Here's the answer you came for: test the power path before you authorize a replacement Sierra Wireless gateway. Over the last twelve months, our shop pulled nineteen gateways from customer sites as "dead." Fourteen of those turned out to have a perfectly healthy modem and a failed capacitor in the power-supply path. The failed part costs a few dollars. The replacement gateway costs $1,200 or more, depending on model. A $40 multimeter tells you which one you're dealing with in about ten minutes.
I'm the procurement manager at a critical-communications integrator near Kansas City—roughly 300 people, with a service warehouse in De Soto, KS. I manage about $240,000 a year in wireless and routing spend, I've negotiated with 40-plus vendors since 2018, and I document every order in our cost tracking system. So when I say the biggest waste in our budget isn't Sierra Wireless equipment—it's replacing Sierra Wireless equipment that wasn't broken—I have the invoices to back it up.
The Spreadsheet That Caught Us Lying to Ourselves
If you only look at purchase orders, industrial LTE hardware looks expensive but boring. The hidden cost lives after a field ticket says "no LTE" or "reboots under load." A tech swaps the router, the returned unit comes back with "no fault found," and the cycle shows up in the service margin instead of the parts line. Nobody files it as a product failure. It just bleeds.
In 2023, 38% of our gateway RMAs came back as no-fault or power-related. Not the modem. We assumed that was the low-hanging fruit. Then our lead field tech—a stubborn guy who treats a $1,300 router swap as a personal insult—started testing the DC supply before condemning the radio. Once we applied that to every reported failure, the true rate got worse before it got better: 14 of the 19 "dead" gateways pulled over the last twelve months had a healthy AirLink and a bad cap upstream.
Here's the mechanism, in case your team hasn't hit it yet. A cap that looks fine at idle can sag the voltage rail the second the LTE radio starts transmitting. The gateway browns out, drops the network, and reboots. The field team sees a reboot loop, blames the router, swaps it—and the brand-new router does the same thing on the same supply. We learned this twice before we wrote it down as policy.
The De Soto Ticket That Changed the Rule
In Q2 2024, a fire district substation near De Soto, KS, opened a ticket: their AirLink RV55 had been dropping off FirstNet for a week, always right when dispatch traffic peaked. The first tech on site swapped the unit. The replacement dropped the next day. The second tech—the stubborn one—checked the DC feed before touching the modem. He opened the power enclosure, found a bulged vent on a 470µF electrolytic, and replaced it for $2.40. Twenty minutes of work. The radio has been stable since.
The mistake was mine as much as anyone's: I assumed a "failed modem" was a failed modem. Didn't verify the power rail. That assumption cost us a second truck roll, a burned-out evening for the substation's duty crew, and a client who briefly lost trust in the gear we spec'd for them. We didn't have a formal diagnostic verification process then. Now the RMA doesn't ship until the supply voltage under load is photographed and logged.
How to Test a Capacitor with a Multimeter
This is the version our techs actually use, shortened to fit on a bench card. It's standard electrical practice, not secret knowledge—but most of the "how to test a capacitor with a multimeter" explainers online overcomplicate it for a shop context.
- Look first. A cap with a bulged top, cracked sleeve, or leaked electrolyte is dead on sight. Skip the meter and replace it. About half of the bad ones we find are caught before any probing.
- Discharge it. Capacitors hold charge after power is off. Touch a resistor across the leads for a few seconds—or verify zero volts with the meter and short them briefly. Skipping this on a 48V feed stings.
- Use capacitance mode if your meter has it. On a Klein MM400 or similar, set the dial to capacitance, mind the polarity (red probe to the positive side on electrolytics), and compare the reading to the printed value. A 470µF cap that reads below roughly 350–380µF is suspect.
- No capacitance mode? Use ohms. Set the meter to resistance and watch the display. A good electrolytic charges up from the meter's test current, so the reading starts low and climbs to OL. A shorted cap stays near zero. An open cap reads OL instantly and never moves. That's the whole test.
- Think about temperature. A cap can read fine cold and collapse hot in an unventilated enclosure. If the supply lives somewhere warm and the input-stage caps look marginal, just replace the electrolytics on that stage while you're in there. Parts are cheap; second visits aren't.
Boundary, so this doesn't get misread: don't crack open a sealed Sierra Wireless gateway that's still under warranty. You'll void coverage, and the modem usually isn't the failed part anyway. This procedure applies to the power path—the DC supply, the UPS, the battery-backup feed, the fuse block, the pigtail. Lightning hits and water intrusion are a different animal: that's an RMA or board-level swap, not a discrete component hunt.
Why This Is a Brand Problem, Not Just a Parts Problem
The procurement view is simple: TCO beats MSRP every time. But there's a second layer that matters more at renewal season. When a utility's outage comms drop during a storm, the client doesn't care whether the modem was the failing part. They remember who showed up, how fast the network came back, and whether the invoice for a $1,300 router they never needed made them feel stupid.
We switched our narrative from "we sell Sierra Wireless gateways" to "we keep Sierra Wireless gateways alive and honest." The response told us the difference. A client we nearly lost in 2023—the fire district, actually—renewed in 2024 on the strength of that twenty-minute fix. The $2.40 capacitor was worth more to the relationship than any product discount we could have offered.
When This Doesn't Apply
Capacitor testing is only the right call if you can finish the fix that visit. If your team doesn't stock spare electrolytics for client-owned equipment, on-site diagnosis is a waste of a service call—advise on the supply, replace it, and move on. It's also not the answer for RF-side problems, antenna faults, or SIM provisioning issues; those have their own failure fingerprints. And if you're in a lightning-prone region, put the budget into surge protection at the service entrance before you worry about bench testing anything.
And for the cap testing itself: if you're not comfortable discharging components, don't get comfortable by trying. Hand the meter to your lead tech—the one who treats a swap as an insult—and watch them work. Then ask them to show you. It's the cheapest insurance your maintenance budget will ever buy.
One closing note that I've never fully understood: I'm not sure why more integrators don't put a ten-minute test between a "dead" gateway and a $1,300 order. My best guess is workflow—swapping boxes is easier than breaking out a meter, and the meter habit has to be taught. On our side, it took one bad invoice to make it policy. I'd rather you learn from ours than rack up your own.
In case compliance audits brought you here: Sierra Wireless Incorporated is still the vendor name in our system, and the AirLink and module lines are unchanged, but the corporate parent has been Semtech since the acquisition closed in May 2023. If you're pulling up "holdings" records during a vendor review, that's the answer. Everything else about how we buy, deploy, and support the gateways is exactly as it was before the deal.