This circuit is a new implementation of a brilliant idea that’s been published before1,2,3,4. It extends the original concept to a multichannel design while using standard components to minimize costs. The concept is based on the relationship between the voltage across a silicon junction and its temperature, when the current flowing through the junction is switched between two values, I1 and I2:
ΔV = (kT/q) ln (I1/I2)
where k is Boltzmann’s constant (1.38 × 10−23 J/°K), T is the Kelvin temperature, and q is the charge of one electron (1.6 × 10−19 C). This relationship shows that the Kelvin temperature is proportional to ΔV, if the current ratio can be controlled accurately, regardless of device-to-device variations. If I1/I2 = 10, the conversion factor between temperature and voltage is 199 µV/°K.
The basic idea is to use a current source that can be switched using a 10/1 ratio. While previous designs were adequate for a single-channel measurement, they could not be easily expanded to a multichannel system. This design achieves its expandability by replacing the current source with a voltage source that’s bootstrapped above the sensor voltage. The source can be shared by as many sensors as needed. The output of the circuit is digitized by an analog-to-digital converter (ADC) and processed by a microcontroller. The microcontroller also controls the input-channel multiplexer and drives the analog switches to control the circuit’s operation.
The simplified schematic shows how this is accomplished (Fig. 1). First, assume the multiplexer is connecting QA to A1’s non-inverting input. During the first phase of the measurement, S1 is positioned to select the 0.25-V reference and S2 is closed. A1 will maintain equilibrium by developing a voltage drop across RA that’s equal to the 0.25-V reference, and the current flowing through sensor QA is equal to 0.25 V/RA. The voltage charge across C1 is equal to the sensor voltage.
For the second phase of the measurement, S1 selects the 2.5-V reference and the voltage drop across RA becomes 2.5 V. This causes the current flowing through QA to increase by a factor of 10. If, at the same time, S2 is opened, the voltage rise at the input of A2 will equal the difference between the two voltage drops across temperature sensor QA as the current increases tenfold.
This voltage rise is proportional to the Kelvin temperature of the sensor. The voltage differential VBE alone is amplified by A2, since the first-phase voltage drop is stored in C1 and is effectively subtracted from A2’s input voltage during the second phase. The various voltages during the two phases of the measurement are summarized in the table. The two phases are repeated for each of the other channels, with each sensor being selected in sequence by the input multiplexer.
As the table shows, the temperature to be measured is proportional to the difference in A2’s output voltage between the two phases. The accuracy of the temperature measurement is determined by the accuracy of the 10/1 current ratio, and by the gain of A2 (which is set by R3 and R4).
The full schematic demonstrates how the circuit can be designed to minimize the effect of component tolerances on the temperature measurements (Fig. 2). The 0.25- and 2.5-V sources are generated from the same reference IC, so the current ratio is set entirely by the R1/R2 ratio, regardless of the actual voltage or drift in the LM385.
The only other source of error in the current ratio is A1’s offset voltage, which must be small with respect to the 0.25-V reference. The values of RA and RB don’t impact the current ratio, and therefore don’t affect the accuracy. The gain of the amplifier stage A2 is determined by resistors R3 and R4.
During the first low-current phase of the measurement, in which the noninverting input of A2 is grounded, its amplified offset voltage can be measured by the ADC and subsequently subtracted from the measurement during the second high-current phase, effectively compensating A2’s offset voltage. R5 and R6 create an additional 10 mV of positive offset to ensure that A2’s output always feeds a positive voltage to the ADC, even if A2’s offset voltage happens to be negative.
In summary, the measurement accuracy is limited by the stability of two resistor ratios, R1/R2 and R3/R4, and by A1’s offset voltage. All of the other component tolerances and drift factors are eliminated by the differential measurement scheme.
- Williams, Jim, “Transistor Sensor Needs No Compensation,” EDN, April 25, 1991 (also published in Linear Technology application note AN-45).
- Woodward, W. Stephen, “Low-Cost Precision Thermometry,” Electronic Design, August 21, 1995, p. 99.
- Woodward, W. Stephen, “Transistor Forms RS-232 Digital Thermometer,” EDN, May 9, 1996.
- Steele, Jerry, “Sense Junction Temperatures Without Calibration,” Electronic Design, May 13, 1998, p. 130.