Custom Wire Harnesses for Small Medical Devices: What Separates a Good Build From a Recall
2026-09-10 15:16Custom Wire Harnesses for Small Medical Devices: What Separates a Good Build From a Recall
A customer sent us a chest lead cable last spring with a complaint we hear often. The ECG trace looked clean on the bench and noisy on the patient. Shielding was correct. Connector was correct. The jacket was ordinary PVC. Every time the patient breathed, the cable dragged across the sheet and the trace jumped.
The fix was not a better connector or a thicker shield. It was a different jacket compound with a conductive layer underneath. Two weeks of arguing about firmware, solved by a cable material nobody had questioned.
We build custom medical cable assemblies for monitoring, infusion, handheld diagnostics, dental, and home care equipment. Most of our customers are small and mid-size device makers in the US and Europe, the kind who file their own 510(k) and need a harness supplier who understands why the paperwork matters. Here is what we have learned, including the parts of this business we deliberately do not take on.
What this guide covers:
Which connectors actually get used in small medical devices, and why
Jacket materials, cure systems, and sterilization compatibility
Microvolt signals, triboelectric noise, and the grounding mistakes that show up as artifacts
IEC 60601-1 leakage and creepage, explained without the standards-speak
Documentation and traceability: what we supply and what stays with you
A spec template for your next RFQ
Table of contents
What counts as a small medical device harness
"Small medical device" is not a technical category. It is a practical one, and it covers a lot of ground. What these products share is that the cable assembly is doing real work: it carries a signal someone is going to make a clinical decision on, or it is handled by a nurse forty times a day, or it gets wiped down with bleach until it cracks.
Where we spend most of our time:
Patient monitoring. ECG lead wires, SpO2 sensor cables, NIBP hose-and-cable assemblies, temperature probes. Short runs, tight noise budgets, frequent handling.
Infusion and enteral feeding pumps. Internal chassis harnesses plus the door and sensor wiring that flexes every time the pump cycles.
Handheld diagnostics. Glucose meters, portable ultrasound, pulse oximeters, otoscopes. Small pitch connectors, overmolded strain relief, drop-test survival.
Surgical and dental. Handpiece cables, foot switch cables, cart wiring. Repetitive autoclave exposure is usually the spec that kills the first design.
Home care. CPAP, oxygen concentrators, mobility aids. Long service life, low cost, and a user who is not an engineer and will yank the cable by the cord.
The common thread: the electrical requirements are modest, and the mechanical and material requirements are not. That is a different engineering problem than a data center harness, and it is why cheap cable assemblies fail in this market even when the continuity test passes.
Connectors used in medical equipment
Medical device designers gravitate toward push-pull and positive-lock connectors, mostly because a cable that separates mid-procedure is a clinical event, not a support ticket. Here is what we build with, and the honest trade-off on each.
| Connector family | Typical pitch / contacts | Current rating | Where it shows up |
|---|---|---|---|
| LEMO 00B / 0B / 1B | 2-14 contacts | Up to 10 A per contact | Patient monitors, ultrasound probes, surgical carts |
| ODU Medi-Snap / Mini-Snap | 2-12 contacts | Up to 8 A | Reusable sensor cables, hybrid signal + power |
| Hirose HR10 / HR25 | 4-20 contacts | 2 A | Handheld handhelds, probe connections |
| DIN 42 802 (touch-proof, 3-pin / 5-pin) | 3 or 5 contacts | 1 A | ECG trunk cables, patient cable to lead wire junction |
| 4 mm banana, safety shrouded | 1 contact | 5 A | Individual ECG lead wires, patient-side terminations |
| Molex Micro-Fit / Mini-Fit | 3.0 / 4.2 mm | 8.5-13 A | Inside the chassis: power distribution, pumps, motors |
| JST SH / ZH / GH | 1.0 / 1.5 / 1.25 mm | 1 A | Sensor boards, display modules, battery connectors |
| USB-C, Micro-USB | - | 3-5 A | Charging, data export, firmware updates |
Three notes from real projects. First, LEMO and ODU both have aggressive counterfeit and grey-market problems. If a quote is 40% below everyone else, ask which distributor the parts came from and get the invoice. We buy through authorized channels and document it.
