[{"data":1,"prerenderedAt":66},["ShallowReactive",2],{"blog-ntc-thermistor-divider-pcb-validation-en":3},{"title":4,"description":5,"date":6,"lastUpdated":6,"category":7,"image":8,"imageAlt":9,"readingTime":10,"wordCount":11,"timeRequired":12,"htmlContent":13,"faq":14,"tags":30,"slug":34,"sourceLocale":35,"availableLocales":36,"jsonld":42},"NTC Thermistor Divider Design and Validation","An NTC-low voltage divider converts thermistor resistance to a voltage, and validation checks the assembled thermal measurement channel and its response.","2026-11-13","design","/assets/blogs/ntc-thermistor-divider-pcb-validation/ntc-divider-cover.svg","Conceptual NTC measurement channel from thermal coupling through the divider and ADC/software to validation of the complete assembled channel",11,2154,"PT11M","\u003Cp>An NTC thermistor channel turns resistance into an electrical measurement and then into a temperature estimate. Its success depends on more than choosing a nominal resistance: the sensor must be thermally coupled to the right object, the ADC must read the divider correctly, and the released software must use the curve for the installed part.\u003C/p>\n\u003Cp>The calculation is hypothetical and does not claim temperature accuracy or represent manufacturer part data.\u003C/p>\n\u003Ch2>Start with the temperature being measured\u003C/h2>\n\u003Cp>Write the \u003Cstrong>measurand\u003C/strong> before selecting the thermistor. Is the channel intended to report ambient air, a battery surface, a component case, or a location on the PCB? A sensor near that location is not automatically at the same temperature.\u003C/p>\n\u003Cp>The requirements should state:\u003C/p>\n\u003Cul>\n\u003Cli>Operating temperature range, allowable error, and required response time.\u003C/li>\n\u003Cli>Thermal contact or separation from the target, including adhesive, enclosure, and airflow.\u003C/li>\n\u003Cli>Nearby heat sources, copper heat paths, coatings, and expected operating power.\u003C/li>\n\u003Cli>Fault response and the system action associated with the reported temperature.\u003C/li>\n\u003C/ul>\n\u003Cp>For example, a board-mounted sensor may measure the local board temperature reliably while under-reading a nearby component hotspot. Calibration at one operating condition cannot necessarily remove a changing thermal gradient.\u003C/p>\n\u003Ch2>Select an exact part and controlled R-T source\u003C/h2>\n\u003Cp>Identify the thermistor by \u003Cstrong>manufacturer and full MPN\u003C/strong>, not just “10 kΩ NTC.” Attach the current, controlled datasheet or manufacturer-approved R-T data to the design record. Confirm that it applies to the ordered tolerance and package.\u003C/p>\n\u003Cp>Record at least:\u003C/p>\n\u003Cul>\n\u003Cli>Reference resistance, its stated temperature, and tolerance.\u003C/li>\n\u003Cli>R-T table or curve, applicable tolerance, and permitted operating range.\u003C/li>\n\u003Cli>B-constant, its two defining temperatures, and tolerance, if used.\u003C/li>\n\u003Cli>Dissipation and thermal time-constant data, including test conditions.\u003C/li>\n\u003Cli>Package, recommended footprint, assembly limits, and lifecycle status.\u003C/li>\n\u003C/ul>\n\u003Cp>\u003Cimg src=\"/assets/blogs/ntc-thermistor-divider-pcb-validation/bead-thermistors.jpg\" alt=\"Examples of bead-encapsulated thermistors with different lead arrangements\">\u003C/p>\n\u003Cp>\u003Cem>These thermistors illustrate package variation; the photograph does not identify the hypothetical example or establish a specific resistance-temperature curve.\u003C/em>\u003C/p>\n\u003Cp>Murata’s \u003Ca href=\"https://www.murata.com/en-us/products/thermistor/ntc/overview/basic/characteristic\">\u003Cstrong>“Basic characteristics of NTC thermistors”\u003C/strong>\u003C/a> explains why resistance must be associated with temperature, why measurement current can cause self-heating, and why thermal behavior depends on the environment. Its family-level guidance is not a substitute for the selected MPN’s datasheet.