The second week of February 2025, I walked into the office at 7:15 a.m. and knew it was going to be one of those weeks. Randy from facilities was standing in my doorway holding the shop compressor's pressure gauge like it was evidence. In the breakroom, the under-counter ice maker had a yellow light on. And my phone was pinging with a temperature alarm from the server closet, where the Johnson Controls AC unit had been cycling all night.
I should back up. I'm the office administrator for a 120-person structural engineering firm. I manage all facilities purchasing — roughly $350,000 a year, give or take, across maybe fifteen vendors. I report to both operations and finance. I'm the one who orders the filters, the parts, the service visits, and occasionally the peace offerings when the coffee machine breaks.
That week, everything broke at once. And it taught me more about HVAC maintenance than any product training I've sat through.
Why Is My Ice Maker Not Making Ice?
The breakroom ice maker is one of those machines everyone ignores until it stops, at which point it becomes the most important piece of equipment in the building. By 9 a.m., three different engineers had asked me "why is my ice maker not making ice?" — and one of them had already Googled it and printed the results. That was a new one.
The internet's answer, as far as I can tell, is consistent: check the water supply line, check the freezer temperature, and check the inlet valve. I did all three. The line was fine. The coil was cold. The inlet valve... I want to say it tested fine, but I'd have to check my notes. The point is, I was ready to order a $150 valve kit based on a YouTube video.
Then I noticed the water filter. It had been in place for eleven months. The manual says to replace it every six. That's on me — I'm usually good about the calendar, but the machine is tucked under a counter and out of sight. I found a replacement filter for $24, swapped it in, and within two hours it was producing its first batch of cubes. No valve. No service call. No $150 anything. (Note to self: put the filter reminder on the shared calendar this time, not just my own.)
The Johnson Controls Sensor That Died of Dust
The server closet was the more serious problem. We run a small data center — two racks, a UPS, and a dedicated cooling unit from Johnson Controls. It's a split system, the commercial kind with the metal enclosure and the flashing LED codes, a unit that handles data center cooling HVAC for a room that was never designed to be a server room in the first place.
Monday morning, the wall controller displayed a sensor fault and the compressor shut down. If cooling stays off, the room hits 90 degrees fast, and our NAS starts making unhappy noises.
I pulled up the Johnson Controls temperature sensor manual online — a twenty-page PDF with a resistance table and a fault isolation chart. If I remember correctly, it says to measure resistance across the sensor leads and compare the reading to the table. At 77°F, a healthy 10k thermistor should read about ten thousand ohms. My meter read open circuit. No continuity at all. The manual was clear: replace the sensor.
I called the local JCI dealer. The pleasant lady on the phone quoted me a $450 trip charge, a $300 sensor, and a two-week lead time because the part had to be "sourced from the regional warehouse." I put the phone down and stared at the budget spreadsheet. We could absorb it. But it felt wrong, and I've learned to trust that feeling.
Here's something vendors won't tell you: if you ask for the fault code history on a commercial controller, a tech will often read it to you over the phone. I called back, asked, and got a log showing the closet temperature had spiked to 96°F over the previous weekend — right in the range where the sensor saturates. The sensor wasn't the root cause. It was just the one that cried out.
Let me rephrase that. The sensor wasn't lying. It was reporting a room that had gotten dangerously hot. The question was why. The answer, it turned out, was dust. The condenser coil was packed with fine construction dust from the renovation two floors down, and the unit had been running hot for weeks until it tripped the safety. I cleaned the coil with a $9 can of coil cleaner, replaced a cheap intake filter, and reset the controller. The sensor read correctly and the cooling kicked back in. ASHRAE's preventive maintenance guidance recommends inspecting filters every 30 to 90 days depending on dust load — and our renovation-adjacent closet clearly sits at the aggressive end of that window.
The dealer's $750 fix would have technically "worked." It also would have solved nothing, because the new sensor would have overheated the same way within a month. Everything I'd read about "sensor fault means replace sensor" was, in our specific case, wrong at the root.
The 'Dead' Air Compressor — and the $12 Capacitor
Then there was Randy's compressor. He'd been asking for a new one for months. The old unit, some no-name brand from 2013, wouldn't build pressure past maybe 20 PSI. Randy wanted a 4-gallon DeWalt air compressor — the 135 PSI pancake model — and honestly, it was a reasonable request. I had it in the cart.
The same HVAC tech who'd helped me with the sensor issue happened to be in the building the next day for a different job. He walked past the shop, looked at the compressor, and said, almost in passing, "bet it's the capacitor." Fifteen minutes and a $12 part later, the machine was humming like it had never failed.
We did still buy the DeWalt — on our timeline, with a proper PO, not as an emergency. The old unit is undersized for what Randy needs, and raising shop air pressure is worth spending on. The difference is we didn't pay a rush fee, and we didn't buy it while the old one was bleeding us dry. The conventional wisdom is that a compressor that won't build pressure is dead and should be replaced. My experience with this one compressor says to check the cheap capacitor before you order the $400 machine.
The Honeywell Home Thermostat and the Shape of Change
After that week, I went home and looked at my furnace's old wall thermostat — a Honeywell, but from the 1990s, with a mechanical dial and a little sliding switch for HEAT/OFF/COOL. It worked fine. I'd been thinking about replacing it for a year.
I ended up putting in a modern Honeywell home thermostat — the smart model with the app. The install took about twenty minutes. Two wires, a base plate, and a whole lot of wondering if I'd wired it backward. (I hadn't.)
I mention it because the contrast is the point. At work, I spent a week decoding commercial building controls — sensor resistance, fault logs, controller resets. Johnson Controls means a building automation stack, and I still can't fully navigate the Metasys interface after six years. At home, a $100 thermostat gives me a diagnostic readout, a filter reminder, and an energy usage graph I didn't ask for but now check every week. That's the industry changing in real time.
What was commercial-grade capability in 2020 is now the default in consumer gear. But that week also reminded me that the fundamentals haven't changed. Filters still clog. Capacitors still fail. Coils still collect dust. The execution has transformed — the hardware is smarter, the manuals are online, the diagnostics are deeper — but the basics? Still the basics.
What I'd Tell Another Administrator
I'm not an HVAC tech. I'm the person who orders the HVAC tech. But after six years of this, I have a few rules that have survived contact with reality:
- Run the diagnostic before you buy the part. Johnson Controls publishes the sensor manual online, with the resistance table and the isolation steps. Take the reading yourself. Then ask what might have caused the part to fail in the first place. Parts don't just die; they're usually killed by something upstream.
- Change the filter before you change the part. The ice maker's water filter, the cooling unit's intake filter, the compressor's air filter — every one of our failures was downstream of something clogged. A $24 filter beats a $300 part every time.
- Ask for the fault code before you authorize the service visit. Get the logged data first, then decide. That one habit saved us about $750 and a week of downtime.
I also ordered four spare water filters for the ice maker. Because I've learned that, in an office of 120 engineers, ice is not a luxury — it's infrastructure.