Second, the DIN 42 802 safety connector is not glamorous and it is not going away. Half the ECG trunk cables we build still use it, thirty years after it was introduced. It is touch-proof, which is the only feature that matters when the connector sits millimeters from a patient's chest.
Third, on disposable patient cables the whole logic changes. Nobody is going to pay for a LEMO connector on a sensor that gets thrown away after three days. Those builds run on molded plugs (overmolded connector shells, two-shot or insert molding) with the contact geometry fixed by the device maker. Overmolding is its own discipline and we wrote a longer piece on it separately if that is your application.
Jacket materials and cure systems
This is where most of the design decisions happen, and where most of the failures originate. People spend weeks on the connector and five minutes on the jacket.
| Jacket material | Temp range | Flex life | Autoclavable | Honest assessment |
|---|---|---|---|---|
| Silicone, platinum-cure | -60 to +200°C | Excellent | Yes, 500+ cycles | Best for patient-contact and repeated sterilization. Expensive, poor tear resistance, needs careful strain relief. |
| Silicone, peroxide-cure | -50 to +200°C | Excellent | Yes, fewer cycles | Cheaper. Also smells, can yellow, and the by-products are a biocompatibility question. Avoid for skin contact. |
| TPU (polyether-based) | -40 to +105°C | Very good | Limited, 30-50 cycles | Toughest option per dollar. Good abrasion resistance. Not a reusable-instrument material. |
| PVC | -20 to +80°C | Fair | No | Fine inside a chassis. Plasticizer migration and DEHP questions make it a poor choice for anything touching skin. |
| FEP / PTFE | -70 to +200°C | Fair to good | Yes | Chemical resistance is unmatched, friction is low, and it is stiff. Soldering is unpleasant which raises cost. |
The thing worth arguing about is the cure system. Plenty of cable gets sold as "medical grade" because it is white and feels soft. There is no standard behind that label. What actually matters for biocompatibility and long-term stability is whether the silicone is platinum-cured or peroxide-cured, and what durometer it is.
Peroxide-cure silicone uses organic peroxides as the crosslinking agent, and the decomposition by-products (mostly volatile acids and ketones) can remain in the compound. In a well-run process they are post-cured out. In a poorly run one they are not, and you find out about it in ISO 10993 testing six weeks later. Platinum-cure has no by-products, cures cleaner, and costs more. If a supplier quotes you an autoclavable cable and cannot tell you the cure chemistry and the durometer off the top of their head, they are reselling something they have never characterized.
Sterilization compatibility
Reusable medical devices are the harder business, because the question is never "does it survive one cycle." It is "does it survive cycle 500." Here is how the methods break down against common jacket materials.
| Method | Typical conditions | Friendly materials | Watch out for |
|---|---|---|---|
| Steam autoclave | 134°C, 3 min, 2-3 bar | Silicone, FEP, PEEK, some PSU | PVC is out. TPU degrades after a few dozen cycles. Repeated cycles dry out and harden silicone over time. |
| Ethylene oxide (EtO) | 37-63°C, 4-12 h | Almost everything | Residual EtO aeration time. Porous materials and long coiled cables hold residue longer. |
| Gamma irradiation | 25-40 kGy | TPU, PE, most silicones | Discoloration is normal and usually harmless. PVC can embrittle. Some polymers crosslink and harden with cumulative dose. |
| H2O2 vapor / plasma | 45-55°C | Silicone, TPU, PVC | Lumen and crevice penetration is limited. Coiled cables may have shadowed areas. |
| Peracetic acid (liquid) | 50-56°C | FEP, PTFE, some silicone | Aggressive on many elastomers and on plated contacts. Confirm compatibility before you commit. |
Two practical points. First, connector plastic matters as much as the cable. We have seen a beautiful silicone cable come back from the field with a stress-cracked polycarbonate housing at cycle 40, because the housing was never evaluated against the disinfectant the hospital actually uses. Polycarbonate and quaternary ammonium compounds are a bad combination, and it shows up as crazing before it shows up as a crack.
