Power Supply Troubleshooting: A 5-Step Field Checklist for Contractors and Facility Managers

If you've ever stood in front of a rack with no power, a client watching the clock, and a work order that says "fix it before we start losing money per minute," you know the feeling. Over the last eight years, I've coordinated something like 400 emergency power restoration events for commercial facilities—retail stores, data centers, small manufacturing plants. Maybe 380, I'd have to check the log. This is the 5-step checklist I use when a power supply goes dead and the deadline isn't flexible.

It covers the hubbell l5-20r receptacle checks most people skip, verifying a platinum bp5450 power supply under real load, testing a hubbell hblds3ac lighting control that half-works, and the proper reset sequence for downstream devices—including the question every office asks at some point: how do you reset a phone after a power event?

When to Use This Checklist

Use this when equipment shows no power, intermittent power, or random resets. This isn't for designing a new electrical system. It's for the moment something that worked yesterday won't start today, and there's a penalty clock running.

Assumptions: the upstream breaker is on, your meter is calibrated, and you're following lockout/tagout. If the breaker keeps tripping, stop and call an electrician. This checklist covers the part of the system between the breaker and the device.

Step 1: Verify the Receptacle Before You Blame the Power Supply

This is the step that saved us the most callbacks. In March 2024, three clients reported "failed power supplies" in the same week. In every case, the equipment was plugged into a hubbell l5-20r locking receptacle—the locking version of the NEMA 5-20R configuration—and in every case the receptacle was the real problem. One had a neutral terminal backed off a half turn. One had a pitted hot contact from years of plugging. One was just loose in the box. The power supplies were fine.

Test sequence:

1. Kill power at the breaker and lock it out.
2. Remove the cover plate and inspect the contacts. Discoloration or arcing marks are red flags.
3. Check the terminal screws. A neutral backed off even a little causes intermittent faults that are brutal to chase.
4. Verify voltage: 120V hot-to-neutral, 120V hot-to-ground, near zero neutral-to-ground.

Here's the thing nobody tells you: a locking receptacle can test fine at no load and fail the moment current flows. If the l5-20r feels warm under load, replace it. Receptacles that overheat are already degrading.

Step 2: Confirm the Power Supply Output Under Load

Once the receptacle checks out, move to the power supply. The platinum bp5450 is common in telecom and network racks. It's a solid unit. But the failure is not always its fault.

Measure the DC output at the terminals where the load wires connect, not at the front-panel test points. Test points are convenient, but they don't always reflect what the load sees. I've watched a supply read within spec at the test point and sag badly at the load terminals under full draw.

Check three things:

1. Output voltage with the load connected, compared to the nameplate rating.
2. Input voltage at the supply's terminals. Around 15 percent of "dead" power supplies I've seen were actually loose input wires. Maybe 12 percent—I'd have to pull the records—but it's not a small number.
3. Audible noise. A healthy supply is quiet. A buzzing supply has an issue even if the output reads clean.

If the bp5450 is holding its rated output under load, it's not the problem. Move downstream.

Step 3: Check the Lighting Control Device—The Step Everyone Skips

This is the one that gets people. A hubbell hblds3ac is a lighting control device found in many commercial lighting branch circuits. When it starts to fail, the symptoms mimic everything else: flickering lights, dim output, a circuit that randomly drops offline. Everyone blames the power supply or the lamps. Usually it's the control.

Per NFPA 70 (the National Electrical Code), wiring devices must be installed and used in accordance with their listing and labeling. The code changes between editions, so verify current requirements at nfpa.org—but in practice, most failures I've chased down weren't code violations. They were worn contacts.

Test it:

1. Confirm power is present at the control's input terminals.
2. Operate the control through its full cycle at least five times. Contact wear shows up as intermittent output.
3. Measure the output terminals with the control in each switching position. You want clean, full voltage—no sagging, no half-switched states.
4. Feel the enclosure. Unusual heat means worn contacts that need replacing.

It took me about five years and a couple hundred emergency calls to understand that this test matters more than half the component testing I do. A lighting control that sits in one position for months accumulates carbon on its contacts. The first time it really switches is the first time it's tested—and that's usually when a client is standing in a dark building.