\u003C/p>\n\u003Cp>A B-constant is associated with a defined temperature pair. A common approximation is:\u003C/p>\n\u003Cdiv class=\"math-block\">R(T)=R(T\u003Csub>0\u003C/sub>) exp[B((1)/(T)-(1)/(T\u003Csub>0\u003C/sub>))]\u003C/div>\n\n\u003Cp>Temperatures in this equation are in kelvin. Use the manufacturer’s specified B interval and check the approximation against the controlled R-T data over the intended range. Even within the interval, the approximation can have residual error; outside it, do not assume the model remains adequate. A table with validated interpolation or part-specific fitted coefficients may be preferable.\u003C/p>\n\u003Ch2>Choose the divider for the range and ADC\u003C/h2>\n\u003Cp>In a \u003Cstrong>thermistor-low\u003C/strong> divider, the fixed resistor connects excitation to the measurement node, and the NTC connects the node to ground. As temperature rises and NTC resistance falls, node voltage falls. Reversing the components reverses the voltage direction and requires different equations.\u003C/p>\n\u003Cp>\u003Cimg src=\"/assets/blogs/ntc-thermistor-divider-pcb-validation/ntc-low-divider-candidate.svg\" alt=\"NTC-low divider with the fixed resistor between excitation and the measurement node, the NTC between the node and ground, and an ideal high-impedance ADC input connected to the node.\">\u003C/p>\n\u003Cp>Murata’s \u003Ca href=\"https://www.murata.com/en-us/products/thermistor/ntc/overview/basic/circuit\">\u003Cstrong>“Temperature Detection Circuit”\u003C/strong>\u003C/a> describes both orientations and shows how additional resistors alter the transfer characteristic. Choose the topology with ADC range, fault detection, protection, and firmware conventions in mind.\u003C/p>\n\u003Cp>For the thermistor-low arrangement:\u003C/p>\n\u003Cdiv class=\"math-block\">V\u003Csub>node\u003C/sub>=V\u003Csub>exc\u003C/sub> * (R\u003Csub>NTC\u003C/sub>)/(R\u003Csub>fixed\u003C/sub>+R\u003Csub>NTC\u003C/sub>)\u003C/div>\n\n\u003Cp>If the normalized measured voltage is \u003Cspan class=\"math-inline\">x=(V\u003Csub>node\u003C/sub>)/(V\u003Csub>exc\u003C/sub>)\u003C/span>, then:\u003C/p>\n\u003Cdiv class=\"math-block\">R\u003Csub>NTC\u003C/sub>=R\u003Csub>fixed\u003C/sub> * (x)/(1-x)\u003C/div>\n\n\u003Cp>The inversion becomes sensitive to voltage errors near the rails, especially as \u003Cspan class=\"math-inline\">x\u003C/span> approaches one. Check the entire temperature range, not just a nominal point.\u003C/p>\n\u003Cp>Choose the fixed resistor by evaluating the exact R-T curve and the design objective: for example, worst-case temperature resolution, power, fault margin, or sensitivity near a control threshold. Matching it to the NTC resistance at one temperature puts the node at midscale there; it is not automatically the best choice across the full range.\u003C/p>\n\u003Ch3>Hypothetical calculation\u003C/h3>\n\u003Cp>Assume the thermistor-low topology, an ideal 3.3 V excitation, a 10 kΩ fixed resistor, and a hypothetical NTC resistance of 10 kΩ at the temperature of interest:\u003C/p>\n\u003Cul>\n\u003Cli>Divider current: \u003Cspan class=\"math-inline\">(3.3 V)/(20 kΩ)=0.165 mA\u003C/span>.\u003C/li>\n\u003Cli>Node voltage: \u003Cspan class=\"math-inline\">(3.3×10)/(10+10)=1.65 V\u003C/span>.\u003C/li>\n\u003Cli>NTC power: \u003Cspan class=\"math-inline\">(1.65\u003Csup>2\u003C/sup>)/(10,000)=0.00027225 W\u003C/span>, or about \u003Cstrong>0.272 mW\u003C/strong>.\u003C/li>\n\u003Cli>For an illustrative ideal 12-bit ADC using a 0–4095 endpoint convention, midscale corresponds to about code 2048 after rounding.\u003C/li>\n\u003C/ul>\n\u003Cp>Actual ADC transfer and rounding rules come from the controller documentation. No temperature or accuracy follows from this example without an exact NTC curve and error analysis.