e-book

An Engineer’s Guide to
Temperature Sensing
Temperature sensor design challenges and solutions

TI.com/temperature
Table of contents

Introduction 3 Fluid temperature measurement

4 RTD replacement in heat meters using
1. Temperature sensor technologies
digital temperature sensors
Temperature sensing fundamentals 5

2. Ambient and board temperature monitoring Simple temperature limit alerts

7
How to protect battery power management
Fundamentals of heat conduction systems from thermal damage 36
Part 1: Temperature sensors: PCB guidelines
for surface mount devices 9 Cold junction compensation
Temperature compensation using
Design guidelines for air high-accuracy temperature sensors 38
temperature measurement
Part 2: Temperature sensors: PCB guidelines Processor die temperature monitoring
for surface mount devices (page 11) 18 High-performance processor die
temperature monitoring 40
Design guidelines for component
temperature measurement Ultra-high accuracy temperature sensing
Part 3: Temperature sensors: PCB guidelines Precise temperature measurements with
for surface mount devices (page 18) 25 the TMP116 and TMP117

3. Temperature sensing application examples 29 Additional resources

Multipoint temperature monitoring

Efficient cold chain management with scalable
high-accuracy temperature sensors 30

Body temperature sensing

Design challenges of wireless patient
temperature monitors 32

An Engineer’s Guide to Temperature Sensing 2 Texas Instruments Incorporated

Introduction

Whether you are designing personal electronics, industrial This e-book’s chapters describe significant temperature
or automotive systems, you must address some of the same challenges, focus on design considerations for applications,
challenges: how to increase performance, add features and and assess trade-offs between temperature accuracy and
shrink form factors. Along with these considerations, it’s application size while considering sensor placement. If you
imperative to carefully monitor temperature in order to ensure have feedback about the topics covered here or any other
safety and protect systems and consumers from harm. temperature monitoring and protection questions, please
Another trend across numerous industries is increased submit them to the Sensors forum on the TI E2E™ design
data processing from more sensors, further emphasizing support forums.
the importance of temperature measurement – not just
to measure system or environmental conditions but to
compensate for other temperature-sensitive components Authors
and maintain both sensor and system accuracy. As an
Sydney Davis, Bryan Padilla, Punya Prakash, Ben
added benefit, accurate temperature monitoring can increase
Kasemsadeh, Aaron Heng, Amit Ashara, Jose Arteaga,
system performance and reduce costs by removing the
Megan Anderson, Nicole Khoury, Brandon Fisher, Jalen Tate
need to overdesign systems to compensate for inaccurate
temperature measurements.

An Engineer’s Guide to Temperature Sensing 3 Texas Instruments Incorporated

Chapter 1: Temperature sensor technologies

Overview • Resistive temperature detectors (RTDs). Because

RTDs are made of pure metal, such as platinum, they can
The demand for temperature sensing in today’s advanced
operate with high accuracy in extremely high-temperature
manufacturing environments has grown to encompass nearly
environments; the downside is that they are one of the most
every application in the automotive, industrial, personal
expensive types of temperature sensors. RTDs operate on
electronics, communications and enterprise sectors. Given
the known correlation between a pure metal’s increase in
the diverse array of applications that require temperature
resistance and an increase in temperature.
monitoring, there are many considerations when choosing the
most efficient type of sensor, including sensor accuracy, size, • Thermocouples. A thermocouple fuses together two
different metals, such as copper and iron, to form a
placement, drift and calibration.
junction. This junction produces a small voltage when
In order to help you select the best sensor for your the temperature changes, which is convertible into a
application, Chapter 1 compares the four most popular corresponding temperature value. Although thermocouples
temperature sensors: can measure the widest temperature range among the four
• Integrated circuit (IC) sensors. Made from silicon sensor types, they are often larger in size and cost more
and able to leverage the highly predictable thermal than IC sensors or thermistors.
characteristics of the silicon p-n junction, IC sensors provide
high accuracy, low power consumption, fast response time
and a compact form factor.
• Thermistors. There are two main categories of thermistors:
negative temperature coefficients (NTCs) and positive
temperature coefficients (PTCs). As temperature rises, the
resistance of an NTC will decrease, while the resistance
of a PTC will increase. NTCs usually require calibration for
greater accuracy, but silicon-based PTC thermistors have a
linear relationship between temperature and resistance and
thus do not require calibration, saving resources and time in
your design process.

An Engineer’s Guide to Temperature Sensing 4 Texas Instruments Incorporated

Temperature sensing fundamentals

Introduction to Temperature Sensing

µA
In embedded systems, there is a constant need for
higher performance and more features in a smaller
form factor. This requires system designers to monitor
the overall temperature to ensure safety and protect
the systems. The trend of sensor data-logging further
drives the need for temperature measurement to not
E
only measure system or environmental conditions, but
to compensate for temperature-sensitive components V
BE
B
and maintain accuracy of the total system.
Thermal Design Considerations
Considerations for efficient thermal monitoring and
protection include: C

i. Accuracy: Sensor accuracy represents how close

the temperature is to the true value. Applications
GND
should consider factors such as linearity and
acquisition circuits across the operating Figure 1. Temperature Dependence of Silicon
temperature range. Bandgap
ii. Size: While the size of the sensor makes an impact
on the design, analyzing the overall circuit can yield KT §I ·
a more optimized design. 'VBE u In¨ C1 ¸
q © IC2 ¹ (1)
iii. Sensor Placement: Package and placement can
impact the response time and conduction path. Given the predictable behavior, ICs offer high linearity
Both are critical for effective thermal design. and accuracy across a wide temperature range (up to
±0.1°C). IC sensors can integrate system functionality
Table 1. Temperature Sensing Technology that offers a small footprint and low power
Comparison consumption. These sensors do operate in a limited
Thermo
temperature range and offer fewer packaging options
IC Sensor Thermistor RTD that can measure off-board temperature when
couple
–55°C to –100°C to –240°C to –260°C to
compared to thermistors.
Range
+200°C +500°C 600°C +2300°C These sensors are typically fully-integrated, and
Accuracy
Good / Calibration
Best Better monolithic sensors and accuracy are designed for the
Best dependant entire system instead of for one element.
Footprint /
Smallest Small Moderate Large Thermistors
Size
Complexity Easy Moderate Complex Complex Thermistors are passive components that change
Linearity Best Low Best Better resistance with temperature. Thermistors fall into two
Point-to- categories, negative temperature coefficient (NTC) and
point, Multi- Point-to- Point-to- Point-to- positive temperature coefficient (PTC).
Topology
drop, Daisy point point point
Chain While thermistors offer a variety of package options for
Low to Low to
onboard and off-board temperature sensing, typical
Price Expensive Expensive implementation requires more system components.
Moderate Moderate
NTC thermistor are non-linear and often bear
IC Sensors increased calibration cost and software overhead. An
exception to this are Silicon-based PTC thermistors.
IC temperature sensors rely on the predictive
temperature dependence of a Silicon bandgap. The The true system accuracy for Thermistors are often
precision current sources the internal forward biased difficult to determine. NTC system error contributors
p-n junction with the resulting ∆VBE that corresponds to include NTC tolerance, bias resistor (Tolerance,
the device temperature. Temperature Drift), ADC (quantization error),
linearization error, reference voltage (Accuracy,
Temperature Drift).

SNOAA25 – January 2019 Temperature sensing fundamentals 5

Submit Documentation Feedback
Copyright © 2019, Texas Instruments Incorporated
 www.ti.com

The temperature of the cold junction must be known to

V
CC derive the hot junction temperature. Here, accuracy is
limited by the fact that there are two systems that have
Thermistor RT
separate tolerances and capabilities interacting with
each other.
Amp ADC
Isothermal Block
Type J

Iron
Copper
+
Th
Figure 2. Typical Thermistor Implementation Tc

Constantan Copper
-
Resistive Temperature Detectors (RTD)
RTDs are temperature sensors made of pure material, TMP
typically in platinum, nickel, or copper, with a highly
predictable resistance-temperature relationship.
V+ Cold Junction Compensation
Temperature Sensor
ISOURCE
Figure 4. Thermocouple With CJC Temperature
Sensor
RL
Thermocouples do not require external excitation, and
RL hence are not impacted by self-heating issues. They
can also support extreme temperatures (>2000°C).
+ While they are rugged and inexpensive,
VMEAS
thermocouples do require an additional temperature
PGA û ADC
sensor for cold-junction-compensation (CJC). They
± tend to be non-linear and are highly sensitive to
RL parasitic junctions where the thermocouple is attached
to the board.
RBIAS Finally, digitizing a thermocouple would be susceptible
RL
to previously discussed ADC errors.
Device Recommendations
Figure 3. Complex 4-Wire RTD Circuit For 40 years, Texas Instruments has manufactured
several IC-based temperature sensors, including:
Platinum RTDs can be highly accurate and very linear • Digital temperature sensors:
across a very wide temperature range up to 600°C. – Highest accuracy temperature sensors
Implementation with these analog sensors involve
complex circuitry and design challenges. Ultimately, – Lowest power with the smallest footprint
the accurate systems involve complex error analysis – LM75 / TMP75 temperature sensors
due to a higher number of contributing components – Multi-channel remote diode temperature
that also impact the overall system size. RTDs also sensors
require calibration during manufacturing followed by an
annual calibration process in the field. • High-accuracy analog temperature sensors
• Linear thermistors
Contributors to the RTD system error include RTD
Tolerance, self-heating, ADC (quantization error) and • Temperature switches or thermostats that offer
references used in the system. integrated hysteresis for enhanced noise immunity
Thermocouples Table 2. Related Documentation
Thermocouples are made of two dissimilar electrical COLLATERAL DESCRIPTION
conductors that form electrical junctions at different
temperatures. A thermocouple produces a Application Report - Layout Considerations for Accurately
Accuracy Measuring Ambient Temperature
temperature-dependent voltage as a result of the
thermoelectric seebeck effect. This voltage translates Application Report -
Methods to Calibrate Temperature
Temperature
to the difference of temperature between the hot Calibration
Monitoring Systems
junction (Th) and the cold junction (Tc).
6 Temperature sensing fundamentals SNOAA25 – January 2019
Submit Documentation Feedback
Copyright © 2019, Texas Instruments Incorporated
Chapter 2: Ambient and board temperature monitoring

Overview Chapter 2 explains methods for improving accuracy while

reducing thermal response time and offers recommendations
Power-hungry electronic components such as central
regarding:
processing units, graphics processing units, field-
programmable gate arrays or voltage regulators heat up • Ambient temperature monitoring. Maintaining tight
during regular operation. This is because all electrical control of environmental conditions or ensuring safe
components have at least a small natural resistivity to current, operating conditions requires high-accuracy ambient
which causes some of the electrical energy to be lost as heat. temperature monitoring. High ambient temperatures can
The heat generated from these components can interfere with cause overheating that leads to hardware damage or
accurate temperature measuring. reduced system efficiency. Selecting the correct layout
technique for your process helps minimize the impact of
Some applications require ambient air temperature heat generated from components on the PCB.
measurements, while others need to measure the temperature
• PCB component temperature monitoring. For PCBs,
of a nearby component on the printed circuit board (PCB).
temperature monitoring helps maintain component safety,
But instantaneous temperature measuring is not possible
reliability and performance. High temperatures can cause
in real-world processes for several reasons. Factors such
hardware damage and electrical hazards and decrease
as the position of the sensor relative to the heat-generating
system efficiency and life spans. High temperatures can
element(s) and the physical characteristics of the sensor itself
also come from multiple sources, including heat-generating
can cause the recorded temperature value to lag behind the
components mounted directly on the PCB or inadequate
actual temperature value. When integrating a temperature system ventilation. Because it’s possible to place TI’s
sensor, you’ll need to minimize the trade-off between meeting small-form-factor temperature sensors in proximity to heat
system requirements and obtaining the optimal thermal sources, measurement accuracy improves.
response time at the lowest possible cost.
Perhaps the most important design decision is layout
technique. A poor PCB layout can lead to inaccurate
temperature readings and eventually system damage and
a shortened board lifetime. Board layout considerations
include the maximum allowable temperature, the operating
environment and how much power (as heat) the board
components will dissipate.

An Engineer’s Guide to Temperature Sensing Texas Instruments Incorporated

 Application Report
SNOA967A – July 2017 – Revised January 2019

Temperature sensors: PCB guidelines for surface mount

devices

Ben Kasemsadeh, Aaron Heng, and Amit Ashara

ABSTRACT
Power hungry electronic components such as CPUs, GPUs, or FPGAs, as well as voltage regulators heat
up during operation. Some applications require ambient air temperature measurements while others need
to measure the temperature of a nearby component on the PCB. Measuring ambient air temperature with
a surface mount technology (SMT) device is challenging due to the thermal influence of other components
within the system. In other systems, in which the temperature of a component needs to be measured,
ambient air temperature can influence and degrade the measurement accuracy.
The system designer needs to make certain design decisions regarding both package type and PCB
layout when integrating a temperature sensor. This application note provides recommendations to system
designers and explains methods for improving the accuracy of the temperature point being measured. The
Recommendations are provided both for air temperature measurements and for component temperature
measurement. The report details layout techniques, device orientation, and best practices for mounting.

Contents
1 Introduction ................................................................................................................... 2
1.1 What Is Heat Conduction? ......................................................................................... 2
1.2 Determining The Dominant Thermal Conduction Path Of Selected Package Types ..................... 3
1.3 Determining Thermal Conduction Through The PCB .......................................................... 7
2 Design Guidelines For Air Temperature Measurement ............................................................... 11
2.1 Ground Plane Considerations ................................................................................... 11
2.2 Partitioning the PCB .............................................................................................. 13
2.3 Isolation Island .................................................................................................... 14
2.4 Perforation.......................................................................................................... 15
2.5 Edge Connector ................................................................................................... 16
2.6 Controlling the Thermal Mass of the PCB ..................................................................... 17
3 Design Guidelines For Component Temperature Measurement ..................................................... 18
3.1 Location............................................................................................................. 18
4 Summary .................................................................................................................... 20
Trademarks
All trademarks are the property of their respective owners.

SNOA967A – July 2017 – Revised January 2019 Temperature sensors: PCB guidelines for surface mount devices 8
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
Introduction www.ti.com

1 Introduction

1.1 What Is Heat Conduction?

There are three methods of heat transfer: heat conduction through solids, heat convection through fluids
and gases, and heat generated by radiation. This report focuses on heat conduction as it dominates the
heat transfer in PCBs and is therefore most relevant to temperature measurements.
Heat conduction is defined as the transfer of heat through a volume or a body. Heat is transferred through
microscopic collisions of particles; the more collisions, the hotter the object is. Heat transfer occurs when
there is a temperature difference between two objects or between different areas of an object, and its rate
depends on the geometry, thickness, and material of the object. Due to the law of equilibrium, heat
transfers from a hotter body to a colder body until the whole system reaches final equilibrium, as shown in
Figure 1. There is no net heat transfer between two objects that are equilibrium temperature. The equation
for heat transfer through conduction is shown in Equation 1
Q T2 T1
kA
t d
where
• Q/t: The rate of heat transfer [J/s]
• k: the thermal conductivity of the material [W/m×K]
• A: Surface of the contact area [m2]
• ΔT: The temperature difference of T1 temperature of one object and T2 temperature of the other [K]
• d: The thickness of the material [m] (1)

Figure 1. Thermal Conduction Model

Thermal conductivity (k) is the measure of a material's capability to conduct heat. It is used to describe
how heat conducts through a material. Metals are highly thermally conductive whereas materials like air,
wool, paper, or plastic are poor conductors of heat. Materials with a very low thermal conductivity, such as
polystyrene foam, act like a thermal insulator.

