA specifically contains assembled components, while “PCB card” can also refer to a bare card-type PCB.

4. Why do PCB cards use gold fingers?
Gold fingers provide durable, corrosion-resistant electrical contacts between the PCB card and a card-edge connector.

5. Can a PCB card include NFC?
Yes. An NFC PCB card can integrate a 13.56 MHz antenna and NFC IC, along with memory, LEDs, or other functions.

6. What should I send a PCB manufacturer for a custom card quote?
Provide Gerber or ODB++, stackup, mechanical drawing, connector and gold-finger specifications, BOM, pick-and-place data, assembly drawing, impedance requirements, test requirements, and order quantity.

PCB cards range from simple controller boards to high-speed plug-in modules, so the word “card” alone is not enough to define a manufacturing requirement. Mechanical interface, assembly scope, connector geometry, stackup, surface finish, and test criteria should be specified together.

For a custom PCB card or circuit card assembly project, send your Gerber files, stackup, mechanical drawings, BOM, pick-and-place files, gold-finger requirements, and test specifications to sales@bestpcbs.com for DFM and manufacturing review.

You may also like

Tags: , , , ,
Posted in best pcb, PCB Assembly / PCBA, PCB Design, PCB Technology | No Comments »

PCB Bow & Twist: IPC Limits, Measurement & Calculation Guide

September 18th, 2026

PCB bow and twist are forms of PCB warpage that can affect solder paste printing, component coplanarity, connector fit, automated handling, and final mechanical assembly. The percentage may look small, but on a large or thin PCB, even a fraction of one percent can produce several millimeters of displacement.

For engineers and buyers, the practical questions are therefore not only “Is the board warped?” but how the deformation is classified, how it is measured, which IPC requirement applies, and whether the board meets the agreed flatness specification.

PCB Bow and Twist measurement on a precision surface plate

Key Takeaways

  • PCB bow is a curved deformation, while PCB twist is a diagonal deformation that lifts one corner out of the plane formed by the other three corners.
  • IPC-6012F specifies a default maximum bow and twist of 0.75% for printed boards using surface-mount components and 1.5% for other printed boards, unless procurement documentation specifies otherwise.
  • IPC-TM-650 2.4.22 is the key test method for determining bow and twist percentage on rigid printed boards, rigid portions of rigid-flex boards, and multiple-board panels.
  • For bow, the measured gap is divided by the corresponding board length or width. For production-method twist, the lifted-corner gap is divided by twice the board diagonal.
  • A 200 × 300 mm SMT PCB with a 0.75% limit allows 1.50 mm bow across the 200 mm direction, 2.25 mm across the 300 mm direction, and approximately 5.41 mm raised-corner gap for the IPC production twist method.
  • Asymmetric stackups, uneven copper distribution, material construction, lamination conditions, and later thermal or mechanical stress can all contribute to PCB warpage.
  • The IPC 0.75% value is an acceptance limit, not necessarily the optimum flatness target for every assembly.

What Are PCB Bow and Twist?

PCB bow and twist are two different forms of deviation from flatness. PCB bow is roughly cylindrical or spherical curvature in which the four corners of a rectangular board remain in one plane, while twist occurs along a diagonal so that one corner lies outside the plane formed by the other three.

Characteristic PCB Bow PCB Twist
Typical shape Arc, hump, or shallow dome Propeller-like diagonal distortion
Corner condition Four corners can remain coplanar One corner lifts relative to the other three
Main measurement reference Board length and width Board diagonal
Production measurement Maximum gap along an edge direction Raised-corner gap

A board can also show a combination of bow and twist. In that case, simply measuring the highest point from a tabletop does not necessarily produce the correct IPC percentage.

The test setup and calculation method need to match the type of deformation being evaluated.

PCB bow compared with PCB twist deformation

What Is the IPC Standard for PCB Bow and Twist?

IPC-6012F specifies that, unless otherwise stated in the procurement documentation, finished rigid printed boards designed in accordance with IPC-2221 have a maximum bow and twist of 0.75% when surface-mount components are used and 1.5% for other printed boards.

Board Application Default IPC-6012F Maximum
Printed board using surface-mount components 0.75%
Other printed boards 1.50%

These percentages are default acceptance requirements, not universal design targets. A customer drawing, procurement specification, or product-specific requirement can call for tighter flatness.

Finished boards are also assessed in their delivered form. If boards are supplied in pallet arrays for assembly, the bow and twist requirement for the array may be agreed separately between the user and supplier.

IPC-TM-650 2.4.22 vs 2.4.22.1: What Is the Difference?

IPC-TM-650 2.4.22 and 2.4.22.1 sound nearly identical, but they report flatness differently.

Test Method Main Purpose Result
IPC-TM-650 2.4.22C Bow and twist of rigid boards, rigid portions of rigid-flex, and multiple-board panels Percentage
IPC-TM-650 2.4.22.1C Maximum vertical displacement of panels, finished rigid boards, and rigid portions of rigid-flex Displacement in mm/in

Method 2.4.22 includes production Go/No-Go procedures for bow and twist, plus a more precise referee procedure for twist. It uses a precision surface plate, feeler or pin gauges, measuring devices, and, for the referee method, additional support and dial-indicator equipment.

Method 2.4.22.1 instead records the maximum vertical displacement of an unrestrained specimen. Its scope applies to laminates at least 0.5 mm [0.020 in] thick and can also be used after etching or thermal stress when agreed between user and supplier.

These methods should therefore not be treated as interchangeable calculations.

