Publish Time: 2026-10-12 Origin: Site
A cold solder joint is one of the few reflow defects that can look almost normal and still fail you. It forms when the solder alloy reaches its liquid state but never gets enough heat, or enough time at that heat, to wet the pad and component termination and grow a sound intermetallic bond. The surface often reads as dull, grainy, lumpy, or poorly wetted, but the real problem is underneath: a weak metallurgical connection that may pass electrical test on the bench and open up under thermal cycling or mechanical shock in the field. Cold joints are almost always a thermal story, and thermal stories are measurable. If you can profile the real board, you can find and close the gap.
This article is written for process engineers, SMT line supervisors, and quality staff who have to stop a cold-joint escape before it reaches a customer. It walks through what a cold joint actually is, the five permitted cause families behind it, how to detect one with the tools you already have, and a profiling-and-control routine you can run without stopping the line. For the wider diagnostic method that covers every reflow defect family, our complete guide to reflow soldering defects explains how to classify a symptom before you change a single setting.
One boundary up front. A dull or grainy joint is not automatically a cold joint. The same appearance can come from excessive time above liquidus, which thickens intermetallic compound, or from an oxygen-rich atmosphere that oxidizes the surface as it cools. That is why this article keeps pushing you back to a measured profile rather than a visual guess.
The fastest way to prevent cold joints is to know precisely what you are looking for and what distinguishes it from its look-alikes. Cold joints get missed because they do not always announce themselves the way a tombstone or a bridge does. They sit quietly on a board that otherwise passes visual inspection, and they fail later.
A cold joint usually presents as a dull or matte surface rather than the smooth, reflective fillet a well-formed joint shows. The solder may look grainy, rough, or lumpy, and the fillet shape is often incomplete or uneven at the edges where wetting should have pulled the alloy smoothly up the termination. In some cases the solder sits on the pad like a separate blob that never truly merged with the surface beneath it.
Appearance alone is a weak inspector, which is why standards such as IPC-A-610 describe acceptance in terms of wetting and fillet geometry rather than shine. A joint can be bright and still poorly bonded, and matte and still acceptable. The visual signature is a flag to investigate, not a verdict.
The danger of a cold joint is that it can be electrically continuous on the test bench and mechanically fragile under real duty. The intermetallic bond never fully formed, so the connection is held by partial wetting and contact rather than a sound metallurgical layer. Under thermal expansion, vibration, or drop shock, that partial contact can crack and open.
This is the most expensive class of defect because it escapes. A board that fails at the oven is a scrap or rework cost; a board that fails in a customer's product is a field return, a warranty claim, and a credibility loss. Cold joints deserve prevention precisely because they are hard to catch at the end of the line.
In the four-family way of thinking about reflow defects, a cold joint is squarely a thermal defect. It is formed during the reflow stage, by how heat was delivered to the joint, rather than by how the paste was printed or where the component was placed. That placement tells you which instrument to reach for: a profiler and a thermocouple on the real board, not the printer.
The same thermal controls that prevent cold joints also bear on other defects. Too much heat or too long above liquidus thickens intermetallic compound and drives dull, grainy joints from the opposite direction. Too little heat, or heat that never reaches the joint, produces the cold joint. The line between the two is a measured window, not a feeling. Everything that follows is about taking the profile measurement on a board that represents production and making sure the heat that reaches the joint is enough, consistently, across every board and every shift.
A reflow oven does not heat a joint directly. It heats air, moves that air across the board, and the board and its components then conduct heat to the joint. Cold joints appear when something in that chain breaks, so it helps to understand the chain before naming a cause.
A typical reflow profile runs through four functional stages. Preheat brings the assembly up without thermal shock and begins activating the flux. Soak, sometimes called thermal equalization, evens out the temperature across the board so heavy and light areas arrive at reflow together and the flux finishes its work. Reflow takes the assembly above the alloy liquidus so the solder melts and wets, and cooling sets the final grain structure.
Cold joints are born in the soak and reflow stages. If soak is too short or too cool, the flux is not fully active when the alloy melts. If reflow peak is too low or time above liquidus too short, the alloy melts but does not wet properly. Both failures land you at the same dull, grainy, weakly bonded joint.
Two numbers decide whether a joint forms correctly: the peak temperature and the time spent above liquidus, often called TAL. Peak tells you whether the alloy got hot enough to melt and flow. Time above liquidus tells you whether it stayed molten long enough to wet the pad, climb the termination, and grow intermetallic compound at the interface.
The relationship is a window, not a single point. Too little of either, and you get a cold or partial joint. Too much of either, and you grow excessive intermetallic compound. The window is set by the paste supplier's recommendation for the specific alloy and should be confirmed on your board rather than assumed from the oven display.