9 Temperature sensors: PCB guidelines for surface mount devices SNOA967A – July 2017 – Revised January 2019
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
www.ti.com Introduction

The materials that are most relevant to thermal analysis of PCBs are copper, FR4, and solder mask.
Copper is an excellent conductor of heat; it conducts heat significantly faster than FR4. Table 1 lists the
thermal conductivities found in PCBs. The higher the value, the more efficient the material is in
transferring heat, which results in a shorter thermal response time. For low k values, the temperature
gradient between the source and the sensor can be significantly large and must be considered carefully
during layout.

Table 1. Material Thermal Conductivity Coefficients Of Selected Materials

Material Thermal Conductivity k [W/(m×K)]
Air 0.0275
Solder Mask 0.245
FR4 0.25
Gold 314
Copper 385
Silver 406

1.2 Determining The Dominant Thermal Conduction Path Of Selected Package Types
Surface mount temperature sensors offer several advantages over sensors with through-hole packages.
Advantages include a smaller package size with a low profile, convenient PCB placement, and ease of
assembly. However, SMT temperature sensors can be difficult to isolate because they have the tendency
to measure the PCB temperature rather than ambient air temperature. Therefore, special layout
techniques need to be employed if the objective of the temperature sensor is to measure the ambient
temperature rather than the PCB temperature. Local analog or digital temperature sensors determine
temperature by measuring their own die temperature. Therefore, it is important to understand the
dominant temperature conduction paths between the die of the temperature sensor and the object or
environment whose temperature is to be determined.
Heat is conducted primarily through the following paths:
1. The Die-attach pad (DAP), if present, provides the most dominant thermal path between the PCB and
the die
2. The leads provide the most significant thermal path if the package type does not include a DAP
3. The mold compound provides an additional thermal path, but due to its low thermal conductivity, any
heat transfer through the mold compound itself is slower than heat transfer through the leads or DAP.
The package type choice determines how quickly the temperature sensor can respond to changes in
temperature. Figure 2 shows the relative thermal response rates of different classes of selected SMT
package types that are used for temperature measurements.

Figure 2. Relative Thermal Response Rate (Typical)

SNOA967A – July 2017 – Revised January 2019 Temperature sensors: PCB guidelines for surface mount devices 10
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
Introduction www.ti.com

Package types without a mold compound (CSP, DSBGA) and packages with a DAP (QFN, DFN) are well
suited if a fast thermal heat transfer from the PCB is desired, while package types without DAP are better
in applications in which slower response rates are desired. A fast thermal response rate allows the
temperature sensor to respond to any temperature changes quickly and therefore provide an accurate
reading.
Sections Section 1.2.2 to Section 1.2.1 show cross sections of commonly used SMT package types for
Texas Instruments' temperature sensors.

1.2.1 Leadless Packages Without Mold Compound (CSP, DSBGA)

Wafer Chip Scale Package (WCSP) leads are Ball Grid Array (BGA) balls processed directly onto the die.
Heat from the BGA balls are directly transferred to the die instead of transferring over the pins or through
a die attach pad, as shown in Figure 3. Generally, this is the package type with the fastest thermal
response because there is no mold to heat up.

Figure 3. Heat Transfer WCSP (LMT70[YFQ], TMP103[YFF], TMP108[YFF]) Package Cross Section

1.2.2 Leadless Packages With Die Attach Pad (QFN, DFN)

Packages with a DAP, such as QFN and DFN packages, have a large exposed surface area through
which heat can transfer quickly. These package types will respond quickly to temperature changes of the
copper plane which the DAP is soldered onto. Because the die sits directly on top of the thermal pad, heat
can transfer rapidly from the thermal pad to the die.

Figure 4. Heat Transfer WSON (TMP116[DRV]) Package Cross Section

11 Temperature sensors: PCB guidelines for surface mount devices SNOA967A – July 2017 – Revised January 2019
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
www.ti.com Introduction

1.2.3 Leadless Packages Without Die Attach Pad (DFN)

Leadless packages without a DAP, such as the 2-pin DFN package of the LMT01 shown in Figure 5,
transfer most heat through the pins itself. A small package of this type can still respond to temperature
change quickly because of its small thermal mass of the mold compound.

Figure 5. Heat Transfer WSON (LMT01[DQX]) Package Cross Section

1.2.4 Packages With Leads (MSOP, SOIC, VSSOP, SOT563, SOT23)

Other packages such as SOIC8, MSOP8, SOT563, and SOT23 transfer most heat through their leads.
The leads transfer 60% to 70% heat to the die thermal sensor.

Figure 6. Heat Transfer MSOP8, SOIC8 (LM75[D], TMP75[D]) Package Cross Section

SNOA967A – July 2017 – Revised January 2019 Temperature sensors: PCB guidelines for surface mount devices 12
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
Introduction www.ti.com

Figure 7. Heat Transfer SOT563 (TMP102[DRL]) Package Cross Section

Figure 8. Heat Transfer SOT23 (LM71A[DBV]) Package Cross Section

13 Temperature sensors: PCB guidelines for surface mount devices SNOA967A – July 2017 – Revised January 2019
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
www.ti.com Introduction

1.3 Determining Thermal Conduction Through The PCB

Power hungry components can generate a significant amount of heat during operation, and a PCB
designer needs to have an understanding of how heat is conducted by the PCB. The layout of the PCB
affects the thermal conductivity and thus the temperature measurement. Understanding the total thermal
resistance of the PCB will help the PCB designer to determine whether it is necessary to use filled or
plated vias, use thicker copper plating or add additional copper layers to disperse heat quicker.

1.3.1 General Thermal Conduction Equation

Thermal resistance can be expressed by the following equation:
L
T
k u ACS
where
• θ is the thermal resistance [K/W]
• k is the thermal conductivity factor [W/(m*K)]
• L is the thermal path length [m]
• ACS is the cross sectional area in which heat is applied [m2] (2)
To calculate the thermal conduction through the PCB, the individual paths can be broken down and
analyzed separately. The main components are the thermal conduction through the PCB (see
Section 1.3.4 and Section 1.3.2), and the conduction through the via (see Section 1.3.5). The most
common materials in many PCB applications are FR4, copper, and soldermask materials. By applying]
Equation 2 to the perpendicular path for the appropriate PCB materials, the longitudinal path, and the
thermal flow through the vias individually, thermal conduction through the PCB can be modeled
accurately.

1.3.2 Longitudinal Thermal Conduction

Figure 9 shows the longitudinal conduction path of a PCB with the direction of the heat flow from the heat
source along the FR4.

Figure 9. Longitudinal Conduction Heat Flow

Figure 9 applies the general thermal conduction equation to a cuboid.

L L
T
k u ACS k u W u t
where
• L is the path length of heat flow
• W is the width and t is the thickness
• W x t = ACS is the cross sectional area where the heat is being applied
• L = W for square (3)

SNOA967A – July 2017 – Revised January 2019 Temperature sensors: PCB guidelines for surface mount devices 14
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
Introduction www.ti.com

1.3.3 Example: Determining The Dominant Longitudinal Thermal Conduction Path

Applying Equation 3 to a 1.6mm thick layer of FR4 of square dimensions (1m × 1m) results in a thermal
resistance of 2,500°C/W, as shown in Equation 4
L 1m o
FR4 &:
k×W×t W
0.25 o
u 1m u 2 u 1.6 u 10 3 m
mu C (4)
Two 1oz (35µm) copper planes of the same PCB would have the thermal resistance of 3,710°C/W, as
calculated in Equation 5
L 1m o
Cu &:
k×W×t W
385 u 1m u 2 u 35 u 10 6 m
m u oC (5)
While thickness of the copper plane is significantly thinner than the thickness of the FR4 layer, the ability
to transfer heat is in the same order of magnitude. This is because the thermal conductivity of copper is
approximately 1,500 times larger than the thermal conductivity of FR4. It can be seen that the copper
planes of the above example PCB transfer heat slightly slower than the significantly thicker FR4 layer.
In the longitudinal direction, several layers of the PCB need to be considered in parallel. Compared to a
PCB without copper floods, it is possible to almost double the heat transfer rate along the plane by adding
two 1oz layers of copper floods.
Because the different layers transfer heat in parallel, the effective total thermal resistance of the cross-
section is 1,494°C/W, as calculated in Equation 6
TCu u TFR4 3710 u 2500
Ttotal TCu || TFR4 1494qC / W
TCu TFR4 3710 2500 (6)

1.3.4 Perpendicular Thermal Conduction

Figure 10 shows the perpendicular conduction heat flow of a PCB.

Figure 10. Perpendicular Conduction Heat Flow

Equation 7 is the perpendicular conduction heat flow equation to determine the thermal resistance of a
material, as specified in Table 1.
t t
T
k u ACS k u W u L
where
• t is the path of heat flow (the heat flows through the thickness of the material) [m]
• W x L = ACS is the cross sectional area where the heat is being applied [m2] (7)

15 Temperature sensors: PCB guidelines for surface mount devices SNOA967A – July 2017 – Revised January 2019
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
www.ti.com Introduction

1.3.5 Thermal Conduction Through A Via

Calculating the thermal conduction path of a via (as shown in Figure 11) can be useful to determine if a
regular via suffices, a larger size or quantity of vias is required, or if vias need to be filled with a
conductive fill. A conductive fill transfers heat faster to the opposite side of the board, but also increases
PCB manufacturing cost.
D0 D1

Figure 11. Conduction Through Via

Equation 8 applies the general thermal conduction equation to a via.

L L
T
k u ACS § § D ·2 § D ·2 ·
k u S u ¨¨ 0 ¸ ¨ 1 ¸ ¸
¨© 2 ¹ © 2 ¹ ¸
© ¹
where
• L is the length of the via (the heat flows through the length of the cylinder) [m]
• D0 is the outer via diameter [m]
• D1 is the inner via diameter [m] (8)

1.3.6 Example: Determining The Dominant Perpendicular Thermal Conduction Path

In the perpendicular direction, a PCB designer may want to compare the thermal resistance of a via with
the equivalent thermal resistance of FR4 to determine if placing additional vias is a useful technique in
transferring heat quickly from one side of the PCB to the other.
The sidewalls of a non-tented, non-filled via with a 0.5mm drill hole size, a sidewall copper thickness of
35µm, and a length of 1.6mm have the thermal resistance of 81℃/W, as shown in Equation 9. Note that
the sidewall thickness of a via is often different from the copper plating thickness and depends on via
dimensions and the manufacturing process of the PCB manufacturer.
L L 1.6 u 10 3
o
&:
k u A CS
Cu
§§ D · 2
§ D1 ·
2
· W §§ 0.5 u 10 · 3 2
§ 0.43 u 10 · ·
3 2

k u Œ u ¨¨ 0 ¸ ¨ ¸ ¸¸ 385 u Œ u ¨¨ ¸ ¨ ¸ ¸
¨© 2 ¹ ¨©
© 2¹ mu C ¹ ¸¹
o
© ¹ ©
2 ¹ © 2
(9)
In order to obtain an accurate result, the thermal resistance of the air cylinder inside the via also needs to
be calculated and considered in parallel with the thermal resistance of the via sidewalls.
L L 1.6 u 10 3 o
&:
k u A CS
Cu
§§ D · 2 § D1 · ·
2
W §§ 0.43 u 10 3 · 2 §0· ·
2
k u Œ u ¨¨ 0 ¸ ¨ ¸ ¸ 385 u Œ u ¨¨ ¸ ¨ ¸ ¸¸
¨ © 2 ¹ ¹¸ m u oC ¨
©© 2 ¹ ©©
2 ¹ © 2¹ ¹
(10)

SNOA967A – July 2017 – Revised January 2019 Temperature sensors: PCB guidelines for surface mount devices 16
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
Introduction www.ti.com

Equation 10 shows that the thermal resistance of the air cylinder is greater than 400,000℃/W. Because it
is approximately 5,000 times as large as the thermal resistance of the via sidewalls, the thermal
conduction contribution of the air has a negligible effect and can be ignored, as is proven by Equation 11.
TCu u Tair 81u 400646
Tvia TCu || Tair 81qC / W
TCu Tair 81 400646 (11)
The air filled drill hole of a via does not contribute much to the heat transfer rate, so almost all of the heat
transfer of a standard via occurs through its sidewalls. However, vias in which the hole is filled with a
different material may benefit from the heat transfer contribution of that material. Some designs require
vias to be filled in order to transfer heat even faster than a normal via. Filled vias should be considered if
even multiple parallel standard vias do not provide a sufficiently fast heat transfer rate to meet system
specifications.
The thermal resistance of 81℃/W for the non-filled via from this example can be compared to the thermal
resistance of a solid FR4 cylinder of equal outer diameter to determine how much more effective a copper
via is in transferring heat from one side of the PCB to the other. Equation 12 shows that the thermal
resistance of an equivalently sized cylinder of FR4 is 32,595℃/W, which is approximately 400 times more
resistive than the thermal resistance of the via.
L L 1.6 u 10 3 o
&:
k u A CS
FR4
§§ D · 2
§ D1 ·
2
· W §§ 0.5 u 10 3 · 2 § 0· ·
2
k u Œ u ¨¨ 0 ¸ ¨ ¸ ¸¸ 0.25 u Œ u ¨¨ ¸ ¨ ¸ ¸¸
¨© 2 ¹ ©2¹ m u oC ¨ © 2¹ ¹
©© ¹
© ¹ 2
(12)
An air filled drill hole contributes negligible thermal transfer. However, because the thermal conductivity of
copper is approximately 1,500 higher than FR4, the via of above dimensions is able to transfer heat to the
opposite side of the PCB through the via sidewalls approximately 400 times faster than an FR4 cylinder of
the same outer diameter. Therefore, placing multiple parallel non-filled vias can be a very effective method
to transfer heat quickly from one side of the PCB to the other within a localized area.

17 Temperature sensors: PCB guidelines for surface mount devices SNOA967A – July 2017 – Revised January 2019
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
www.ti.com Design Guidelines For Air Temperature Measurement

2 Design Guidelines For Air Temperature Measurement

For some applications, it is desirable to design for a low thermal conductivity between the PCB and the
temperature sensors. This is desired when the ambient air temperature is to be measured, rather than the
temperature of the PCB or nearby components.
Examples of such applications include
• Thermostat ambient air temperature measurement
• Indoor and outdoor weather stations
• Wireless sensor node
In such systems, slow changes in air temperature are of interest, while heat sources such as a the heat of
a nearby processor would result in an inaccurate reading. The guidelines in sections Section 2.1 to
Section 2.6 docato-extra-info-title Edge connector Section 2.5.

