Cool-Plate Temperature Hub for EV Battery Cooling
← Back to: High-Voltage Energy & Safety
This page walks you through everything you need to build a cool-plate temperature hub: when a hub is justified, how to choose NTC vs RTD, design the AFE, reference, ADC and isolation, add safety diagnostics and finally turn all of that into concrete IC choices and RFQ fields.
What Is a Cool-Plate Temperature Hub?
When you start adding more temperature points to your EV battery pack or power electronics cool plate, the harness quickly turns into a tangle of wires. Each NTC or RTD looks small on the drawing, but every sensor needs routing, protection and diagnostics back to the ECU.
A cool-plate temperature hub is a local measurement node that gathers multiple NTC or RTD channels—typically 8, 12 or 16—onto a single board. It combines sensor AFEs, low-drift references, multi-channel ADCs and isolation so that the upstream cool plate only exposes short sensor stubs while the downstream BMS or thermal ECU receives clean, digital temperature data.
In a typical EV, the hub sits close to the cool plate, module or pack enclosure on the high-voltage side, and reports into a thermal ECU, BMS or body controller. One vehicle may use several hubs so each cool plate or thermal zone can be monitored and diagnosed as a block instead of as individual loose sensors.
You typically consider a cool-plate temperature hub when:
- You have more than 4–6 temperature points on a cool plate or module and direct harness routing to the ECU is no longer practical.
- The sensors sit in a noisy or high-voltage environment and you want an isolated measurement boundary near the cool plate.
- Functional safety or OBD requires consistent open/short diagnostics and self-test across all temperature points.
- You want a single thermal map per zone instead of scattered ADC channels inside the central ECU.
- You plan to reuse the same thermal hub across multiple vehicle programs or pack variants with different channel counts.
System Architecture for Multi-Channel Sensing
Once you decide to use a cool-plate temperature hub, the next step is to define the signal chain. At a high level it always follows the same flow: NTC or RTD sensors feed a front-end AFE, which relies on a stable reference, drives a multi-channel ADC, crosses an isolation barrier and finally lands in a BMS or thermal ECU.
In small packs, this entire chain may live on a single board. In larger battery systems or vehicles with multiple cool plates, several hubs are deployed in parallel, each monitoring a thermal zone and forwarding digitized temperatures over a short, isolated interface to the central controller.
Choosing Between NTC and RTD on Cool Plates
Many teams treat temperature sensors almost like a belief system: some “always use NTCs”, others insist that “a proper EV must use RTDs”. On a cool plate, that kind of brand loyalty is less useful than a clear look at your real constraints: temperature range, accuracy, long-term drift, mechanical integration and harness length.
A cool-plate temperature hub usually monitors a mix of “ordinary thermal housekeeping” points and a few truly critical locations near high-voltage busbars or power modules. In practice this means the right answer is rarely “all NTC” or “all RTD”, but a combination chosen by zone: NTCs where cost and density matter most, RTDs where stability and traceability dominate.
How NTC and RTD compare for cool-plate sensing:
| Criteria | NTC | RTD |
|---|---|---|
| Cost | Very low sensor cost, simple divider AFE. | Higher device and AFE cost per channel. |
| Accuracy | Adequate for most cool-plate ranges when calibrated. | Higher absolute accuracy and easier linearisation. |
| Drift / ageing | More sensitive to tolerance spread and long-term drift. | Better long-term stability for 10+ year lifetimes. |
| Temperature range | Comfortable for typical EV coolant and pack ranges. | Handles wider ranges with predictable behaviour. |
| Wiring | Two-wire is simple but sensitive to long harness resistance. | 3/4-wire options compensate line resistance on long runs. |
| Mechanical integration | Wide choice of probe styles that fit into cool-plate pockets and surfaces. | Often packaged as more “instrument-grade” sensors with higher unit cost. |
Practical selection rules for cool-plate hubs:
- If the cool plate only needs general thermal housekeeping, harness runs are short and cost pressure is high, a pure NTC array is usually sufficient.
- If specific locations sit near HV busbars or power modules and must remain within tight limits over 10+ years, use RTDs at those critical points.
- If harnesses are long or pass through noisy, high-current areas, RTDs with 3-wire or 4-wire connections make line resistance easier to control.
