You fire up a cabinet that’s been sitting in storage for three years. The screen lights up, but the graphics flicker. Sound cuts in and out. You swap the ROM chips, reseat connectors, and check the power supply voltages—all normal. Then you notice something: the problem gets worse after five minutes of play, but improving slightly when you let it cool for ten minutes. That pattern—degradation with heat, partial recovery with cooling—is the unmistakable fingerprint of cold solder joints.
Cold solder joints are one of the most frustrating failures in vintage arcade and game board repair because they’re nearly invisible to the naked eye and don’t show up on a multimeter. A joint that looks perfectly acceptable under a magnifying glass can fail intermittently under the thermal cycling that happens during normal operation. The connection appears solid, but the electrical contact is marginal at best—especially when the board heats up and the copper and solder expand at slightly different rates.
For decades, technicians had only two options: carefully reflow every joint by hand (time-consuming and risky) or buy expensive industrial equipment. But thermal imaging cameras have dropped dramatically in price. A decent thermal camera now costs less than a quality oscilloscope—and it reveals exactly which joints are failing before you touch a soldering iron. In this article, I’ll walk you through how thermal imaging actually detects cold joints, what you’re looking for, how to use the equipment properly, and how to distinguish real failures from normal thermal variation.
Why cold solder joints happen—and why they’re hard to see
A proper solder joint forms when solder melts and flows across both the component lead and the copper pad, creating a metallurgical bond. The solder cools and solidifies with a shiny, cone-shaped surface. But a cold joint occurs when solder doesn’t flow properly—either because the iron wasn’t hot enough, the joint wasn’t cleaned, the board wasn’t preheated, or the iron was applied for too short a time. Instead of flowing and wetting the pad, the solder just sits there in a grainy, dull ball.
The problem is that a cold joint can still carry current—just not reliably. The resistance across the joint is higher than normal, and more importantly, the mechanical connection is weak. As the board heats during operation, different materials expand at different rates. Copper expands. The component lead expands. The solder expands. The board material expands. But they all expand by different amounts (measured in parts per million). When everything cools back down, the stresses reverse. After hundreds of thermal cycles, the cold joint cracks microscopically. Current flow becomes intermittent. The resistance jumps from a few milliohms to several ohms—or open circuit.
Visually, this is almost impossible to detect without magnification, and even then, you’re looking for something that might be barely visible. The joint surface might look slightly dull or grainy rather than shiny, but in a complex board with hundreds of joints, finding the bad ones this way is like searching for a specific grain of sand.
How thermal imaging reveals what resistance measurements can’t
Here’s where thermal imaging becomes powerful: a joint with high electrical resistance generates heat. It’s basic Joule heating—the power dissipated in a resistor is proportional to I²R. A cold joint has elevated resistance. When current flows through it, it generates heat. A normal joint (resistance ~1-5 milliohms) barely warms up. A cold joint (resistance ~100 milliohms to several ohms, or intermittently open) can reach 10-20°C above the ambient temperature of the board—sometimes more.
The beauty of thermal imaging is that it shows you this heat distribution in real time. You run the board under load (the arcade game running, the test program executing), and the thermal camera displays the board’s surface temperature in a false-color image. Good joints disappear into the background. Failing cold joints light up as hot spots—literally warmer than everything around them.
This only works if current is actually flowing through the joint. A cold joint on a signal line that carries little current might not generate detectable heat. But joints on power distribution, ground return paths, and heavily loaded signal traces will show up clearly. This is why thermal imaging works best for finding failures on power planes, power delivery traces, and the common ground connections that carry return current from logic chips.
The physics of thermal distribution on a circuit board
To interpret a thermal image correctly, you need to understand what normal heat distribution looks like on a vintage arcade board.
Component-generated heat is the baseline. Active components—CPU chips, RAM, ROM, logic ICs, and power regulators—generate heat when they operate. Vintage boards typically have one or two obvious heat sources: the CPU (usually warm, 50-70°C above ambient), the power regulator (often the hottest spot), and sometimes specific logic chips. This heat spreads across the copper traces and plane layers, creating a gradient. Traces immediately connected to hot components run warmer than traces across the board.
Heat flows along copper paths. Copper is an excellent thermal conductor (second only to silver). Heat generated by a warm chip spreads along the copper traces connected to it, following the path of least thermal resistance. This means you’ll see temperature gradients along traces, higher near the source, lower as you move away from the chip.
Multi-layer boards conduct heat vertically. Vintage arcade boards are typically 2-4 layers. Power planes and ground planes conduct heat extremely efficiently vertically through the board. This makes the board more thermally uniform than you might expect. Heat from one side can conduct through the planes and appear on the other side.