2.1 Ground Plane Considerations

Due to the higher thermal conductivity of copper, running solid ground planes between other ICs and the
sensor will cause undesired heat transfer which should be avoided. It is best to avoid copper planes near
the temperature sensor that are connected to the copper planes of other ICs, as shown in Figure 12

Figure 12. GND Plane Does Not Extend To Temperature Sensor (Top View)

For even better results, create a separate a small copper plane on both sides of the sensor as shown in
Figure 13 and Figure 14 , and add several vias to thermally link the top and bottom planes together.
Because of the low thermal conductivity of the solder mask compared to copper (see Table 1), it is
advised to create a solder mask cut-out around the copper plane. This will allow the sensor to respond to
ambient air temperature measurements significantly faster than in systems in which the copper plane is
coated by solder mask. Add a physical gap between the plane around the sensor and the planes of the
rest of the PCB. Hatched GND planes in the main section of the PCB further reduce heat flow from other
ICs to the sensor.

SNOA967A – July 2017 – Revised January 2019 Temperature sensors: PCB guidelines for surface mount devices 18
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
Design Guidelines For Air Temperature Measurement www.ti.com

Figure 13. Hatched GND Plane (Top View)

Figure 14. Exposed Sensor GND Pad (Bottom View)

19 Temperature sensors: PCB guidelines for surface mount devices SNOA967A – July 2017 – Revised January 2019
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
www.ti.com Design Guidelines For Air Temperature Measurement

2.2 Partitioning the PCB

The temperature sensor should be in an area of the PCB that is as far away from the main heat
generating ICs, as shown in Figure 15. This can be achieved by placing the sensor in a corner of the
PCB, away from other components. Doing so will minimize the effect that other components on the board
have on the temperature reading.

Figure 15. Sensor Placed In PCB Corner

SNOA967A – July 2017 – Revised January 2019 Temperature sensors: PCB guidelines for surface mount devices 20
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
Design Guidelines For Air Temperature Measurement www.ti.com

2.3 Isolation Island

If feasible, a partial router trace around the temperature sensor creates an isolation island which greatly
reduces heat transfer from the main heat source to the sensor. Heat transfer is reduced because the
thermal conductivity of air compared is significantly lower than the thermal conductivity of FR4. An
example of an isolation island is shown in Figure 16.

Figure 16. Isolation Island Significantly Reduces Heat Transfer From Main Heat Source To Temperature
Sensor

21 Temperature sensors: PCB guidelines for surface mount devices SNOA967A – July 2017 – Revised January 2019
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
www.ti.com Design Guidelines For Air Temperature Measurement

2.4 Perforation
As an alternative to the isolation island discussed in Section 2.3, it is possible to add a perforation around
the section with the temperature sensor, as shown in Figure 17. Doing so greatly minimizes the amount
heat transfer through the FR4 material. An example of a perforated PCB is the TMP116 Evaluation
module.

Figure 17. Perforation Reduces Heat Transfer From Heat Source To Temperature Sensor

SNOA967A – July 2017 – Revised January 2019 Temperature sensors: PCB guidelines for surface mount devices 22
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
Design Guidelines For Air Temperature Measurement www.ti.com

2.5 Edge Connector

A miniature PCB that contains only the temperature sensor and is mounted using an edge connector to
the main PCB is a highly effective method for avoiding significant heat transfer from the main PCB to the
temperature sensor. The edge connector should ideally be mounted at a location away from major heat
sources on the main PCB so that radiated heat from ICs does not interfere with the temperature reading.
This technique is illustrated in Figure 18

Figure 18. Dedicated Sensor PCB With Edge Connector

23 Temperature sensors: PCB guidelines for surface mount devices SNOA967A – July 2017 – Revised January 2019
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
www.ti.com Design Guidelines For Air Temperature Measurement

2.6 Controlling the Thermal Mass of the PCB

The thermal mass is a material's ability to store heat energy. A material with a high thermal mass will
respond to temperature fluctuations more slowly than one with a lower thermal mass. To keep the thermal
mass of the PCB as small as possible, it is advised to use a thin PCB (e.g. 0.8mm rather than the
standard 1.6mm FR4 thickness), or even place the temperature sensor on a flex PCB as shown in
Figure 19. When combined with either one of the techniques for reducing PCB surface area (see
Section 2.3 and Section 2.4), a thin PCB can correspond to changes in air temperature much more rapidly
than a large, thick PCB with a high thermal mass.

Figure 19. Dedicated Sensor Flex PCB With Edge Connector

SNOA967A – July 2017 – Revised January 2019 Temperature sensors: PCB guidelines for surface mount devices 24
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
Design Guidelines For Component Temperature Measurement www.ti.com

3 Design Guidelines For Component Temperature Measurement

In many scenarios, system designers want to monitor the die temperature of a power hungry IC such as a
MCU, GPU, ASIC, FPGA, DSP, or CPU in order to dynamically adjust its performance, control fan speed
in the system, or initiate a safe system shutdown. Using a remote temperature sensor such as TMP46x,
TMP43x or TMP411 is the preferred method for monitoring temperature of ICs as long as they have a
suitable integrated temperature diode. Guidance on product selection for remote temperature sensors is
available here. Application note Optimizing Remote Diode Temperature Sensor Design contains details on
compensation techniques and layout practices for remote temperature sensors.
If the IC does not contain a suitable temperature sense diode or if a remote sensor cannot be used for any
other reason, then a local sensor or an external diode can be used instead. The following sections provide
guidelines on obtaining the most accurate measurement and fastest response of a local sensor that
monitors the temperature of another IC.

3.1 Location
The sensor location should be chosen to be as close as possible to the heat source that is to be
monitored. Avoid any perforations or slits in the PCB between the IC and the temperature sensor, as they
will reduce the thermal response.

3.1.1 Bottom-Side Mounting

If possible, mount the temperature monitor on the bottom side of the PCB, directly below the heat source,
as shown in Figure 20. As explained in section Section 1.3.5, vias are a highly effective method for
transferring heat quickly from one side of the PCB to the other because of the superior thermal
conductivity of copper compared to FR4. Therefore, using as many parallel vias as feasible or using filled
conductive vias to transfer heat from the heat source to the temperature monitor creates a fast thermal
equilibrium between the two ICs. A QFN or DFN package with a DAP further helps to decrease the
thermal resistance path between the vias and the sensor die.

Figure 20. Sensor Mounted On Opposite Side Of Heat Source; Multiple Vias Ensure Fast Heat Transfer

25 Temperature sensors: PCB guidelines for surface mount devices SNOA967A – July 2017 – Revised January 2019
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
www.ti.com Design Guidelines For Component Temperature Measurement

3.1.2 Ground Plane Considerations

If it is not practical or cost effective to place the temperature sensor on the opposite side of the heat
source, place it on the same side as close as possible to it, as shown in Figure 21. The most effective way
of creating a thermal equilibrium between the heat source and the temperature monitor is through the
ground plane. Use a solid ground plane that extends from the heat source to the temperature sensor.

Figure 21. Shared GND Plane Helps With Thermal Equilibrium

SNOA967A – July 2017 – Revised January 2019 Temperature sensors: PCB guidelines for surface mount devices 26
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
Summary www.ti.com

4 Summary
When designing a PCB with a temperature sensor, the system designer needs to consider if the objective
is to measure ambient air temperature or to monitor the temperature of a nearby power hungry IC. This
application note discusses the background and layout techniques for both objectives. For ambient air
temperature measurements, physical isolation between the sensor and heat generating components on
the same PCB is critical. Additionally, consideration of thermally conductive paths such as GND planes
play an important role to ensure that nearby components do not cause false ambient temperature
readings. In contrast, measuring the die temperature of ICs requires careful consideration of sensor
location and a path with high thermal conductivity to create a fast thermal equilibrium between the sensor
and the heat generating IC.

27 Temperature sensors: PCB guidelines for surface mount devices SNOA967A – July 2017 – Revised January 2019
Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
www.ti.com Revision History

Revision History
NOTE: Page numbers for previous revisions may differ from page numbers in the current version.

Changes from Original (July 2017) to A Revision ........................................................................................................... Page

• Corrected Figure 8 ........................................................................................................................ 6

• Changed Equation 4 ...................................................................................................................... 8
• Changed Equation 5 ...................................................................................................................... 8
• Changed Equation 9 ...................................................................................................................... 9
• Changed Equation 10..................................................................................................................... 9
• Changed Equation 12 ................................................................................................................... 10

SNOA967A – July 2017 – Revised January 2019 Revision History 28

Submit Documentation Feedback
Copyright © 2017–2019, Texas Instruments Incorporated
Chapter 3: Temperature sensing application examples

Overview • Simple temperature limit alerts. In some applications,

continuous temperature collection is unnecessary. However,
Temperature sensing is crucial in a wide variety of
it’s critical for systems to stay above or below a temperature
applications for:
threshold to prevent thermal damage. TI’s temperature
• Protection: instant notifications when a system crosses a switches and digital temperature sensors enable simple
thermal limit in order to prevent damage or maintain safety. temperature monitoring to detect when a temperature
• Monitoring and controlling system temperature: to take crosses a limit with hysteresis. An external resistor (either pin-
appropriate corrective actions. programmable, factory-programmed or software-configurable
over I2C) can select threshold trip points.
• Calibration: to compensate for temperature-sensitive
• Cold junction compensation. Temperature drift factors into
sensors and applications while maintaining high accuracy
correcting temperature shifts in a system. Temperature
and functionality.
affects everything from passive components (resistors,
Chapter 3 discusses several temperature sensing application capacitors) to active components (amplifiers, data converters,
examples and the implementation of TI’s temperature sensor references, clocks) to optical components
portfolio in these applications: (intensity changes, spectral shifts, sensitivity, noise).
• Multipoint temperature monitoring. It might be Cold junction compensation increases the temperature
necessary to place temperature sensors in multiple accuracy of a thermocouple. TI’s highly linear, high-accuracy
locations to monitor systems and subsystems. For example, integrated circuit (IC) temperature sensors provide feedback
temperature sensors are primarily placed on PCBs next to correct temperature shifts in precision systems and
to components that need measuring, or mounted to heat condense the number of additional components.
sinks, in the cavity beneath a socketed processor, or daisy- • Processor die temperature monitoring. The effective
chained together in a cable. Designers can also use remote monitoring of processor die temperature requires high-
temperature sensors to monitor the temperature across a temperature accuracy and optimal sensor placement.
P-channel N-channel P-channel bipolar junction transistor Integrated thermal diodes or external temperature sensors
and collect measurements from several PCB locations are more accurate alternatives to built-in low-accuracy
simultaneously. internal sensors. Remote temperature sensors have inputs
• Body temperature sensing. Understanding a patient’s for direct connections to the processor’s integrated diodes.
temperature is a critical first step in clinical diagnosis and If there are no integrated diodes, it is possible to set high-
an important concern for athletes. Wearable temperature accuracy IC sensors external to the processor. These
monitors have design trade-offs between power usage, alternatives maximize temperature accuracy and maintain
size, system performance (radio frequency, accuracy), and high reliability over time.
comfort. Beyond ultra-high accuracy, industry trends favor • Ultra-high accuracy temperature sensing. Temperature
compact, wearable form factors. TI’s temperature sensors monitoring is perhaps one of the most important parameters
(up to 0.1°C accuracy) meet American Society for Testing to know for a system, yet it can be difficult to measure
and Materials E1112 standards for medical thermometers accurately. These challenges are due to factors that can
and can help keep battery-powered wearables compact influence the accuracy of a temperature sensor like sensor
and comfortable. self-heating, calibration drifts and distance from heat sources.
TI’s temperature sensors were designed to overcome these
• Fluid temperature measurement. Many metering and
challenges and provide accurate temperature readings.
industrial processes must have either a direct measure
Engineers can leverage these devices to more easily
of a fluid’s temperature or use temperature data for
incorporate temperature sensing into existing system
compensation to more accurately calculate the volumetric
designs.
rate of flow. Fluid temperature monitoring requires small
sensors (to reduce resistance in the flow) that consume
low power (for cases with flammable fluids). TI’s low-power
analog and digital solutions have an accuracy comparable
to Class-AA resistive temperature detectors, draw as little
as 6.3 μW and reduce the amount of cabling compared to a
four-wire resistive-temperature detector configuration.

An Engineer’s Guide to Temperature Sensing Texas Instruments Incorporated

www.ti.com TI Tech Note

Application Brief
Efficient Cold Chain Management With Scalable High­
Accuracy Temperature Sensors

From producers to consumers, it is important that temperature ranges. Based on the application, there
perishable items, especially food and medicines, are different topologies that may be deployed. In the
reach the end consumer in fresh and viable condition, point to point topology like that represented in Figure
so as to maintain their nutrients and efficacy. To 2 , a single microcontroller (MCU) is connected to a
ensure quality and product safety, manufacturers temperature sensor that may be an analog output or a
specify the temperatures at which the items must be digital output sensor. This is useful when managing a
transported and stored. pallet of goods or a temperature controlled container
during shipping.
Before reaching the consumer at their local grocery,
perishable produces like fruits, vegetables or frozen The TMP116 temperature sensor operates from -55°C
meals, spend a significant time in transportation and to +125°C with 0.2°C accuracy from -10°C to +85°C.
on the shelves of large refrigeration units as shown in It contains I2C and SMBus-interface communication,
Figure 1. The same is true of pharmaceuticals such as well as an integrated EEPROM memory. There is
as vaccines. Thus it becomes crucial that these items no calibration required for the TMP116, and minimal
be maintained at the correct temperature. current is consumed which minimizes self-heating.
The TMP116 is typically found in applications with a
Cold chain management and Good Distribution
heavy focus on high accuracy.
Practices (GDP) ensure that the right conditions
are met during every phase of the life cycle of Wireless Communication

packaged and perishable items. At the same time it

ensures that anytime a possible excursion outside the
storage temperature is about to occur, an appropriate + Temperature
Battery MCU
action can be taken by the operator either during
_ Sensor

transportation or during storage to ensure that there

is as little wastage as possible. x

Data Logger

Figure 2. Point-to-Point Topology

When sensing multiple locations like display in

refrigerators or in reefer containers, the cost of a
single MCU is too high to be implemented multiple
times in the entire system. In such cases, the
most common topologies that are used are the star,
daisy chain (Figure 3) or shared bus, with one
MCU being the host controller for multiple sensors.
A star topology allows easy fault isolation if one
Figure 1. Typical Grocery Aisle branch fails and may use both analog and digital
output temperature sensors, but has a higher cost of
Cold Chain Topology implementation as the controller peripheral count is
Use of temperature sensors with gauges and simple higher because of which the system cannot scale very
analog sensors, have been quite common for a well and the cost of assembly and cable itself.
long time. However, with advances in semiconductor On the other hand, with shared bus, the scalability
technology and the fact that most cold chain can easily be addressed with digital temperature
management is done in the temperature range of – sensor that share the line and may be individually
40°C to +10°C, integrated temperature sensors are addresses using in band addressing like the case of
the best option for cold chain management in these