Second, actual hospital reprocessing is rougher than the label on the machine. In the field, devices get wiped with whatever is on the cart: 70% IPA, 0.5% bleach, quat wipes, sometimes all three in one shift. If your instructions for use say a specific agent, someone will ignore it. Design for the worst plausible chemical and 500 cycles, not the ideal one and 20.
Microvolt signals and triboelectric noise
ECG sits around 1 mV. EEG runs in the tens of microvolts. At those levels the cable is not a passive part, it is a sensor, and it can generate its own signal whether you want it to or not.
The mechanism that catches people is triboelectric noise. When the jacket and the insulation layers rub against each other inside a flexing cable, charge separates and you see a voltage. The patient moves, the cable flexes, the trace spikes. Cheap PVC lead wires are notorious for this and firmware engineers get blamed for it regularly.
How to reduce it:
Use a conductive layer between the jacket and the shield, or a semi-conductive extruded layer over the insulation. This gives the separated charge somewhere to go besides your input amplifier.
Pick a low-friction jacket compound. Silicone is not the lowest friction material, but it is far better than a stiff PVC that grabs and releases in bursts.
Add aramid or polyester filler so the internal conductors cannot slide past each other freely. Movement inside the cable is the noise source.
Reduce the movement itself with strain relief and proper cable management at both ends.
Shielding is the next layer. For patient-connected leads we generally run 90% or higher braided coverage, with foil plus drain for lower-frequency electrostatic protection. Spiral shields give high coverage on paper but open up when the cable flexes, which is exactly when you need them.
Grounding is where designs quietly go wrong. The shield should terminate at one end only, usually the equipment end, with the patient end floating. Ground both ends and you have built a loop antenna that picks up every fluorescent ballast and electrosurgical unit in the room.
Two safety components belong in this section because they live in the same circuit. Defibrillation protection resistors, typically around 10 kΩ in series with each ECG lead, need to be energy-rated for the pulse they will actually see. Under-rated resistors become fuses during a defib event, and a fused lead wire reads as a lead-off alarm during a code. And lead-off detection depends on the DC path staying within a resistance window, so a marginal crimp or a resistor drifted out of tolerance produces alarms that look like software problems.
IEC 60601-1: leakage, creepage, and hipot
IEC 60601-1 is the standard your device will be tested against, and cable assemblies cause more failures against it than most designers expect. You do not need to memorize the tables. You do need to know which three things bit the last person.
Patient leakage current. Type BF applied parts have a patient leakage limit of 100 µA in normal condition. Type CF, which is what an ECG or an intracardiac application requires, is 10 µA. Those are small numbers, and insulation thickness, cable length, and the presence of a shield all feed into them. A cable that passes continuity with flying colors can still blow the leakage budget because it was built with single-layer insulation at a thinner wall than the spec called for.
Creepage and clearance. The required separation between conductors depends on working voltage and pollution degree. In a tiny connector or a molded plug, those distances are hard to hold, and the failure usually traces back to a soldered splice or a stripped-and-tinned wire where the insulation stops too far back. This is a drawing-review problem. It is much cheaper to catch it on paper than at a test house appointment you booked six weeks out.
Hipot. Dielectric strength testing on applied-part isolation, typically in the 1500 VAC range depending on the insulation class, run on 100% of production for some customers and sampled for others. The interesting failures are not dead shorts. They are marginal parts that passed at the factory and broke down after handling, which is why we hipot after final assembly and after any rework, never before.
If you want one sentence of advice from this section: bring your 60601-1 reviewer into the DFM stage. The cheapest place to fail a standard is a conference call. The most expensive place is a test lab, with a shipping date already sold to a distributor.
Biocompatibility and material compliance
Any material with patient contact needs biocompatibility data, and the relevant standard is ISO 10993. For a cable assembly the applicable parts are usually -5 for cytotoxicity, -10 for sensitization and irritation, and -23 for irritation testing on devices with skin contact.
Here is what we can and cannot give you, because this is a place where suppliers overpromise and device makers get burned.