How Do You Measure PCB Bow?

PCB bow is measured by placing the board on a precision flat surface with the convex side facing upward and measuring the gap created by the curvature. A bow and twist PCB check should classify the deformation before selecting the measurement method.

For the IPC-TM-650 2.4.22 production method:

  1. Measure the board length L and width W.
  2. Place the board on a precision surface plate, convex side upward.
  3. For the edge being checked, apply enough pressure at both corners of that edge to bring them into contact with the datum surface.
  4. Insert a feeler or pin gauge between the PCB and surface plate.
  5. Determine the largest gauge that fits for the length and width directions.
  6. Record these measurements as RL and RW.
  7. Calculate bow percentage separately for length and width.

IPC calculates bow in the corresponding board direction. The denominator is not automatically the PCB diagonal.

This distinction matters because using the diagonal would produce a lower calculated percentage and could incorrectly classify an out-of-tolerance board as acceptable.

PCB bow measurement using a precision surface plate and feeler gauge with length width and gap dimensions

How Do You Measure PCB Twist?

For production testing, PCB twist is measured by placing three corners of the board against a flat datum surface and measuring the gap beneath the remaining lifted corner.

The IPC-TM-650 2.4.22 procedure is:

  1. Measure the board diagonal and record it as D.
  2. Place the PCB on the surface plate.
  3. Position it so that three corners contact the surface.
  4. If necessary, restrain only one corner to establish three-point contact.
  5. Insert a feeler or pin gauge under the remaining lifted corner.
  6. Find the largest gauge that fits without lifting the other three corners.
  7. Record that gap as R.
  8. Calculate the twist percentage.

The production formula contains a factor of two because constraining one corner against the surface plate approximately doubles the observed vertical twist deflection.

If three corners cannot be brought into contact by restraining only one corner, the referee method should be used instead of forcing the production procedure.

PCB twist measurement showing board diagonal D and raised corner gap R on a precision surface plate

What Is the PCB Bow and Twist Formula?

IPC-TM-650 2.4.22 uses different formulas for bow and production-method twist. A PCB bow and twist calculator must use the correct board direction or diagonal for the deformation being measured.

Bow in the length direction:

BowL (%) = (RL / L) × 100

Bow in the width direction:

BowW (%) = (RW / W) × 100

Where:

  • RL = maximum measured gap in the length direction
  • RW = maximum measured gap in the width direction
  • L = PCB length
  • W = PCB width

Twist — production method:

Twist (%) = (R / (2 × D)) × 100

Where:

  • R = maximum gap under the raised corner
  • D = PCB diagonal

The factor of two should not be removed when using this IPC production measurement method.

For a rectangular PCB, the diagonal is:

D = √(L² + W²)

How Do You Calculate PCB Bow and Twist?

Consider a 200 × 300 mm PCB that will use SMT components. Using the default IPC-6012F limit of 0.75%, the maximum allowable production-measurement gaps can be calculated before inspection.

The PCB bow and twist percentage must always be evaluated against the board dimensions and the agreed acceptance limit.

PCB dimensions

  • Width = 200 mm
  • Length = 300 mm
  • Bow/twist limit = 0.75%

First calculate the board diagonal:

D = √(200² + 300²) = 360.56 mm

The allowable gaps are:

Calculation Formula Maximum Gap
Bow across 200 mm direction 200 × 0.75% 1.50 mm
Bow across 300 mm direction 300 × 0.75% 2.25 mm
Twist production gap 2 × 360.56 × 0.75% 5.41 mm

Now assume the measured 300 mm-direction bow gap is 1.80 mm:

Bow = (1.80 / 300) × 100 = 0.60%

That result is below 0.75%.

If the measured lifted-corner gap for twist is 4.00 mm:

Twist = (4.00 / (2 × 360.56)) × 100 ≈ 0.55%

That result is also below 0.75%.

This example shows why a physical gap in millimeters cannot be judged by itself. The same displacement produces a different bow or twist percentage depending on board dimensions.

What Causes PCB Bow and Twist?

PCB bow and twist usually result from uneven internal stress rather than one isolated defect. The source can enter during design, material preparation, fabrication, or later thermal processing.

Evaluating bow and twist in PCB production starts with identifying where uneven stress entered the process.

PCB Design

Common design-related contributors include asymmetric stackups, unequal copper weights above and below the centerline, large differences in local copper density, large cutouts, and panel breakaway areas that do not reflect the copper distribution of the finished board.

Materials

Glass-fabric orientation, core and prepreg construction, CTE mismatch, mixed laminate systems, and moisture condition can affect dimensional stability.

PCB Manufacturing

Press conditions, incomplete cure, incorrect prepreg construction, uneven heating or cooling, solder-mask cure, and hot-air solder leveling can introduce or release stress.

Assembly and Handling

Mechanical loading, storage, reflow, fixtures, conveyor support, and asymmetric component mass can further change the board shape.

A useful way to think about these mechanisms is that some stress is built into the PCB during lamination, while other deformation is introduced later by handling or thermal processing.

Why Does Copper Distribution Affect PCB Warpage?

Copper and dielectric materials respond differently to temperature and processing stress, so an unbalanced copper pattern can create unequal forces through the PCB thickness.

The problem can occur in two ways:

  • Layer-to-layer imbalance: one side of the stack has heavier copper or larger plane areas than its mirrored layer.
  • Local imbalance: one region contains dense copper while another region contains very little copper.