\u003C/p>\n\u003Cp>For this topology, an open NTC connection tends toward the excitation rail and a short to ground tends toward zero, subject to loading and wiring details. Define diagnostic thresholds with tolerances and valid extreme temperatures included; test other faults separately.\u003C/p>\n\u003Ch2>Account for excitation, ADC, and settling\u003C/h2>\n\u003Cp>If the divider excitation and ADC reference track the \u003Cstrong>same actual source\u003C/strong>, their shared steady-state voltage scale can cancel in the ideal normalized ratio. That does not cancel every error. Source drops, differences between the two paths, reference and excitation transients, ADC gain or offset, noise, leakage, and incomplete settling can still affect the result.\u003C/p>\n\u003Cp>Use the selected ADC documentation to check:\u003C/p>\n\u003Cul>\n\u003Cli>Input voltage range, reference requirements, and power-up behavior.\u003C/li>\n\u003Cli>Acquisition time and permitted source impedance, including the divider’s Thevenin resistance over temperature.\u003C/li>\n\u003Cli>Multiplexer switching, sample-capacitor charge demand, and first-conversion behavior.\u003C/li>\n\u003Cli>ADC offset, gain, nonlinearity, quantization, noise, and reference sensitivity.\u003C/li>\n\u003Cli>Input and protection-device leakage across temperature.\u003C/li>\n\u003Cli>RC filter settling, startup delay, and the effect of any capacitor at the ADC pin.\u003C/li>\n\u003C/ul>\n\u003Cp>A capacitor can provide local charge and filter noise, but it also changes settling time. If excitation is switched to reduce power, allow both electrical settling and the relevant thermal behavior before accepting a sample. Verify the sequence on hardware.\u003C/p>\n\u003Cp>Averaging can reduce some uncorrelated noise. It does not remove a persistent offset, a thermal gradient, model residual, or correlated interference.\u003C/p>\n\u003Ch2>Check self-heating and thermal response\u003C/h2>\n\u003Cp>The R-T data used for sensing generally describes resistance under sufficiently low measurement power that self-heating is negligible. A powered divider warms the NTC; because it is an NTC, that warming changes its resistance and apparent temperature.\u003C/p>\n\u003Cp>For the thermistor-low divider:\u003C/p>\n\u003Cdiv class=\"math-block\">P\u003Csub>NTC\u003C/sub>=(V\u003Csub>exc\u003C/sub>\u003Csup>2\u003C/sup> * R\u003Csub>NTC\u003C/sub>)/((R\u003Csub>fixed\u003C/sub>+R\u003Csub>NTC\u003C/sub>)\u003Csup>2\u003C/sup>)\u003C/div>\n\n\u003Cp>With fixed excitation and fixed resistance, this power is greatest at \u003Cspan class=\"math-inline\">R\u003Csub>NTC\u003C/sub>=R\u003Csub>fixed\u003C/sub>\u003C/span>, if that resistance occurs in the operating range. Evaluate the actual range and excitation limits.\u003C/p>\n\u003Cp>A datasheet dissipation constant may help make an initial estimate, but it is conditional on the test environment. PCB copper, solder lands, airflow, mounting, and coating can change heat removal. Compare readings at different excitation powers on the assembled product, allowing sufficient settling, and assess the resulting temperature shift.\u003C/p>\n\u003Cp>Likewise, a catalog thermal time constant is not the finished product’s response time. The commonly cited 63.2% step-response point depends on the sensor and its environment. Measure response using the intended interface and define the test step and acceptance criterion.\u003C/p>\n\u003Ch2>Build an error budget before calibration\u003C/h2>\n\u003Cp>The following matrix is a starting structure, not a set of universal allowances. Enter values or bounds for the actual design over temperature and operating conditions.