Solder joints conduct heat both ways. A good solder joint is thermally conductive. It transfers heat from the component lead (if the lead is warm) down to the pad and the copper it’s connected to. A cold joint, with poor metallurgical contact, has much higher thermal resistance. This is the key: heat from current flowing through the joint gets trapped at the joint itself rather than spreading efficiently into the surrounding copper.
So when you image a board, you’re looking for anomalies: a hot spot that doesn’t align with a known heat source, or excessive heat concentrated at a specific solder joint when the surrounding area is much cooler. This is the signature of a cold joint with high resistance.
What you need: thermal camera selection and settings
You don’t need industrial-grade thermal imaging equipment for this work. A mid-range thermal camera—something in the $200-500 range—is perfectly adequate. Look for these specifications:
Resolution: At least 160×120 pixels, ideally 320×240. This gives you enough detail to pinpoint individual components and joints. Entry-level FLIR cameras and comparable units from other manufacturers meet this spec.
Temperature range: Most vintage arcade boards run between 30°C and 80°C. You want a camera rated for at least 0-100°C, preferably wider. Many cameras have switchable ranges; make sure it covers this band without loss of sensitivity.
Thermal sensitivity (NETD): This measures the smallest temperature difference the camera can detect, typically in milliKelvins. Aim for less than 100 mK (0.1°C). This lets you see the subtle temperature differences that identify marginal joints. Lower (better) values make marginal failures more obvious.
Emissivity adjustment: Circuit boards, especially with solder mask, have varying emissivity (how much infrared they emit). Most thermal cameras let you adjust this. PCB solder mask is around 0.95; bare copper is lower (0.05-0.1). Some cameras auto-adjust, others require manual input. This matters because reading the absolute temperature accurately requires correct emissivity. However, for finding relative hot spots, you can often leave it at the default and just look for anomalies.
In terms of specific products: FLIR offers the Lepton-based cameras and the E4 compact model, both under $400. Seek Thermal makes budget options. Xintest and other Chinese thermal cameras have improved significantly and often cost $150-300. The real differentiator is usability and software, not raw specifications for this application.
Setting up for reliable imaging
Establish baseline temperature. Before you image a running board, measure the ambient temperature of your workspace and let the board sit at rest for 15-20 minutes. Record this baseline. You’re going to compare joint temperatures against this baseline and against the surrounding copper and traces. If the room is 22°C and a joint is running 60°C, you’ve got a 38°C delta. That’s significant.
Run a known test under consistent load. Fire up the arcade game or run whatever test program loads the board. Ideally, use something that exercises the board consistently and generates repeatable, sustained heat. A game running in an attract loop works well because it’s predictable. Let the board run for 10-15 minutes before imaging, giving the thermal distribution time to stabilize. Cold joints behave worse as the board heats up, so this is when you’ll see the clearest signal.
Control the angle and distance. Thermal accuracy drops significantly if you’re not looking straight at the surface. Aim for perpendicular to the board. Distance matters too—stay within the camera’s minimum working distance (usually 20 cm to 1 meter, depending on the model). Don’t zoom in so close that your hand shadows the area or your body heat affects the reading.
Avoid reflections and ambient thermal interference. Windows, lights, and nearby hot objects reflect infrared and confuse the reading. Work away from direct sunlight if possible. If you’re in a bright room, consider using a cardboard hood to shade the board from ambient light sources. Your own body heat can also affect readings if you’re leaning over the board—be aware of where you’re standing relative to what you’re measuring.
Reading the thermal image: what you’re actually looking at
Once you’ve got a thermal image, interpretation is straightforward if you know what to expect.
The power regulator (or regulators) will be the hottest components on the board. Voltage regulators are inherently inefficient—they dissipate power as heat to regulate the output voltage. Older discrete regulators (sometimes using separate transistors, diodes, and resistors) are typically hotter than modern integrated regulators. The regulator and the surrounding copper will form the warmest region on the board. This is normal.
The CPU will be the second-warmest region. Large-scale integration chips like the 68000 CPU or Z80 CPU generate significant heat proportional to clock frequency and the amount of active logic. You’ll see a warm halo around these chips.
ROM and RAM chips are usually cooler than the CPU. These chips dissipate less power than processors. They may be warmer than ambient, but typically 15-30°C above ambient rather than 40-50°C above.
Trace and via temperature should be fairly uniform across regions. The copper traces connecting logic should be roughly uniform in temperature, with gradients as you move away from heat sources. A trace should not suddenly jump 20°C at a single joint unless something is generating heat there.