What we supply: material certificates from the compound manufacturer, with the exact grade identified. For most medical-grade silicone and TPU compounds, the raw material supplier already holds ISO 10993 test reports, and we pass those through along with a declaration that we used that specific grade in your lot. We also provide RoHS and REACH declarations and, where relevant, a statement of no intentional DEHP or latex.
What stays with you: the biocompatibility conclusion for the finished device. This is not a paperwork technicality. If we use an adhesive, a primer, a heat-shrink, or a molded overmold that introduces a new chemistry, the finished assembly may need its own testing even though every input material is certified. Your notified body or FDA reviewer will want to see the linkage from the material certificate to your device file, and that linkage is your submission, not our invoice.
Two material questions worth asking early. Is the compound lot-traceable back to the manufacturer, so you can prove which grade went into which device lot. And is the compound free of animal-derived ingredients, which some device makers need for regulatory reasons and which is a surprisingly common surprise.
How we build a medical harness
The process is close to what we run on any precision harness, with the documentation load turned up. Here is how it goes from a drawing or a sample.
Step 1: DFM and standards review (free, 1-2 days). An engineer checks wire gauge against current and against the leakage budget, stripping length against the terminal spec, creepage and clearance against the working voltage, and material selection against the sterilization method you named. We also confirm the connector sourcing channel in writing. Anywhere from a quarter to a third of the drawings we receive have at least one item worth fixing, and most of them are wire gauge or material choice rather than anything exotic.
Step 2: Sample build (5-10 days). We hand-build 5-20 pieces with the exact materials and connectors specified, then run continuity, hipot, pull force, and dimensional checks. Crimp cross-section photos and a material certificate come with the samples if you want them.
Step 3: Approval and tooling setup. Once you sign off, we set up the dedicated applicator for the terminal, program the cutting and stripping equipment, and run first article inspection with the full dimensional report.
Step 4: Production. Mid-size runs (500-5,000 pcs) typically ship in 12-18 days, longer than our industrial work because of the in-process documentation. During the run we hold 2-hour crimp height checks, pull tests per lot, and 100% continuity testing.
Step 5: Final inspection and documentation package. Every piece gets a continuity test and a full visual inspection before packing. The documentation package goes out with the shipment: material certs, test data, lot traceability, and a certificate of conformity.
One process rule we do not bend on: no change without written approval. On industrial harnesses, we sometimes substitute an equivalent connector and tell the customer afterward to save them a week. On medical work that is not acceptable to anyone, including us. If a part is on allocation or a compound grade is discontinued, we tell you, and we wait. You have a device file and a submission that references a specific bill of materials, and a silent substitution puts that at risk.
Testing and QC on every batch
| Test | Method | Typical pass criteria |
|---|---|---|
| Continuity and pin mapping | Automated tester, 100% of pieces | No opens or shorts; mapping matches the golden sample |
| Hipot / dielectric strength | Per IEC 60601-1, 100% or sampled per agreement | No breakdown at the specified test voltage |
| Patient leakage current | Leakage meter, sampled per lot | Within the Type BF or Type CF limit for the applied part |
| Crimp pull force | Digital pull tester, sampled per batch | Meets the terminal datasheet value |
| Strain relief pull | Axial pull at the connector, sampled | No conductor damage or jacket retraction below spec |
| Flex / bend cycle | Test rig on design validation samples | Electrical continuity held through the specified cycle count |
| Crimp height | Micrometer, in-process every 2 hours | Within terminal tolerance, trend recorded |
| Visual inspection | 100% of pieces, IPC/WHMA-A-620 Class 2 plus medical cosmetics | No nicked strands, no flux residue, no jacket damage, no discoloration |
| Dimensional check | Calipers and pin gauges against drawing | Per drawing tolerance, typically ±2 mm on length |
| Label and barcode verification | Scan verification, 100% of packs | Lot code legible and traceable to production records |
Something worth saying about overmolded and molded-plug builds: they fail in different places than crimped assemblies. A crimped harness fails at the crimp. A molded plug fails at the transition between the hard molded shell and the flexible jacket, because that is where the bending stress concentrates. We test that junction specifically, with an axial pull plus a bend cycle, rather than relying on a connector-level pull test that never touches the weak point.
