When I talk about pack intrusion, I mean any unwanted liquid that finds its way into or around the battery pack: rainwater, coolant, saltwater, car-wash spray, muddy water or condensate. The real concern is not the liquid itself, but the electrical and safety consequences when it creates creepage paths, leakage currents or local hotspots.

For engineering and safety work, I split water ingress into three practical levels: thin condensation film, localized puddles at low points or joints, and full flooding or submersion after deep water events. Each level drives a different response in the BMS and powertrain safety concept.

Typical field scenarios I keep in mind:

• Vehicle crossing deep puddles or flooded streets, with the pack partially submerged.
• High-pressure car wash jets directed at the pack bottom and HV connectors.
• Impact or curb hit that cracks the pack housing and lets water or coolant seep in.
• Long-term parking in humid environments causing slow condensation inside the pack.

I also map these scenarios onto risk and safety expectations:

Safety standards such as ISO 26262, ISO 6469, UL 2580 and IP ratings from IEC 60529 all shape how I justify water ingress sensing. IP67 or IP69K protection alone only claims resistance to water entry under specified conditions; it does not tell me whether the pack has actually stayed dry in real life. Intrusion sensing is my way to observe what really happened over the lifetime.

In my own architecture, I treat water ingress sensing as the function that sees the liquid itself. Residual or leakage detection sees the electrical consequence of degraded insulation. Pack fire and off-gassing sensing see the thermal and chemical consequences. HVIL monitoring sees covers and service plugs. Keeping these roles separate avoids overlap between pages and makes the safety case easier to explain.

My goal with water ingress sensing is simple: convert the physical presence of liquid into an electrical signal that is measurable, diagnosable and safe. Instead of asking “is it wet or dry?”, I ask “does the impedance between the sensing electrodes sit in a range that I can detect, test and classify over the vehicle lifetime?”.

Under dry conditions, the gap between electrodes mainly contains air and plastic, so the impedance is very high and only a tiny leakage or parasitic capacitance is visible to the front end. Once water, coolant or a contaminated liquid bridges the gap, the equivalent resistance drops and the capacitance profile changes. The AFE monitors this shift using excitation, filtering and either ADC measurements or comparators with thresholds.

Different liquids behave differently, and I keep that in mind when planning ranges:

In practice I think in three sensing “styles”:

• Resistive sensing – using a divider or constant-current drive and reading voltage. It is simple and intuitive, but long-term corrosion and polarization can shift the apparent resistance and hide real ingress.
• Capacitive / touch-like sensing – relying on changes in capacitance and field distribution when liquid appears near the electrodes. It is useful when the sensor sits behind an insulating layer, but it is more sensitive to EMC and layout details.
• Multi-frequency impedance probing – exciting the sensor at two or more frequencies to distinguish different liquids or to separate resistive and capacitive components. This is more complex, but attractive for high-end platforms.

High-impedance instrumentation front-ends and low-leakage operational amplifiers are often used to sense the tiny currents and voltages without loading the electrodes. On top of the measurement itself, I also care about the failure modes: open or shorted sensor lines, slow drift from corrosion, or an AFE that saturates silently. A simple wet/dry switch cannot tell these cases apart or support periodic self-test.

All of this sets up the requirements for the actual impedance AFE: input range, measurable impedance window, excitation frequency and amplitude, input protection and diagnostics. The next section turns these principles into a concrete signal chain with comparators, ADCs and safety MCU interfaces.

Once I know how water or coolant changes the impedance between electrodes, I need a signal chain that can reliably excite, sense and classify those changes. In practice my architecture starts at the sensing electrodes, injects a small excitation, measures the resulting voltage or current through an impedance front-end, filters the signal and finally feeds ADCs or comparators that report to a safety MCU or BMS.

A typical path looks like this: electrodes and wiring harness, a low-voltage AC or DC excitation driver, an impedance front-end built around operational amplifiers or an instrumentation front-end, an analogue filter or rectifier, and then either an ADC channel or a comparator bank. From there, the safety MCU applies thresholds, timing and diagnostics, before forwarding fault information through a galvanic isolation stage to the main vehicle network.

I usually consider three implementation styles for the AFE:

• Simple divider + comparator – a resistive divider with the sensor in series or parallel and a comparator watching the node. This is attractive on low-cost platforms, but has limited diagnostic coverage and little tolerance to drift.
• AC excitation + synchronous detection – a switching driver and an op amp front-end that measure the sensor response at a chosen frequency, often with demodulation or rectification. This is much better against electrode polarization and long-term corrosion.
• Multi-channel impedance AFE – a dedicated front-end that can excite and read several sensing points, sharing the ADC and diagnostics. This is useful for large packs with multiple low points and connector areas.