Look for isolated hot spots that don’t align with known components. If you see a region that’s significantly warmer (10-20°C hotter than the surrounding area) and there’s no active component there, you’ve likely found a cold joint. The hot spot will typically be small—the size of a solder joint or slightly larger—and isolated to one area of the board.
Be specific about what “significant” means in your context. If the ambient temperature is 22°C, the board baseline is 40°C, and you see one joint at 65°C while everything around it is 42°C, that’s a 23°C anomaly. That’s a cold joint. If you see a spot at 45°C in an area where the baseline is 42°C, that might be marginal—but if it’s clearly isolated and reproducible across multiple test runs, it’s worth investigating.
The diagnostic workflow: step-by-step imaging procedure
Here’s the systematic approach to finding cold joints on a vintage arcade board:
Step 1: Prepare the board and baseline image
Power on the arcade cabinet or bring up the board on a test rig. Allow it to run for 5 minutes to reach initial steady-state temperature. Take a reference photograph with normal lighting so you can cross-reference component locations later. Allow the board to run for another 10 minutes so the thermal distribution fully stabilizes. Record the ambient room temperature.
Step 2: Acquire the full thermal image
Position the thermal camera perpendicular to the board surface, filling the frame with the PCB. Start from the top edge and capture the entire board systematically. If the board is larger than your camera’s field of view, capture overlapping images of sections: top half, bottom half, left half, right half, etc. This takes 5-10 minutes but ensures you don’t miss anything. Keep the board running throughout.
Save all images with metadata (timestamp, ambient temperature recorded). You’ll reference these later.
Step 3: Identify anomalies and mark locations
Look at each thermal image systematically. Identify any bright spots (hot areas in the false-color scale) that are isolated and not directly under a known heat source. Use the reference photograph to identify the component or trace at that location. Mark these anomalies—write down their location in board coordinates or reference nearby components. Ask yourself: is there a good reason this spot should be warm? Is there a current-drawing component there, or an active trace? Or is this an isolated hot spot with no obvious heat source?
Typically, on a working board, you’ll find these obvious hot zones: the power regulator (hottest), the CPU (second hottest), possibly a ROM or RAM chip if they’re heavily used, and gradients along power traces leading away from the regulator. Anything outside this pattern is suspicious.
Step 4: Narrow down cold joint candidates with close-up imaging
Once you’ve identified suspicious areas, zoom in on them with the thermal camera (if it has zoom—many do, though digital zoom is less useful than optical). Get a close-up thermal image of the area surrounding the suspected joint. If your camera allows, capture a few frames as you move the camera slightly; this helps you triangulate the exact location of the hot spot.
If you have marginal anomalies that are borderline (3-5°C above surrounding area), apply slight mechanical stress to the board while imaging. Gently flex the board (not excessively; you’re not trying to break anything), or tap the suspected area lightly. A cold joint under stress sometimes shows an increase in resistance and a temporary temperature rise. If the thermal signature changes noticeably under mechanical perturbation, that’s additional evidence.
Step 5: Cross-reference with electrical testing
Once you’ve identified suspect joints, stop the board and let it cool to ambient. Then, using a multimeter set to resistance (ohms) mode, measure the resistance across any suspicious joints you found. A good solder joint on a trace or component lead should read nearly zero ohms (typically 0.001-0.01 ohms, which multimeters display as 0). A cold joint might show 0.1-1 ohm, or the resistance might be unstable (fluctuating as you move the probe).
This is a secondary confirmation step. You’ve already got strong visual evidence from thermal imaging. The resistance measurement confirms that there really is elevated resistance at that joint.
Step 6: Document and prioritize
Compile your findings: a list of suspected cold joints with their locations on the board, the thermal anomaly (temperature delta from baseline), and any resistance measurements. Prioritize them by severity: the hottest spots first, as these are the most likely to fail first under thermal cycling.
At this point, you have a repair roadmap. You know exactly which joints need attention rather than reflowing the entire board.
Interpreting edge cases and confusing signals
Heat spreading vs. hot joints. Sometimes you see a warm area that’s larger than a single joint—maybe a 2-3 cm region. This might be heat spreading from a hot component through copper traces or planes, not a single cold joint. Look for the peak of the thermal anomaly (the hottest point). If it’s centered on a trace junction or component lead, you’ve got a cold joint. If it’s more diffuse and centered on a component or spreads gradually away from a hot chip, it’s heat distribution.