\u003C/p>\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Error contributor\u003C/th>\n\u003Cth>How to assess it\u003C/th>\n\u003Cth>Primary owner\u003C/th>\n\u003C/tr>\n\u003C/thead>\n\u003Ctbody>\u003Ctr>\n\u003Ctd>NTC R-T tolerance\u003C/td>\n\u003Ctd>Apply the exact part’s guaranteed curve limits across the range\u003C/td>\n\u003Ctd>Component engineer\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>Conversion model\u003C/td>\n\u003Ctd>Compare table interpolation or fitted model with controlled R-T data\u003C/td>\n\u003Ctd>Firmware engineer\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>Fixed resistor\u003C/td>\n\u003Ctd>Include initial tolerance and temperature coefficient at operating temperature\u003C/td>\n\u003Ctd>Circuit engineer\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>Excitation and reference\u003C/td>\n\u003Ctd>Analyze shared and nonshared paths, drops, and transients\u003C/td>\n\u003Ctd>Circuit engineer\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>ADC transfer and noise\u003C/td>\n\u003Ctd>Use controller limits and measured channel behavior\u003C/td>\n\u003Ctd>Circuit and firmware engineers\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>Acquisition and filtering\u003C/td>\n\u003Ctd>Calculate and test settling after switching and startup\u003C/td>\n\u003Ctd>Circuit engineer\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>Leakage and wiring\u003C/td>\n\u003Ctd>Evaluate input, protection, contamination, and connector effects\u003C/td>\n\u003Ctd>Circuit and layout engineers\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>Self-heating\u003C/td>\n\u003Ctd>Bound power and measure the assembled thermal effect\u003C/td>\n\u003Ctd>Validation engineer\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>Thermal coupling\u003C/td>\n\u003Ctd>Compare the target and sensor under representative conditions\u003C/td>\n\u003Ctd>Mechanical and validation engineers\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>Calibration residual and drift\u003C/td>\n\u003Ctd>Evaluate independent check points and applicable stability evidence\u003C/td>\n\u003Ctd>Validation engineer\u003C/td>\n\u003C/tr>\n\u003C/tbody>\u003C/table>\n\u003Cp>Express each contribution in a common unit, usually temperature error at defined conditions. Where necessary, convert electrical error to temperature using the local transfer slope. Avoid counting the same tolerance twice—for instance, when a manufacturer’s curve limit already includes reference-resistance variation.\u003C/p>\n\u003Cp>Use a justified worst-case combination for bounded effects, or a statistical combination only when distributions, correlations, and confidence assumptions are supported. Keep thermal gradients and time-dependent effects visible rather than hiding them in a single nominal number.\u003C/p>\n\u003Ch2>Release the layout and assembly interface\u003C/h2>\n\u003Cp>Place the thermistor to serve the measurand. Document its distance from heat sources, its contact path to the target, and any copper intended to conduct or isolate heat. A layout change that looks electrically harmless can change the thermal measurement.\u003C/p>\n\u003Cp>Keep the sensitive node short and away from switching signals. Control contamination and moisture where leakage matters. Specify the footprint, land pattern, solder process, orientation where relevant, coating, connectors, and inspection requirements. For a remote sensor, include wire resistance, connector integrity, and the fault behavior of the complete harness.\u003C/p>\n\u003Cp>The broader \u003Ca href=\"/en/blog/circuit-board-components/\">circuit board components\u003C/a> guide provides context for documenting component choices.\u003C/p>\n\u003Ch2>Calibrate and validate the assembled channel\u003C/h2>\n\u003Cp>Separate \u003Cstrong>electrical conversion\u003C/strong> from \u003Cstrong>thermal measurement\u003C/strong> during validation.\u003C/p>\n\u003Col>\n\u003Cli>Substitute known resistances at the sensor interface to test divider conversion, ADC behavior, fault thresholds, and firmware independently of thermal coupling.\u003C/li>\n\u003Cli>Test the assembled sensor against a traceable temperature reference with documented uncertainty. Place and couple the reference so it represents the intended measurand.