- A common strategy is to use one or two “golden RTDs” per cool plate for reference and a surrounding ring of NTCs for thermal density.
AFE Design for NTC/RTD Arrays
Selecting NTCs or RTDs is only the first step. The real challenge in a cool-plate temperature hub is turning eight, twelve or more sensors into clean, repeatable signals for a shared ADC. As channel count rises, the analogue front end must deal with reference loading, line resistance, filtering and EMC without exploding the BOM or PCB area.
In practice, most cool-plate hubs fall into three AFE families: simple divider networks for dense NTC arrays, constant-current RTD front ends for long harnesses and safety-critical points, and multiplexed differential front ends that can mix sensor types and handle more demanding environments.
Typical AFE families for cool-plate arrays:
- Divider-based NTC AFE – multiple NTCs share a reference rail in simple resistor dividers. BOM and PCB footprint are low, but line resistance and reference drift affect all channels.
- Constant-current RTD AFE – precision current sources drive RTDs with 3-wire or 4-wire connections. It scales well to long harnesses and demanding accuracy, at the cost of higher complexity.
- Multiplexed differential front end – a mux and differential amplifier feed a common ADC input, enabling mixed NTC/RTD arrays and better control of common-mode range and noise.
Matching AFE topology to channel count and wiring:
| Channel / scenario | Preferred AFE topology | Typical use case |
|---|---|---|
| ≤ 4 NTC, short harness | Simple divider-based AFE | Small module, ECU close to the cool plate. |
| 8–12 NTC, mid-length harness | Divider AFE with buffered reference and mux | Medium pack cool plates with moderate noise. |
| 4–8 RTD, long harness, safety-critical | Constant-current RTD AFE with 3/4-wire | HV bus and power-module plates with tight limits. |
| Mixed NTC + RTD, long harness | Multiplexed differential front end | Combined pack and inverter cool-plate zones. |
Common AFE pitfalls to watch for:
- Overloading a shared reference rail with too many NTC dividers, slowing the response and amplifying drift across all channels.
- Ignoring line and contact resistance in long harnesses, which silently shifts measured temperature over time.
- Choosing mux and filter time constants without considering settling time, so previous channels bleed into the next measurement.
- Routing sensor returns without a clear reference strategy, inviting EMC issues in high dv/dt, high di/dt environments.
Low-Drift References and Accuracy Budget
It is tempting to treat cool-plate temperature accuracy as a sensor problem, but the reference and ADC form the foundation for every channel. If the reference drifts or the ADC has significant gain and linearity error, the entire thermal map shifts together and the cooling strategy is tuned to the wrong absolute temperature.
A realistic accuracy budget for a cool-plate temperature hub needs to separate out the main contributors. Instead of asking whether the NTC or RTD itself is “good enough”, you define a total error target and then allocate slices of that budget to the reference, ADC, sensor tolerance and wiring so that no single block silently dominates the system.
Typical error sources in a cool-plate temperature hub:
- Reference drift and tolerance — initial error and ppm/°C drift of the voltage reference that shifts all channels together.
- ADC gain, offset and INL/DNL — how closely the digital code tracks the true input voltage across the full range.
- Sensor tolerance and ageing — NTC or RTD curve accuracy, lot-to-lot spread and long-term drift in the harsh thermal environment.
- Wiring and contact resistance — long harnesses, connectors and contact quality that perturb the effective resistance seen by the AFE.
- Noise and filtering choices — bandwidth and averaging settings that trade update rate against measurement jitter.
Example accuracy budget for a ±1 °C cool-plate target:
| Error source | Budget share (°C-equivalent) |
|---|---|
| Reference drift and tolerance | ≈ 0.3 °C |
| ADC gain / INL / quantisation | ≈ 0.3 °C |
| Sensor tolerance and ageing | ≈ 0.3 °C |
| Wiring and contact resistance | ≈ 0.2 °C |
| Combined total (RSS or worst-case sum) | ≈ 1.0 °C |
This example is illustrative rather than prescriptive. The key idea is to make the reference and ADC visible entries in the budget instead of leaving them as unquantified “infrastructure” error.
