Sensor Stamping Parts: Housings, Contacts and Springs

The first batch of pressure transmitters I watched fail in the field back in the nineties didn't die from a bad chip. They died from the metal wrapped around the chip. Three failure modes show up again and again in sensor stamping programs, and all three start in the stamped envelope rather than the electronics. Seal failure lets condensate crawl past the housing seam and short the board. Signal drift creeps in when the contact spring relaxes past its elastic limit and the baseline wanders. Probe oxidation builds a dielectric film on the sensing tip and the output goes noisy before the unit is even a year old. None of these are bugs you patch in firmware. They are born on the stamping press, in the alloy choice and the tolerance stack, long before the sense element is ever placed.

This article walks the envelope part by part - housing, contacts, springs, shielding - and then the decisions that decide warranty cost: alloy, plating, tolerance, and the program structure behind them. The numbers below are the ones that show up on real sensor prints and real supplier audits, and each section ends with the question to ask before tooling starts.

The Snapshot

• Sensor stampings live in a 0.1 to 3.0 mm thickness envelope, with critical features held to ±0.02 mm and general features to ±0.05 mm.
• Three alloys carry most programs: 304 stainless for housings in ordinary atmospheres, 316L when chlorides or chemicals are present, and C17200 beryllium copper for contacts, probe tips, and springs.
• IP67 is won in the drawing and the weld; IP69K is won in flange flatness - a 0.03 mm bow on the sealing face breaks the seal under high-pressure washdown.
• Gold over a nickel barrier, 0.5 to 3.0 µm with a 1.27 µm minimum on wear zones, is the standard signal-path finish.
• Spring features must never operate beyond 75% of the yield point, or permanent set reads as sensor drift on the controller.
• High-speed progressive dies on Aida presses run up to 300 SPM with ±0.005 mm positioning, and strip enters the press up to 650 mm wide.

What "Sensor Stamping" Actually Covers

We are not talking about structural brackets or body panels. Precision sensor stamping is the 0.1 to 3.0 mm thick envelope around a sensing element - the housing that keeps the world out, the contacts that carry the signal, the springs that hold mechanical preload, and the shielding can that keeps RF noise off the die. The parts are small, the volumes are high, and the tolerance budget is brutal. A typical program runs ±0.02 mm on critical features and ±0.05 mm on the rest. Miss that by a hair and the O-ring won't seat, the lid won't close, or the spring won't return.

The family divides into four stamped groups, each with its own material logic, tolerance logic, and dominant failure mode. Knowing which group a part belongs to tells you which drawing callouts matter and which supplier questions to ask.

Part group	Typical material	Thickness	Critical feature	Dominant failure mode
Housing (drawn or folded)	304, 316L stainless	0.3 to 2.0 mm	Sealing face flatness, draw depth	Seal leakage, pitting corrosion
Contacts and probe tips	C17200 beryllium copper	0.1 to 0.8 mm	Contact force, plating zone	Oxidation film, contact resistance drift
Springs and latches	C17200, C5210 phosphor bronze	0.1 to 0.4 mm	Spring rate, elastic limit	Permanent set, fatigue fracture
Shielding cans and ground frames	304, C17200	0.1 to 0.3 mm	Seam continuity, coplanarity	RF leakage, solder joint stress

Housings: Sealing, Ratings, and the 304 vs 316L Decision

For most industrial housings I spec 304 stainless as the default. It carries 18% chromium and 8% nickel, anneal tensile around 505 MPa, and density 8.0 g/cm³. That buys solid corrosion resistance in ordinary humidity and mild atmospheres at a sane cost. When the sensor sits in a marine, chemical, or chloride environment, 304 isn't enough - pitting starts at the weld line and creeps inward. That's where 316L earns its premium. The 2% molybdenum and the low-carbon chemistry (under 0.03% C) push pitting resistance well up, and the L grade welds without sensitization. Tensile runs 485 to 620 MPa depending on temper, density 8.0 g/cm³, thermal expansion about 15.9 µm/m·°C.

The sealing story is where stamping meets rating. A drawn 316L housing with a laser-welded seam and a precision-formed gasket groove is what gets you to IP67 - dust-tight and immersion-proof to 1 meter for 30 minutes. Push the application into food, pharma, or mobile equipment washdown and you need IP69K: survival under 80°C water at 100 bar from four sweeping nozzles at 14 to 16 liters per minute. That rating is won or lost in the flange geometry and the stamping flatness of the sealing face, not in a rubber gasket alone. A 0.03 mm bow on the flange is enough to break the seal under high-pressure spray, and no amount of gasket compression hides it.