When I pick or design an impedance AFE, I work through a short checklist:

• Measurement capability – input voltage and current range, the impedance window I need to distinguish (for example from a few kΩ down to tens of kΩ, up to dry-state MΩ), excitation voltage and frequency, noise floor and latency.
• Input protection and robustness – ESD and surge withstand on the electrode lines, tolerance of reverse polarity, short-to-battery and short-to-ground behaviour, and how gracefully the front-end recovers from these events.
• Diagnostics – ability to detect open sensors, shorted sensors, stuck-at-wet or stuck-at-dry conditions, AFE saturation, reference failures and any abnormal readings that indicate corrosion or wiring damage.
• System interfaces – number of channels, ADC interface options, comparator outputs for fast actions, GPIOs for self-test injection, and isolation strategy towards the main pack or vehicle controller.

Instrumentation front-ends and low-leakage operational amplifiers are at the heart of this chain. They let me observe small impedance changes over long harnesses without loading the sensor, while still supporting the self-test and diagnostic modes that functional safety requires.

Even the best impedance AFE will not deliver much value if the sensing electrodes are in the wrong place. When I plan pack intrusion sensing, I start by mapping likely ingress paths and low points: where water or coolant is most likely to collect after a deep puddle, a car wash or a housing crack. Those locations drive how many sensing points I need and how I route the harness.

Typical placement rules I use:

• Add sensing points near the lowest regions of the pack floor where liquid can pool and stay, rather than locations that only see brief splashes.
• Watch joints, seams and fasteners where the housing is more likely to open a leak path after impact or long-term vibration.
• Place sensors close to HV connectors and service plugs, where water ingress can quickly lead to creepage along insulation surfaces.
• Consider areas around cooling plates and channels where coolant leaks would combine electrical and chemical risks.

Mechanical and electrical trade-offs come together in this step:

On the harness side, I treat intrusion sensing as a safety-relevant network: dedicated connector positions, clear pin assignment, protected routing and strain relief. The sensing harness should not be an afterthought piggybacked onto HV cabling. At the same time, I avoid turning this page into a full HVIL or thermal layout guide: temperature sensors, gas sensors and lid switches are covered in their own topics.

In short, I let the physics of where liquid actually goes drive the sensor layout, and I make sure that mechanical sealing, electrical safety and EMC constraints are designed together instead of fighting each other.

Detecting water is only half of the job. I also need to know that the sensing chain itself is alive and trustworthy. A water ingress function that silently dies in the field is worse than not having one at all, because the rest of the safety case still assumes the signal is available.

My first step is to design explicit self-test modes. At the simplest level, I use pull-up or pull-down networks and comparators to detect open and short circuits on the sensor lines. If the measured voltage sits in an impossible region, I know the harness or electrodes are damaged. On higher-end platforms, I periodically inject a known excitation into the impedance front-end and check that the measured response lands in a narrow, expected window. This proves that the instrumentation front-end, operational amplifiers, ADC and reference rails are still behaving.

In my diagnostic planning I typically cover three classes of checks:

• Structural checks – detect open or shorted sensor lines, swapped pins, stuck-at-dry and stuck-at-wet behaviours using pull-up / pull-down networks and comparator thresholds.
• Functional checks – periodic known-signal injection and comparison against an expected impedance window, often referenced to a “dry” baseline channel.
• Plausibility checks – cross-checks against other measurements such as IMD, residual leakage, temperature and vehicle state to decide whether a reading is physically plausible.

I also translate raw readings into graded alarm levels:

To avoid nuisance behaviour, I add hysteresis and debounce in both the analogue and digital domains. True ingress events evolve over seconds or minutes, while most noise is much faster. I take advantage of that difference: minimum-duration timers, filtered averages and threshold bands help avoid toggling between wet and dry states or flapping between alarm levels.

Finally, I define a reporting path that the software and diagnostics teams can implement: sensor and AFE readings enter the safety MCU, which classifies the state and raises DTCs, then passes fault flags through a galvanic isolation stage onto the HV or chassis CAN bus. From there, the BMS, body controller or gateway can show HMI messages and apply power limits. Treating water ingress as a safety-related signal means I also apply ISO 26262 data integrity and supervision measures to this path.

From a functional safety perspective, pack intrusion sensing is one of the earliest indicators that something has gone wrong at the housing or cooling level. It does not replace insulation monitoring, residual leakage detection or off-gassing sensors, but it gives me a physical trigger before many of the electrical effects show up. I treat it as a supporting input for safety goals that protect against electric shock and uncontrolled thermal events.