Isolated ADCs and Digital Interfaces
A cool-plate temperature hub usually lives close to the battery pack or power electronics, on the high-voltage side of the system. To keep the BMS or thermal ECU in the low-voltage logic domain, you must draw a clear isolation boundary somewhere along the measurement path, either inside the ADC or between the ADC and the controller interface.
Temperature itself is a slow signal, but the isolation choice is not only about bandwidth. It needs to withstand continuous working voltage, transient events and EMC stress for the full vehicle lifetime. That is why many cool-plate hubs use isolated ΣΔ ADCs purpose-built for precision measurement on the high-voltage side, or pair a local ADC with a digital isolator and a short SPI or I²C link into the ECU.
Key parameters when selecting an isolated ADC for cool-plate hubs:
- Resolution and noise-free code — cool-plate temperature needs usable resolution at the degree or sub-degree level, so effective bits and noise-free code width matter more than headline resolution.
- Conversion rate and latency (ΣΔ vs SAR) — ΣΔ ADCs bring high resolution and filtering with modest latency, which is acceptable for thermal signals, while SAR ADCs suit simpler, faster channels.
- Isolation rating and lifetime — working voltage, surge capability and insulation rating must align with the pack design and survive the full thermal and electrical stress profile of the vehicle.
Integrated isolated ADC vs discrete ADC plus isolator:
| Approach | Strengths | Trade-offs | Typical use case |
|---|---|---|---|
| Integrated isolated ADC | Single device handles measurement and isolation, simplifying BOM and layout. | Less flexibility in ADC family choice and scaling to unusual channel counts. | Platform BMS designs and reusable cool-plate hub modules. |
| Discrete ADC + digital isolator | Freedom to pick ADC resolution, interface and channel count independently of the isolator. | More PCB area and EMC work, plus additional components and interfaces to validate. | Custom packs, mixed sensing tasks or reuse of existing ADC platforms. |
Whichever option you choose, the digital side typically speaks SPI, I²C or a daisy-chain link into the BMS or thermal ECU, so interface planning should be aligned with the overall pack communication architecture.
Calibration, Diagnostics and Functional Safety Hooks
A cool-plate temperature hub is not just a collection of ADC channels. It becomes part of the safety chain: it informs when to derate, when to protect the pack and when to log faults. That means calibration, fault detection and redundancy need to be designed in at the hub level instead of being left to “software corrections” later.
When you plan a safety-relevant hub for an EV pack, it helps to treat each feature as an explicit hook into the functional safety concept: how it is calibrated, how sensor faults are caught, how redundant channels are cross-checked and how the hub reports its own health to the BMS or thermal ECU.
When designing a cool-plate temperature hub, check whether you have covered these hooks:
- Calibration strategy (factory vs in-field) — is there a clear plan for factory calibration of Vref, ADC and a subset of channels, and can the system apply in-field offset or gain trims if needed?
- Sensor open / short detection — can the AFE and ADC distinguish between normal temperature extremes and true open-circuit or short-to-rail conditions on each channel?
- Redundant channels at critical locations — do HV busbars, power-module interfaces or other safety-critical spots use dual NTCs, RTDs or mixed pairs so that a single sensor failure does not leave the system blind?
- Self-test and plausibility checks — does the hub support internal self-test of its ADC and references, and does the BMS apply plausibility rules such as maximum gradient or minimum rate of change over time?
- Diagnostic reporting to BMS / thermal ECU — are fault flags, status bits and channel-level error information exposed in a way that can be mapped into DTCs and OBD freeze frames?
Typical fault modes and how a hub can detect them:
| Fault mode | Detection method | IC / system feature used |
|---|---|---|
| Sensor open circuit | Measured input voltage rails out of the valid temperature range for that sensor type. | AFE open-detect comparators, ADC window limits and threshold-based diagnostics. |
| Sensor short to ground or rail | Input clamped near Vref or GND, independent of expected cool-plate behaviour. | Threshold comparators and plausibility checks across multiple channels. |
| Stuck or frozen sensor | Temperature reading does not move over time while neighbouring sensors change as expected. | Rate-of-change and gradient checks in the BMS or thermal ECU. |
| ADC / reference failure | Self-test modes produce incorrect codes or reference-switched measurements disagree. | ADC self-test routines, internal reference multiplexing, built-in test patterns. |
| Redundant channel mismatch | Dual sensors at a critical point disagree beyond a configured tolerance window for a sustained period. | Cross-check logic in the hub MCU or BMS, with channel status flags and error counters. |
This list does not replace a full safety analysis, but it forces each fault mode to be mapped to a specific detection mechanism and IC feature instead of being left implicit.
