It was a Tuesday in March 2023. I'd just wrapped up a system redesign for a client's cleanroom HVAC setup—six zones, all needing tight differential pressure control. The spec called for Dwyer DP sensors, which made sense. Their Magnehelic gauges are basically industry standard, and the price was… well, reasonable enough for the budget we had. I submitted the order: six Dwyer 616W-2-NIST transmitters. Felt good about it.
A month later, the client called. Pressure readings were drifting. Not a little, either—we were seeing variations of ±0.8% of full scale on a system that needed ±0.25%. I'd spec'd a sensor with ±0.5% accuracy. It met the spec on paper. But in the field, it wasn't even close. That's when I learned how much I didn't know about reading datasheets.
The Setup: Why I Chose Dwyer
Look, I'm not here to trash Dwyer. Their products are solid for a lot of applications. The 616W series is a workhorse—good temperature stability, decent long-term drift specs, and the price is hard to beat for a NIST-traceable transmitter. For general HVAC monitoring? Perfectly fine.
But here's the thing I didn't fully appreciate: accuracy specs aren't just numbers. They're context-dependent. The 616W has a stated accuracy of ±0.5% for ranges above 1 in. w.c. But that's at ideal conditions: 20°C, stable pressure, no vibration. Our cleanroom had temperature swings of ±3°C across the day. The ductwork had minor mechanical vibrations from the fans. And we needed a 4-20 mA loop that could handle long cable runs.
I had looked at the spec sheet, but honestly? I'd glossed over the fine print. The combined error (accuracy + thermal drift + span stability) wasn't explicitly stated in the main table. I had to dig into the manual to find it. And by then, the order was already placed.
The Moment It Went Wrong
The first sign of trouble came during commissioning. The readings looked okay on startup—within 1% of our reference pressure standard. But after the system had been running for a few hours and the temperature stabilized higher, the drift started.
We had six units installed. Three of them were consistently reading high by about 0.6% FS. One was reading low. Two were within spec. There was no pattern I could diagnose without pulling the sensors and bench-testing them. That's when I made the call to the client: we need to re-evaluate the sensor selection.
"I think we need a different product," I said. "The Dwyer is close, but not close enough for this application."
The client wasn't thrilled. They'd already approved the Dwyer based on my spec. Now I was asking for change order approval. The cost to replace all six sensors, including labor for re-installation and re-commissioning: $4,800. Plus a 3-week delay because the replacement sensors had a longer lead time.
That $4,800 figure is burned into my memory. It wasn't just the replacement parts. It was the lost time, the credibility hit with the client, and the embarrassment of having to admit I'd made a rookie mistake.
The Pivot: Finding the Right Sensor
After that disaster, I did what I should have done the first time: I actually compared the total error band across multiple manufacturers, including Dwyer's own higher-end options. For a cleanroom application needing ±0.25% accuracy across a temperature range of 10-40°C, here's what I found:
- Dwyer 616W-2-NIST: ±0.5% accuracy at 20°C. Thermal drift adds ~0.02% per °C. Total error band over 10-40°C: approximately ±0.8% FS. Not sufficent.
- Dwyer 646B-0-NIST: Higher-end transmitter, ±0.25% accuracy. Better thermal compensation. Total error band closer to ±0.4% FS. Closer, but still tight.
- Alternative brand (Setra, now part of Gems): ±0.25% accuracy, but with a tighter thermal spec (0.01% per °C) and better long-term stability. Total error band under ±0.3% FS.
We ended up using the Setra ASM for that project. It cost about 25% more per sensor, but the total installed cost was actually lower than the Dwyer when you accounted for the fact that we didn't need to over-spec and add contingency. Lesson: The cheapest sensor is rarely the cheapest solution.
"An informed customer asks better questions and makes faster decisions." That's something I tell my clients now, partly because it's true, but also because I wish someone had told me that before I placed the order.
The Real Lesson: Reading Past the First Page
So what did I learn? Three things, which I now use as a checklist before I spec any sensor:
- Don't just look at accuracy—look at the total error band. Accuracy at 20°C tells you almost nothing about real-world performance. Thermal drift, span stability, and hysteresis all matter. Ask for the combined spec across your operating temperature range.
- Installation matters more than you think. For DP sensors, a big one is straight pipe requirements. If you're using a pitot tube or averaging sensor, you need sufficient straight run to get a stable reading. We had a 90-degree elbow too close to one of the sensing points—that alone added noise that made the Dwyer's accuracy irrelevant. We fixed that during the re-installation.
- Don't assume 'NIST traceable' means high accuracy. It just means the calibration is traceable to a national standard. A ±1% sensor can be NIST traceable. High accuracy is a separate, more expensive feature.
Honestly, I still use Dwyer products. Their Magnehelic gauges are fantastic for visual indication. Their flow meters work well in ductwork. But for applications where tight accuracy over a temperature range is critical, I now know to look beyond the brand name and dig into the datasheet. Sometimes the right choice is a different Dwyer model—their 646 series is good—but sometimes it's a different brand entirely.
Closing Thoughts
That $4,800 mistake was expensive, but it made me a better engineer. I now spend an extra 30 minutes per sensor spec reviewing the fine print. I've caught 12 potential mismatches this year using that checklist. I'd rather spend 10 minutes explaining options to a client than deal with a field failure later.
If you're spec'ing DP sensors for a cleanroom, hospital, or any temperature-sensitive application, don't trust the first number on the datasheet. Do the math for your actual conditions. And if you're not sure, ask the manufacturer for the total error band over your specific temperature range. They'll usually provide it. If they can't or won't, that's a red flag.
Save yourself the $4,800. Read the fine print.