Wall thickness on drawn sensor housings typically runs 0.3 to 1.2 mm, and the draw depth in a single station is limited by the alloy's draw ratio. When the depth exceeds that limit the die stages the draw - a pre-draw station, then the finish draw - and every extra stage adds a station and a tolerance step. Two details separate a housing die that holds its seal from one that doesn't. First, form the gasket groove in the same die as the flange: a formed groove keeps concentricity with the sealing face, while a groove machined as a secondary operation re-opens the tolerance stack and adds a deburring risk on the sealing lip. Second, coin the flange in a dedicated station. Flange flatness problems are usually not a material problem - they are a stripping and blanking sequence problem, or a draw that pulls the flange edge in unevenly. A coining station flattens the face and stabilizes it against coil-to-coil variation.

Rating	Test condition	What the die must deliver
IP67	Dust-tight; immersion to 1 m for 30 min	Drawn housing, laser-welded seam, formed gasket groove
IP69K	80°C water at 100 bar from four nozzles at 14-16 L/min	All of IP67 plus sealing-face flatness held to the 0.03 mm class

Contacts and Probe Tips: Why C17200 Wins the Signal Path

The signal path is the part you cannot afford to let drift. For probe tips, contact springs, and terminal clips I reach for beryllium copper C17200 before anything else. After solution heat treatment and precipitation hardening it hits 1,100 to 1,300 MPa tensile while keeping roughly 22% IACS electrical conductivity - a combination phosphor bronze and most stainless simply cannot match. That conductivity means lower contact resistance and less self-heating at the tip, which matters when your sensor is trying to resolve microvolt-level signals against a noisy supply.

Gold plating is the second half of the oxidation story. A bare C17200 tip grows an oxide and sulfide film in a humid environment within months, and that film is exactly the dielectric layer that makes your output noisy. A 0.5 to 3.0 µm hard gold flash over a nickel barrier stops the film before it starts. The nickel gives a diffusion barrier; the gold gives a chemically inert contact surface with contact resistance measured in milliohms. For probe tips that see repeated mating cycles, I spec a minimum of 1.27 µm gold on the wear zone. Below that, the plating wears through to nickel inside the warranty window and the drift comes back to haunt the service team.

Contact force is the hidden spec on every sensor contact print. The force the tip applies to the mating surface sets the contact resistance budget, and it comes from the geometry stamped into the beam - thickness, width, and bend radius. Because force scales steeply with thickness, a beam cut 0.03 mm over spec can push the force band out of the mating connector's range. That is why contact programs are run on progressive dies with in-die sensing and SPC on the strip, not corrected at a bench. On the plating side, selective reel-to-reel plating puts gold only where the tip actually wipes and leaves the solder tail in tin - a cost decision that matters when a sensor program runs millions of units. The same selective process, at the same 2 to 8 µm capability band used for terminal programs, is what keeps the unit price sane.

Springs: Elastic Limit Is the Whole Game

Every sensor that relies on mechanical preload - a pressure diaphragm return spring, a connector latch, a battery contact - is betting its life on the spring staying inside its elastic region. C17200 is my go-to here for the same reason as contacts: high yield with good conductivity. The design rule I give customers is blunt - never let the operating deflection exceed 75% of the yield point, or you get permanent set and the baseline walks. A spring that relaxes 0.1 mm over ten thousand cycles reads as drift on the controller, and the board gets blamed for a mechanical problem it didn't cause.

The stamping tolerance on spring features is where ±0.02 mm lives or dies. A spring tongue cut 0.03 mm too thick changes the spring rate by more than the sensor's full-scale allowance in some designs. We hold those features on high-tonnage progressive dies with in-die sensing and SPC on the strip, because post-stamp correction of a spring rate isn't practical at volume. You set the rate at the die, not at the bench.

Two production details decide whether the design survives contact with the press. First, the spring should be stress-relieved after forming - a low-temperature heat treatment that removes the residual stress the bend put in and stabilizes the rate against relaxation in service. Second, the raw strip temper must be locked into the purchase spec and verified on the mill certificate, because spring rate moves with yield strength. A spring stamped from a coil at the soft end of a temper band and a coil at the hard end are different parts that share a part number. Lot-to-lot variation in the incoming coil is the quietest way to turn a good die into a drift complaint.