SNOAA33A – MARCH 2019 – REVISED JANUARY 2022 Efficient Cold Chain Management With Scalable High-Accuracy Temperature 30
Submit Document Feedback Sensors
Copyright © 2022 Texas Instruments Incorporated
TI Tech Note www.ti.com

I2C bus or out of band signaling using chip select communication interface of SMAART Wire ™ uses
which is the case with SPI. However, reliable power UART bus which is a standard peripheral on almost
delivery and signal integrity over a long chain may be all MCU, making it easier develop software, than
a concern. using bit-banged approach. At the same time, with
a daisy chain implementation, it makes it easier to
The daisy chain does not require out of band signaling
identify the location of a cable break, which enables
and rather uses in band addressing scheme. As each
easy maintenance and overall system reliability.
stage of the chain acts as a buffer for the next
chain, signal integrity may be maintained over longer
distances.
Wireless Communication

+ Temperature Temperature
Battery _ MCU Sensor 1 Sensor 2

Figure 4. Eye Diagram for TMP107

Temperature
Data Logger Sensor N The current consumption of the TMP107 when
performing temperature conversions with an active
bus communication is typically 300 µA. It has a
Figure 3. Daisy Chain of Temperature Sensors
shutdown current of 3.8 µA in low power mode.
Irrespective of which stage of cold chain management With a wide operating voltage of 1.7 to 5.5 V, the
is being monitored, electronic systems provide a low current consumption makes it ideal for battery
unique advantage of not only logging the temperature operated systems during transport phase of cold
of the pallet or refrigeration unit, but also providing chain management. At the same time the baud rate
thresholds that generate alert above a certain can be increased for more real time update as may be
threshold. Such events can be visually communicated the case when storing frozen food items.
through audio or visual alerts like a buzzer of flashing Additionally, TMP107 allows the configuration and
LED, but also can be integrated into cloud services temperature limits to be stored to its internal non-
using either wired or wireless MCUs, allowing round volatile memory. This enables the device to be
the clock monitoring and data logging. automatically configured on power up, eliminating
Daisy chain topology in cold chain management the need for individual device configuration to be
performed making the system operational faster. It
The TMP107 is a digital output temperature sensor also has 8 EEPROM locations providing up to 128 bits
that supports a total of 32 daisy-chained devices of EEPROM to store user information or calibration
and is ideal for replacing NTC thermistors in cold information.
chain management, because of its high accuracy and
ease of system wide scalability without the need to In conclusion, different types of temperature
add additional MCUs. The TMP107 has a maximum measurement solutions can be used to efficiently
accuracy specification of ±0.4°C in the range of – track temperature in cold chain applications. The
20°C to +70°C and ±0.55°C in the range of –40°C to TMP116 could be ideal for monitoring a single
+100°C with a temperature resolution of 0.015625°C. location and has accuracy suitable for both produce
and pharmeceuticals, while the TMP107 has the
With an automated address assignment, the TMP107 right combination of accuracy, power consumption
allows system developers to write software without and features to support a battery-based cold chain
the need to assign the address at each sensor node management system where multi-point monitoring is
as the system is scaled by adding additional sensor needed.
node. At the same time, with the use of a push-pull
communication IO, the system is made more resilient Table 1. Device Recommendations
Device Optimized Parameters Performance Trade-Off
against the noise affecting the temperature value over
long cables. This allows for data transfers over span 0.4°C accuracy and 32 0.4°C accuracy may not be
TMP107
devices in daisy chain enough for pharmeceuticals
lengths of 1000 feet between adjacent devices in the
16 devices in daisy chain and
chain. TMP144 Small form factor
1°C accuracy

Figure 4 shows the signal integrity of the TMP116 0.2°C accuracy across No daisy chain capability
temperature
communication interface at 9600 bps. The

31 Efficient Cold Chain Management With Scalable High-Accuracy Temperature SNOAA33A – MARCH 2019 – REVISED JANUARY 2022
Sensors Submit Document Feedback
Copyright © 2022 Texas Instruments Incorporated
Design Challenges of Wireless Patient Temperature
Monitors

element and the other devices.

Introduction
2. Minimize thermal mass surrounding the
Monitoring patient vitals in a clinical environment has temperature sensing element for faster response.
traditionally been a job for expensive and heavily
calibrated systems requiring patients to be tethered to 3. Provide good thermal contact between the patient
bedside monitors. Wireless patient monitoring systems and the sensing element to minimize the
provide for both patient comfort and clinical temperature gradient between the sensor and the
convenience, provided they can still be made to target.
operate within strict medical standards. Optimizing Thermal Isolation & Mass
In the case of wearable temperature monitors, there For optimal thermal mass and isolation, the
are many design tradeoffs that must be considered for recommendations in Layout Considerations for
power consumption, size, system performance (in Wearable Temperature Sensing should be followed.
terms of both RF and accuracy), and patient comfort. Figure 1 shows an example of this for a skin
Thinner flexible batteries provide greater comfort but temperature monitoring system. The TMP117
may require more careful power management. measuring temperature is extended from the rest of
Smaller, lower cost, designs suffer in terms of thermal the PCB using a narrow arm to minimize thermal
isolation and RF performance. Optimal solutions for conductance from the rest of the board. Figure 2
long term monitoring must make good use of board shows the stack up for the same 2-layer flex PCB.
area to improve accuracy and signal integrity, while Using a flex-board also helps to reduce total thermal
trying to keep current consumption as low as possible. mass, which improves thermal response time of the
System designers will have to balance these patient monitor. Copper fills between the top and
requirements alongside the comfort and experience of bottom side of the board should be ommitted to avoid
the patient. drawing heat away from the TMP117, and increasing
Standard Compliance for Thermometers the thermal mass.

The governing standards for intermittent electrical

patient thermometers are given in ASTM E1112 and
ISO-80601-2-56. For standard compliant clinical
temperature measurement applications under ASTM
E1112, human body temperature monitors must
produce readings within ± 0.1 °C accuracy, and must
read and display temperature from a minimum of
35.8°C to 41.0°C. At a bare minimum, any temperature
monitoring design should include a sensing element
able to meet these requirements after calibration.
Figure 1. TMP117 (U1) on Flex PCB, using an
TI recommends using the TMP117 ultra-high-accuracy extended arm isolates the IC from being influenced
digital temperature sensor for this purpose. The device by heat from other devices.
itself has better than 0.1°C accuracy from 25 to 50 °C
and requires no calibration to exceed the requirements Thermal Contact
of both ASTM E1112, and ISO-80601. Additionally, the
TMP117’s low overall current draw and one-shot mode Reliable measurement of the patient's skin
are ideal for battery operated applications. The digital temperature requires good thermal contact between
I2C output of this device also greatly simplifies system the patient to be monitored and the sensing device.
design when compared with RTD or thermistor based This works in conjunction with the thermal isolation
solutions. from the rest of the board to ensure that the
temperature being reported is as close as possible to
Layout Considerations the actual skin temperature of the patient. With the
Even with an appropriate sensing element, ensuring TMP117, a solid copper pour and contact vias can be
total system accuracy will still require care in terms of used to provide a thermal path from the underside of
layout. For monitoring skin temperature an ideal layout the board as shown in Figure 3. The pad contacts the
will do all of the following: wearer's skin directly, and ensures the primary heat
source for the device is from the person to be
1. Maximize thermal isolation between the sensing monitored.

SNOAA07 – August 2018 Design Challenges of Wireless Patient Temperature Monitors 32

Submit Documentation Feedback
Copyright © 2018, Texas Instruments Incorporated
 www.ti.com

System Power
Power requirements will vary based on overall system
design, but most wireless patient monitors will need to
have enough energy storage for several years of shelf
life, and at least 48-72 hours of active life. Coin-cell
batteries can easily exceed these requirements for
energy capacity, but they are entirely rigid and may be
uncomfortable to device wearers. In the case of
patches which are not intended to be reused, a coin-
cell based solution can also be extremely wasteful.
An alternative option for energy storage is to use thin-
film, flexible batteries. Due to small storage capacities,
using these batteries will require that total system
power consumption be minimal. For only intermittent
temperature monitoring, systems powered with flexible
batteries can exceed the requirements for multiple
years of shelf-life and 48-72 hours active time. The
Figure 2. Example flex layer Stack, thickness design trade off between current consumption and
should be minimized to reduce thermal mass. additional features must be made by the system
designer.
Making System Tradeoffs
If the system is required to meet the requirements of
ASTM E1112 and ISO ISO-80601-2-56 following the
recommendations on layout is essential, but there are
other system design considerations to be made. For
patient comfort, non-temperature-sensing devices and
the RF region should be kept in as small an area as
possible. Keeping the populated region of the board
compact will reduce the portion of the monitor which
feels rigid to the wearer.
For RF communication, any wireless protocol that can
be made to work on a flex PCB is acceptable. Since
most wearable patient monitors will want to keep
Figure 3. Copper pour underneath TMP117 (Left), power consumption low, a BLE wireless
topside layout for TMP117 (U1, Right). The vias
underneath the TMP117 and the copper pour communication link is recommended. If the information
provide a thermal path between the patient's skin being transmitted from the monitor is only temperature,
and the device. the monitor can be configured to broadcast the
temperature reading alongside it's pairing ID. Sending
the information in this manner removes the
Self-Heating requirement for an actual connection to be made and
Regardless of the choice of sensing element and maintained, and will reduce system power
proper layout, the stringent accuracy requirements for consumption even further.
medical thermometers will require that device self- To get more information on these topics, or for general
heating be taken into consideration. Some self-heating tips when measuring temperature, please see the
will always be present from the resistive losses of the additional resources linked to in Table 1.
chosen sensing element. The TMP117, may be
configured for one-shot mode conversion and be kept Table 1. Related Materials
in shutdown mode between successive reads, to
minimize self-heating. Individual temperature readings Document Type Description
(using a configurable number of averaged readings) Wearable Temp-Sensing Layout
can be triggered using the one-shot feature of the Application Report Consideration Optimized for Thermal
device. Human body temperature will not Response
conventionally exhibit change on the order of seconds, Layout Considerations for Wearable
Tech Note
so taking these readings at 10 to 60 second intervals Temperature Sensing
is sufficient to monitor patient temperature over long Precise Temperature Measurements with
Tech Note
periods. This method has the added benefit of TMP116
extending the active-battery life of the system.

33 Design Challenges of Wireless Patient Temperature Monitors SNOAA07 – August 2018

Submit Documentation Feedback
Copyright © 2018, Texas Instruments Incorporated
RTD replacement in heat meters using digital temperature
sensors

A heat meter is a device which measures thermal RTDs in Heat Meters

energy by measuring the flow rate and the change of
Solid-state heat meters are gaining popularity in heat
temperature between the inlet and outlet of the
energy billing for residential and industrial users.
system. These devices are very commonly found in
These meters come with both flow measurement on
both industrial plants to measure boiler outputs and in
either the inlet or outlet pipe and a pair of matched
residential heating and cooling systems to measure
RTD temperature probes on both inlet and outlet
the heat delivery.
pipes. Figure 2 shows a block diagram of a heat meter
Since the measurement of thermal energy requires system using RTDs.
both flow rate and temperature, it is imperative that
they be measured accurately. Accuracy is important
because an inaccurate measurement can result in
under billing or over billing.
To avoid measurement errors, a resistance
temperature detector (RTD) is commonly used. An
RTD is a passive component whose resistance
changes with temperature. RTDs are made using
metals—such as platinum, copper, or nickel—and
support a wide temperature range (approximately
Figure 2. Fluid Temperature Measurement With
–200°C to +850°C). The accuracy of a Platinum RTD RTD
is defined by its class. The IEC/EN60751 standard
defines four RTD classes—Class C, B, A, and
AA—where Class C is the least accurate and Class Low-power and high-accuracy RTDs in heat meters
AA is the most accurate. The lower accuracy classes are desired because heat meters are standalone,
will have a larger temperature range. For example, a battery-powered systems in most residential units. The
Class C thin-film RTD covers the temperature range of system’s ability to quickly wake from power-off mode,
–50°C to +600°C, while a Class AA thin-film RTD sample the RTD temperature, and return to power-off
covers the temperature range of 0°C to +150°C. mode extends battery life and minimized energy
consumption.
Most RTD applications use a current source to excite
the RTD element and create a voltage difference However, these system require that the RTDs are well-
across the RTD, as shown in Figure 1. This voltage is matched and have matched traces to read the
proportional to the resistance of the RTD and the differential measurement correctly. At the same time,
excitation current. The voltage potential is amplified, the system cost and complexity requires careful design
converted to a digital output by an ADC, and then fed consideration and costly calibration. High-accuracy
into an MCU where a lookup table is used to convert digital temperature sensors like the TMP117 can
the digital output to temperature. provide a cost-optimized, yet equally accurate
replacement for RTDs.
Replacing RTD With TMP117 Digital Temp Sensors
The TMP117 is a digital temperature sensor designed
for low-power, high-accuracy applications. The device
provides a 16-bit temperature result with a resolution
of 0.0078°C, along with a typical factory-calibrated
accuracy of ±0.1°C across –20°C to +50°C with a
maximum accuracy specification of ±0.3°C over the
temperature range of –55°C to +150°C, which exceeds
the accuracy of a Class AA RTD in the same range.
Figure 3 illustrates the accuracy specification of the
Figure 1. Basic RTD Circuit TMP117 vs an RTD. The graph clearly shows that
accuracy specification for the TMP117 can easily
SPACER outperform that of a Class AA RTD.
SPACER

SNOAA28 – March 2019 RTD replacement in heat meters using digital temperature sensors 34
Submit Documentation Feedback
Copyright © 2019, Texas Instruments Incorporated
 www.ti.com

The TMP117, with a comparable accuracy as the

0.5
TMP117 Class AA thin-film RTD while consuming only a
0.4 Class AA RTD fraction of the power required for temperature
0.3 measurement, is designed for a variety Heat Meter
Temperature Error (qC)

0.2 applications.
0.1 Figure 4 shows the same block diagram of Heat Meter
0 system using the TMP117 to replace the RTDs. By
-0.1
using the TMP117 instead of an RTD, designers can
simplify both their software and system architecture to
-0.2
save time, board space, and costs.
-0.3
-0.4
-0.5
0 20 40 60 80 100 120 140
Temperature (qC) Accu