\u003C/li>\n\u003Cli>Use controlled points across the specified range, including endpoints and intermediate check points. Allow adequate soak and record gradients and stability.\u003C/li>\n\u003Cli>Test startup, excitation changes, airflow or enclosure conditions, and temperature transitions relevant to use. Distinguish steady-state error from lag.\u003C/li>\n\u003Cli>Compare results with predeclared acceptance limits, accounting for reference uncertainty and any guard band.\u003C/li>\n\u003C/ol>\n\u003Cp>Define whether calibration is electrical, thermal, or both. Record calibration points, method, reference identification, coefficients, storage precision, and firmware version. A one-point correction can remove an offset at its calibration condition; it does not establish accuracy across the range. Two-point calibration also does not guarantee correction of curve shape or changing thermal gradients. Verify at independent points.\u003C/p>\n\u003Cp>Specify production-test coverage separately from engineering validation: which units are checked, which conditions are exercised, and how failures are dispositioned.\u003C/p>\n\u003Cp>For a broader view of which fault conditions a production test can detect, use the \u003Ca href=\"/en/blog/pcba-test-coverage-matrix/\">PCBA test coverage matrix\u003C/a>. Define the relevant sensor-channel faults and supporting evidence for the actual test setup.\u003C/p>\n\u003Ch2>Control release and substitutions\u003C/h2>\n\u003Cp>Assign named people or functions before release; a suggested ownership checklist is:\u003C/p>\n\u003Cul>\n\u003Cli>\u003Cstrong>System owner:\u003C/strong> approves measurand, limits, fault actions, and acceptance criteria.\u003C/li>\n\u003Cli>\u003Cstrong>Component owner:\u003C/strong> approves exact MPN, controlled R-T source, tolerances, and supply status.\u003C/li>\n\u003Cli>\u003Cstrong>Circuit and layout owners:\u003C/strong> approve ADC timing, error budget, placement, footprint, and assembly notes.\u003C/li>\n\u003Cli>\u003Cstrong>Firmware owner:\u003C/strong> controls topology equations, model, diagnostics, coefficients, and versioning.\u003C/li>\n\u003Cli>\u003Cstrong>Validation owner:\u003C/strong> approves reference traceability, test evidence, calibration procedure, and production coverage.\u003C/li>\n\u003Cli>\u003Cstrong>Change-control owner:\u003C/strong> records approvals and decides whether revalidation is required.\u003C/li>\n\u003C/ul>\n\u003Cp>Treat a substitute as a new engineering decision. Equal resistance at 25 °C does not establish equal curve, B interval, tolerance, thermal response, package coupling, or assembly compatibility. Review the complete channel and repeat affected tests before approval.\u003C/p>\n\u003Cp>The \u003Ca href=\"/en/blog/pcb-component-substitution-approval/\">PCB component substitution approval checklist\u003C/a> helps document the exact alternate, its approval scope, and the affected build. When an approved change moves into production, the \u003Ca href=\"/en/blog/pcba-engineering-change-order/\">engineering change order checklist\u003C/a> addresses the baseline, effectivity, and disposition of affected material.\u003C/p>\n\u003Cp>For a PCB project inquiry, use the \u003Ca href=\"/en/quote/\">quote page\u003C/a> and state the documented requirements; responsibilities and scope should be agreed for the project.\u003C/p>\n\u003Cdiv data-component=\"BlogQuickQuoteInline\">\u003C/div>\n\n\u003C!-- faq:start -->\n\u003Ch2>FAQ\u003C/h2>\n\u003Ch3>Is a thermistor with the same nominal resistance a direct replacement?\u003C/h3>\n\u003Cp>Not necessarily. Compare the exact R-T curve and tolerances, operating range, package, thermal behavior, and assembly requirements, then assess whether the channel needs revalidation.\u003C/p>\n\u003Ch3>Should the fixed resistor always equal the NTC’s resistance at 25 °C?\u003C/h3>\n\u003Cp>No. Equality gives midscale when the NTC is at that resistance. Select the fixed resistor against the actual temperature range, ADC behavior, power, and measurement objective.