Lead-free SAC alloys, such as SAC305, wet more slowly than tin-lead and are more sensitive to oxidation during the molten window. The solidus temperature of SAC305 is around 217 degrees Celsius, though the exact value should always be confirmed with the alloy supplier's data sheet because small composition differences move it. Because lead-free solders wet more slowly, they need a well-formed soak and a sufficient reflow window to pull the fillet up cleanly.
This is why cold joints became more common after the industry moved to lead-free: the process window comfortable for tin-lead is tighter for SAC, and a profile merely adequate before can leave joints under-wetted afterward. Treating lead-free as "the same profile, slightly hotter" manufactures cold joints.
The first and most direct cause family is simply not enough heat at the joint for long enough: a profile where peak temperature sits below what the alloy needs, or time above liquidus is shorter than the paste supplier recommends. It is the cause engineers suspect first, and often correctly, but it is worth separating the ways it happens.
The most obvious version is a recipe whose peak was set too low for the alloy in use, often when carried over from a tin-lead product or trimmed by an engineer trying to reduce other defects. Because the alloy liquidus is fixed, a peak that does not clear it with adequate margin leaves the solder only partially molten.
The correction is straightforward in principle: raise the peak into the supplier's recommended band. In practice, the hard part is confirming the band for your specific paste and then verifying on the board that the joint actually reaches it. A number typed into the oven recipe is never the same as a number measured at the pad.
Even with an adequate peak, a short dwell above liquidus starves the wetting process. The alloy melts, but it does not have the seconds it needs to spread, reduce the remaining oxides, and form the intermetallic layer. The result is a joint that melted and then froze before it finished bonding.
This is common when conveyor speed is increased to lift throughput without revisiting the profile. The peak may look acceptable on a fast pass, but the time above liquidus collapses. Slow the conveyor or raise the reflow zone temperature to restore the dwell, then re-profile to confirm.
Soak is where the flux does its cleanup work before the alloy melts. If soak is too short or too cool, the flux is not fully active when reflow begins, and the surfaces carry oxides that the molten solder cannot wet through. The joint then looks cold even when the peak and TAL were, on paper, sufficient.
Tuning soak is a balance. Too little, and flux is ineffective. Too much, and the flux is spent before reflow, which also hurts wetting and can leave residues. The right soak is specific to the paste chemistry, another reason the data sheet belongs on the line.
A recipe can look correct while the board sees too little heat. Zone heater aging, a conveyor speed edited without re-profiling, or a different program loaded by mistake all make the real profile drift below the recipe. The recipe is an input; the measured profile is the truth.
The discipline that prevents this is to profile the real board at the real conveyor speed on a schedule and after any change. A stored recipe is a starting point, not a guarantee; when cold joints appear, the first move is always to measure what the board actually experienced.
The second cause family is thermal mass. A board is not uniformly easy to heat. Big copper planes, large components, and thick layers soak up heat and lag behind the thin, lightly populated areas, so when the profile was built on a light area the heavy area never reaches the joint-forming window.
Large copper pours and ground planes act like heat sinks attached to the joint. As the solder tries to melt and wet, the surrounding copper draws heat away faster than the air can supply it, so the pad temperature lags the air temperature. A joint sitting on a heavy plane can read cold while everything around it looks fine.
This is a classic reason a board passes profile on a test coupon but still produces cold joints at specific locations: the coupon did not carry the same copper distribution as the product. The remedy is to profile at the actual heavy-copper locations and, where the margin is tight, give the profile enough soak and peak to bring those areas into the window.
Large packages and heavy components have their own thermal mass. A big connector or a power device reaches reflow temperature later than a small resistor on the same board, so if the profile is timed for the small part, the large part is still catching up when the board exits the reflow zone. Its joints form late and incompletely.
The pattern to look for is cold joints clustered under or near large components while the rest of the board looks clean, which signals thermal-mass lag rather than a global profile problem. The fix is to shift the profile so the heavy parts reach the window, accepting a longer or hotter cycle on thin areas that should still stay inside tolerance.
The way to catch thermal mass as a cause is to attach thermocouples to both a heavy location and a light location and watch the spread. A large delta between them at the moment of peak, or a heavy-area trace that never reaches the joint-forming window, tells you the mass is the problem.
Once you see it, the levers are soak length, peak temperature, conveyor speed, and zone airflow. Pulling any of those in the right direction brings the heavy areas up. The mistake to avoid is raising the peak blindly, which can push the light areas into excessive time above liquidus and trade a cold joint for an over-aged one.
The third cause family is the oven's ability to deliver heat. Even a perfect recipe cannot help if the heater zones cannot reach or hold the required temperature at the conveyor speed and load you are running. Insufficient heater-zone power shows up most when throughput climbs past what the machine was sized for.
Every reflow oven has a practical limit on how much board mass it can heat per minute. Push the conveyor speed up, load denser panels, or run heavier boards, and the zones may not replace the heat the boards carry away. The air temperature drops, the board sees less, and cold joints appear at the far edge of the oven's capacity.