Figure 3. TMP117 vs. RTD Class-AA Accuracy

The TMP117 features a shutdown mode where the

device aborts the active running conversion and enters
a low-power shutdown mode where it typically
consumes 250 nA of current. It can perform quick,
15.5-ms temperature conversions using the one-shot
conversion mode an active current as low as 3.5 μA
for a duty cycle of 1 Hz. After completing a one-shot Figure 4. Fluid Temperature Measurement With
TMP117
conversion, the device returns to low-power shutdown
mode.
Overall, the high-accuracy TMP117 temperature
Also, the device features an offset register that sensor with digital interface, fast conversion, and
automatically applies a user-defined offset to the extremely low-power shutdown mode eliminates the
measurement results prior to an MCU read. As the need for multiple narrow-tolerance discrete
TMP117 provides additional simplicity over an RTD, it components and integrated devices, which can save
eliminates the need for costly calibration, external PCB space, complexity, and cost in Heat Meters.
circuitry, matched traces, and Kelvin connections
easing the system designers task for accurate COLLATERAL DESCRIPTION
measurement. ±0.1°C Accurate Digital Temperature
TMP117
Finally the TMP117 features a fast mode (400 kHz) Sensor with Integrated NV Memory
I2C communication. These specs make the TMP117 TMP116
±0.2°C Accurate Digital Temperature
excellent for low-power consumption requirements, as Sensor with Integrated NV Memory
well as quick power on-off cycling necessary in Heat RTD Class-AA Replacement With High-
Meter systems. Application Report Accuracy Digital Temperature Sensors in
Field Transmitters
Replacing Resistance Temperature
Application Report
Detectors With the TMP116 Temp Sensor
Precise Temperature Measurements with
Application Report
TMP116

35 RTD replacement in heat meters using digital temperature sensors SNOAA28 – March 2019
Submit Documentation Feedback
Copyright © 2019, Texas Instruments Incorporated
How to protect battery power management systems from
thermal damage

Introduction
To protect battery management systems (BMS) from
With a growing demand for portable personal thermal damage, either discrete or integrated
electronics, battery power management systems are temperature-sensing solutions are used. A discrete
becoming increasingly important for stringent design, solution consists of a thermistor, a comparator, and a
reliability, and safety requirements. Nowadays, voltage reference as shown in Figure 1. This approach
customers expect their personal electronics to have a provides real-time thermal protection without
longer battery life, a shorter charge time, and a smaller interrupting the control processing system. Since
form factor. The increased charge and discharge battery applications require protection at both hot and
currents, as well as the smaller form factor, make the cold temperatures, a temperature window comparator
battery packs vulnerable to thermal damage. In is better solution. An example of this output is
addition, different battery technologies have different displayed in Figure 2. In this example, the trip points
charging and discharging requirements that are are set to 60°C and 0°C with a 10°C hysteresis. Note
sensitive to temperature as shown in Table 1. that the Set Output High (SOH) is a system diagnostic
Typically, batteries can be discharged over a wider test feature that allows the user to force the output
temperature range, but the charge temperature is high independent of the temperature. The specific
limited. Note: fast charging can be done safely if the implementation depends on the application
cell temperature is kept between 10°C and 40°C. requirements:
These temperature limits are tied to the battery cell
chemistry due to its temperature dependent chemical • Features
reaction. If charged too quickly, the cell pressure can • Cost
build up and may lead to venting and reduced battery • Footprint
life. If the operating temperature is too high, cell
degradation can occur and may result in thermal • Power
runaway and explosion. On the other hand, if the • Accuracy
temperature is too low, irreversible cell chemical Some of the key features that customers typically look
reactions can occur and shorten battery life. Thus, for are hysteresis, trip point programmability, trip test,
battery temperature monitoring is very critical for qualifications (like automotive and UL, for example),
battery management systems. output type, channel count, and supply voltage range.
Table 1. Common Charge and Discharge
Temperature Limits for Various Batteries
TH = 60°C

Battery 50°C
Charge Temperature Discharge Temperature
Type Measured
Temperature
Lead Acid –20°C to 50°C –20°C to 50°C
10°C
NiCd, NiMH 0°C to 45°C –20°C to 65°C 5°C
TL = 0°C

Li-ion 0°C to 45°C –20°C to 60°C

Thermal Protection Solutions

Changes hysteresis
HYSTSET1 from 10°C to 5°C

HYSTSET0

Forces OUT pin high

SOH regardless of temperature

OUT

Time

Figure 1. Example Discrete Implementation of a Figure 2. Example Temperature Window

Temperature Switch Comparator Output Behavior

SNOA995A – June 2018 – Revised January 2019 How to protect battery power management systems from thermal damage 36
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
 www.ti.com

Discrete Solutions detection

It is quite common to see discrete implementations of • Various threshold programming options (resistor,
a temperature switch using a Negative Temperature pin programmable, factory preset)
Coefficient (NTC) thermistor since the use of these • Some parts also offer analog output
devices are well established. Furthermore, thermistor
solutions are often considered low-cost. However, due The highly integrated sensor lowers the solution cost,
to the demanding requirements of thermal protection which enables redundancy in safety applications.
like guaranteed performance, discrete solutions often TI provides a broad portfolio of temperature
prove to be challenging and costly. Some of the key switch/thermostats like the TMP303 and TMP390. The
challenges when designing a discrete thermal TMP303 uses a window comparator and offers design
protection solution are accuracy, reliability, and flexibility through an extra small footprint (SOT-563),
efficiency. low power (5 μA maximum) and low supply voltage
Due to the non-linear nature of NTC thermistors, capability (as low as 1.4 V). No additional components
maintaining an accurate trip point at high or low are required for operation and can function
temperatures is difficult without using precision independent of microprocessors or microcontrollers.
components that can increase system cost. Calibration Seven trip points are available through different device
is also not practical in these hardware-based switching options which can be programmed at the factory to
applications. In addition, discrete implementations any desired temperature. For applications that require
require multiple components to work together, which different values, contact your local TI representative.
can decrease system reliability. Lastly, NTC discrete The TMP390, as shown in Figure 3, is a resistor-
solutions dissipate a significant amount of power at hot programmable dual-output temperature switch with two
temperatures because the NTC resistance decreases internal comparators and two outputs. The TMP390 is
significantly. offered in the same small package, has ultra-low
IC Solutions: Temperature Switch/Thermostats power (1 μA maximum) and low supply voltage
capability (1.62 V). Both the hot and cold trip points
can be configured at any desired temperature window
with hysteresis options between 5°C and 30°C, using
just two resistors. The separate hot and cold trip
outputs generate independent warning signals to be
interpreted by the microprocessor.
For alternative device recommendations, refer to
Table 2. To learn more about batteries, PCB
guidelines, and protection, refer to the reference
material in Table 3.

Table 2. Alternative Device Recommendations

Device Optimized Parameters Performance Trade-Off
TMP708 Resistor Programmable Reduced Accuracy
Pin-programmable Increased power
TMP302
Figure 3. TMP390 Block Diagram temperature switch consumption
Two internal
comparators. Two Increased power
Integrated temperature switches, like the TMP303 and LM56
overtemp outputs and consumption
TMP390, are becoming more popular with battery one analog output
power management systems. These devices typically
have a temperature sensor, comparator, and voltage Table 3. Related Documentation
reference fully integrated in a single chip. These
reduce design complexity by autonomously making Literature Number Title
decisions providing real-time thermal protection BU-410: Charging at High and Low
without interrupting the control processing system. The Web Link
Temperatures
key advantages of these sensors are as follows: Temperature Sensors: PCB Guidelines for
SNOA967
• Autonomously enable thermal protection Surface Mount Devices
independent of control unit SNOA996
Protecting Control Systems From Thermal
Damage
• Zero software
• Guaranteed temperature accuracy for trip point with
hysteresis
• Simple & cost effective over / under temperature
37 How to protect battery power management systems from thermal damage SNOA995A – June 2018 – Revised January 2019
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
Temperature compensation using high-accuracy
temperature sensors

Field transmitters are used extensively in factory Because the thermocouple measures a differential
automation and control-to-sense process parameters temperature, the temperature at the cold junction must
like temperature, pressure, and flow rate, to name a be known to determine the temperature at the hot
few. The sensors used in the field transmitters are junction—this process is known as cold junction
mostly analog sensors that must be sampled compensation (CJC).
accurately using an analog front-end (AFE).
Platinum RTDs are popularly used to measure the
Due to the operational conditions that arise from the temperature of the cold junction due to their high
placement of the field transmitters, it may be subjected accuracy. Figure 2 shows a block diagram of a cold
to wide varying temperature conditions and hence, junction-compensated thermocouple system using an
require some form of temperature compensation. RTD reference.
Traditionally, accurate temperature sensors like
Platinum resistance temperature detector (RTD) are
used in such temperature compensation systems.
RTDs are most commonly used temperature-sensing
components in industrial applications where high
accuracy and longevity is desired.
Most RTD applications use a current source to excite
the RTD element and create a voltage difference
across the RTD, as shown in Figure 1. This voltage is
proportional to the resistance of the RTD and the
excitation current. The voltage potential is amplified,
converted to a digital output by an ADC, and then fed Figure 2. CJC Using RTD System Block Diagram
into an MCU where a lookup table is used to convert
the digital output to temperature. However, while the system block diagram looks quite
simple, there are a lot of considerations like noise,
self-heating, and placement considerations that must
be carefully analyzed. Also, RTDs are sensitive to
trace routing and the trace lengths must be matched.
A lot of these considerations may be optimized to
reduce complexity and cost by replacing RTDs with
temperature-sensing ICs like the TMP117 digital
temperature sensor.
Replacing RTD With TMP117 Digital Temp Sensors
The TMP117 is a digital temperature sensor designed
for low-power, high-accuracy applications. The device
Figure 1. Basic RTD Circuit provides a 16-bit temperature result with a resolution
of 0.0078°C, along with a factory-calibrated
RTDs in Cold Junction Compensation Systems performance of ±0.1°C across –25°C to +50°C or
±0.3°C across the full operating range of –55°C to
Thermocouples are temperature-sensing devices that +150°C, which exceeds the accuracy of a Class-AA
cover a large temperature range and are made by RTD. Figure 3 depicts the results of an oil bath
joining wires of two different metals. Due to seebeck experiment conducted on the TMP117. The graph
effect, the output voltage is approximately proportional shows that TMP117 can meet the accuracy of a thin
to the difference in temperature between the hot film Class AA RTD for a CJC application.
junction and cold junction.

SNOAA27 – March 2019 Temperature compensation using high-accuracy temperature sensors 38

Submit Documentation Feedback
Copyright © 2019, Texas Instruments Incorporated
 www.ti.com

As described earlier, the TMP117 is comparable in

0.5
TMP117 accuracy to the Class AA thin-film RTD and consumes
0.4 Class AA RTD a fraction of the power of a PT100 RTD, when used in
0.3 CJC applications. Figure 4 shows the block diagram
Temperature Error (qC)

0.2 for CJC systems using the TMP117 to replace the

RTDs. The systems using the TMP117 eliminates the
0.1
need for additional components, such as delta-sigma
0 ADCs, programmable gain amplifiers, and RC filters,
-0.1 than the systems using RTD elements reducing the
-0.2 overall system cost. At the same time it reduces
-0.3
complex layout considerations because of a digital
read out.
-0.4
-0.5
0 20 40 60 80 100 120 140
Temperature (qC) Accu

Figure 3. TMP117 vs. RTD Accuracy

The TMP117 features a shutdown mode where the

device aborts the active running conversion and enters
a low-power shutdown mode where it typically
consumes 250 nA of current mitigating the effects of
self-heating. When triggered by an MCU, the TMP117 Figure 4. CJC Using TMP117 System Block
can perform quick, 15.5-ms temperature conversions Diagram
using the one-shot conversion mode with active
current as low as 3.5 μA for a duty cycle of 1 Hz. The
output is a direct temperature read which does not COLLATERAL DESCRIPTION
require any linearization. After completing a one-shot ±0.1°C Accurate Digital Temperature
TMP117
conversion, the device automatically returns to low- Sensor with Integrated NV Memory
power shutdown mode. This simplifies the software ±0.2°C Accurate Digital Temperature
implementation versus an RTD, eliminating the need TMP116
Sensor with Integrated NV Memory
for calibration, external circuitry, matched traces, and RTD Class-AA Replacement With High-
Kelvin connections. Application Report Accuracy Digital Temperature Sensors in
Field Transmitters
The TMP117 also features a fast mode (400 kHz) I2C
Replacing Resistance Temperature
communication and an offset register that Application Report
Detectors With the TMP116 Temp Sensor
automatically applies a user-defined offset to the
Precise Temperature Measurements with
measurement results prior to an MCU read. These Application Report
TMP116
specs make the TMP117 excellent for low-power
consumption requirements in CJC applications for field
transmitters.

39 Temperature compensation using high-accuracy temperature sensors SNOAA27 – March 2019

Submit Documentation Feedback
Copyright © 2019, Texas Instruments Incorporated
High-Performance Processor Die Temperature Monitoring

complex circuitry that impact the accuracy of the

Introduction
integrated sensor. In such systems, the accuracy is at
Power management in high-performance processors, ±5°C. Systems designed with such lower accuracy are
such as CPUs, GPUs, ASICs, and FPGAs, is typically not able to maximize their performance their thermal
complex. With thermal monitoring, these systems can design limit.
not only initiate a safe system shutdown, but also

xxxx
leverage temperature to dynamically adjust Thermal design limit
performance. To enhance system reliability and

xxxx
Safety margin
maximize performance, it is often desirable to monitor
processor temperature. Higher temperatures can

xx

Performance
activate a cooling fan, modify a system clock, or,
should the thermal threshold of the processor be

1oC Accuracy
exceeded, quickly shut down the system completely.
Applications such as infotainment, ADAS, servers,

4oC Accuracy
notebooks, and aerospace and defense systems can
take advantage of these thermal monitoring
techniques.