During multilayer pressing, copper density also affects resin flow and local pressure. During later thermal cycles, different copper distributions can contribute to unequal expansion and stress.

A symmetrical stackup with similar copper weights and coverage on corresponding layers reduces this risk. Panel breakaway areas should also avoid extreme copper-density differences compared with the PCB itself.

Copper filling or thieving may help balance sparse areas, but it should still respect electrical clearance, impedance, creepage, and signal-integrity requirements.

PCB warpage comparison showing asymmetric stackup and uneven copper versus symmetric stackup and balanced copper

How Can PCB Bow and Twist Be Reduced Before Fabrication?

The most effective time to control PCB warpage is before the stackup and panel design are frozen.

Useful DFM checks include:

  • Keep the multilayer stackup symmetric around the centerline.
  • Use matching copper weights on mirrored layers where practical.
  • Balance copper coverage across opposing sides.
  • Review large copper-free or low-density regions.
  • Keep core and prepreg construction mechanically balanced.
  • Review glass orientation where material construction makes it relevant.
  • Consider board thickness relative to size and mechanical support.
  • Balance copper in panel rails and breakaway areas.
  • Review mixed-material stackups for CTE and lamination compatibility.
  • Define any tighter flatness requirement on the drawing before production.

For EBest Circuit projects, bow-and-twist risk can be reviewed during DFM together with stackup construction, copper distribution, panelization, material selection, and finished thickness.

How Is Bow and Twist Controlled During PCB Manufacturing?

Manufacturing control focuses on preventing uneven stress from being built into or released from the panel.

Important process controls include:

  • Laminate and prepreg storage
  • Glass direction and ply construction
  • Controlled lamination temperature and pressure
  • Resin-flow and cure control
  • Symmetrical panel construction
  • Copper plating balance
  • Solder-mask and legend curing
  • HASL or other high-temperature processing
  • Panel support during handling
  • Controlled cooling
  • Final flatness inspection

Lamination is particularly important because once intrinsic deformation is locked into the multilayer structure, later flattening may not provide a permanent correction.

Final inspection should be performed on the board in the required delivered form and against the applicable customer or IPC limit.

When Should a PCB Use a Tighter Flatness Limit Than IPC 0.75%?

A tighter requirement may be appropriate whenever the standard 0.75% bare-board limit does not provide enough mechanical or assembly margin.

Examples include boards with:

  • Large BGA or fine-pitch area-array packages
  • Long board-edge connectors
  • Press-fit connector fields
  • Tight card-guide or chassis interfaces
  • Thin, large-format PCBs
  • High-density SMT on both sides
  • Precision optical or sensor assemblies
  • Customer-defined coplanarity requirements

The correct tighter value depends on the product. It should not be assumed that every BGA board requires 0.5%, 0.3%, or another generic number.

IPC-6012F allows procurement documentation to specify requirements other than the default 0.75% or 1.5% limits.

For tight-flatness projects, the designer and PCB manufacturer should agree on the measurement method, delivery form, panel condition, temperature condition, and acceptance limit before fabrication.

Does IPC-TM-650 Bow and Twist Apply to an Assembled PCBA?

IPC-TM-650 2.4.22 is primarily a bare printed-board flatness test. Its scope covers individual rigid boards, rigid portions of rigid-flex boards, and multiple-board panels, but it does not address all special conditions of populated assemblies such as component weight, placement, edge supports, and connectors.

An assembled PCBA can behave differently because reflow temperature, component packages, heat sinks, connectors, fixtures, and solder joints all influence deformation.

For temperature-dependent board flatness, separate methods may be used to evaluate local board shape through a simulated reflow cycle. These do not replace room-temperature bow and twist inspection of the bare PCB.

Bare-board bow/twist and assembled-PCBA warpage should therefore be treated as related but different engineering problems.

FAQ About PCB Bow and Twist

1. What is the maximum PCB bow and twist allowed by IPC?
IPC-6012F specifies a default maximum of 0.75% for printed boards using surface-mount components and 1.5% for other printed boards, unless procurement documentation requires a different limit.

2. How is PCB bow percentage calculated?
PCB bow percentage is calculated from the maximum measured gap relative to the corresponding board length or width.

3. How is PCB twist percentage calculated?
PCB twist percentage is calculated from the raised-corner gap and the PCB diagonal using the IPC-TM-650 production method.

4. Is PCB bow the same as PCB twist?
No. Bow is a curved deformation in which the four corners can remain coplanar, while twist is a diagonal deformation that raises one corner relative to the other three.

5. What commonly causes excessive PCB warpage?
Asymmetric stackups, uneven copper distribution, material construction, lamination stress, incomplete cure, mechanical handling, and later thermal processes can all contribute.

6. Can a customer specify a tighter bow and twist limit?
Yes. A customer can define a tighter requirement in the procurement documentation, and that agreed requirement takes precedence over the general IPC default for the project.

PCB bow and twist should be controlled from stackup design through final inspection rather than treated only as an end-of-line flatness problem. Symmetry, copper balance, material construction, lamination, panel design, and thermal processing all influence the final result.

For PCB projects with tight flatness requirements, EBest Circuit can review the Gerber files, stackup, copper distribution, panelization, finished thickness, assembly requirements, and target bow/twist limit before production. Send the project data to sales@bestpcbs.com for DFM review.