Thermal Management Use Cases and Placement Patterns
A cool-plate temperature hub only makes sense when you see where it lives in a real pack. Different EV architectures place hubs in different ways: a single hub watching the main pack cold plate, several hubs dividing a large pack into zones, or shared hubs that also monitor inverter or on-board charger cold plates.
Thinking in terms of coverage area, harness runs and module boundaries helps you choose how many hubs you need and where to mount them, instead of starting from an arbitrary channel count.
BEV Main Pack Cool Plate — single hub per pack region
In many battery-electric vehicles, a main traction pack uses one or more large cold plates under the cell stack. A temperature hub mounted on a serviceable electronics island can collect all NTC or RTD channels for a given region with short harness runs.
- Coverage vs channel count — one hub typically covers a sub-area of the pack rather than the entire pack, balancing thermal resolution and harness complexity.
- Harness routing — sensor wires should follow structural members and coolant lines, avoiding long parallel runs with HV busbars where possible.
- Serviceability — hub placement should allow replacement without disturbing the pack structure, while keeping sensor connectors accessible and protected.
Shared Hub for Pack and Power Electronics Cold Plates
Some platforms share a cooling loop between the traction pack and power electronics such as the inverter or on-board charger. In these cases a single hub can sometimes monitor both the pack cold plate and a nearby power electronics plate.
- Domain separation — even when signals come from one hub, channels should be tagged so software can distinguish pack temperatures from inverter or OBC temperatures.
- Thermal latency — shared coolant loops mean different plates may heat and cool at different rates, so sampling strategies and logging should preserve timestamps and channel identities.
- Error containment — if a shared hub fails, you need a clear view of which parts of the thermal system are affected and whether additional redundancy is required for critical zones.
Modular or Split Packs with Multiple Hubs
Large packs, packs split across front and rear axles, or modular designs often benefit from a hub-per-module pattern. Each module or cold plate has its own hub, and the BMS aggregates all hub outputs into a global view of pack temperature.
- Hub-per-module pattern — assigning a dedicated hub to each module simplifies mapping between module IDs and temperature channels, and eases field replacement.
- Daisy-chain or multi-drop interface — multiple hubs can be connected via daisy-chain SPI, multi-drop links or several short buses feeding a central BMS controller.
- Scaling and reuse — the same hub hardware can be reused across small and large packs, simply changing how many modules and hubs are populated in the system.
IC & Reference Selection Map Across Major Vendors
This section maps the main signal-chain functions of a cool-plate temperature hub—sensor AFEs, references, ADCs, isolators and temperature monitor ICs—to example families from the major automotive IC vendors. The goal is not to list every part number, but to highlight which product lines are typically used to build 8, 12 or 16 channel hubs and how they fit into the system.
Use this as a starting point when you prepare a short list for each project. Once you know which vendor families match your topology and accuracy needs, you can drop in the exact part numbers and reference designs from their documentation or application notes.
AFE and ADC options for multi-channel temperature hubs:
| Vendor | Typical AFE / ADC role | Why it fits cool-plate hubs |
|---|---|---|
| TI | Precision ΣΔ ADCs with built-in mux and RTD/NTC-friendly front ends. | Ready-made reference designs for 8–16 channels, strong support for isolated measurement in BMS and thermal systems. |
| ST | ΣΔ and SAR ADCs integrated in automotive MCUs or as standalone converters. | Good fit when you want to host the hub on a small MCU with on-chip ADC and external NTC/RTD AFEs. |
| NXP | Automotive MCUs with multi-channel ADCs and safety features. | Useful for hub designs that integrate local diagnostics and safety checks directly into the controller. |
| Renesas | Low-noise ΣΔ ADCs and current/voltage AFEs used in BMS and thermal sensing. | Good choice for high-channel-count hubs that reuse BMS-style AFEs and ΣΔ ADC chains. |
| onsemi | Mixed-signal front ends and ADCs used in automotive sensing and motor control. | Attractive when the cool-plate hub is co-located with pumps, valves or e-compressors that already use onsemi analog front ends. |
| Microchip | MCUs with ADCs and external ADC families for precision sensing. | Helpful for flexible, cost-sensitive hubs where you want to mix low-cost MCUs with higher-grade external ADCs. |
| Melexis | Sensor-oriented signal-conditioning ICs and ADCs for automotive environments. | Well-suited to hub designs that emphasise robust sensing in harsh thermal and EMC conditions around the pack. |
For each project, you can narrow this list to one or two vendors that match your preferred ADC topology and safety concept, then drill down into their 8/12/16-channel reference designs.
