Here's a message I get more often than you'd think, usually left as a voicemail after 10 p.m.: “We need a thermal camera. Do you think it can see through this cabinet?”
I'm a principal field engineer at a test-and-measurement firm that does a lot of emergency work with National Instruments (NI) equipment. I've handled 200-plus rush diagnostics over the last eight years—same-day setups, overnight logging, “can you be on site by shift change” requests. The “can thermal cameras see through walls” question shows up in some form in a surprising number of them.
So let's answer the literal version first. Then I'll explain why it's usually the wrong question entirely.
Short answer: no thermal camera sees through walls—not even FLIR
Thermal cameras measure infrared radiation emitted from surfaces. The detector array translates that radiation into the temperature map you see on screen. Typical thermal cameras work in the long-wave infrared range, roughly 7.5 to 14 micrometers.
At those wavelengths, materials like drywall, plywood, brick, and steel enclosures are opaque. The radiation from an object behind a wall simply doesn't reach the camera. That's the detail most people miss: what actually reaches the camera is infrared radiation emitted by the wall's own surface.
That's why the phrase “see through walls” is misleading. A FLIR thermal camera is a brilliant tool, but it will always show you the surface temperature of whatever material is in front of it. Point it at a closed electrical panel, and you get the temperature of the panel door. You won't see an overheating contactor hiding inside. If heat has been conducting through the metal for a while, a warm patch on the door might suggest a general area to investigate, but it won't identify which component is failing.
You'll notice no reputable manufacturer markets a thermal camera as “see-through.” There's a reason for that: per FTC advertising guidelines (ftc.gov), marketing claims must be truthful, non-misleading, and substantiated. “Sees through walls” would be hard to substantiate because it isn't what the technology does.
What if you truly need to inspect live equipment inside a closed panel? Install an IR window—a port that is transparent to infrared—and aim the camera through that. It's standard practice for switchgear maintenance, and it works. No X-ray vision required.
Why “can it see through walls?” is the wrong lens
Over the years, I've come to see that question as a symptom. When a machine is down and production is screaming, our brains want root causes to look like something. We picture a blackened relay, a swollen capacitor, a visibly blown fuse. If we can just see inside that wall, we'll finally know, right?
The trouble is that a lot of mechanical and electrical faults aren't visible at a particular moment in time. They're intermittent. They happen once every few hours, under load, at a particular operating temperature, or only after the machine has been running for a while. The fault is less like a broken part and more like a bad pattern: a signal that drops out at the wrong time, a tolerance stack that drifts when the machine heats up, a supply voltage that dips under load.
In my role triaging urgent requests, I ask the same three questions before we dispatch: How much time do we have? Is the fault reproducible? What data has already been recorded? The answer to the third question is often “nothing.” And that's the true bottleneck—not the lack of a magic camera, but the lack of recorded data.
Point tools are great. They just don't capture patterns.
Every technician's kit has point-in-time tools. I carry a Fluke T6-1000 Pro digital clamp meter myself—it's excellent for finding out if a wire is drawing current right now without breaking the circuit. A FLIR thermal camera is the right call for finding a loose connection right now. And a photoelectric sensor is great for knowing whether a part is present right now.
But none of them tells you what was happening at 11:03 p.m., the moment the line faulted. A clamp meter reads what a conductor is doing at the instant you clamp it. A photoelectric sensor tells you, through its discrete output, whether an object is present at an instant. A thermal image gives a snapshot of surface temperatures at an instant. Snapshot measurements are useful, but if the fault is random and brief, an instant is exactly what you don't need.
That's the layer most people miss: the problem isn't visibility, it's time. You need a record—many signals, on one timeline—so you can see the correlation that explains the failure.
The cost of looking for a camera instead of a logger
Last March, a plant manager called about a packaging line. The line kept stopping after it had been running steadily for an hour or two. Two interlock relays had already been replaced on his side, and an emergency contractor had charged him $4,200 to clean every connection in the control cabinet, but the stops continued. He asked if we had a FLIR thermal camera he could rent to “look through the cabinet door.”
I wish I could say I convinced him immediately. In hindsight, I should have insisted on a logger from the very first call. But the manager was tired and wanted a visual smoking gun. We did both, actually. We showed him why a thermal camera couldn't see through the steel enclosure, then set up a proper measurement. It took about 90 minutes: a small National Instruments data logger, a CompactDAQ chassis with a thermocouple module and a digital input module, one thermocouple on the conveyor frame, and one tap into the existing photoelectric sensor that confirmed bottle presence at the infeed. (Should mention: we supplied the logger on a rush basis from our own lab bench; there wasn't time to order anything.) We also spot-checked the control feeder with a Fluke T6-1000 Pro digital clamp meter—clean supply, no voltage dip.
The overnight dataset was boring, which is exactly what we wanted. The logger captured three stoppages. Every one occurred after the frame temperature climbed above roughly 43°C. Once the frame got hot, thermal expansion shifted the photoelectric sensor's bracket just enough that its beam lost the reflector. The PLC read “no bottle present” and stopped the line. When the line cooled, the alignment came back, so maintenance could never find a fault on cold checks.
The surprise wasn't an overheating component in the cabinet. It was how obviously the fault pattern popped out once the photoelectric sensor signal, frame temperature, and stop command were on the same timeline. And the fix wasn't a big capital project: a repositioned bracket and a spring-loaded mount. The total cost was roughly $80 in parts and one afternoon of labor.
Before that setup, the plant had spent about $9,000 in replacement relays, contractor visits, and overtime—not counting downtime on a line they estimated at $1,200 per hour.
I don't have hard data on industry-wide misdiagnosis rates, but based on our own 200-plus rush diagnostics, my sense is that at least a quarter of “mystery failures” involve this exact pattern: a fault that only appears under specific conditions over time, while everyone is looking for something visibly broken in a single snapshot. I wish I had tracked that metric more carefully; the anecdotal evidence alone is enough to make me default to logging first whenever a customer says “intermittent.”
The fix: log before you “look through”
If your question is literally about seeing inside a closed electrical panel, the practical answer is an IR window and a good thermal camera. If the suspected fault is a one-time, always-visible issue and you can open the panel safely, a thermal camera can finish the job in minutes. In those cases I'll tell you honestly: skip the data logger; a camera is enough.
But if your fault is intermittent—if it happens under load, after warm-up, once a shift, or only when something else is also true—then the right tool isn't a camera at all. It's a synchronized data logger. Before you spend money chasing an impossible “see-through” image, spend 90 minutes setting up voltage, current, temperature, and one or two process sensor signals on the same clock.
When we build those setups quickly, we use a National Instruments data logger, usually a CompactDAQ with the specific C-series modules for the signals involved. NI has a proper support structure around it, it's easy to scale from four channels to forty, and the software puts all channels on one timeline without drama. In an emergency, boring is good. The one time we needed help wiring a PNP photoelectric sensor into a digital input on short notice, National Instruments support picked up and walked us through it in under 20 minutes. That kind of response matters when a customer's line is down and the clock is loud.
Bottom line: no, thermal cameras can't see through walls—FLIR's included. But once you stop expecting X-ray vision, you get to the more useful problem: you don't need to see through the wall. You need to see the pattern. And that's a data logging problem, not a camera problem.