Shielding Cans: Keeping the RF World Off the Die

The shielding can is the quiet part of the program nobody notices until EMC testing fails. For the stamped shielding can family specifically, the geometry rules below decide whether the can is a shield or an antenna. A five-sided drawn can, seam-welded to the stamped ground frame, gives you a Faraday box around the sensitive front end. Shielding effectiveness depends less on the metal than on seam continuity - any gap longer than a tenth of the wavelength you're fighting becomes an antenna instead of a wall. For most industrial sensor bands a 0.15 mm formed lip with a welded seam holds 30 to 40 dB attenuation through the gigahertz range. C17200 works here too when you want the can to double as a spring contact to the lid; 304 is fine when it's pure shielding and you ground it through the weld.

Drawn and folded cans are different economic animals. A drawn can is one piece - no side seams to leak - but the draw depth is limited and the tooling is heavier. A folded can blanks a flat pattern and folds the walls, which opens side-seam joints that must be welded or soldered, but it runs faster and the tooling is cheaper. For volume programs the folded can usually wins on cost; for the highest frequencies the drawn can wins on seam continuity. The ground frame under the can needs coplanarity held on the solder feet, because a foot lifted 0.05 mm above the board plane is a solder joint that cracks in thermal cycling.

Material and Plating Selection

The alloy table below is the one I put in front of every sensor customer before tooling starts. It is a selection table, not a menu - the environment and the signal path decide the row, and the row decides everything downstream.

Alloy	What it is for	Key property	Watch out for
304 stainless	Housings, shields, frames in ordinary atmospheres	18% Cr, 8% Ni; anneal tensile ~505 MPa	Work-hardens in deep draws; pitting at weld lines in chlorides
316L stainless	Marine, chemical, pharma, food housings	2% Mo, low carbon; pitting resistance, weldable without sensitization	Costs more than 304; slightly lower CTE (15.9 µm/m·°C)
C17200 beryllium copper	Contacts, probe tips, springs, spring-loaded shields	1,100-1,300 MPa hardened tensile; ~22% IACS conductivity	Needs precipitation-hardening discipline; higher cost
C5210 phosphor bronze	Spring contacts where C17200 is over-spec	Stable spring temper, solderable, good fatigue	Lower conductivity than C17200
C26000 brass	Low-cost contacts and frames with mild demands	Cheapest, easy to stamp at high speed	Relaxes under load and heat; tarnishes

Plating selection follows the same logic. The finish has to match the duty cycle of the surface it protects - a wipe zone, a solder joint, and a corrosion barrier are three different jobs that belong to three different finishes.

Finish	Typical band	Where it belongs	Caution
Hard gold over nickel	0.5-3.0 µm; 1.27 µm minimum on wear zones	Probe tips, wipe zones, low-level signal contacts	Expensive - use selective plating zones
Silver	2-8 µm	RF and high-frequency paths where conductivity rules	Tarnishes in sulfurous atmospheres
Tin	2-8 µm	Solder tails, low-cost contact finish	Creeps and oxidizes in hot service
Nickel	2-8 µm	Barrier under gold; hard-wearing surface	Hard - can wear the mating surface

Three rules resolve most plating debates. Rule one: if the surface carries signal current through a mechanical contact, it is a gold or silver surface. Rule two: if the surface is soldered, it is a tin surface. Rule three: if the surface does neither, nickel alone is usually enough. The reel-to-reel selective process used for terminal plating applies each finish only where it earns its cost, which is why a sensor with a gold wipe zone and a tin solder tail does not carry the price tag of a fully gold-plated part.

The Temperature Drift You Can't Stamp Out - But Can Design Around

Thermal expansion is the one variable stamping can't remove, only manage. The housing, the contact, and the sensing element expand at different rates, and that mismatch is your temperature drift. 304 runs about 17.3 µm/m·°C, 316L about 15.9, C17200 about 16.7, and the ceramic or silicon sense element is often under 4. That spread forces a differential expansion across the bond line every time the unit heats from -40 to 125°C, and it shows up as apparent offset. The fix is in the design, not the press: float the sense element on a compliant member, or match the contact alloy's CTE to the substrate so the bond sees shear instead of peel. Gold's temperature coefficient of resistance sits near 0.0034 per °C, small enough that a gold-plated contact path adds negligible drift next to the mechanical mismatch.