This is a quiet failure because it often tracks with production pressure rather than with a setting change. The recipe did not move; the load did. The tell is cold joints that worsen as line speed increases or as board weight grows, and that improve when you slow down or lighten the load. Heat is also lost at every opening, conveyor gap, and cold-board entry, so a zone recovering from a large load swing may never reach setpoint before the board moves on; watching zone stability under load, not just at idle, shows whether capacity is the bottleneck.
Choosing an oven against the product mix is the durable fix for a power-limited line. A machine with more heating zones and stronger convection gives a wider process window, which matters most for lead-free alloys, heavy copper, and mixed assemblies. An overview of SMT reflow oven solutions groups configurations by application and process window so you can see which strengths match your boards.
For lines running dense, thermally demanding products, a modular hot air reflow oven in the Lyra series, with a multi-zone heating chamber, PC and Siemens PLC control, optional nitrogen, and an updated cooling zone, gives process engineers room to hold soak and peak on heavy boards without overcooking light ones. The point is to buy the control you need, not the longest specification sheet.
The fourth cause family is subtler because it is a measurement problem that masquerades as a process problem. The profile says the board is fine, the joints say otherwise, and the gap is that the thermocouple was never measuring the board that matters.
It is tempting to profile a bare test coupon or a scrap board and call it representative. But a coupon lacks the copper distribution, component mass, and layer stack of the real product, so its trace can sit in the window while the actual assembly lags below it. Cold joints then appear at exactly the locations the coupon could not see.
The rule is to profile the real assembly wherever possible, including its heaviest thermal mass and its most heat-starved joint. A coupon is a useful quick check, but it is not evidence that the product is in window. Even on the real board, placement decides everything: a thermocouple stuck to thick copper or a bare spot, or poorly adhered and lifted off the surface, reads warmer than the failing joint, so the trace looks fine while the joint was never in the window.
A representative routine has four habits. Profile the real product, not a coupon. Attach sensors at the failing joints with solid contact. Record the conveyor speed and load with the trace. Re-profile after any change to recipe, speed, product, or zone maintenance.
Done consistently, this routine makes the cause families visible instead of hidden. Most cold-joint problems that "made no sense" turn out to be a profile taken somewhere the problem was not, or at a condition the line had already left behind.
The fifth permitted cause family is paste. Even with perfect heat, solder that has lost its flux activity or picked up moisture may not wet properly, so a joint can look cold because the surfaces were never cleaned enough to bond.
Solder paste is a suspension of alloy powder in flux, with a working life. Past its recommended storage life, or after too long open to air, the activators that remove oxides lose strength. The alloy may melt on schedule, but it cannot wet a surface it cannot clean, so the fillet forms incompletely and reads as a cold joint.
The signature that separates this from a pure thermal cause is that raising the peak does not fully fix it. If more heat helps only a little, or the dull grainy appearance persists, the flux activity is the likely limiter; the fix is paste freshness and handling, not more peak temperature.
Paste also changes between printing and reflow. Left open too long, or exposed to humid air, it loses volatiles and picks up moisture, both of which degrade wetting. The open-time limit between print and reflow exists for exactly this reason, and it is one of the most commonly exceeded limits on a busy line.
Defects from open time and moisture tend to worsen toward the end of a batch or shift, because boards have sat longer and paste has been open longer. If cold-looking joints track with time since printing rather than with oven settings, the paste clock is the suspect, not the profile.
The controls here are unglamorous but effective. Rotate paste stock by date, respect the open-time limit, control storage humidity, and treat a new paste lot as a process change that earns a first-article check. A first-article verification after a paste change catches behavior differences before they reach the product.
Paste condition also belongs in the defect log. When cold joints appear, note the paste lot, the open time, and the storage conditions alongside the profile data. The combination usually points to the real cause faster than either record alone.
Detecting cold joints and keeping them out takes more than a profile. It takes detection tools that catch marginal joints, an understanding of what the acceptance standards actually require, and a routine that holds the process in window across shifts. This section pulls those pieces together.
Visual and AOI inspection catch the obvious dull and grainy joints, but a cold joint can be subtle, so a 2D image alone is not enough. X-ray can reveal incomplete wetting and voids at the interface that suggest a joint never properly formed, and a cross-section under a microscope is the definitive check for the intermetallic layer and fillet geometry.
The practical approach is layered: AOI for the visible signature, X-ray for the hidden interface on area-array and hidden joints, and periodic cross-sectioning on a sample to confirm the metallurgy. A closed loop feeding those findings back to the profile and paste record is what keeps cold joints from slipping through.
