4.0 Application Hints

The LM82 can be applied easily in the same way as other

integrated-circuit temperature sensors and its remote diode

sensing capability allows it to be used in new ways as well.

It can be soldered to a printed circuit board, and because the

path of best thermal conductivity is between the die and the

pins, its temperature will effectively be that of the printed

circuit board lands and traces soldered to the LM82’s pins.

This presumes that the ambient air temperature is almost the

same as the surface temperature of the printed circuit board;

if the air temperature is much higher or lower than the

surface temperature, the actual temperature of the of the

LM82 die will be at an intermediate temperature between the

surface and air temperatures. Again, the primary thermal

conduction path is through the leads, so the circuit board

temperature will contribute to the die temperature much

more strongly than will the air temperature.

To measure temperature external to the LM82’s die, use a

remote diode. This diode can be located on the die of a

target IC, allowing measurement of the IC’s temperature,

independent of the LM82’s temperature. The LM82 has been

optimized to measure the remote diode of a Pentium II

processor as shown in Figure 9. A discrete diode can also be

used to sense the temperature of external objects or ambient

air. Remember that a discrete diode’s temperature will be

affected, and often dominated, by the temperature of its

leads.

Most silicon diodes do not lend themselves well to this

application. It is recommended that a 2N3904 transistor

base emitter junction be used with the collector tied to the

base.

A diode connected 2N3904 approximates the junction avail-

able on a Pentium microprocessor for temperature measure-

ment. Therefore, the LM82 can sense the temperature of this

diode effectively.

4.1 ACCURACY EFFECTS OF DIODE NON-IDEALITY

FACTOR

The technique used in today’s remote temperature sensors

is to measure the change in V

points of a diode. For a bias current ratio of N:1, this differ-

ence is given as:

where:

•

η is the non-ideality factor of the process the diode is

manufactured on,

•

q is the electron charge,

•

k is the Boltzmann’s constant,

•

N is the current ratio,

•

T is the absolute temperature in ˚K.

The temperature sensor then measures

∆V

to digital data. In this equation, k and q are well defined

universal constants, and N is a parameter controlled by the

temperature sensor. The only other parameter is

η, which

depends on the diode that is used for measurement. Since

∆V

η and T, the variations in η

cannot be distinguished from variations in temperature.

Since the non-ideality factor is not controlled by the tempera-

ture sensor, it will directly add to the inaccuracy of the

η from part to part. As an example, assume a temperature

temperature of 25 ˚C and the process used to manufacture

accuracy of the temperature sensor at room temperature will

be:

T

.

The additional inaccuracy in the temperature measurement

caused by

η, can be eliminated if each temperature sensor is

calibrated with the remote diode that it will be paired with.

4.2 PCB LAYOUT FOR MINIMIZING NOISE

In a noisy environment, such as a processor mother board,

layout considerations are very critical. Noise induced on

traces running between the remote temperature diode sen-

sor and the LM82 can cause temperature conversion errors.

The following guidelines should be followed:

1.

Place a 0.1 µF power supply bypass capacitor as close

as possible to the V

capacitor as close as possible to the D+ and D− pins.

Make sure the traces to the 2.2nF capacitor are

matched.

2.

The recommended 2.2nF diode bypass capacitor actu-

ally has a range of 200pF to 3.3nF. The average tem-

perature accuracy will not degrade. Increasing the ca-

pacitance will lower the corner frequency where

differential noise error affects the temperature reading

thus producing a reading that is more stable. Con-

versely, lowering the capacitance will increase the cor-

ner frequency where differential noise error affects the

temperature reading thus producing a reading that is

less stable.

10129715

FIGURE 9. Pentium or 3904 Temperature vs LM82

Temperature Reading

www.national.com

15