Measurement Error
Figure 2. Performance of a System Can be
Enhanced Though High-Accuracy Thermal
Monitoring

Sensor Placement & Accuracy

The thermal behavior of processors is monitored either
through an integrated temperature sensor or thermal
diode, or alternatively through an external temperature
sensor. In some cases, both are used to maximize
performance and boost reliability of the system.
Integrated Temperature Sensor: BJT
Some high-performance processors include a bipolar
transistor (BJT) for temperature sensing. This has a
Figure 1. Motherboards With High Performance very predictable transfer function that is dependent on
Processors Typically Need Heak Sinks temperature. Remote temperature sensors use this
principle to measure the die temperature. The most
common BJT found in CMOS processes is a PNP.
Design Considerations in Die Temperature
Monitoring
1x 10x
For effective temperature monitoring, there are two Integrated PNP Transistor-Connected Configuration I1 I2
design considerations for high-performance
processors: (i) Temperature accuracy (ii) Sensor RS RS
Placement. The temperature accuracy of the
Internal and PCB Series Resistance
processor has a direct correlation to the sensor RS RS
Measured VBE 1x (IC1)
and VBE 10x (IC2)
location.
Processor performance can be maximized to drive the
overall system to its thermal design limit through high- Processor, FPGA, or
accuracy temperature monitoring. While most ICs have ASIC
built-in temperature sensors due to variations across
wafers and other various lots, the accuracy is not Figure 3. ΔVBE Measurement With Two Currents
consistent. High-performance processors include

SNOAA29A – February 2019 – Revised October 2019 High-Performance Processor Die Temperature Monitoring 40
Submit Documentation Feedback
Copyright © 2019, Texas Instruments Incorporated
 www.ti.com

The design of such systems can be challenging due to

the noise and error caused by BJT process variations. Hole drilled in heatsink
Thermal diode error sources can be from Ideality
Factor variation, series resistance, noise injection, and a

Beta Compensation. Processor/FPGA/GPU

Socket b c
Ideality Factor Variation— BJT-thermal-diode
characteristics are dependent on process PCB

geometry and other process variables. If the Figure 4. Sensor Placement for High-Performance
Ideality Factor n is known, the n-factor register Processor Temperature Monitoring
can be used to correct the n-factor error.
Alternatively, software calibration methods can i. Location a: Sensor located in a hole drilled into
be used to correct this in the desired the center of the heat sink of the microprocessor
temperature range. provides close proximity. The heat sink can be
Series Resistance— Due to the current source, any clipped to the processor or attached with epoxy,
resistance in the signal path appears as a and generally sits on top of the processor. These
voltage offset. Modern remote temperature typically require long leads, and as the heat sink-to-
sensors employ a series resistance cancellation microprocessor thermal connection degrades, the
algorithm that removes temperature error due to sensor data becomes incorrect.
resistance up to 1–2 kΩ. This enables robust, ii. Location b: Another potential location is in the
accurate measurements, even when coupled cavity beneath a socketed processor, with which
with RC Filters. assembly is straightforward. Given the sensor is
isolated from the airflow, ambient temperature has
Noise Injection— EMI or inductive coupling into the minimal impact on the sensor reading. Additionally,
remote junction PCB traces can cause error if the heat sink gets detached from the processor,
when diode traces run in parallel with high- the sensor indicates an increase in processor
frequency signal lines carrying high currents. temperature. That said, given the sensor
Tracing remote temperature sensors need to placement, the temperature differences between
consider this during board design. the sensor and the processor case are 5°C to
10°C.
Beta Compensation— Thermal Transistors
integrated into an FPGA or processor may have iii. Location c: The sensor can be mounted on the
Beta < 1. A remote temperature sensor with circuit board next to the MPU. While this is simple
Beta Compensation is specifically designed to to implement, the correlation between sensor
work with these transistors, and to correct temperature and MPU temperature is much
temperature measurement errors associated weaker.
with them. The Beta Compensation feature Device Recommendation
provides no benefit when used with a typical
discrete transistor. Footprint size is a factor in selecting a local
temperature sensor. The TMP112 with a 1.6 x 1.6 mm
Device Recommendation can be used in close proximity of the processor and
TMP46x devices offer high-accuracy temperature gain a good accuracy of 0.5°C. This can maximize
measurements to monitor up to eight remote BJTs, as performance when compared to a 5°C to 20°C that is
well as the local temperature. Many commercial typically observed in temperature sensors integrated
applications can benefit from such multi-channel within processors.
remote sensors. The TMP451-Q1 offers an automotive
qualified high-accuracy remote and local temperature Table 1. Key Collaterals
sensor for automotive applications. COLLATERAL DESCRIPTION

External Temperature Sensor Remote Temperature Optimizing Remote Diode Temperature Sensor
Sensor Design
Some processors have built-in temperature sensor. Local Temperature Temperature Sensors: PCB Guidelines for Surface
While the location is ideal, the built-in sensors make Sensor Mount Devices
them less accurate due to variations across the wafers Board Temperature
Monitoring Board Temperature
Monitoring
and different lots. Additionally, it is essential to trim the
processor based on a reference; this reference is
compared with die temperature to adjust the
coefficients. Given the complex circuitry, the
processors incur self-heating that builds an error. This
error increases with increases in temperature.

41 High-Performance Processor Die Temperature Monitoring SNOAA29A – February 2019 – Revised October 2019
Submit Documentation Feedback
Copyright © 2019, Texas Instruments Incorporated
 Application Report
SNOA986A – April 2018 – Revised August 2019

Precise Temperature Measurements With the TMP116 and

TMP117

Mihail Gurevitch

ABSTRACT
Engineers must carefully consider the overall system design when designing high-precision temperature
measurement applications. This application note provides recommendations on how to design a precise
temperature measuring system based on the TMP116 and TMP117 temperature sensors. By following this
application note, the user should be able to design a precise measuring system which adheres to the
performance specifications of the TMP116/117.

Contents
1 Introduction ................................................................................................................... 2
2 TMP116 and TMP117 Device Differences ............................................................................... 2
3 PCB Considerations ......................................................................................................... 3
4 Measuring Solid Surface Temperature ................................................................................... 3
5 Measuring Human Body Temperature ................................................................................... 5
6 Measuring Still Air Temperature .......................................................................................... 6
7 Measuring Moving Air Temperature ...................................................................................... 6
8 Measuring Thermal Resistance in Different Environments ............................................................ 8
9 Soldering to PCB ............................................................................................................ 9
10 Self-Heating ................................................................................................................. 11
11 Self-Heating Estimation Example ........................................................................................ 12
12 Supply Voltage Change .................................................................................................. 14
13 Data Averaging ............................................................................................................. 14
14 Summary ................................................................................................................... 15

List of Figures
1 Simplified Schematic of Temperature Flow During Solid Surface Measurement ................................... 3
2 PCB Layout Example for Rigid Surface Temperature Measuring ..................................................... 5
3 Moving Air Temperature Measurements Noise. Air Speed 0.5, 1 and 2 Meter/Sec. Averaging 8 Samples
Per Reading. 5 Consecutive Measurements at Room Temperature. ................................................. 7
4 PCB Layout Example for Air Temperature Measuring .................................................................. 8
5 Printed-Circuit Boards Used ............................................................................................... 9
6 Soldering Shift at +25ºC and Supply 3.3 V With Thermal Pad Soldered on a Rigid PCB. ...................... 10
7 Soldering Shift for TMP116/117 Without the Thermal Pad Soldered to the PCB. +25ºC, V = +3.3 V .......... 10
8 Device Consumption Power vs Temperature and Part Supply Voltage in Continuous Conversion Mode.
No Pauses Between Conversions, No I2C Bus Activity. ............................................................. 11
9 Supply Current vs. Pin Input Voltage and Device Supply Voltage for Any Digital Pin Input Cell. ............... 12
10 Device Supply Current vs. I2C Bus Clocking Frequency and Supply Voltage. Part is in Shutdown Mode,
but SCL, SDA, and ADD0 Pins are Under Constant I2C Data Flow. ............................................... 12
11 TMP116/117 Coupon Board Self-Heating Effect vs. Time and Supply Voltage in Still +25ºC Air. ............. 13
12 The TMP116/117 Sampling Distribution for 3 Different Oil Bath Temperatures and 3.3-V Supply Voltage.
No Data Averaging. ........................................................................................................ 14
13 The TMP116/117 Sampling Distribution for 3 Different Supply Voltages at +25ºC. No Data Averaging. ...... 15
14 Temperature Sampling Noise With 8, 32, and 64 Internal Averages. Temperature +25ºC and V = +3.3 V. .. 15

SNOA986A – April 2018 – Revised August 2019 Precise Temperature Measurements With the TMP116 and TMP117 42
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
Introduction www.ti.com

Trademarks
All trademarks are the property of their respective owners.

1 Introduction
There are many system factors which can negatively affect the precision of temperature measurements,
and these must be addressed to achieve a high accuracy. The main parameters that affect measurement
precision with the corresponding source of their control are:
• The accuracy of the temperature sensor itself as its accuracy, stability, and repeatability, are set by the
manufacturer and out of the designer’s control.
• The system engineer controls the supply voltage range and noise, the sensor conversion mode, the
system power consumption, the data sampling rate, the communication bus voltage, the I2C bus
frequency, and data flow over it.
• The PCB designer controls the mounting and position of the sensor on the PCB, the temperature
resistance between the sensor and the measured object, and the temperature “leakage” from the
sensor to surrounding air.
These parameters are important for precise temperature measurements and must be analyzed during the
system design. The purpose of this article is to provide recommendations to the system designer, based
on experience obtained in part characterization and device use in real applications.
When using the TMP116/117 for precise temperature measurements, there are a few critical
considerations that must be accounted for by the system designer:
• Proper PCB sensor location and orientation in the system. The proper location must provide the
precise temperature measurement with minimal offset and minimal time delay.
• Proper device electrical and communication interface mode, which can minimize measurement noise,
minimize part self-heating and ensure measurements stability.
• Proper PCB material and thickness, PCB mounting, and PCB layout. All these should provide a
minimal temperature difference between the sensor and the measured object, and should minimize
sensor response time when an object temperature is changing.

2 TMP116 and TMP117 Device Differences

The TMP116 and TMP117 have a similar internal schematic, register map, and electrical characteristics.
The main differences between two devices are shown in Table 1.

Table 1. Parameter Differences Between the TMP116 and TMP117

PARAMETER TMP116 TMP117
Ensured precision at room temperature ±0.2 ±0.1
(°C)
Temperature range (°C) –40 to +125 –55 to +150
Supply voltage range (V) 1.9 to 5.5 1.8 to 5.5
Shut down current at +25ºC (+125ºC) (µA) 0.25 (3) 0.15 (0.8)
Typical PSRR (m°C/V) 10 6
Package DRV-6 DRV-6 and WCSP
Thermal mass (mJ/°C) 5.1 5.1 and 0.8 (WCSP)
Price on Jun 2019 (1 Ku) ($) 0.99 1.6

Additionally, the TMP117 has a register to compensate the temperature offset and a reset bit in the
configuration register. Both parts are in the same 6-pin DRV package, but the TMP117 also has a smaller
WCSP-6 package version with a 1.5-mm × 1.0-mm × 0.5-mm die size. All conclusions found for either
device listed in this application note will apply to both the TMP116 and TMP117.

43 Precise Temperature Measurements With the TMP116 and TMP117 SNOA986A – April 2018 – Revised August 2019
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
www.ti.com PCB Considerations

3 PCB Considerations
There are two main tasks in temperature measurements: measuring air (gas) temperature and measuring
temperature of a solid surface. A liquid temperature measurement usually falls in one of the above,
because the sensor is often placed inside a metallic probe for liquid measurements. These two different
tasks dictate two different approaches to device mounting. However, in all cases, these common rules
must be applied:
• To get the manufacturer ensured measurement precision, the 0.1-µF bypass capacitor must be placed
no more than 5 mm (200 mils) away from the device
• To avoid possible heat influence coming from the pullup resistor on the SDA pin and the pullup resistor
on the SCL and ALERT pins (if present), the pins must be placed at least 10 mm (400 mils) away from
the device.
• If there is a risk that the board may bend during PCB mounting, all efforts to prevent the mechanical
tension on the device package must be taken. Guard holes in the PCB around the part can help in this
case.

4 Measuring Solid Surface Temperature

Measuring the temperature of a solid surface is the most common type of temperature measurement, and
the standard approach is to make a tiny rigid PCB, solder the device on one side of the PCB, and attach
the opposite side of the PCB to object surface. Figure 1 shows a cross-sectional view on how to mount
the TMP116/117 sensor to a PCB, along with a simplified schematic of the thermal processes required for
surface temperature measurement.

Figure 1. Simplified Schematic of Temperature Flow During Solid Surface Measurement

On this schematic:
• Tobj is the measured object temperature.
• Tair is the environment temperature (typically air).
• Ts is the sensor temperature.
• Rso is the thermal resistance between the sensor and the object.
• Rsa is the thermal resistance between the sensor and the air (environment).
• Ps is the averaged power dissipated by the sensor during the measurement.
• Mt is the combined thermal mass of device, plus the surrounding PCB area.

SNOA986A – April 2018 – Revised August 2019 Precise Temperature Measurements With the TMP116 and TMP117 44
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
Measuring Solid Surface Temperature www.ti.com

The most important formula for the precise temperature measurement is:
§ Tobj Tair ·
Tofs ¨ ¸ u R so
© R so + R sa ¹ (1)
Ts Tobj Tofs
where
• Tofs is a temperature offset between the measured object and sensor. (2)
Equation 1 shows that the sensor temperature offset is zero only in two cases: if Rso is zero or Rsa is
infinite. If there is a difference between Tobj and Tair (and despite all efforts to make Rsa >> Rso), however,
there will always be some offset between sensor and object temperature. This shift will increase when the
difference between Tobj and Tair is larger, or when Rsa becomes smaller and approaches to Rso value.
Let's calculate temperature offsets for two metallic object temperatures (+50°C and +100°C) where still air
temperature stays the same +25°C and the temperature resistance from sensor to object surface
assumed from line 6 of Table 2 (140m°C/mWt). Let's also assume that the temperature resistance
between sensor and air is equal to line 1 of Table 2 (300m°C/Wt). For object temperatures +50°C and
+100°C, the measurement offset, according to Equation 1, will be 7.9°C and 23.8°C accordingly, which is
not acceptable for precise measurements. TI recommends to use a thinner PCB with a better layout, and
cover the top surface of PCB with thermal isolating foam. The best solution to avoid temperature leakage
to the surrounding air may be to make a cave-kind cavern in the object body and put the PCB of the
sensor inside it, but this kind solution is not always available.
If the sensor temperature shift from the object temperature is too big and cannot be ignored, a system
calibration is needed. In some cases, it should be done for different combinations of Tobj and Tair. This
happens because Rsa is not a linear parameter, and instead depends on the air speed, air moisture, air
temperature, PCB orientation, and so on. All this makes the Rsa value estimation very difficult to find.
However, by making Rso as small as possible and Rsa as big as possible, it would be much easier to
minimize the temperature shift.
Another important aspect is when designers can trust the sensor readings, like when the object
temperature changes from Tobj1 to Tobj2. To estimate or understand the process of this temperature change,
we can use a Gaussian formula for an ideal case. In reality, the object temperature rarely changes
instantly, and therefore the sensor follows the object temperature slower than Equation 3 shows.
t / tr
Ts Tobj1 Tobj2 Tobj1 u e
where
• t is a time passing from beginning object temperature change.
• tr is a response time. (3)
tr R so u M t (4)
Here we can assume that Rsa >> Rso and ignore the temperature leakage to environment. According to
formula, to have minimal measurement delay, it is important to have a small response time (tr), which
means the Rso and Mt should be kept at minimal value, especially if the object temperature changes fast.
Because the device is dissipating some power during the measurements, the sensor is heating itself. The
self-heating temperature shift Tsh is calculate by Equation 5.
Tsh Ps u R so (5)
The influence of self-heating on measurement precision is discussed in Section 10.
The following are the recommendations for systems measuring rigid surface temperature:
• Use a PCB with minimal thickness.
• The side of the PCB that makes contact with the surface to be measured should be covered with an
exposed copper layer (and not covered with a solder mask). To prevent copper oxidation, a gold or
melted solder paste cover should be used.
• To improve thermal contact to the surface, consider adding a thermal conducting paste or sticky
thermal film between the surface and the PCB.