You may also like

Tags: , , , ,
Posted in best pcb, PCB Design, PCB Technology, Quality | Comments Closed

PMIC Power Management Integrated Circuit: Functions, Types & PCB Design Guide

September 18th, 2026

A PMIC, or Power Management Integrated Circuit, manages the different voltage rails required by processors, memory, communication devices, sensors, and other electronic circuits. Instead of building every rail from separate regulators and control ICs, a PMIC can combine several power functions into one coordinated device.

This approach is increasingly common in embedded systems, edge AI hardware, automotive electronics, IoT products, industrial equipment, and other boards with complex power trees. However, using a PMIC does not eliminate power-design work. Rail sequencing, current capacity, external components, PCB layout, thermal management, assembly, and testing still determine whether the final system operates reliably.

PMIC power management integrated circuit supplying CPU DDR I/O sensors and USB power rails

Key Takeaways

  • PMIC stands for Power Management Integrated Circuit, a chip that can generate, regulate, sequence, monitor, and protect multiple power rails in one electronic system.
  • A PMIC may integrate buck converters, boost converters, buck-boost converters, LDOs, battery charging, power-path control, sequencing, monitoring, and fault protection.
  • PMICs are especially useful when processors, FPGAs, DDR memory, sensors, and peripherals require multiple voltages with controlled startup and shutdown timing.
  • A PMIC is not the same as a single voltage regulator or DC-DC converter. It usually coordinates several power functions rather than controlling only one rail.
  • PMIC selection should consider input voltage, rail count, output current, efficiency, quiescent current, switching frequency, sequencing, digital interfaces, package, and thermal limits.
  • PCB layout strongly affects PMIC performance. High-current loops, switching nodes, inductors, feedback routing, decoupling, grounding, and thermal vias require careful placement.
  • Most PMICs still need external components such as inductors and capacitors, even when several regulators and control functions are integrated into one IC.

What Is a PMIC Power Management Integrated Circuit?

A PMIC, or Power Management Integrated Circuit, is an IC designed to generate, regulate, distribute, sequence, monitor, and protect the power rails used by an electronic system.

A simple product may need only one regulated voltage, but a modern processor-based PCB can require several rails such as:

  • 0.8–1.0 V processor core
  • 1.1 V memory
  • 1.8 V I/O
  • 3.3 V peripherals
  • 5 V USB or auxiliary circuits

Instead of using one independent regulator for every rail, a PMIC can combine several power converters and control functions in one package.

The term PMIC can also be used broadly by semiconductor distributors for many power-management IC categories. In practical embedded design, however, “a PMIC” usually refers to a more integrated device that manages several power domains or system-level power functions.

What Functions Can a PMIC Integrate?

A PMIC can combine several power-conversion and control blocks that would otherwise require multiple ICs and supporting components.

PMIC Function What It Does
Buck converter Steps voltage down efficiently
Boost converter Steps voltage up
Buck-boost converter Regulates when input may be above or below output
LDO Provides low-noise regulated voltage
Sequencer Controls rail startup and shutdown order
Battery charger Manages charging current and voltage
Power-path controller Selects or manages available power sources
Voltage supervisor Monitors rails and generates reset/fault signals
Load switch Connects or disconnects downstream loads
Protection circuit Handles overcurrent, overvoltage, undervoltage, or thermal faults

Not every PMIC includes all of these functions. A processor-oriented PMIC may focus on several buck converters, LDOs, sequencing, and monitoring, while a battery-powered device may place more emphasis on charging, power-path control, and low quiescent current.

This integration reduces the number of separate control ICs, but external inductors, capacitors, resistors, and sometimes MOSFETs may still be required.

PMIC functional blocks including buck boost buck-boost LDO charger power path supervisor and protection

How Does a PMIC Work in a Multi-Rail Power Tree?

A PMIC sits between the main power source and the different voltage domains required by the system.

A simplified power tree might look like:

12 V / 5 V / Battery Input → PMIC → 0.9 V CPU core / 1.1 V DDR memory / 1.8 V I/O / 3.3 V sensors and peripherals / 5 V USB or auxiliary load

Each rail can have different voltage, current, noise, and startup requirements. The PMIC coordinates these rails instead of treating them as independent power supplies.

A switching converter is normally used where efficiency matters, while an LDO may supply a lower-current rail that needs less noise. Some designs also allow rails to be reprogrammed through I²C, SPI, or stored configuration.

For processor and FPGA projects, EBest Circuit can review the power tree together with PMIC placement, external power components, current paths, and PCB stackup during DFM before fabrication.

PMIC multi-rail power tree showing 12 V input and CPU DDR I/O sensor and USB outputs

How Does PMIC Power Sequencing Work?

Power sequencing controls the order and timing in which voltage rails turn on and off.

Processors, FPGAs, DDR memory, and other complex ICs may require one rail to stabilize before another starts. Applying the wrong voltage first can cause excessive current, startup failure, latch-up, or undefined device behavior.

A simplified startup sequence might be:

Core rail → Memory rail → I/O rail → Peripheral rail

A PMIC may control this using:

  • Programmable delay times
  • Soft-start ramps
  • Power-good signals
  • Enable outputs
  • Voltage monitoring
  • Reset generation
  • Fault shutdown
  • Controlled reverse sequencing during power-down

Some PMICs use fixed factory sequencing, while others allow the sequence to be programmed through registers or nonvolatile configuration.

Power-down behavior matters as much as startup. Certain processors require rails to fall in a defined order so that I/O pins, memory, or analog sections are not left biased incorrectly during shutdown.

PMIC power sequencing timing diagram with core memory I/O peripheral rails soft start delay power good and shutdown

PMIC vs Voltage Regulator vs DC-DC Converter: What Is the Difference?

