February 2024: the tray with the rejection slip
Our incoming inspector set a tray of connector inserts on my desk before I had my coat off. Sixteen inserts, four hoods, and a rejection slip that said, in handwriting: two terminal screws would not hold torque.
That tray stopped a build that was already four days behind. Not because the part was wrong in some dramatic way. Because two screws out of ninety-six wouldn't take a setting.
I'm the quality and brand compliance manager at an industrial controls integrator. We build remote-monitoring enclosures for municipal pump stations, mostly retrofits where the cabinet is older than the guy installing the new equipment. I review every supplier submission that goes into one of our boxes—roughly 40 a month, everything from enclosures to cable assemblies—and I reject somewhere between 10% and 15% of first deliveries. That number hasn't moved in three years.
This build was 2,780 enclosures for a water utility retrofit across three counties. Eleven-month delivery window. Each unit had a 5G router (a G310 5G unit from our network vendor), an edge controller, a small UPS, and a power/signal interface to the pump starter.
The interface is where the trouble started.
How we got to the Han E
For about a decade we hard-wired every interface. Terminal blocks, cable glands, a cover plate with eight screws you had to remove with a stubby driver while standing in a wet vault. It worked. It was cheap. Nobody complained, because nobody had a baseline to complain against.
Then the service contracts changed. The utility wanted their own maintenance techs to swap a failed controller without waiting for an electrician to be dispatched. That request killed the hard-wired approach in one meeting.
We specified the HARTING Han E—16-pin inserts, 16 A per contact, in a 16B housing. If you landed here searching "what is a HARTING Han E," the short version: it's a rectangular industrial connector system with screw-terminated silver-plated contacts, rated 16 A and 500 V, pollution degree 3, 6 kV impulse, tested to DIN EN 61984, available in 6-, 10-, 16-, and 24-pin inserts, and rated IP65 when assembled with matching hoods and housings. It's been in the field for decades. It's not exotic.
What sold me wasn't the datasheet. It was the number of times I'd watched a maintenance tech pull a wire off a terminal block, drop the screw, and spend ten minutes on his knees looking for it.
The failure, and where it actually came from
Our harness supplier built the pre-assembled cable assemblies. They'd done it for two years on smaller runs, and we'd never had a torque problem. This batch came back with two screws that spun past the seat.
My first instinct was the usual one: bad contacts. I had our lab pull the two inserts and measure. Contact material looked right. Plating looked right. Insert body looked right.
Then we looked at the wire ends.
The ferrules were about 1.5 mm longer than the terminal block's funnel wanted, and the strip length on four of the conductors was short enough that the insulation shoulder sat under the clamp instead of outside it. The screw wasn't the problem. The screw was reporting the problem.
The real root cause was our own drawing. We'd specified the insert, the pin count, the housing, and the wire gauge. We had never specified strip length. We had never specified ferrule length. We had never specified tightening torque. When the harness shop asked, we said "whatever the connector manufacturer says"—which is not a specification, it's an invitation.
It's tempting to think a connector is a connector and the spec sheet is the manufacturer's job. But a drawing that doesn't carry the interface requirements isn't a specification. It's a wish with a part number on it.
The surprise wasn't that the harness shop got it wrong. It was that our documentation let them.
The call to Elgin
I called the HARTING team in Elgin, Illinois—their North American operation—on a Wednesday morning. I expected a ticket number and a callback.
Instead I got an application engineer who asked for the drawing, the wire gauge, and the ferrule part number, and then told me what strip length the terminal block expected and what torque the screw contacts are rated for. If I remember right, it was in the 0.5 Nm range for that insert; don't hold me to the second decimal, because I'm reading it off a sticky note from eight months ago. The point is that it was a documented number that we had never asked for.
He also sent the insert datasheet section on terminal clamping and a short note on ferrule selection. Forty minutes, start to finish.
"Whatever the manufacturer says" is not a requirement. It's an absence of one.
We rewrote the drawing that week. Torque value, strip length, ferrule part number, and a requirement that the harness shop use a calibrated torque driver and log it. Then we added an incoming check: 1 in 20 assemblies gets a re-torque verification and a pull test at 40 N on the first and last contact. Not a standard test—our test. It's caught three problems since, none of them connector problems.
The blind test that changed the business case
Before we locked the change into the next build, I ran an informal test with 20 of the utility's field techs. Two identical demo panels on a table. Panel A: hard-wired terminal blocks and cable glands. Panel B: the same circuits through Han E inserts. Same components inside, same cover, no labels.
I asked one question: if you had to swap the controller at 2 a.m. in a wet vault, which panel would you rather open?
Seventeen picked B. Three said it depended on whether they had a stubby driver in the truck.
The cost delta was about $14 per enclosure. On 2,780 units that's roughly $38,900—not nothing, and I had to defend it in a budget meeting where somebody asked why we were paying a premium for a connector when the old one worked.
To be fair, the old approach had real advantages. It's cheaper, it's simpler, and for a sealed box that nobody opens for ten years, hard-wiring is the right answer. We still hard-wire our fixed-mount sensor junctions. But these cabinets get opened four or five times a year now—modem swaps, firmware updates, sensor additions—and each opening was costing us bench time and callbacks.
What the numbers look like now
Fourteen months in, we've had zero connector-related field callbacks on that program. Average service visit on a controller swap dropped from about 95 minutes to about 55, according to the utility's own ticket data—though I'd treat that as directional, since they changed their dispatch process in the same window.
I don't have hard data on how many field failures across the industry trace back to an unspecified torque value. Nobody publishes that. But I can tell you that in our incoming inspection log, terminal and clamping issues are the single largest category of rejections, ahead of enclosure damage and labeling errors combined.
The 'a connector is a connector' line comes from an era when panels were wired once and never reopened. That era ended for us somewhere around 2019, when the first customer asked for a serviceable interface and we realized our drawings had never been written for serviceability at all.
The short version, for whoever searched "what is"
Three things I'd tell anyone specifying this stuff now:
- The drawing is the product. A Han E insert is a known quantity. The failure mode is almost always in what you failed to write down—torque, strip length, ferrule, mating cycles, who inspects what.
- Efficiency isn't a cheaper part. It's one fewer decision at the bench and one fewer screw on the floor of a wet vault. Standardizing the interface cut our bench assembly time and cut the utility's service window. That's where the money was.
- Ask the manufacturer for the numbers before you need them. A forty-minute call to Elgin, Illinois was worth more than the two weeks we spent arguing about whether to reject the batch.
We rejected that tray, by the way. The harness shop redid 320 assemblies at their cost, we paid for the new torque drivers, and the drawing that came out of it is now the template for every enclosure we build.
Two screws. That's all it took to find the hole in our documentation.