Acceptance of a soldered joint is defined by reference standards rather than by shine. IPC-A-610, the widely used assembly acceptance document, judges joints on wetting, fillet geometry, and soundness rather than surface brightness, so a dull joint is not automatically rejected and a bright one not automatically accepted. The specification and its criteria are maintained by IPC at https://www.ipc.org.
For your own line, the useful move is to agree acceptance criteria with the customer and then hold the process to them with measurement, not opinion. A joint that meets the wetting and fillet requirements is good whether or not it sparkles, and keeping the standard prevents both over-rejection of acceptable matte joints and under-rejection of genuinely cold ones.
A routine that actually prevents cold joints has five moves. First, profile the real board at the real conveyor speed with sensors at the suspect joints. Second, compare the measured peak and time above liquidus to the paste supplier's window. Third, change only one variable if a correction is needed, then re-profile. Fourth, confirm the fix with a fresh trace and, where doubt remains, a cross-section. Fifth, record the recipe, speed, paste lot, and result so the fix survives shift changes.
This routine overlaps with the method in our complete guide to reflow soldering defects and pairs well with the controls for neighboring defects. The same profile discipline that prevents cold joints also supports reducing BGA voiding and preventing solder bridging, because all three live or die on a verified, representative profile.
Equipment decisions shape how often you fight cold joints. More heating zones, stronger convection, and stable cooling give you a wider window, the margin that absorbs variation in paste, copper, and component mass. A board that warps during reflow also changes paste-to-component contact and can leave corner joints under-formed, so the guidance on PCB warpage during reflow is worth reading alongside this one.
The buying logic is to match the machine to the product. Dense, lead-free, mixed-mass boards need zoning and convection headroom; simple, light boards do not. A modular hot air reflow oven with multi-zone heating, optional nitrogen, and an updated cooling zone is the kind of configuration that gives engineers room to hold soak and peak on demanding products. Match the control you buy to the defects you actually see.
Some cold-joint problems resist the routine because the cause spans design, material, and process at once. When the heaviest copper areas stay out of window no matter how you tune the recipe, or when paste and profile both look correct but joints still under-wet, a second set of eyes on the board construction and the oven behavior is worth more than another degree of peak.
I.C.T supplies complete SMT turnkey lines, including the Lyra series reflow ovens, with factory planning, installation, training, and overseas on-site technical support. If you are chasing a cold-joint pattern that will not close, our process engineers can review your profile data and board construction and help separate the thermal, mass, and paste causes so the fix lands the first time.
9.1 What exactly is a cold solder joint?
A cold solder joint is one where the alloy melted but never got enough heat or time to wet the pad and component and form a sound intermetallic bond. It usually looks dull, grainy, or lumpy with a poor fillet, and it can be mechanically weak even when it passes an electrical test on the bench. The defining feature is incomplete wetting, not necessarily a visible gap.
9.2 How is a cold joint different from a dull joint caused by too much heat?
They can look similar but come from opposite problems. A cold joint is under-heated or under-dwelled, so the solder never fully wets. A dull joint from excess heat or time above liquidus has over-grown intermetallic compound and a coarsened grain, which also reads as matte. The only reliable way to tell them apart is the profile trace: short time above liquidus points to cold, long time above liquidus points to over-aging.
9.3 Can a cold joint be caused by the solder paste rather than the oven?
Yes. Paste that has aged past its activity, or that has been open too long and picked up moisture, can lose the flux strength needed to clean the surfaces and wet the joint. The alloy may melt on schedule, but the joint still forms incompletely and looks cold. If raising the peak does not fix the appearance, suspect the paste lot, the open time, and the storage conditions.
9.4 Why do cold joints often appear only under specific components?
Because those components or the copper around them carry more thermal mass and lag behind the rest of the board during reflow. Large packages, connectors, and heavy copper planes reach the joint-forming window later or never, so cold joints cluster under them while thin areas look fine. Profiling the actual heavy locations is how you confirm it.
9.5 Does nitrogen reflow prevent cold solder joints?
Nitrogen helps wetting and reduces oxidation, which can widen the process window for lead-free alloys, but it does not fix a joint that never reaches the required peak temperature or time above liquidus. Cold joints are primarily a heat-delivery and dwell problem. Nitrogen is a support, not a substitute, for a verified profile on the real board.
9.6 How often should I re-profile to keep cold joints away?
Re-profile whenever you change the product, the paste lot, the conveyor speed, or the oven after maintenance, and on a regular schedule even when nothing appears to change. A profile is a snapshot of a specific condition, and the line drifts. A stored recipe that is never checked is the most common source of cold joints that "made no sense."
I.C.T designs and supplies complete SMT assembly lines, including the Lyra 622, 733, and 933 series reflow ovens, with turnkey planning, installation, training, and overseas on-site technical support. If a cold-joint pattern on your line will not close, our process engineers can review your profile data and board construction and help pin down whether the cause is heat, thermal mass, oven capacity, profiling, or paste.