A voltage regulator or DC-DC converter normally focuses on one power-conversion function, while a PMIC can coordinate several power rails and supervisory functions.

Feature PMIC Voltage Regulator DC-DC Converter
Typical rail count Multiple Usually one Usually one
Voltage regulation Yes Yes Yes
Buck/boost conversion Often several Device dependent Main function
Sequencing Often Usually limited Usually limited
Monitoring Often integrated Basic or limited Basic or limited
Battery management Possible Rare Rare
Digital programming Often available Less common Device dependent
System-level control Strong Limited Limited

An LDO is a voltage regulator. A buck converter is a DC-DC regulator. Both may be part of a PMIC.

The distinction is therefore mainly about integration and system coordination, not whether the IC can regulate voltage.

A design with one 3.3 V rail may not need a PMIC at all. A processor board with six rails, sequencing, reset logic, fault monitoring, and low-power modes is a much stronger PMIC application.

PMIC vs Discrete Power Design: When Does Integration Make Sense?

A PMIC becomes more attractive as the number of power rails and coordination requirements increase.

PMIC advantages include:

  • Smaller PCB area
  • Fewer separate control ICs
  • Integrated sequencing
  • Central fault monitoring
  • Programmable rail control
  • Reduced system-level design complexity
  • Coordinated low-power modes

A discrete design can still be the better choice when:

  • Only one or two rails are needed
  • Rail requirements change frequently
  • Different regulator vendors must be second-sourced
  • Heat needs to be distributed across the PCB
  • One rail requires an unusual converter topology
  • Independent replacement or qualification matters

A PMIC can also create sourcing concentration because several rails depend on one component. If that device becomes unavailable, replacing it may require more redesign than replacing one discrete regulator.

The decision should therefore consider not only board area and BOM count, but also lifecycle, sourcing, thermal distribution, and firmware configuration.

Discrete power design compared with integrated PMIC design on PCB

Key Specifications When Selecting a PMIC

Selecting a PMIC starts with the complete power tree, not simply the number of output channels. Input conditions, rail voltage, peak current, transient response, efficiency, sequencing, thermal limits, and control interfaces should all be checked against the processor and system requirements.

The ranges can vary significantly between PMIC families. For example, TI’s TPS65219-Q1 accepts up to 5.5 V, integrates 3 buck converters and 4 LDOs, provides up to 3.5 A from its highest-current buck, and switches at up to about 2.3 MHz. ST’s STPMIC1L also accepts 2.8–5.5 V, provides two buck converters rated up to 2 A, and operates at 2 MHz.

Specification Check Item Typical PMIC Specifications
Input voltage range Normal input, tolerance, startup and transient conditions STPMIC1L: 2.8–5.5 V; TPS65219-Q1: up to 5.5 V
Output rail count Number of buck, boost and LDO rails required TPS65219-Q1: 3 buck + 4 LDO
Output voltage range CPU core, DDR, I/O and peripheral voltages TPS65219-Q1: 0.6–3.4 V; STPMIC1L BUCK1: 0.5–4.2 V depending on mode
Output current Continuous, peak and startup current on each rail TPS65219-Q1: 3.5 A + 2 A + 2 A buck capability
Voltage accuracy Processor/DDR rail tolerance over line, load and temperature STPMIC1L buck regulation: typically around the ±1% to ±1.5% class, depending on rail and condition
Efficiency Efficiency at real operating loads, not only peak value Modern buck stages commonly operate in the 80–90%+ range, depending on VIN, VOUT and load
Quiescent current Standby power and battery-life impact TPS65219-Q1: about 250 µA typical
Switching frequency Inductor size, ripple, EMI and switching loss STPMIC1L: 2 MHz; TPS65219-Q1: up to about 2.3 MHz
Sequencing Power-up/down order, delay, ramp and power-good behavior Programmable sequencing is available on many processor PMICs
Protection OCP, OVP, UVLO, thermal shutdown and fault reporting Feature set varies by PMIC
Digital control I²C, SPI or other runtime configuration TPS65219-Q1 and STPMIC1L support I²C
Package / thermal path PCB area, exposed pad and heat transfer STPMIC1L: 4 × 4 × 1 mm VFQFPN-28

Protection features should also match the application. Useful functions can include:

  • Overcurrent protection (OCP)
  • Overvoltage protection (OVP)
  • Undervoltage lockout (UVLO)
  • Thermal shutdown
  • Short-circuit protection
  • Soft start
  • Output discharge
  • Power-good and fault reporting

Finally, do not select a PMIC from electrical specifications alone. Package size, exposed thermal pad, junction temperature, PCB copper area, thermal vias, external inductors and capacitors all affect how much usable power the PMIC can deliver in the finished board.

Where Are PMICs Used?

PMICs are common wherever electronic systems require several controlled voltage domains in limited PCB space.

Typical applications include:

  • Smartphones
  • Tablets
  • Wearables
  • IoT devices
  • Edge AI hardware
  • Embedded Linux systems
  • FPGA boards
  • Automotive infotainment
  • ADAS electronics
  • Industrial control systems
  • Medical devices
  • Networking equipment
  • Cameras and imaging systems

A smartphone PMIC may manage the processor, memory, display, camera, radio, battery, and always-on rails.

An FPGA or edge-computing board may use a PMIC to supply core, auxiliary, I/O, memory, and transceiver voltages in a defined sequence.