Tolerances and the Tolerance Cascade

A sensor housing drawing that arrives with only a title-block tolerance is a gamble. The features that decide field behavior deserve explicit callouts, and the table below is the set I look for first on a sensor print. If a feature is missing from the drawing, it will be missing from the inspection plan, and it will be the one that fails in the field.

Feature	Typical tolerance	Held by	Verified by
Sealing-face flatness	0.03 mm class	Coining station	Optical flatness gauge
Datum-to-feature positions	±0.02 mm	Pilot-registered stations	CMM or vision gauging
Spring tongue thickness	Strip tolerance class	Incoming coil cert, in-die check	Micrometer at first article
Bend angles	±1° where function allows	Die angle plus overbend compensation	Angle gauge, CMM
Plating thickness	Per zone spec	Plating line process control	XRF per lot

A sensor program is a tolerance cascade, not a list of independent numbers. Housing flatness drives the seal, the seal drives the IP rating, contact yield drives the drift, and plating thickness drives oxidation life. You cannot tighten one link without checking the effect on the next - which is why the tool and die design review happens before the die steel is ordered, not after the first bad lot.

How a Precision Stamping Program Actually Runs

A sensor stamping program is a tolerance cascade, not a single part. Housing flatness drives the seal, the seal drives the IP rating, contact yield drives the drift, and plating thickness drives the oxidation life. You cannot tune one without watching the others. We qualify every lot on the strip - tensile of the incoming coil, plating thickness by XRF, flatness and critical-feature dimensions by vision gauging - before a single part ships. That's how a part that looks identical on the print behaves for ten years in the field instead of failing in the first monsoon season.

The equipment behind sensor programs starts with 21 presses: Aida high-speed presses in the 25 to 80 ton range running up to 300 SPM with ±0.005 mm positioning, dedicated EV-class terminal presses with automatic feeding, and 45 to 110 ton presses for larger housings and heavier frames. Strip enters the press up to 650 mm wide and from 0.05 to 3.0 mm thick. The tool room builds and maintains the dies in-house with wire EDM, CNC machining centers, and tooling grinders, and the quality lab runs CMM, optical measurement, and in-line vision systems on every program. That combination - high-speed progressive tooling, in-house die control, and vision-based inspection - is what holds ±0.02 mm features at volumes where a bench inspection plan would be a bottleneck.

The program flow is the same discipline as any precision stamping program: a DFM review of the tolerance stack before tooling, a rapid prototyping run to validate form and force, then tool build, first article, PPAP, and production. The difference is where the attention goes. On a sensor program, the prototype stage exists to prove the spring rate and the seal, not just the dimensions - and the production stage carries SPC on the features that decide those two behaviors.

What to Ask Before You Approve a Sensor Stamping Supplier

• Which features carry SPC, and what is the reaction plan when one drifts past 90% of its tolerance band?
• How is plating thickness verified (XRF), and how is the plating zone registered to the die?
• Does first article cover dimensions only, or also flatness, plating, and spring rate?
• How are incoming coil lots verified against the temper spec on the mill certificate?
• What is the containment procedure when an in-process check fails?
• Can they show Cpk history on a similar sensor part from a previous program?

Volume, Tooling, and Program Economics

Sensor programs run the volume gamut. Prototype runs of a few thousand units validate the design, pilot runs of tens of thousands shake out the process, and production runs climb into the millions of units a year. The tooling decision changes with the volume: a single-up die is right for a pilot, while a multi-up progressive die - two, four, or more cavities across a 650 mm strip - is what makes a million-unit sensor program affordable. A four-up die running at 200 SPM, for example, produces 48,000 parts per hour, and the tooling cost amortizes across the whole program quantity. That is why the volume forecast belongs in the RFQ, and why a program quoted as a one-off prototype and a program quoted against two million units are different economic animals. The same logic applies to plating: selective zones and a reel-to-reel process keep the per-part cost flat at volume, while a full-plate specification multiplies cost by the entire surface area for no measurable field benefit.

Build the Envelope Before You Build the Board

If you are specifying a sensor and the housing, contacts, and springs are an afterthought, you are building the failure modes in from day one. Get the stamped envelope right - the alloy, the seal geometry, the plating, the spring design - and the electronics have a fighting chance. Bring us your drawing or even a rough concept and we will walk the tolerance stack with you, pick the 304 or 316L or C17200 that fits the environment, and quote the progressive die high speed stamping program against your volume. Thirty years of field failures say the metal around the chip decides the warranty, not the chip. Send the drawing for a DFM review within one business day, and ask the same questions from the checklist above of any supplier you are considering.