The Project That Almost Got Away
In February 2024, our team in Colorado Springs was under the gun. A smart‑agriculture client had ordered 200 prototype boards that would integrate a Semtech LoRa module (SX1276) into a slick clear phone design. The deadline was March 1st. The contract? $15,000. My role as senior application engineer was to handle the technical integration and sign off on the hardware.
I've been doing this for about six years. I've personally made (and documented) eight significant mistakes that collectively wasted roughly $12,000. After the third disaster, I started keeping a checklist. But this time — I didn't use it.
The First Sign of Trouble
Assembly went smoothly. We flashed the firmware and powered up the boards. The LoRa modules came alive, but the signal strength was about 8 dBm lower than expected. Not a dealbreaker for the client, but the datasheet promised –2 dBm max. Something was off.
I started with the power rail. The decoupling capacitors on the module's VDD line — two 10 µF ceramics — were the usual suspects. If one was bad, the module might not get clean power.
I grabbed my multimeter, switched it to capacitance mode, and touched the leads across one cap. It read 9.8 µF. “Looks fine.” The other read 10.1 µF. I shrugged and moved on to a different suspect.
That was my first mistake.
The Rabbit Hole
Three days later, I'd swapped antennas, re‑flowed the whole board, and even ordered a new batch of modules (that cost $600 plus rush shipping). Nothing fixed the signal drop. Our project manager was getting nervous — the client had already pushed back once.
Then a colleague from our group in the bay area called. “Did you check the ESR on those caps?”
I hadn't.
Let me rephrase that: I hadn't even considered that a capacitor could show correct capacitance but have high equivalent series resistance (ESR), which kills the power filtering at RF frequencies. That's exactly what had happened. The 10 µF reading was a surface illusion — the ceramic had degraded, raising its ESR from 0.1 Ω to nearly 5 Ω. The LoRa module couldn't draw current fast enough.
People assume a multimeter capacitance reading is straightforward — touch the leads and read the number. The reality is that without proper discharge, isolation, and an ESR check, you can get dangerously misleading results. My cheap handheld meter didn't even measure ESR.
The Right Way to Test a Capacitor with a Multimeter
I'm not gonna turn this into a full tutorial, but here's the quick version I now have taped to our workstation:
- Disconnect and discharge. Remove the cap from the circuit (or at least lift one leg). Short the leads with a resistor to safely drain any stored charge.
- Set the meter correctly. Most meters have a dedicated capacitance mode; some require you to select a range manually. If the reading seems too perfect (e.g., exactly 10.0 µF every time), suspect the meter's range limits.
- Check ESR. A dedicated ESR meter or an LCR bridge is better. For RF circuits like LoRa modules, ESR above 1 Ω on a ceramic cap is a red flag.
- Use a known good capacitor as a reference. Test a new one from the same batch first.
If I remember correctly, our replacement caps measured 10 µF with an ESR of 0.15 Ω. That's what we should have seen all along.
The Cost of Certainty
It was February 26th. We needed 200 good caps by March 1st. Our regular distributor quoted 5‑day lead time — that would be March 2nd, one day late. The client had made it clear: miss the deadline and the order could be canceled.
I placed a rush order with a premium supplier. The caps themselves cost $45. The rush fee — $280. Total: $325. I have mixed feelings about rush‑order premiums. On one hand, $280 for overnight shipping feels like gouging when the parts weigh less than a stamp. On the other, that $280 saved a $15,000 contract. The alternative was missing the deadline, losing the client, and probably damaging our group's reputation.
In March 2024, we paid $280 extra for guaranteed delivery. The alternative would have been missing a $15,000 event.
That experience cemented my belief: in urgent situations, the certainty of delivery is worth a premium. Uncertain cheap options aren't cheaper if they fail — they cost time, credibility, and future business.
What I Wish I'd Known
I don't have hard data on how many prototype delays are caused by capacitor misdiagnosis. But based on our last 20 integration projects, I'd guess at least three had similar issues that went unnoticed for days. We've now caught seven potential errors using a pre‑flight checklist in the past 18 months.
The biggest lesson: test capacitors properly the first time. Don't assume the multimeter tells the whole story. And when time is tight, pay for delivery certainty — it's cheaper than the alternative.
If you're in Colorado Springs and working with LoRa modules, drop by our lab. I'll show you our capacitor testing procedure. And I'll probably tell you this story again.