Automotive PMICs may add watchdogs, safety monitoring, protected outputs, and communication with the main processor. The underlying architecture changes by application, but the common requirement is coordinated management of several power domains.

What PCB Layout Rules Matter for a PMIC?

PMIC PCB layout is critical because switching converters combine high current, fast switching edges, sensitive feedback nodes, and thermal dissipation within a small area.

Important layout rules include:

  • Place input capacitors close to VIN and power ground.
  • Minimize high-di/dt switching loops.
  • Keep inductors close to switching pins.
  • Keep the SW node compact.
  • Route feedback away from noisy switching nodes.
  • Use wide copper for high-current paths.
  • Provide low-impedance ground return paths.
  • Place output capacitors close to their rails.
  • Use thermal vias under exposed pads where required.
  • Avoid routing sensitive analog signals through switching-current paths.

The input capacitor, switching FETs, inductor, and output capacitor form critical current loops. Increasing their loop area raises parasitic inductance and can worsen ringing, EMI, and voltage ripple.

Grounding also needs to follow the PMIC vendor’s recommended layout. Analog ground, power ground, thermal pad connections, and feedback return paths should not be rearranged casually because the schematic appears electrically equivalent.

EBest Circuit can review PMIC placement, high-current copper, thermal vias, power planes, component spacing, and assembly requirements during PCB DFM.

PMIC PCB layout with input capacitor short hot loop inductor feedback routing and thermal vias

What Common PMIC Design Mistakes Cause Power Problems?

PMIC problems often come from system integration rather than a defective IC.

Common mistakes include:

  • Selecting a rail with insufficient output current
  • Ignoring processor startup or transient current
  • Using the wrong startup sequence
  • Incorrect power-down timing
  • Poor input or output capacitor placement
  • Choosing the wrong inductor
  • Routing feedback close to the SW node
  • Making switching hot loops too large
  • Inadequate thermal vias
  • Excessive copper loss on high-current rails
  • Ignoring PMIC configuration or OTP settings
  • Using the wrong I²C address
  • Forgetting power-good or reset timing requirements
  • Assuming every rail can run at maximum current simultaneously

Thermal design is another frequent issue. A PMIC supplying several rails can dissipate significant heat even when individual converters operate efficiently.

The exposed thermal pad, PCB copper area, via array, airflow, ambient temperature, and neighboring heat sources all affect junction temperature.

Startup testing should therefore be performed with the real processor, memory, and peripherals rather than only checking the PMIC outputs with no load.

Common PMIC PCB design mistakes including long hot loop noisy feedback poor thermal design and bad capacitor placement

How Should a PMIC Design Be Prepared for PCB Assembly and Testing?

PMIC assembly requirements depend strongly on the package. QFN, BGA, and WLCSP devices need different stencil, placement, inspection, and rework strategies.

For QFN PMICs, review:

  • Exposed thermal pad
  • Paste aperture ratio
  • Thermal via layout
  • Solder-mask definition
  • Void control
  • Package orientation

For BGA or WLCSP packages, additional concerns include:

  • Fine-pitch pad geometry
  • Via-in-pad requirements
  • Solder-mask registration
  • Board warpage
  • X-ray inspection
  • Rework capability

A production test plan should verify more than static output voltage. Useful checks can include:

  • Rail startup sequence
  • Rail shutdown sequence
  • Output-voltage accuracy
  • Ripple
  • Load response
  • Power-good timing
  • Reset timing
  • Fault response
  • I²C/SPI communication
  • Thermal behavior under load

SPI can verify solder-paste deposition before placement, while AOI helps inspect visible joints and nearby passives. X-ray is particularly useful for hidden BGA/WLCSP joints and exposed-pad voiding where relevant.

PMIC PCB assembly and testing flow with SPI AOI X-ray and functional test equipment

FAQ About PMIC Power Management ICs

1. What does PMIC stand for?
The PMIC full form is Power Management Integrated Circuit.

2. What is PMIC in electronics?
The PMIC meaning in electronics is a chip that generates, regulates, sequences, monitors, and protects one or more system power rails.

3. Is a PMIC chip the same as a voltage regulator?
No. A voltage regulator normally controls one rail, while a PMIC chip can integrate several regulators plus sequencing, monitoring, protection, charging, or digital control.

4. Does a PMIC need external components?
Usually yes. Depending on the architecture, a PMIC may require external inductors, input and output capacitors, resistors, sense components, and sometimes external MOSFETs.

5. Why is power sequencing important in a PMIC?
Processors, memory, and I/O rails may need to start and stop in a defined order. Incorrect sequencing can cause startup failure, excessive current, or undefined system behavior.

6. Can one PMIC power a processor, memory and peripherals?
Yes, if the PMIC provides the required number of rails, voltage ranges, output currents, sequencing, and total thermal capacity.

Planning a PCB Around a PMIC?

A PMIC can simplify a complex power tree, but its performance still depends on the surrounding PCB. High-current loops, inductors, decoupling, feedback routing, thermal pads, sequencing signals, and assembly quality all need to be considered together.

EBest Circuit supports PMIC-based PCB fabrication, DFM review, fine-pitch SMT assembly, component sourcing, SPI, AOI, X-ray inspection, and functional testing. For a new power-management PCB project, send your Gerber files, BOM, schematic, stackup, and power requirements to sales@bestpcbs.com for engineering review.