Reference, temperature monitor and isolation building blocks:
| Vendor | Low-drift reference / monitor role | Isolation / interface role |
|---|---|---|
| TI | Low-drift voltage references and multi-channel temp monitors that anchor the hub accuracy budget. | Isolated ΣΔ modulators and digital isolators for SPI / I²C links into the BMS. |
| ST | Voltage references and supervisor ICs used to stabilise ADC rails and monitor hub supply health. | Digital isolators and high-side interfaces for hubs close to HV domains. |
| NXP | Integrated monitors in MCUs and PMICs to supervise temperature and supply rails around the hub. | Interface devices for SPI, CAN and LIN that bridge the hub to the vehicle network. |
| Renesas | Precision references and supervisors optimised for BMS and sensing applications. | Isolated data-converter interfaces and digital isolators suited to HV pack environments. |
| onsemi | Temperature monitor ICs combined with power-stage drivers for pumps and valves. | Isolation and high-side interface ICs that align with traction inverter and pack domains. |
| Microchip | General-purpose references and temperature sensor ICs that can be paired with external ADCs. | Digital isolators and interface bridges for SPI / I²C and automotive communication buses. |
| Melexis | Automotive temperature sensor SoCs that integrate sensing, conditioning and diagnostics. | Interfaces aimed at sensor-rich environments, complementing hub-centric architectures. |
The exact choice depends on your functional safety requirements, preferred vendor ecosystem and whether the same reference, monitor and isolation parts will be reused in other pack or inverter modules.
BOM & Procurement Checklist for Temperature Hubs
When an RFQ or BOM line only says “multi-channel temperature board”, most suppliers will respond with generic designs that are not tuned for cool-plate sensing or automotive safety. The checklist below turns the technical design work from this page into concrete fields you can put into a specification, so that quotations reflect the real requirements of a cool-plate temperature hub.
You can copy the left column into your RFQ template and fill the right column with project-specific values. This makes it clear that you need an automotive cool-plate hub, not just any temperature measurement board.
RFQ / BOM fields to define for a cool-plate temperature hub:
| Field in RFQ / BOM | What to specify |
|---|---|
| Number of temperature channels | Required channel count (e.g. 8 / 12 / 16) and expected expansion margin for future variants. |
| Sensor type and placement | NTC only, RTD only or mixed; typical sensor curve, wiring length and cold-plate regions covered. |
| Accuracy and long-term drift target | Total temperature error budget (e.g. ±1 °C over life) and whether factory / in-field calibration is expected. |
| Reference and ADC architecture | Preference for discrete reference + ADC, integrated isolated ADC, ΣΔ vs SAR and any vendor families you want to reuse. |
| Isolation domain and safety level | Which side of the isolation boundary the hub lives on, working voltage, insulation rating and target safety level (e.g. ASIL concept). |
| Diagnostics and functional safety hooks | Required open/short detection, self-test coverage, redundant channels and how diagnostic bits should be exposed to the BMS or ECU. |
| Communication interface | SPI, I²C or daisy-chain protocol, data rate, bus length and connector style matching the pack or ECU requirements. |
| Redundancy concept | Which locations need dual sensors or dual channels, and whether comparison is handled in the hub or in the BMS. |
| Operating environment | Temperature, vibration and coolant exposure at the hub mounting location, plus expected lifetime and service strategy. |
| Mechanical form factor | Board size, connector orientation, mounting points and clearance to nearby HV components on the pack or module. |
| Preferred vendor ecosystem | Any preferred semiconductor vendors for AFE, ADC, reference and isolation so that suppliers can align with your existing platforms. |
Filling in these fields turns a vague “multi-channel temperature measurement” request into a precise cool-plate hub specification, so suppliers can respond with architectures and IC selections that match your thermal and safety goals.