\u003C/p>\n\u003Ch3>Does a shared supply and ADC reference eliminate supply error?\u003C/h3>\n\u003Cp>It can cancel the ideal shared steady-state voltage scale in a ratiometric calculation. Path differences, transients, settling, ADC errors, leakage, and noise still need evaluation.\u003C/p>\n\u003Ch3>Can software averaging correct self-heating?\u003C/h3>\n\u003Cp>No. Averaging may reduce some noise, but it does not remove a persistent temperature shift caused by measurement power. Limit and measure self-heating in the assembled environment.\u003C/p>\n\u003Ch3>How many calibration points are enough?\u003C/h3>\n\u003Cp>There is no universal count. Choose points from the required range, model, tolerance, and acceptance criteria, and use independent check points to verify the result.\u003C/p>\n\u003C!-- faq:end -->\n\u003Csection class=\"related-links\" aria-label=\"Related\">\u003Ch3>Related links\u003C/h3>\u003Cul>\u003Cli>\u003Ca href=\"/en/blog/circuit-board-components/\">Circuit Board Components: Common PCB Parts, What They Do and How to Identify Them\u003C/a>\u003C/li>\u003Cli>\u003Ca href=\"/en/blog/pcba-test-coverage-matrix/\">PCB Assembly Test Coverage: How to Build a Coverage Matrix\u003C/a>\u003C/li>\u003Cli>\u003Ca href=\"/en/blog/pcb-component-substitution-approval/\">PCB Component Substitution: An Approval Checklist\u003C/a>\u003C/li>\u003Cli>\u003Ca href=\"/en/blog/pcba-engineering-change-order/\">PCB Assembly Engineering Change Orders: A Release Checklist\u003C/a>\u003C/li>\u003Cli>\u003Ca href=\"/en/quote/\">quote page\u003C/a>\u003C/li>\u003C/ul>\u003C/section>",[15,18,21,24,27],{"question":16,"answer":17,"answerText":17},"Is a thermistor with the same nominal resistance a direct replacement?","Not necessarily. Compare the exact R-T curve and tolerances, operating range, package, thermal behavior, and assembly requirements, then assess whether the channel needs revalidation.",{"question":19,"answer":20,"answerText":20},"Should the fixed resistor always equal the NTC’s resistance at 25 °C?","No. Equality gives midscale when the NTC is at that resistance. Select the fixed resistor against the actual temperature range, ADC behavior, power, and measurement objective.",{"question":22,"answer":23,"answerText":23},"Does a shared supply and ADC reference eliminate supply error?","It can cancel the ideal shared steady-state voltage scale in a ratiometric calculation. Path differences, transients, settling, ADC errors, leakage, and noise still need evaluation.",{"question":25,"answer":26,"answerText":26},"Can software averaging correct self-heating?","No. Averaging may reduce some noise, but it does not remove a persistent temperature shift caused by measurement power. Limit and measure self-heating in the assembled environment.",{"question":28,"answer":29,"answerText":29},"How many calibration points are enough?","There is no universal count. Choose points from the required range, model, tolerance, and acceptance criteria, and use independent check points to verify the result.",[31,32,33],"NTC thermistor","temperature sensing","PCB design","ntc-thermistor-divider-pcb-validation","en",[35,37,38,39,40,41],"de","it","cn","fr","ru",{"blog":43,"breadcrumb":52},{"@context":44,"@type":45,"headline":4,"description":5,"image":8,"url":46,"datePublished":6,"dateModified":6,"timeRequired":12,"keywords":47,"articleSection":7,"author":48,"publisher":51},"https://schema.org","BlogPosting","https://hilpcb.com/en/blog/ntc-thermistor-divider-pcb-validation/","NTC thermistor, temperature sensing, PCB design",{"@type":49,"name":50},"Organization","HILPCB",{"@type":49,"name":50},{"@context":44,"@type":53,"itemListElement":54},"BreadcrumbList",[55,60,64],{"@type":56,"position":57,"name":58,"item":59},"ListItem",1,"Home","https://hilpcb.com/",{"@type":56,"position":61,"name":62,"item":63},2,"Blog","https://hilpcb.com/en/blog/",{"@type":56,"position":65,"name":34,"item":46},3,1794758597813]