You may also like

Tags: , , , ,
Posted in best pcb, Electrical Components, PCB Design, SMT Technology | No Comments »

TQFP Package: Dimensions, Pin Counts, LQFP/QFN Comparison & PCB Guide

September 17th, 2026

The TQFP package remains widely used for microcontrollers, mixed-signal ICs, communication devices, motor-control ICs, and other components that need moderate to high I/O counts without moving to BGA packaging. Their exposed gull-wing leads also make them attractive when visual solder inspection, prototyping, and rework matter.

The main challenge is that “TQFP64” or “TQFP100” does not uniquely define a package. Two ICs can have the same TQFP pin count but different body sizes, lead pitches, or overall dimensions. For PCB designers, the correct workflow is therefore part number → package drawing → land pattern, not pin count → generic footprint.

TQFP package overview showing thin body gull-wing leads visible solder joints and common pin counts

Key Takeaways

  • TQFP stands for Thin Quad Flat Package, a surface-mount IC package with gull-wing leads extending from all four sides.
  • TQFP is a package family, not one fixed footprint. Pin count alone does not determine body size, lead pitch, or PCB land pattern.
  • Common examples include TQFP32, TQFP44, TQFP48, TQFP64, TQFP100, and TQFP144, but dimensions can vary between semiconductor manufacturers.
  • For example, NXP lists both 10 × 10 mm and 7 × 7 mm TQFP64 packages, showing why “TQFP64” is not enough information for footprint selection.
  • TQFP and LQFP are closely related QFP variants, but package names alone do not guarantee footprint compatibility.
  • Compared with QFN, TQFP uses visible gull-wing leads, which makes solder-joint inspection and rework easier but requires more PCB area.
  • PCB footprints should always be built from the exact manufacturer package drawing, including pitch, body size, overall lead span, lead width, lead length, and package orientation.

What Is a TQFP Package?

TQFP stands for Thin Quad Flat Package, it is a surface-mount package with leads extending from all four sides of a thin molded body. It belongs to the broader QFP, or Quad Flat Package, family. The leads are formed into a gull-wing shape so they can sit on PCB pads and be soldered by standard SMT reflow processes.

Typical TQFP characteristics include:

  • Leads on four sides
  • Gull-wing lead shape
  • Surface-mount assembly
  • Relatively low package profile
  • Fine lead pitch
  • Visible solder joints
  • Pin counts from a few dozen to well above 100

A TQFP package is commonly used when the IC needs more I/O than SOIC or similar two-sided packages can provide but the design does not require BGA-level interconnect density.

How Is a TQFP Package Constructed?

A TQFP package normally contains a silicon die mounted inside a molded plastic body and electrically connected to an external lead frame.

Its basic construction includes:

  • Silicon die
  • Die attach material
  • Bond wires or equivalent internal interconnection
  • Copper-alloy lead frame
  • Mold compound
  • Gull-wing leads
  • Pin 1 orientation mark

The lead frame carries signals from the silicon die to the external terminals. After molding, the leads extend outward from all four sides and bend downward toward the PCB.

The package body itself is smaller than the total installed footprint because the leads extend beyond the molded body. This distinction matters during placement and land-pattern design.

Exploded TQFP package construction showing mold compound silicon die bond wires lead frame gull-wing leads and pin 1

What Dimensions Define a TQFP Package?

TQFP package dimensions include several mechanical values, not just the molded body width.

Parameter What It Describes
D × E Molded body length and width
HD × HE or overall D × E Total span including leads
A Overall package height
A1 Standoff above the PCB seating plane
e Lead pitch
b Lead width
L Gull-wing lead length
N Total lead count

Body size and overall lead span are not the same dimension. A PCB footprint designed only from the stated body size can miss the additional space occupied by the gull-wing leads.

TQFP package dimension drawing showing body size overall lead span pitch height lead width and lead length

TQFP32, 44, 48, 64, 100 and 144: What Are the Common Dimensions?

TQFP packages appear in many pin counts, but the following values should be treated as representative package examples rather than universal dimensions.

Package Example Representative Body Size Representative Pitch
TQFP32 7 × 7 mm 0.80 mm
TQFP44 10 × 10 mm 0.80 mm
TQFP48 7 × 7 mm 0.50 mm
TQFP64 7 × 7 or 10 × 10 mm examples 0.40 or 0.50 mm examples
TQFP100 14 × 14 mm 0.50 mm
TQFP144 16 × 16 mm example 0.40 mm

The package name should therefore never replace the exact IC mechanical drawing.

Representative TQFP32 TQFP44 TQFP48 TQFP64 TQFP100 and TQFP144 package examples

Does the Same TQFP Pin Count Always Mean the Same Footprint?

No. The same TQFP pin count does not guarantee the same body size, lead pitch, lead span, or PCB footprint.

TQFP64 is a clear example. Different vendors offer 64-pin packages in multiple body sizes and pitches, so a CAD library entry named only “TQFP64” is not sufficiently specific for production.

Before reusing a footprint, compare:

  • Body dimensions
  • Overall lead span
  • Lead pitch
  • Lead width
  • Lead length
  • Package height
  • Pin 1 orientation

A footprint can look plausible on screen and still be completely incompatible with the physical component.

TQFP64 comparison showing same 64 pin count with 7 by 7 mm and 10 by 10 mm packages and different pitch

TQFP vs QFP: What Is the Difference?

QFP is the broader Quad Flat Package family, while TQFP is a thinner-profile member of that family.

Both use:

  • Leads on four sides
  • Gull-wing terminals
  • Surface-mount assembly
  • Similar general soldering methods

The difference is mainly in the mechanical outline and package profile. TQFP should therefore not be treated as a package technology completely separate from QFP; it is better understood as a thinner mechanical implementation within the same four-sided leaded package concept.