Traceability and documentation
For a device maker, the harness is a component in a regulated file. That changes what "quality" means. A perfectly good harness with sloppy paperwork is a defect.
What our records cover, per lot: the wire spool lot and compound certificate, connector lot and authorized-distributor invoice, production date and operator, crimp height trend data, pull test results, hipot and continuity results, and a certificate of conformity. We keep these for the period your quality agreement specifies, and we will not destroy them on a whim.
Two things we do that are less common. First, physical lot separation. Material from a new spool never gets mixed into an in-process batch, even when the specification is identical. If you have to investigate a field complaint, you want to be able to point at one spool of wire and one connector lot, not a bin of blended material.Second, we send a product change notification before we change anything that touches your BOM, and we wait for your written approval. That includes connector manufacturer, compound grade, and the factory thatmadethe wire. It costs us lead time and it is the single thing our medical customers value most, because a PCN they can process on their schedule is not a finding in their next audit.
Six failures we see from other suppliers
These are the complaints that come through the door, and what the fix usually costs.
1. The strain relief fails before the wire does. The cable is yanked by the cord, the jacket retracts, and the conductors break inside where nobody can see it. Fix: proper crimp ferrule or overmold at the transition, plus a bend cycle test to at least 10,000 cycles on design validation samples.
2. "Medical grade" silicone that is peroxide-cured. It passes a visual inspection, then fails sensitization testing, or it yellows and smells after a few months in the field. Fix: specify platinum-cure in writing, and get the compound certificate with the grade name on it.
3. Cable dies at autoclave cycle 80. The jacket stiffens, the connector housing crazes, or the printed label turns to a grey smear. Fix: run a 500-cycle validation on the actual assembly, with the actual disinfectant, before you freeze the design. It is a slow test and it is much cheaper than a field correction.
4. Triboelectric noise on a moving lead. The trace looks perfect on a stationary patient and noisy on a walking one. Fix: semi-conductive layer, low-friction jacket, filler to restrict internal movement, and single-point shield termination.
5. Leakage current fails at certification. Usually it is insulation wall thickness, a soldered splice with too little creepage, or a shield terminated on both ends. Fix: review against 60601-1 at the drawing stage, not at the test house.
6. Reverse pin mapping on a hand-assembled small connector. The classic. A 1.0 mm pitch connector looks symmetric under a bench lamp. Fix: 100% automated continuity testing against a golden sample, never a visual check alone, and polarized housings wherever the series offers them.
A spec template for your RFQ
Sending this once saves two rounds of email and usually a week of lead time.
| Parameter | What to specify | Example |
|---|---|---|
| Function | What the assembly does and which device it goes into | ECG trunk cable, reusable, 3-lead |
| Connector end 1 | Series, contact count, housing and contact part numbers | LEMO 1B, 6 contacts, or overmolded plug to your drawing |
| Connector end 2 | Same, or the patient-side termination | DIN 42 802 5-pin, touch-proof |
| Applied part classification | Type B, BF, or CF | Type CF |
| Jacket | Material, cure system if silicone, durometer, color | Platinum-cure silicone, 60 Shore A, gray |
| Conductors | AWG, strand count, insulation material, color code | 26 AWG, 19 strands, 3 conductors plus drain |
| Shielding | Type, coverage, where it terminates | 95% braid, equipment end only |
| Sterilization | Method and required cycle count | Steam autoclave, 500 cycles |
| Biocompatibility | ISO 10993 parts required and contact duration | -5, -10, -23, prolonged skin contact |
| Electrical limits | Leakage target, hipot voltage, max resistance per conductor | 10 µA patient leakage, 1500 VAC hipot |
| Length and tolerance | Overall length and which points it is measured between | 1200 mm ±10 mm, end to end |
| Flex and handling | Static or moving, cycle requirement, drop test if handheld | Handled continuously, 100,000 bend cycles at the plug |
| Documentation | Which records you need with the shipment | Material certs, test data, lot traceability, CoC |
| Quantity | Prototype, first production lot, annual volume | 20 samples, 2,000 first lot, 15,000/yr |
| Target price | Optional, and it genuinely saves time | Under $11/unit at 2,000 |
If you have only a sample and no drawing, that is workable. We identify the connector family, measure the jacket and conductor construction, test the sample electrically, and send you a dimensioned drawing with a material recommendation before we build anything. Be aware that reverse engineering a cable cannot tell you the cure system or the compound grade, so we will propose a specific medical-grade material and let your regulatory team decide whether it is acceptable.