In a typical safety concept, one safety goal ensures that the driver and service personnel are not exposed to dangerous HV potentials, even if the pack is damaged or flooded. Another goal ensures that thermal runaway is either prevented or contained. Water ingress sensing can contribute to both by requesting HV isolation actions and by helping classify the root cause behind insulation or thermal alarms.

I keep the roles of different safety mechanisms clearly separated:

I then look at failure modes from a functional safety point of view:

• Sensor stuck dry – a fault causes the system to report “no ingress” even when liquid is present. This is mitigated by structural and functional checks, and by treating implausible combinations with IMD or leakage alarms as a fault.
• Sensor stuck wet – a short or corrosion makes the system believe there is always ingress. Here I rely on cross-checks, hysteresis, reference channels and the ability to downgrade the signal to avoid permanent Level 3 actions.
• Harness faults – cut, pinched or flooded harness segments can create misleading readings. Pull-up / pull-down schemes, line supervision and connector sealing are part of the mitigation.
• AFE saturation or rail sticking – the impedance front-end or comparator rail-to-rail behaviour hides real changes. Watchdog ranges and periodic self-test help detect this.

On the action side, I design water ingress as an input into a wider safety decision. For example, Level 1 ingress may only log DTCs. Level 2 ingress combined with moderate insulation degradation might trigger BMS power limits and driver warnings. Level 3 ingress combined with strong IMD or leakage alarms, or with off-gassing signals, is a clear “must stop driving” case where contactors are opened and HV re-enable is blocked until the pack is inspected.

Water ingress sensing is not the only barrier in the safety concept, but without it many events would only be visible once they have already become electrical or thermal faults. I treat this page as the place where the physical reality of liquid meets the structured world of safety goals, ASIL decomposition and coordinated HV shutdown.

Once the sensing concept is clear, I map it onto real IC families. I am not looking for a generic “water level switch”, but for automotive-grade building blocks that can implement excitation, impedance measurement, diagnostics and safe reporting. That usually means combining an impedance front-end, comparators, a safety-capable MCU or system basis chip, and isolation devices if the signals cross HV domains.

I think of the signal chain in terms of IC categories rather than single part numbers: excitation drivers for the electrodes, instrumentation or operational amplifiers for the impedance front-end, comparator banks for fast thresholds, multi-channel touch or impedance controllers when I have many sensing points, safety MCUs or system basis chips with watchdog and CAN/LIN, and digital isolators or isolated ADCs when the ingress sensing lives on the pack side of an isolation barrier.

At a high level, I distribute roles across the main automotive vendors like this:

My selection flow usually starts from reference designs and application notes rather than product catalogs: liquid level sensing, capacitive touch in harsh environments, fault-tolerant sensing harnesses and automotive safety MCUs. Even if the example application is a tank or a touch key, the same AFE and comparator techniques often transfer to pack intrusion sensing with only a different electrode layout.

I also draw a clear line to keep this page focused. Dedicated IMD ICs, HVIL controllers and residual current devices belong in their own topics. Here I only map building blocks that directly implement water ingress sensing or its immediate diagnostics and reporting.

When I talk to suppliers or build a BOM for pack intrusion sensing, I translate the concept into concrete fields that can go into a request for quotation. The goal is to make it clear that I am not asking for a simple “humidity sensor”, but for an automotive-capable solution that supports impedance sensing, diagnostics and the safety case.

My checklist usually includes:

I also call out a few common traps for myself and for procurement:

• Buying a generic humidity or water level sensor without an automotive impedance AFE and diagnostics is not enough for a safety-relevant EV pack.
• Choosing devices without self-test hooks or diagnostic outputs makes it very hard to close the functional safety loop and prove adequate diagnostic coverage.
• Ignoring liquid type, temperature and ageing effects in the specification can result in thresholds that work in the lab but drift badly in the field.
• Leaving interfaces vague (“some digital output”) creates integration friction for the BMS and body electronics teams.

When I send out a request for quotation, I make it explicit that this is an EV pack water ingress monitoring function with safety implications, not a generic industrial level switch. I ask vendors for any EV-oriented reference designs, application notes and, where available, FMEA examples that cover both the AFE and the sensing harness. The answers I get back often shape my final IC choices and wiring strategy.

Recommended topics you might also need

• Pack Fire / Off-gassing Sense
• HVIL Monitor
• Insulation Monitoring Device (IMD)
• Battery Junction Box (BJB)

In this FAQ I summarise how I plan, implement and justify pack water ingress sensing in a real EV project, from deciding whether I really need it, through impedance and placement choices, to diagnostics, safety arguments and the way I talk to suppliers when I send out a request for quotation.