TQFP vs LQFP: What Is the Difference?

In a TQFP vs LQFP package comparison, both are closely related low-profile QFP variants. Their naming conventions can overlap enough that engineers should compare actual mechanical drawings instead of relying on the acronym.

Feature TQFP LQFP
Full name Thin Quad Flat Package Low-Profile Quad Flat Package
Leads Gull-wing Gull-wing
Mounting SMT SMT
Typical profile Thin Low profile
Common pitch range 0.4–0.8 mm 0.4–0.8 mm
Footprint compatibility Only if full outline matches Same rule

Never assume that TQFP and LQFP are footprint-compatible just because the pin count is the same.

TQFP vs QFN: Which Is Easier for PCB Assembly?

In a TQFP vs QFN package comparison, TQFP is generally easier to inspect and rework because its gull-wing leads and solder joints are visible around the package perimeter.

Assembly Factor TQFP QFN
External leads Gull-wing Leadless bottom pads
Solder-joint visibility High Limited
AOI access Good Bottom joints less visible
X-ray requirement Usually not essential Often useful
Manual rework Easier More difficult
PCB area Larger Smaller
Fine-pitch defect risk Bridging / opens Insufficient wetting / voids
Thermal pad Package dependent Common on many QFNs

QFN can save significant PCB area and often offers a shorter electrical and thermal path, while TQFP remains attractive when inspection access and rework matter.

TQFP versus LQFP versus QFN package comparison

How Should a TQFP PCB Footprint Be Designed?

A TQFP PCB footprint should be created from the exact semiconductor manufacturer’s land-pattern or package drawing, not from a generic pin-count template.

Check these dimensions first:

  • Lead pitch
  • Lead width
  • Lead length
  • Molded body size
  • Overall lead span
  • Pin 1 orientation
  • Seating-plane information
  • Package tolerances

The PCB land pattern must also provide suitable solder fillets around the gull-wing leads. Toe, heel, and side fillets depend on the package geometry and the chosen land-pattern standard.

Also include clear pin 1 marking, solder-mask clearance, silkscreen that does not overlap pads, component courtyard, pick-and-place origin, and adequate neighboring-component clearance.

TQFP PCB footprint design showing pad length pitch solder mask clearance courtyard and pin 1

What SMT Assembly Problems Are Common with TQFP Packages?

The most common TQFP assembly defects involve fine-pitch leads, solder-paste volume, placement accuracy, and lead coplanarity.

Typical issues include:

  • Solder bridging
  • Insufficient solder
  • Open joints
  • Bent leads
  • Lifted leads
  • Lead coplanarity problems
  • Component misalignment
  • Wrong orientation
  • Contamination around fine-pitch pads

Solder bridging is especially common when pad geometry, stencil aperture, paste volume, or placement is poorly controlled. Fine-pitch packages leave little margin between adjacent solder deposits.

A robust SMT process may use solder paste inspection, accurate placement, controlled reflow, AOI, microscope inspection, and X-ray when hidden structures also require it.

Common TQFP SMT assembly defects including solder bridge open joint bent lead and misalignment

When Is TQFP a Good Package Choice?

TQFP is a practical choice when the design needs a moderate or high pin count while keeping solder joints visible and accessible.

It works particularly well for:

  • Microcontrollers
  • Industrial control ICs
  • Motor-control devices
  • Mixed-signal ICs
  • Communication controllers
  • Prototype and low-to-medium-volume assemblies
  • Products where rework access matters

TQFP may be less attractive when the design requires extremely high I/O density, minimum PCB area, very short high-speed interconnects, exceptional thermal dissipation, or package sizes smaller than exposed-lead QFP can provide.

For many industrial, automotive-control, instrumentation, and embedded applications, the additional board area is acceptable because the visible leads simplify inspection and troubleshooting.

FAQ About TQFP Packages

1. What does TQFP stand for?
TQFP stands for Thin Quad Flat Package. It is a surface-mount IC package with gull-wing leads on all four sides.

2. Is TQFP a surface-mount package?
Yes. TQFP is designed for surface-mount PCB assembly and is commonly soldered using SMT reflow.

3. What is the typical TQFP64 package size?
A common TQFP64 example is 10 × 10 mm with 0.5 mm pitch, but 7 × 7 mm TQFP64 packages also exist. Always check the exact IC datasheet.

4. What is the typical TQFP100 package size?
A common TQFP100 example is 14 × 14 mm with 0.5 mm lead pitch. Always verify the actual manufacturer drawing.

5. Is TQFP the same as LQFP?
No. They are closely related QFP variants, but the package profile and mechanical outline can differ. Some specific TQFP and LQFP parts may share a footprint only when all relevant dimensions match.

6. Can every TQFP with the same number of pins use the same PCB footprint?
No. Pin count alone does not define body size, pitch, lead span, or land pattern. A TQFP64, for example, can exist in multiple mechanical outlines.

If you are moving a TQFP-based design into fabrication or assembly, EBest Circuit can review the component package drawing, footprint, pad geometry, stencil requirements, orientation, BOM, and PCBA manufacturability before production. Send your Gerber files, BOM, placement files, and component datasheets to sales@bestpcbs.com for DFM review.

You may also like

Tags: , , , ,
Posted in best pcb, Electrical Components, PCB Design, SMT Technology | No Comments »

« Older Entries