What we take on, and what we do not
This part matters more in medical work than in any other market we serve, because the cost of a mismatch lands on both of us.
What we are good at: custom cable and harness assemblies for Class I and Class II devices, made from materials your team has already qualified or is willing to qualify. Tight tolerances, small pitch connectors, molding and overmolding, microvolt-signal integrity, autoclave-rated materials, and the documentation package that goes with all of it. Mid-volume production, generally 500 to 50,000 pieces a year per part number.
What we do not do, and will tell you so on the first call: implantables, anything requiring a cleanroom-packaged sterile barrier that has to survive shelf life, finished-device assembly or sterilization, and device-level regulatory submissions. If your project needs a contract manufacturer with an ISO 13485 certified quality system and a cleanroom, we are the wrong supplier, and we would rather lose the quote than appear in your audit findings.
On ISO 13485 specifically, since every medical RFQ asks: ask us for our current certificate status in writing, and ask the same question of any supplier you are evaluating. A harness factory that is ISO 9001 certified but describes itself as "ISO 13485 compliant" is telling you something. Read the scope statement on the certificate, not the logo on the website.
Request a free DFM and standards review
Send us your drawing, a sample photo, or a sample itself. We will confirm the connector family, flag leakage, creepage, and material risks before you commit to tooling, and send a quotation with sample lead time. If your project is outside what we do well, we will say so within 48 hours.
Email: [sales5@xmkehan.com]
Frequently asked questions
Q1: Can you build a custom medical device wire harness in small quantities for clinical trials?
Yes, and a good share of our medical work starts there. We build 5-50 pieces for design verification and clinical builds, with the same material documentation as a production lot. Clinical trial quantities can be priced as a separate line so the cost does not distort your production quotation.
Q2: Are you ISO 13485 certified?
Ask us for the current certificate and read the scope statement, which is the honest answer to this question for any supplier. Separately, we will tell you plainly which parts of a medical build we handle well and which need a contract manufacturer with a cleanroom and a device-level quality system. We would rather decline a program than create a supplier finding for you.
Q3: Silicone or TPU for a reusable sensor cable?
If it gets autoclaved repeatedly, silicone, and specify platinum-cure. If it is wiped down rather than steamed, or if abrasion is the main threat, TPU performs well at a lower cost. The decision usually comes down to sterilization method and cycle count, not to the cable itself.
Q4: How do you handle the biocompatibility documentation?
We supply material certificates from the compound manufacturer identifying the exact grade, along with the supplier's ISO 10993 reports for that grade where they exist. The biocompatibility conclusion for the finished device stays with you, and it should. If our process introduces a new chemistry through adhesive, primer, or overmold, that needs its own evaluation, and we will flag it before you commit to a design.
Q5: What is your lead time for medical assemblies?
Samples run 5-10 days. Production lots of 500-5,000 pieces typically ship in 12-18 days after sample approval, which is longer than our industrial lead times because of the in-process documentation and additional testing. Larger programs are scheduled and confirmed in writing before you commit.
Q6: Can you match a competitor's cable exactly so we can avoid re-qualifying?
We can match the mechanical and electrical construction precisely. What we cannot reliably determine from a sample is the compound grade and cure system. We will propose a specific medical-grade material with certificates and let your regulatory team judge whether it is equivalent. If the answer is no, you have saved yourself a deviation instead of discovering it in an audit.
Q7: Do you support RoHS, REACH, and latex-free requirements?
Yes. We provide RoHS and REACH declarations as standard, and statements covering latex and intentional DEHP on request. Animal-derived ingredient declarations are also available, since some device makers need that for regulatory reasons and it is not something you want to discover late in a submission.