Request a Quote
FAQs – Cool-Plate Temperature Hub Planning & Selection
Each answer reflects how I would think through a real project: when a hub is justified, how I pick NTC versus RTD, what ADC and reference performance I actually need, how I treat isolation and diagnostics and which fields I put into an RFQ so suppliers know I want a true cool-plate hub.
When do I really need a dedicated cool-plate temperature hub?
I plan to use a dedicated hub when the channel count grows beyond four to six sensors, harness runs become long or I want clean isolation between the high-voltage pack and the low-voltage ECU. A hub also makes sense when I need centralised diagnostics, redundancy and a reusable module across several pack variants.
How do I decide between staying on ECU ADC channels and moving to a hub?
I stay on ECU ADC channels when sensors sit close to the ECU, isolation is already solved and the channel count is low. I move to a hub when sensors spread across the pack, wiring becomes messy, EMC risk rises or when I want one standard module I can drop into multiple projects.
How should I choose between NTC and RTD sensors on a cool plate?
I treat NTCs as the cost-optimised default and reserve RTDs for locations where lifetime stability and traceable accuracy really matter. For example, dense cell regions and critical busbar joints may justify RTDs, while less critical plate sections and coolant inlet or outlet points can often use automotive-grade NTCs successfully.
Can I mix NTC and RTD sensors on the same hub and still keep things manageable?
Yes, I can mix NTC and RTD sensors on one hub as long as I plan the front-end topologies and calibration strategy in advance. I group similar sensors together, keep reference and wiring schemes clear and document which channels are used for long-life monitoring versus simple over-temperature protection or trend tracking.
What ADC resolution do I really need for a cool-plate hub?
Instead of chasing theoretical resolution, I back-solve from my temperature accuracy target and noise budget. For many hubs, an effective twelve to fourteen bits over the temperature range is enough, provided I use averaging and a stable reference. If I want sub-degree resolution across many channels, higher effective bits become useful.
Why does the voltage reference matter so much for long-term accuracy?
I think of the reference as the anchor for every channel. If the reference drifts with temperature or age, every NTC or RTD reading moves together, even if the sensors themselves are perfect. That is why I treat reference selection, layout and decoupling as a top-level decision, not an afterthought.
Where should I put the isolation boundary for a cool-plate temperature hub?
I usually place the isolation boundary between the hub and the BMS or thermal ECU, keeping the sensors, AFE and ADC on the high-voltage side. That way, I can optimise measurement quality close to the pack while using a robust digital link with a clear working voltage and insulation rating requirement.
How do I think about isolation rating and safety level for the hub?
I start from the pack voltage, creepage and clearance requirements and the safety concept of the overall system. Then I pick isolation components that meet working voltage and surge margins with headroom. If the hub feeds safety functions, I align the isolation choice and diagnostics with the target ASIL level documented for the project.
What is a practical calibration strategy for a temperature hub?
I assume at least one factory calibration step that ties the reference, ADC and a few representative channels to known points. If the project needs very tight drift over life, I add in-field calibration hooks, such as reference comparisons or golden sensors, and make sure calibration data and versioning are visible to the BMS or ECU.
Which diagnostics and functional safety hooks should I insist on?
I expect open and short detection on each channel, self-test capabilities for the ADC and reference and at least basic plausibility checks between neighbouring sensors or redundant pairs. I also want clear status bits and error counters that the BMS can log, so faults turn into actionable diagnostic trouble codes instead of vague symptoms.
How should the hub interface with my BMS or thermal ECU?
I prefer a simple, well documented digital interface such as SPI, I2C or a daisy-chain link with clear timing and error reporting. I check that the clock rates and cable lengths match my EMC budget and that the protocol exposes channel data, status bits and configuration in a way my software team likes.
What should I include in an RFQ or BOM when I ask suppliers for a hub?
In my RFQ I spell out channel count, sensor types, accuracy and drift targets, isolation domain, diagnostics expectations, interface choice and environmental limits around the pack. That way suppliers understand I want an automotive cool-plate temperature hub with safety hooks, not just a generic laboratory temperature board with a lot of ADC channels.
