A micro-controller for LMT87 temperature sensor?

Date: 2 Sept 2016
A temperature sensor designed for control of irrigation has peculiar requirements. The transducer is located on the air-only side of a metallic disk in moist soil. The disk seals the end of a hollow insulating cylinder from which a ring-shaped heater sends out a few-second heat pulse from its location a few cm from the cylinder end.
What is odd about this application is that the temperature rise T(t) at the transducer arrives after 10+ s, attains a MaxT of only about 1 C [a condition that keeps moisture from moving away from the heater], and lingers for nearly an hour because the decay of T(t) is nearly a power law. Thus sampling T(t) faster than 1 Hz is a waste, whereas most electronic temperature sensors today are designed to respond to T(t) much faster than that. For the best estimate of how a moist soil conducts heat, the measurements of T(t) that are crucial are those at elapsed times between 10 and 2000 s.
A small complication is that the ambient temperature of the measurement, T0, drifts. Its time scale is by a few C in half a day to 20 C over months.
1) Suppose that the transducer is the CMOS IC LMT70. Two conditions need to be met: a) For one T(t) pulse the resolution of the ADC that follows must resolve about 0.01 C meaningfully; b) The system should accept to operate according to condition (a) without operator interference even though T(0) changes (slowly) by about 20 C.
Given the slope for the LMT70, dV/dT = -5.19 mV/C, condition (b) requires that the system (notably the ADC) operates normally if the output voltage from the LMT70 shifts by 100 mV. Yet in one "measurement," condition (a) requires that a temperature shift of 0.01 C, i.e., for voltage shifts of only 0.01*5190 uV ~ 50 uV, the ADC must output a distinct code.
The ADS1146, ADS1147, and ADS1148 devices are precision 16-bit analog-to-digital converters (ADCs) that include many integrated features to decrease
system cost and component count for sensor measurement applications. The devices feature a low-noise, programmable gain amplifier, a precision delta-sigma ADC with a digital filter that settles in a single cycle, and an internal oscillator. The ADS1146 device supports one differential input, e.g., the LMT70 transducer mentioned above, while the ADS1147 and ADS1148 devices support 2 and 4 differential inputs - something not needed for the soil moisture sensor.
According to Table 1, pg 15 of the ADS1146 description (SBAS453G, the 2016 revision), at 5 samples per sec and the low gain settings of 1 [2], the equivalent input noise is 62 uV [31 uV], which is encouragingly close to the above condition (a) of 50 uV that is set by the desired resolution of 0.01 C. According to LMT70 data presented by Michael Wong in his description (tidubt8, pg 3) of its use in a skin temperature application, its accuracy (i.e., deviation from the straight line with slope -5.19 mV/C) is 0.13 C between 20 and 42 C, which is encouraging for condition (b).
2) Suppose that an LMT87 is the temperature transducer. It can meet the condition (a) for 0.01 C resolution because the slope dV/dT is increased from -5.19 mV/C to -13.6 mV/C. That was achieved by building in amplifiers like those in the LMT94022 that give multiple gains and class-AB analog output. In the LMT87 they are used in push-pull to enable an output current of 7 mA without extra parts. Further, its two possible housings have effectively 3 leads. One housing (the TO-92) has 3 leads mounted in parallel from one side. That arrangement permits the LMT87 to be glued on the metallic disk that is needed for the irrigation controller across the end of the hollow cylinder. In that way the ADC and the needed power supplies can all be at a distance, near the heater ring around the cylinder, where those heat sources will not affect the LMT87 temperature.
The data document for the LMT87 (at www.ti.com/lmt87-ca) shows that the LMT87 would be used with a generic ADC [in Fig. 12 on pg 12]. Details of the circuit controlling the ADC sampling are omitted.
Fig. 12 for the LMT87 shows a low-pass filter before the sampling switch. A Table on pg. 12 suggests values of a series resistor that would be needed if the capacitor value becomes large, e.g., 800 Ohms with a 1 uF capacitor. That RC time constant is 800 us, which cuts off noise below 200 Hz. Sampling faster than 2 Hz is not needed for the irrigation controller application. Thus larger R and C values are needed, e.g., larger by 10x each for sampling at 2 Hz.
3) Is another temperature transducer suited to ADC use more desirable?
4) A reader more expert than I am is needed to suggest a specific ADS system, probably some MSP430 micro-controller.
Charles KenKnight, PhD
<chaskenk@yahoo.com>