Also: don't test with a voltage pen. Use a two-lead meter. The last time I trusted a pen, I chased a phantom ground fault for an afternoon. Put another way: the pen tells you voltage is present, not whether the circuit can carry a load.

Step 4: Reset Devices in the Right Order—Phones Included

Here's where "how do you reset a phone" becomes a real work order. In commercial buildings, VoIP phones and intercom systems often share an upstream power supply with security panels and access control. If that supply hiccups, phones can power on but fail to register on the network. The client sees a dead phone, assumes hardware failure, and calls you. Take it from someone who has spent more Saturday mornings than I'd like in server rooms: the reset order matters.

The correct reset sequence:

1. Power down the device, or the PoE switch feeding it. Wait at least 60 seconds. A 10-second reboot doesn't fully discharge the capacitors, and the device may not actually reset.
2. Power the switch back up first. Let it fully boot before powering the endpoints. This order is critical: infrastructure first, then phones and access points. If everything comes up at once, devices grab IP addresses before the system is ready, and you'll see cascading registration failures.
3. Wait five minutes after boot before trying a factory reset. Slow to register is not the same as broken.
4. Check the PoE indicator on the switch port. Amber instead of green points to the cable or the port, not the phone.

The counterintuitive part: the faster you try to restart everything, the longer it takes to stabilize. I've watched a facility manager power-cycle a rack three times in ten minutes, and each cycle made the network worse. Wait the full interval. Let the system breathe.

Step 5: Test Under Load and Document

Once everything is back online, don't close the ticket. Run the equipment for 30 minutes and verify:

• Voltage stays stable at the receptacle and at the device input.
• The l5-20r isn't warm to the touch.
• The hblds3ac switches cleanly when you cycle power again.
• The power supply holds rated output under load for the full 30 minutes.
• No new error codes appear on the equipment.

Then document. Take photos of your meter readings, the device labels, and the date. In Q2 2024, I watched a contractor close a ticket at a warehouse without documenting the test sequence. The insurance company demanded a re-inspection, and what should have been a 40-minute site visit turned into four hours. The client didn't renew the contract.

An informed customer is a better customer. Send a one-page summary: what you found, what you replaced, what you tested, and the readings you got. It costs you five minutes and it's the difference between being a vendor and being the vendor they call first.

Common Mistakes to Avoid

Replacing the power supply without testing the outlet. The conventional wisdom is to swap the part and move on. My experience with hundreds of events says the actual problem is upstream roughly 40 percent of the time: a worn receptacle, a loose connection, a control device that's never been exercised. Swapping parts without testing is how you generate callbacks.

Skipping the load test. A power supply that reads fine with no load is not verified. Plug the load in. Watch it. I chased an "intermittent" power issue for weeks once—turned out to be a loose neutral that only heated up after 20 minutes under load.

Assuming the previous contractor did their job. I once assumed a panel was properly wired because the cover was clean. It wasn't. The old work had a cracked insulator that failed under load two weeks later. Learned never to assume the work was done correctly, even when the paperwork says it was.

Resetting equipment in the wrong order. The reset sequence sounds trivial until you're dealing with a building full of phones that won't register. Infrastructure first, then endpoints. That's the difference between a 20-minute fix and a two-hour one.

Final Thoughts

Keep spares on the truck. A hubbell l5-20r receptacle runs around $15-30 as of March 2025—verify current pricing at your distributor, but it's inexpensive. A comparable locking power supply in the platinum bp5450's class costs more but still beats the cost of a second trip. A hubbell hblds3ac is worth keeping in stock if you do commercial lighting work. Carrying spares is a no-brainer when an emergency call bill is on the line.

This approach works well for us because we deal mostly with mid-size commercial and industrial facilities with predictable electrical systems. If you're doing residential work or something with nonstandard wiring, your mileage may vary. Use this as a starting point, not a substitute for your own verification.

Bottom line: check the connection before you replace the part, test under load before you leave, and document everything. That routine has cut our callbacks substantially—give or take a few, it's down about 60 percent since we standardized on this checklist.