45 Precise Temperature Measurements With the TMP116 and TMP117 SNOA986A – April 2018 – Revised August 2019
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
www.ti.com Measuring Human Body Temperature

• Place additional vias to connect copper layers on both sides of the PCB. Generally, a via has 400
times less thermal-resistance than the same area of regular PCB material. Using a filled via further
decreases the thermal resistance.
• If the PCB internal layers are not used under the device, it is recommended to create internal copper
polygons under the sensors to reduce the PCB side-to-side thermal resistance.
• To increase the temperature resistance to surrounding air, minimize the amount of copper wires on top
of the board.
• To increase thermal resistance to surrounding air, the sensor and the PCB surface exposed to the air
must be covered with thermal-isolating foam, film, or at least with some stain. This protection is
especially important for precise measurements when air around the sensor is moving.
• To minimize the convection air influence, the PCB should be located horizontally and out of any air
flow.
• Soldering the device's thermal pad (TP) to the PCB may be a good choice only for systems which
undergo calibration. The negative aspects of TP soldering are described in Section 9. If the TP is
soldered, it should be connected to ground or left floating. Connecting the package TP to a voltage
other than system ground can lead to permanent device damage.
Figure 2 shows an example of a PCB layout for surface temperature measuring.

(1) Alert pin is not used and grounded. I2C bus pullup resistors are located on master board.
Figure 2. PCB Layout Example for Rigid Surface Temperature Measuring

5 Measuring Human Body Temperature

When making human body temperature measurements, it is important to understand the two cases that
may affect the performance of the system.
The first case is when the thermometer is exposed to the surrounding air temperature before it is pressed
to the body. The goal is to make precise body temperature measurements in the shortest amount of time
when the sensor temperature is changing rapidly at the beginning of measurement. In this case, the
minimal combined thermal mass will allow the sensor to reach a body temperature in the shortest amount
of time. Take care to avoid temperature “leakage” from the sensor to surrounding air. TI recommends to
have a temperature stabilization check before a measurement report is done. As an example, the
stabilization check can be to verify that the temperature didn’t change more than 0.2°C during the last 5
seconds. It is easy to achieve a good thermal contact to the object in this case, and therefore there is less
need to worry about the sensor self-heating. The conversion mode with a small standby time is
recommended.
• Use rigid PCB with minimal thickness to minimize the sensor-to-body thermal resistance.
• Cover the PCB side that makes direct contact with the body with a copper plane. Remove the solder

SNOA986A – April 2018 – Revised August 2019 Precise Temperature Measurements With the TMP116 and TMP117 46
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
Measuring Still Air Temperature www.ti.com

masks above the planes. To avoid oxidation, cover the exposed copper plane with gold or a melted
soldering material.
• Use a bypass capacitor with minimal dimensions to reduce thermal mass.
• Place pullup resistors away from the sensor.
• Depending on the design, cover the sensor and top side of PCB with a thermal-isolating compound.
The second case is a monitoring case where sensor attached to the body for a long period of time. In this
scenario, the temperature is changing very slowly and samples are taken less frequently (like once every
16 sec). It is easy to make a good thermal contact to a body and minimize temperature leakage. Bigger
sensor thermal mass may be useful as a low-frequency filter working to reduce temperature fluctuation
(noise). This can reduce the averaging number down to 1 during sampling, which lowers the power
consumption and extends the battery life. Bigger sensor thermal mass also reduces device self-heating
during conversions.
• Use a flexible PCB to make better temperature contact to the body.
• Cover the PCB side that makes direct contact with the human body with a copper plane. Remove the
solder masks above the planes. To avoid oxidation, cover the exposed copper plane with gold or a
melted soldering material.
• To make PCB maximal flexible and to increase PCB reliability, use the smallest size capacitor and
place the pullup resistors away from the sensor.
• To prevent temperature leakage and protect device contacts from oxidation, cover the top side of the
board with a thermal-isolating protection compound.

6 Measuring Still Air Temperature

The main feature of still air measurement is that the temperature changes slowly (usually less than a
degree per minute), and this is primarily due to air convection. When temperature change is slow, it is not
critical to have minimal thermal mass for sensor and surrounding PCB. Even with increased thermal mass,
the sensor will be able to follow the air temperature with minimal lag. The thermal resistance between slow
moving or still air and the PCB (including mounted sensor) is very high. Therefore, the designer should try
to improve the thermal contact with the air while simultaneously excluding any heat transfer from other
heat sources located on the same PCB. Due to the slow temperature change, there is no need to keep the
device running continuously. The update rate of one sample per second or less may be sufficient for most
use cases. When the device spends most of the time in standby or in shutdown mode, the power
consumption is minimal and self-heating is negligible.
• Place the PCB vertically. This will improve convection air flow and reduce dust collection over time.
The layer of accumulated dust works as a thermal-isolating barrier between the air and the PCB.
• To make a better PCB thermal contact with the air, place copper planes on both sides of PCB.
• Remove the solder masks above the planes. To prevent oxidation, cover the exposed copper plane
with melted soldering material or gold.
• Thermal isolation is required to avoid thermal coupling from the heat sources through the PCB. Use air
gaps between the sensor and PCB heat sources, if needed.

7 Measuring Moving Air Temperature

The main feature of moving air temperature measurement is high thermal noise, which is coming from the
temperature fluctuation inside air stream. Figure 3 shows the measurement noise of the room air flow,
which is moving along the rigid coupon board with a mounted TMP117 sensor at different speeds. As
seen in the graph, the measurements are still noisy even with an internal averaging of 8 temperature
samples.

47 Precise Temperature Measurements With the TMP116 and TMP117 SNOA986A – April 2018 – Revised August 2019
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
www.ti.com Measuring Moving Air Temperature

Figure 3. Moving Air Temperature Measurements Noise. Air Speed 0.5, 1 and 2 Meter/Sec. Averaging 8
Samples Per Reading. 5 Consecutive Measurements at Room Temperature.

The standard approach to reduce the noise is to increase the sample average number, but an alternate
method is to increase the sensor thermal response time in Equation 4.
Increased response time works as a low-pass filter, and it reduces the measurements noise. Knowing
response time tr, the designer can calculate the filter 3db cut-off frequency, Fc=1/tr. However, it is difficult
to estimate the effective combined thermal mass and effective thermal resistance between the sensor and
moving air, due to its dependents of many non-linear factors.
Moving air provides a good thermal contact to the sensor, and there can be a rare case where the sensor
can have the same temperature as a measured object. Low thermal resistance to moving air also
minimizes the device self-heating effect.
• Because moving air temperature usually has a lot of fluctuations, the PCB increased thermal mass can
reduce measurement noise. Therefore, it is acceptable in these cases to use a PCB with increased
thickness.
• Place the PCB vertically along air flow. This makes air flow smooth and prevent air “shades”.
• Design PCB soldering pads bigger than usual, especially the package corner pads. This will improve
the thermal contact from package to air.
• Cover both side of unused board space with a copper layer,
• Use a PCB with thicker copper layers, if possible. This improves thermal conductivity along the PCB,
and it allows better “average” temperature fluctuations from different parts of the board.
• If air (or gas) is expected to contain moisture or includes some corrosive components, the device pins
must be protected by a stain to avoid corrosion or moisture accumulation on the pins.

SNOA986A – April 2018 – Revised August 2019 Precise Temperature Measurements With the TMP116 and TMP117 48
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
Measuring Thermal Resistance in Different Environments www.ti.com

Figure 4 is an example of a PCB layout for air temperature measuring.

(1) Alert pin is not used and grounded. I2C pullup resistors are located on master board.
Figure 4. PCB Layout Example for Air Temperature Measuring

8 Measuring Thermal Resistance in Different Environments

As mentioned earlier, the thermal resistance between the sensor and measured object is a parameter
sensitive to PCB layout, board mounting. and environment condition. This parameter is not easily
calculable upfront. A more practical way is to measure the thermal resistance between the sensor and
object, and the sensor and environment, in already designed system. Knowing Ps, Rso, Rsa and using
Equations 1-5, it is possible to estimate the measurement error and sensor response time for different
temperatures and apply a necessary system correction. To measure Rso or Rsa, the system designer may
do the following:
• The environment or object temperature is fixed and well controlled.
• When TMP116/117 temperature is stabilized the device temperature T1 is read using minimal
conversion power. Single shot mode, which makes P1 power almost zero, is the best choice.
• The TMP116/117 an average consumption power is increased in any possible way. The simplest way
is to increase the supply voltage from min to max and switch to conversion mode without the standby
time. This will be the device power P2.
• When the device internal temperature is stabilized after power increase, the temperature reading T2 is
taken.
Now the designer can calculate the thermal resistance Rsx, which is Rso or Rsa:
T2 T1
R sx
P2 P1 (6)
In this measurement, it is assumed that the object (environment) temperature is stable during the test and
is not changed due to sensor self-heating.
Using this method, data about thermal resistance between part mounted on a coupon board (CB) to a
different kind of environment has been collected. The 2-layer coupon boards used in the experiments
have board size of 21 mm × 11 mm, a board thickness from 6 to 64 mil (0.15 to 1.62 mm), and an
identical layout. See Figure 5. Each CB has surface mounted 0.1-µF bypass capacitor and 6 contact pins.
Table 2 shows the thermal resistance from the TMP116/117 to a different environment.

Table 2. Thermal Resistance Between TMP116/117 to Differentiate Environment. The CB is 21 mm ×

11 mm.
ENVIRONMENT THERMAL COMMENTS
RESISTANCE
(m°C/mWt)
Still Air. 260-320 For all CB thickness and all CB orientation

49 Precise Temperature Measurements With the TMP116 and TMP117 SNOA986A – April 2018 – Revised August 2019
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
www.ti.com Soldering to PCB

Table 2. Thermal Resistance Between TMP116/117 to Differentiate Environment. The CB is 21 mm ×

11 mm. (continued)
ENVIRONMENT THERMAL COMMENTS
RESISTANCE
(m°C/mWt)
Moving Air along CB. 0.5 M/Sec 236 64 mill (1.62 mm) rigid CB (1)
Moving Air along CB. 0.5 M/Sec 190 6 mill (0.15 mm) flex CB (1)
Moving Air along CB. 2 M/Sec 200 64 mill (1.62 mm) rigid CB (1)
Moving Air along CB. 2 M/Sec 156 6 mill (0.15 mm) flex CB (1)
CB pressed to flat copper surface. 140 64 mill (1.62 mm) rigid CB. Thermal conductive paste between PCB
Device thermal pad is not soldered. and copper is used.
CB pressed to flat copper surface. 75 64 mill (1.62 mm) rigid CB. Thermal conductive paste between PCB
Device thermal pad is soldered. and copper is used.
Oil bath. 40 64 mill (1.62 mm) rigid CB. Oil is under intensive circulation (1).
WCSP die. CB pressed to flat 160 32 mill (0.81 mm) rigid CB
copper surface.
WCSP die. CB pressed to flat 160 6 mill (0.18 mm) flexible CB
copper surface.
(1)
The decision to solder or not solder the thermal pad does not make a significant difference.

9 Soldering to PCB
Soldering the TMP116/117 to a PCB can create significant package stress and degrade the absolute
accuracy. The measuring error of a TMP116/117 device in an oil bath before and after soldering often
shows an increase in the error, especially on rigid PCBs with the thermal pad soldered. This soldering shift
can be significant for precise measurements. Figure 5 shows the boards used in soldering shift tests. All
measurements were made in an oil bath.

Figure 5. Printed-Circuit Boards Used

In Figure 5, Board A is the socketed board used for testing loose devices prior to soldering. Board B is a
flexible PCB, and board C is a rigid PCB. Both used for testing devices after soldering.
Figure 7 shows the impact of soldering for 16 devices soldered to a rigid coupon boards. In Figure 7, parts
were measured in an oil bath at +25ºC with a 3.3-V supply before and after soldering. In this case, the
package thermal pad was also soldered to the coupon board. The average soldering shift in the example
is around 20mºC, but for device #4456, it reaches 50mºC. According to our research, the soldering shift is
not predictable, can be positive or negative, and, in the worst case, can reach ±100mºC.

SNOA986A – April 2018 – Revised August 2019 Precise Temperature Measurements With the TMP116 and TMP117 50
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
Soldering to PCB www.ti.com

130
120 Loose Units in sockets
110 Coupon Boards
100

Temperature Error (mC)

90
80
70
60
50
40
30
20
10
0
-10
-20
-30
3393
3386
3387
3081
3735
3734
3385
3080
3047
3737
4048
4050
3740
3739
4456
4439
D001
Device ID

Figure 6. Soldering Shift at +25ºC and Supply 3.3 V With Thermal Pad Soldered on a Rigid PCB.

Furthermore, the soldering shift can be different for different temperatures, which makes it even less
predictable.
The main reason for the soldering shift is mechanical tension coming to the silicon die through the
package from the PCB and the hardened solder. When the temperature drops in the reflow oven, the
solder hardens and fixes the thermal pad and package pin locations. But package material continues to
contract, and because the solder and the rigid PCB have different contraction coefficients than device
package, it creates the mechanical tension which leads to package bending and therefore creates
tensions in the silicon die. However, when the package thermal pad is not soldered, the bending forces
are applied only to the package pins, which have much less mechanical contact to the silicon die.
Figure 7 shows the effects of soldering when the thermal pad is not soldered to the PCB. In this case, the
accuracy shift is much less and the worst-case offset is only 15 mC.
20
15
10
Temperature Error (mC)

5
0
-5
-10
-15
-20
-25 Loose Units
Coupon Boards
-30
1770
1773
1771
1772
2109
2115
1769
2111
2419
1459
2421
2420
1460
1765
2418
2107

D003
Device Number

Figure 7. Soldering Shift for TMP116/117 Without the Thermal Pad Soldered to the PCB. +25ºC, V = +3.3 V

The reasonable question is: when the thermal pad is not soldered, by how much will the thermal
resistance between the sensor and the PCB going to increase? In conducted experiments, the device was
soldered to a rigid coupon board 11-mm × 22-mm × 1.1-mm size with no vias under the part and a copper
radiator was attached to the opposite side of PCB. (The silicon thermo conductive paste between copper
radiator and PCB back side was applied). The measurements showed that not soldering the package
thermo pad increased the thermal resistance from 75 to 140ºC/Wt. By knowing the thermal resistance and
device thermal mass Mt = 5.1 mJ/ºC, it is possible to calculate the sensor thermal response time with
Equation 4.

51 Precise Temperature Measurements With the TMP116 and TMP117 SNOA986A – April 2018 – Revised August 2019
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
www.ti.com Self-Heating

The calculated response time values are 0.39 and 0.72 seconds and measured response time matched
the calculated values. Because the device package thermal mass is extremely small, the thermal
response time is also very small and even the 0.72 second value, when the thermal pad is not soldered,
satisfies most users applications.
Here are the recommendations on how to minimize the soldering shift in the TMP116/117 parts:
• To maintain device manufacturer precision, in case the system calibration is not planned, TI highly
recommends not to solder the package thermal pad to avoid a soldering shift.
• Use the standard reflow oven soldering process with a maximum temperature to +250ºC for one
minute.
• Manual soldering is not acceptable because it creates additional stress on the device package,
resulting in soldering shift as large as ±150mºC.
• Using a flexible PCB with thickness less than 6 mil (0.15 mm) creates minimal mechanical tensions
and minimal soldering shift even in the case when the thermal pad is soldered.
• When using a flexible PCB with thickness more than 6 mil (0.15 mm), the thermal pad must not be
soldered. The flexible PCB minimizes the thermal mass and thermal resistance, which may improve
measurement precision.