Multiple marginal joints in the same region. Vintage boards sometimes have solder quality issues affecting a whole row of joints—perhaps a hand-soldering task that was rushed. You might see multiple smaller hot spots in proximity. Thermally, each individual joint appears 5-15°C above the surrounding baseline, but they’re clustered. This suggests a process issue on a particular area of the board, not a rare single failure.
Thermal cycling behavior changes the signature. If you image the board immediately after power-on, the thermal profile is different than after 15 minutes of steady operation. Cold joints are under more stress and show higher resistance when the board is hot. You might capture an image at 5 minutes and see a marginal signal, then image again at 20 minutes and see the same joint as a bright hot spot. This changing signature as the board heats up is diagnostic of a cold joint, not a normal component. Log this pattern if you observe it.
Distinguishing cold joints from component pin overheating. Sometimes a component’s pins run hot because the component itself is dissipating heavily, not because of a cold joint. The entire component is warm. With cold joints, typically the joint area is hot but the immediate surrounding area (the trace or pad) is not significantly warmer than the baseline. Look for the temperature delta to be localized to the joint itself, not spread across the entire component.
Bypass capacitors and their thermal signature. Older boards sometimes use bulk electrolytic capacitors on power rails instead of modern low-ESR ceramics. These capacitors have internal resistance and dissipate heat, especially if the power supply is rippling or unstable. An old electrolytic capacitor can run quite warm—sometimes 30-50°C above ambient. This is usually not a cold joint, though you might find cold joints soldering the capacitor to the board. Zoom in on the capacitor leads specifically. Are the leads or the solder points themselves the hot spots, or is it the capacitor body? If the body is hot and the leads are normal, the capacitor has high ESR and probably needs replacement (this is a separate issue from cold joints, though related to power supply aging).
Ground planes and thermal uniformity. Multi-layer boards with ground planes conduct heat very efficiently. This makes thermal imaging less effective on some boards because heat spreads quickly and temperature gradients are subtle. If your thermal image shows very uniform temperatures across the entire board with only the obvious hot components standing out, either the board is well-designed thermally (good news), or the resolution of your thermal camera is insufficient to pick up the smaller anomalies. In this case, move to resistance measurement and visual inspection under magnification as your primary diagnostic tools.
When to move from thermal imaging to repair
Once you’ve identified suspect joints using thermal imaging, the next step is repair. However, recognize the limitations of this approach: thermal imaging identifies high-resistance joints, but not every solder joint failure is thermal. Some cold joints carry low current and don’t generate measurable heat. Some failures are cracks that develop intermittently and don’t show up consistently under steady-state operation.
After you’ve thermally imaged the board and identified the suspicious joints, you have a decision to make: reflow only the identified joints, or reflow the entire board?
Targeted reflowing is faster and lower risk. You heat only the suspicious joints, using a soldering iron or hot air rework tool. The advantage is minimal thermal stress on the rest of the board and faster repair time. The disadvantage: other marginal joints you didn’t catch will fail later. On a 20-30 year old board, this is likely.
Full-board reflow is more thorough. You reflow every joint on the board, either using a heat gun or hot air rework station, or by reflowing sections methodically with an iron. This catches all marginal joints—the ones generating heat and the ones carrying low current and not showing on thermal imaging. The advantage: you’ve got more confidence in long-term reliability. The disadvantage: more time, higher thermal stress on fragile components, and higher risk of damage if you don’t control the heating carefully. Vintage boards with old capacitors, brittle traces, and heat-sensitive plastics need careful temperature control.
My recommendation for a board that’s intermittently failing with thermal characteristics: start with targeted reflowing of the thermally identified joints. Test thoroughly. If the board passes several days of continuous operation without issues, you’ve solved the problem. If it fails again, consider full-board reflow or replacement of known-problematic components. This is a practical middle ground—you get fast, focused repairs without over-stressing aged components.
Reflowing technique matters. Whether you use a soldering iron or hot air, the goal is the same: heat the joint to 220-230°C for long enough for solder to melt and flow. This typically takes 3-5 seconds with an iron, or 10-15 seconds with a heat gun (which you move across the board area). Too much heat damages components. Too little heat doesn’t fully reflow the joint. If you’re not confident in your hand-soldering skills, a heat gun on a low setting over a specific area, carefully controlled, is safer than an iron.
What thermal imaging can’t tell you—and what to do instead
Thermal imaging is powerful, but it’s not a complete diagnostic tool. Here’s what it misses:
Cracks that don’t generate heat. A joint that’s fractured but makes intermittent contact carries current only part of the time. When current flows, it generates heat. But if the fracture is intermittent and the joint only carries current briefly before the crack opens again, the average temperature rise might be subtle. You’ll see a marginal thermal anomaly, not a screaming hot spot. Complementary visual inspection under 10-30x magnification is necessary. Look for stress cracks around component leads, especially on components that vibrate during operation or have been flexed mechanically.
