REV. B

TMP12

–8–

The circuit of Figure 8 provides approximately 1

°C of resolution.

The two potentiometers divide VREF by 2, and the 8-bit

potentiometer further divides VREF by 256, so the resolution is

Resolution

VREF

N

V

mV

=

÷ =

÷ =

2

2

25

2

28

49

.

.

where VREF is the voltage reference output (Pin l of the TMP12)

and N is the resolution of the AD8402. Since the VPTAT has a

slope of 5 mV/K, the AD8402 provides 1

°C of resolution. The

adjustment range of this circuit extends from VREF/2 (i.e.,

1.25 V, or –23

°C) to VREF – 1 LSB (i.e., 2.5 V – 4.9 mV, or

226

°C). The VPTAT is therefore

VPTAT

V

Digital Count

mV

=+

×

()

125

49

..

where Digital Count is the value sent to the AD8402 which caused

the setpoint 1 output to go low.

A third way to measure the VPTAT voltage is to close a feed-

back loop around one of the TMP12’s comparators. This causes

the comparator to oscillate, and in turn forces the voltage at the

comparator input to equal the VPTAT voltage. Figure 9 is a

typical circuit for this measurement. An OP193 operational

amplifier, operating as an integrator, provides additional loop-gain

to ensure that the TMP12 comparator will oscillate.

Understanding Error Sources

The accuracy of the VPTAT sensor output is well characterized

and specified; however, preserving this accuracy in a thermal

monitoring control system requires some attention to minimiz-

ing the various potential error sources. The internal sources

of setpoint programming error include the initial tolerances and

temperature drifts of the reference voltage VREF, the setpoint

comparator input offset voltage and bias current, and the hysteresis

current scale factor. When evaluating setpoint programming

errors, remember that any VREF error contribution at the com-

parator inputs is reduced by the resistor divider ratios. Each

comparator’s input bias current drops to less than 1 nA (typ) when

the comparator is tripped. This change accounts for some setpoint

voltage error, equal to the change in bias current multiplied by

the effective setpoint divider ladder resistance to ground.

The thermal mass of the TMP12 package and the degree of

thermal coupling to the surrounding circuitry are the largest

factors in determining the rate of thermal settling, which ultimately

determines the rate at which the desired temperature measure-

ment accuracy may be reached. (See graph in TPC 2.) Thus,

one must allow sufficient time for the device to reach the final

temperature. The typical thermal time constant for the SOIC

plastic package is approximately 70 seconds in still air. There-

fore, to reach the final temperature accuracy within 1%, for a

temperature change of 60

°C, a settling time of five time con-

stants, or six minutes, is necessary. Refer to TPC 3.

External error sources to consider are the accuracy of the external

programming resistors, grounding error voltages, and thermal

gradients. The accuracy of the external programming resistors

directly impacts the resulting setpoint accuracy. Thus, in fixed-

temperature applications, the user should select resistor tolerances

appropriate to the desired programming accuracy. Since setpoint

resistors are typically located in the same air flow as the TMP12,

resistor temperature drift must be taken into account also. This

effect can be minimized by selecting good quality components,

and by keeping all components in close thermal proximity.

Careful circuit board layout is essential to minimize common

thermal error sources. Also, the user should take care to keep

the bottom of the setpoint programming divider ladder as close

to GND (Pin 4) as possible to minimize errors due to IR voltage

drops and coupling of external noise sources. In any case, a 0.1

µF

capacitor for power supply bypassing is always recommended at

the chip.

Safety Considerations in Heating and Cooling System Design

Designers should anticipate potential system fault conditions

that may result in significant safety hazards that are outside

the control of and cannot be corrected by the TMP12-based

circuit. Governmental and industrial regulations regarding

safety requirements and standards for such designs should be

observed where applicable.

Self-Heating Effects

In some applications, the user should consider the effects of self

heating due to the power dissipated by the open-collector out-

puts, which are capable of sinking 20 mA continuously. Under

full load, the TMP12 open-collector output device is dissipating

PV

A

mW

=

06

0 020

12

..

which in a surface-mount SOIC package accounts for a tempera-

ture increase due to self-heating of

∆TP

W

C W

C

DISS

JA

=×

=

×

°

=

°

θ

0 012

158

1 9

./

.

This increase is for still air, of course, and will be reduced at

high airflow levels. However, the user should still be aware that

self-heating effects can directly affect the accuracy of the TMP12.

For Setpoint 2, self-heating will add to the Setpoint temperature

(that is, in the above example, the TMP12 will switch the Setpoint

2 output off 1.9

°C early). Self-heating will not affect the tempera-

ture at which Setpoint 1 turns on, but will add to the hysteresis.

Several circuits for adding external driver transistors and other

buffers are presented in the following sections of this data sheet.

These buffers will reduce self-heating and improve accuracy.

Buffering the Voltage Reference

The reference output VREF is used to generate the temperature

setpoint programming voltages for the TMP12. Since the hyster-

esis is set by the reference current, external circuits which draw

current from the reference will increase the hysteresis value.

The on-board VREF output buffer is typically capable of 500

µA

output drive into as much as 50 pF load (max). Exceeding this

load will affect the accuracy of the reference voltage, could

cause thermal sensing error due to excess heat buildup, and

may induce oscillations. External buffering of VREF with a

low-drift voltage follower will ensure optimal reference accu-

racy. Amplifiers that offer low-drift, low-power consumption,

and low cost appropriate to this application include the OP284

and members of the OP113 and OP193 families.

With excellent drift and noise characteristics, VREF offers a

good voltage reference for data acquisition and transducer exci-

tation applications as well. Output drift is typically better than

–10 ppm/

°C, with 315 nV/Hz (typ) noise spectral density at 1 kHz.