10 Self-Heating
To achieve the best measurement accuracy, the TMP116/117 part is specially designed to dissipate
minimal power and minimize the part temperature change due to self-heating. In typical conditions (supply
voltage is 3.3 V, 8 samples average, one data collection per second), the TMP116/117 dissipates 53 uWt
at +25°C. However, when operating with a higher supply voltage and taking more frequent measurements,
the power dissipation can increase to almost 1 mWt. Figure 8 shows the power dissipation as a function of
the device temperature at different voltage supplies.
1.1
1.8 V 4.1 V
1
2.6 V 4.8 V
0.9 3.3 V 5.5 V
0.8
0.7
Power (mWt)

0.6
0.5
0.4
0.3
0.2
0.1
0
-75 -50 -25 0 25 50 75 100 125 150
Temperature (qC) D011

Figure 8. Device Consumption Power vs Temperature and Part Supply Voltage in Continuous Conversion
Mode. No Pauses Between Conversions, No I2C Bus Activity.

The power consumption in user measurements is usually significantly less than 1 mWt, but to make the
most accurate measurement and reduce any influence of self-heating, all efforts to reduce the dissipation
power must be taken. Here are recommendations on how to reduce the device power consumption:
• Use the minimal supply voltage acceptable for the system. This is especially important when the
device is in continuous conversion mode without the pauses.
• Use one-shot conversion mode or use a conversion cycle mode where the device goes into standby
after a conversion.
• Use pullup resistors larger than 5 kΩ on the SDA, SCL, and ALERT pins. Place resistors at least 10
mm from the TMP116/117 to reduce any influence from the resistor's heat dissipation.
• Ensure that the SCL and SDA signal levels are below 10% and above 90% of the device supply
voltage. If the SCL, SDA, and ADD0 pin input voltages are close to ground or device supply level, the
current going through the digital pin input cell is low, which minimizes the sensor heating (see
Figure 9). Remember that the I2C bus voltage can go up to 6 V and is not limited by the applied supply
SNOA986A – April 2018 – Revised August 2019 Precise Temperature Measurements With the TMP116 and TMP117 52
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
Self-Heating Estimation Example www.ti.com

voltage.
• Avoid heavy bypass traffic on the I2C bus. Remember that the intensive communication to other
devices on the same bus increases the TMP116/117 supply current, even if the device is in shutdown
mode (see Figure 10).
• Use the highest available communication speed. To increase the SCL and SDA rising edge speeds,
use a bus pullup voltage higher than the device supply voltage.
650
600 5.5 V
4.4 V
550 3.3 V
500 2V
450
ISUPPLY (uA)

400
350
300
250
200
150
100
50
0
0 10 20 30 40 50 60 70 80 90 100
VIN/V+ (%) D005

Figure 9. Supply Current vs. Pin Input Voltage and Device Supply Voltage for Any Digital Pin Input Cell.

100
5.5 V
90 4.5 V
80 3.3 V
1.9 V
Supply Current (uA)

70
60
50
40
30
20
10
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
I2C Bus Frequency (MHz) D006

Figure 10. Device Supply Current vs. I2C Bus Clocking Frequency and Supply Voltage. Part is in
Shutdown Mode, but SCL, SDA, and ADD0 Pins are Under Constant I2C Data Flow.

11 Self-Heating Estimation Example

The self-heating impact can be calculated by the simple formula below:
Tsh P u Rt
where
• Tsh is a temperature offset due to sensor self-heating.
• P is an averaged power dissipated by the sensor.
• Rt is a combined temperature resistance to the environment. (7)
This implies that another way to reduce the self-heating is to reduce the thermal resistance to the
measured object. On the contrary, the larger the thermal resistance between the sensor and measured
object, the larger the self-heating influence on measurement precision. Below are listened cases when the
self-heating effect can be ignored:
• The desired measurement precision is worse than ±0.2ºC.

53 Precise Temperature Measurements With the TMP116 and TMP117 SNOA986A – April 2018 – Revised August 2019
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
www.ti.com Self-Heating Estimation Example

• The system calibration takes care of self-heating and all other effects.
• The device average consumption power is less than 0.1 mWt.
• The thermal resistance between the sensor and measured object is small.
In this list, the most difficult parameter to estimate is the thermal resistance between the sensor and the
environment. The estimation is difficult because it depends on many poorly controlled factors. Here is a
recommendation on how to estimate the device object thermal resistance in a real application environment
and then calculate a possible self-heating temperature rise for a worst case scenario. The idea is to
measure the self-heating for some fixed supply voltage and fixed environment temperature, and then
extrapolate results over an entire voltage and temperature range.
Figure 11 shows an example of the self-heating effect on positioning the coupon boards horizontally in a
"still air box", with a TMP116/117 placed on top of the board. At time zero, the device is switched from
shutdown mode to continuous conversion mode with a 64 sampling averaging and no pauses between
conversions. There is no heating from the I2C bus activity because the data reading happens only once
per second. The temperature change on Figure 11 happens only due to device dissipated power and
following self-heating. Let's calculate the thermal resistance between the part and its environment.
80
5.5 V
70 5V
4V
Temperature Change (mC)

60 3V
1.9 V
50

40

30

20

10

0
0 10 20 30 40 50 60 70 80 90
Time (seconds) D007

Figure 11. TMP116/117 Coupon Board Self-Heating Effect vs. Time and Supply Voltage in Still +25ºC Air.

For example, assume the customer test was done with a 3-V supply and air temperature +25ºC. We see
the device temperature stabilized after 90 seconds with 40mºC self-heating value. According to Figure 8,
the consumption power for this mode is 0.36 mWt for a 3-V supply. So, the thermal-resistance between
the device and surrounding air is Rt = 40mºC/ 0.36 mWt = 111C/Wt. Now, knowing the thermal resistance,
it is possible to calculate the self-heating offset for other situations. For example, if the air temperature is
+125ºC and the supply voltage is 4 V according to Figure 8, the dissipated power would be 0.65mWt and
self-heating temperature offset would be Tsh = 111C/Wt x 0.65mWt = 72m°C. The 80-second long settling
time here is associated with stabilization time of air convection process in the “still air box”. if the box size
changes, the self-heating and stabilization time will also change.
As a reminder, this example above is a worst-case scenario where the thermal resistance between the
device and environment is high and device is continuously converting. It does demonstrate, however, that
self-heating can occur and must be considered when trying to achieve the best precision. If the
experiment is repeated with moving air, the self-heating offset will be much smaller and could become
negligible. But in all cases, the recommendation is the same: minimize the device dissipated power.
The easiest way to minimize the dissipated power is to limit the rate at which the temperature is sampled.
If we used device default mode (8 sample averaging with sampling rate 1 Hz) in the example above, the
average supply current would be 16 µA, the dissipated power would be only 48 µWt, and the self-heating
would only be 5.3m°C, which is less than sensor resolution and is negligible.

SNOA986A – April 2018 – Revised August 2019 Precise Temperature Measurements With the TMP116 and TMP117 54
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
Supply Voltage Change www.ti.com

12 Supply Voltage Change

Precise measurements usually mean that supply voltage has minimal noise that does not change during
the measurements. In some battery systems, however, the voltage can change significantly with battery
aging. The TMP117 has excellent (almost zero) electrical PSRR, and the supply voltage change has no
effect on the precision of the readings. The only case where a system designer must take precautions is
when the device dissipates some heat in continues conversion mode without the standby. If the thermal
resistance to the object is significant, the supply voltage change from the maximum to minimum can
create sensor self-heating offset change (so named self-heating PSRR). Standard recommendations of
minimizing device average dissipated power and minimizing thermal resistance to the object applies in this
case. For the TMP116, the typical electrical PSRR is around 10mºC/V, and should be considered if the
supply voltage changes. The best recommendation for precise temperature measurements is simple:
stabilize the supply voltage at minimum system acceptable level.

13 Data Averaging
The TMP116/117 can be configured to take multiple measurements and provide the resultant average as
the result. Figure 12 and Figure 13 show the output temperature distribution with no averaging for 3
temperatures, and no averaging for different supply voltages. In all these cases, the standard deviation of
the readings is about 1 LSB, and data distribution covers an area approximately of six neighboring codes,
which match the ±3 sigma rule. This leads to the important conclusion that sensor internal noise is the
same for whole temperature range –55ºC to +150ºC, and the whole supply voltage range 1.9 V (1.8 V) to
5.5 V. Based on this data, the sensor internal noise without averaging in ideal bath condition can be
estimated as ±25m°C.
50
40qC (St. Dev. = 1.12)
45 25qC (St. Dev. = 1.01)
125qC (St. Dev. = 1.05)

40

35
Appearing in %

30

25

20

15

10

0
-4 -3 -2 -1 0 1 2 3 4
Data Distribution (LSB) D008

Figure 12. The TMP116/117 Sampling Distribution for 3 Different Oil Bath Temperatures
and 3.3-V Supply Voltage. No Data Averaging.

55 Precise Temperature Measurements With the TMP116 and TMP117 SNOA986A – April 2018 – Revised August 2019
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
www.ti.com Summary

50
V = 1.9 V (St. Dev. = 0.91)
45 V = 3.3 V (St. Dev. = 1.01)
V = 5 V (St. Dev. = 0.96)

40

35
Appearing in %

30

25

20

15

10

0
-4 -3 -2 -1 0 1 2 3 4
Data Distribution (LSB) D009

Figure 13. The TMP116/117 Sampling Distribution for 3 Different Supply Voltages at +25ºC.
No Data Averaging.

The TMP116/117 provides an internal mechanism for averaging 8, 32, and 64 consequent samples
controlled by the configuration register. As shown in Figure 14, even the 8 samples averaging reduces the
internal noise distribution to a theoretical minimum of 2 LSB. This means that if the measured temperature
changes slowly and has no temperature fluctuations, the supply voltage is stable and has no glitches, and
there is no heavy bypassing traffic on I2C bus, the 8 samples averaging is enough to neutralize the
internal sensor noise and provide stable temperature readings. However, if the measured conditions are
far from ideal, higher averaging numbers are recommended.
80
Average 8 (St. Dev. = 0.51)
Average 32 (St. Dev. = 0.56)
70 Average 64 (St. Dev. = 0.61)

60

50
Appearing in %

40

30

20

10

0
-4 -3 -2 -1 0 1 2 3 4
Data Distribution (LSB) D010

Figure 14. Temperature Sampling Noise With 8, 32, and 64 Internal Averages.
Temperature +25ºC and V = +3.3 V.

14 Summary
The TMP116/117 provides excellent precision, small power consumption, extremely small thermal mass,
and averaging tools with wide temperature and supply range. To achieve best performance, system
designers must follow the recommendations in this application note and product data sheets.

SNOA986A – April 2018 – Revised August 2019 Precise Temperature Measurements With the TMP116 and TMP117 56
Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
Revision History www.ti.com

Revision History
NOTE: Page numbers for previous revisions may differ from page numbers in the current version.

Changes from Original (April 2018) to A Revision .......................................................................................................... Page

• Added references to the TMP117 device .............................................................................................. 1

• Added Measuring Solid Surface Temperature section ............................................................................... 3
• Added Measuring Human Body Temperature section ............................................................................... 5
• Added Measuring Still Air Temperature section ....................................................................................... 6
• Added Measuring Human Body Temperature section ............................................................................... 6
• Added Measuring Thermal Resistance in Different Environments section ........................................................ 8
• Removed Devices Temperature Error Change Due to Device Soldering for 6 Different Temperatures. V = +3.3 V
graph ...................................................................................................................................... 10
• Added recommendations on how to minimize the soldering shift in the TMP116/117 parts................................... 10
• Changed Device Consumption Power vs Temperature and Part Supply Voltage in Continuous Conversion Mode
graph ...................................................................................................................................... 11
• Added Supply Voltage Change section ............................................................................................... 14

57 Revision History SNOA986A – April 2018 – Revised August 2019

Submit Documentation Feedback
Copyright © 2018–2019, Texas Instruments Incorporated
Additional Resources

Temperature sensors overview Remote temperature sensor calibration tool

Explore TI’s portfolio of temperature sensors for any need. Use this calibration tool to optimize performance across a
selection of temperature sensors.
Temperature sensors parametric quick
search tool Thermistor design tool
Filter TI’s selection of temperature sensors to find the right TMP6 thermistor design tool with look-up tables, performance
solution for your design. comparisons and code examples.

TI Precision Labs – temperature sensors SysConfig system configuration tool

Learn the basics of temperature sensors, terminology, key An intuitive and comprehensive collection of graphical utilities
specifications and design tips in TI’s video training curriculum, for configuring pins, peripherals, radios, subsystems and other
with on-demand courses and tutorials. components.

Important Notice: The products and services of Texas Instruments Incorporated and its subsidiaries described herein are sold subject to TI’s standard terms and conditions of sale.
Customers are advised to obtain the most current and complete information about TI products and services before placing orders. TI assumes no liability for applications assistance,
customer’s applications or product designs, software performance, or infringement of patents. The publication of information regarding any other company’s products or services does not
constitute TI’s approval, warranty or endorsement thereof.

All trademarks are the property of their respective owners.

© 2022 Texas Instruments Incorporated SLYY161A

 IMPORTANT NOTICE AND DISCLAIMER
TI PROVIDES TECHNICAL AND RELIABILITY DATA (INCLUDING DATA SHEETS), DESIGN RESOURCES (INCLUDING REFERENCE
DESIGNS), APPLICATION OR OTHER DESIGN ADVICE, WEB TOOLS, SAFETY INFORMATION, AND OTHER RESOURCES “AS IS”
AND WITH ALL FAULTS, AND DISCLAIMS ALL WARRANTIES, EXPRESS AND IMPLIED, INCLUDING WITHOUT LIMITATION ANY
IMPLIED WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT OF THIRD
PARTY INTELLECTUAL PROPERTY RIGHTS.
These resources are intended for skilled developers designing with TI products. You are solely responsible for (1) selecting the appropriate
TI products for your application, (2) designing, validating and testing your application, and (3) ensuring your application meets applicable
standards, and any other safety, security, regulatory or other requirements.
These resources are subject to change without notice. TI grants you permission to use these resources only for development of an
application that uses the TI products described in the resource. Other reproduction and display of these resources is prohibited. No license
is granted to any other TI intellectual property right or to any third party intellectual property right. TI disclaims responsibility for, and you
will fully indemnify TI and its representatives against, any claims, damages, costs, losses, and liabilities arising out of your use of these
resources.
TI’s products are provided subject to TI’s Terms of Sale or other applicable terms available either on ti.com or provided in conjunction with
such TI products. TI’s provision of these resources does not expand or otherwise alter TI’s applicable warranties or warranty disclaimers for
TI products.
TI objects to and rejects any additional or different terms you may have proposed. IMPORTANT NOTICE

Mailing Address: Texas Instruments, Post Office Box 655303, Dallas, Texas 75265
Copyright © 2022, Texas Instruments Incorporated