Dry joints on signal lines with low current. A signal trace (clock, address bus, control signal) might carry very little continuous current. A cold joint on the trace might cause intermittent logic errors but barely register thermally. This is why arcade boards sometimes show random glitches that thermal imaging alone doesn’t explain. You need to combine thermal imaging with visual inspection and, if necessary, with multimeter resistance testing to validate the quality of signal path connections.
Age-related failures of other components. Thermal imaging shows hot spots from resistance. But many failures on old arcade boards are not resistance-related. Failing electrolytic capacitors, drifting resistors, and logic chips with internal leakage cause trouble but don’t show up as isolated hot spots. A comprehensive diagnosis of a failing arcade board usually requires thermal imaging plus visual inspection plus component-level testing.
Intermittent failures that depend on specific conditions. Some cold joints only fail under specific combinations of temperature, load, and vibration. If you thermal image the board running a simple game that doesn’t fully load the CPU or RAM, you might not see the failure. The joint that fails when the board is running a graphics-heavy game might look fine during a simple test program. This is why knowing the failure mode (when does the board fail? under what conditions?) is important before you start imaging. Run the test that reproduces the failure, then image while that failure is most likely to occur.
Building a practical thermal imaging diagnostic workflow
Here’s the realistic approach for someone maintaining or restoring vintage arcade boards:
Step 1: Establish baseline normal behavior. If you have access to a known-good board of the same type, image it while running a standard test to see what normal thermal distribution looks like. Photograph it and keep the reference. Then, when you image a problem board, you have a clear baseline for comparison.
Step 2: Reproduce the failure condition. If the board has a known failure (graphics glitches, sound dropout, intermittent freezes), run it in a way that triggers the failure. Image while the failure is happening or immediately before it occurs. This is when cold joints are under maximum stress and most likely to show thermal anomalies.
Step 3: Image systematically and carefully. Don’t rush through the board. Take multiple images, overlapping if necessary, to cover the entire surface. Include the underside if accessible, as solder joints on through-hole components can fail on either side.
Step 4: Identify the top 3-5 suspect joints. Don’t try to reflow every slightly warm spot. Focus on the clearest anomalies—the ones that are 15°C or more above baseline, or show reproducible temperature changes under repeated testing.
Step 5: Repair and test incrementally. Fix the most obvious joints first, retest, and observe if the failure is resolved. Log the results. This teaches you what’s actually broken versus what’s just hot for normal reasons.
Over time, this workflow becomes fast. A skilled technician can thermal image a board and identify the likely culprits in 20-30 minutes, rather than spending an hour reflowing dozens of joints blindly.
Decision framework: thermal imaging in your repair practice
Should you invest in a thermal camera for arcade board repair? Here’s the honest math:
If you repair arcade boards regularly (more than once per month), a thermal camera is worth the $200-400 investment. It cuts diagnosis time dramatically and increases your confidence in repairs. You’ll recover the cost in labor savings within a year if you’re doing this professionally or semi-professionally.
If you’re a casual hobbyist repairing one or two boards per year, thermal imaging is less essential. You can rely on careful visual inspection under magnification, resistance testing, and systematic reflowing of the most likely failure points. The time savings don’t justify the equipment cost.
If you’re trying to decide between a thermal camera and other repair tools, prioritize according to your actual repair portfolio. A well-stocked repair toolkit with iron, solder, magnification, and measurement tools comes first. A thermal camera is a multiplier—it makes your existing skills more effective, but it doesn’t substitute for soldering ability or electronics knowledge.
If you’re buying a thermal camera, avoid the extremes. Sub-$100 cameras are often frustratingly low-resolution (80×60 pixels) and slow to update. Cameras over $1,000 have features you don’t need for this application (very wide temperature ranges, extreme sensitivity, video recording). The sweet spot is $250-400 for a camera like the FLIR E4 or similar competitive products—sufficient resolution, adequate sensitivity, and practical software.
One final practical point: thermal imaging is a diagnostic tool, not a replacement for good repair practices. A hot spot on a thermal image tells you where to look, but it doesn’t tell you whether the joint needs reflowing, the component needs replacement, or the board design is just thermally poor. Always combine thermal imaging with visual inspection, component identification, and functional testing. The combination of these tools—thermal imaging plus a good eye—is what catches failures that any single method would miss.