Temperature measurement maybe TC1047A

[asrock70]

I need to heat and measure the temperature of several components within one device.
Measured temperature up to 90C, required resolution 0.1C, required size so that even with the supply cable it fits into the cable eye,cable length up to 50 cm. A relatively cheap and robust solution allowing the measurement of at least 4 temperatures, evaluable with something like the STM32F303 I normally use thermocouples or DS18B20.
the state of the semiconductor market makes evaluation ICs for thermocouples expensive and the DS18B20 is too big in TO-92.
For a while, I was thinking about an NTC thermistor, as I am mainly concerned with temperatures above 60C and there the thermistor's characteristics are quite unsuitable. I don't think it's appropriate.
Another idea is the TC1047A. The SOT23 case with SMD 100nF and the three-wire supply cable fits into the cable eye, it is not a problem to cover it with epoxy. Output is 10mV/C 1.5V for 100C, If use internal PGA in MCU x2 have for 3,3V Vref max mesured teperture 115C and output 20mV/C if LSB of ADC is 0,8mV ,that's resolution 0,04C, 0,08 without PGA.
So much theory, how practical, what the lead wire will do, etc. Alternatively, what do you use as temperature sensors in similar cases?

 

[KlausST]

Hi,

I often use LM73 in SOT-23-6. I2C interface, several usable addresses.

Klaus

 

[wwfeldman]

if you use a device that gives current proportional to temperature, (like AD590)
the cable length is doesn't mater.

AD590 can be expensive

a resistor (or resistor and op amp) converts the current to voltage to
simplify(?) measuring

 

[andre_luis]

Measured temperature up to 90C, required resolution 0.1C,

Generally speaking, the wider the desired range, the worst the resolution. Such a high resolution is found in sensors designated for the medical area, given that the human body has a very strict range. You will find few options to meet this requirement, luckly Digikey have one that seems to fit: S-5851AAA-I6T1U.

Keep in mind anyway that I2C is not a bus designed to opperate offboard, therefore you'll need to reduce a lot the bus data clock to avoid issues.

​

 

[stenzer]

Hi,

with respect to space constrains as well as accuracy requirement the AS6221 from ams [1], might fit. It provides an accuracy of about +-0.1°C at 90°C (+-3 sigma).

[1] https://ams.com/temperature-sensors

BR

 

[asrock70]

Thank you all for the responses.

>Using the I2C bus.
It also occurred to me though, even though I2C is a bus primarily for PCB communication.
I was thinking about the MCP9800. They have a fixed address, Microchip makes versions A0 to A7, but realistically you won't get more than A0 and A5. The LM73 has a three-state address input and even on aliexpress you can cheaply buy nak 0 and 1 versions, i.e. 6 addresses
I2C max capacitnce for standard modede is 400pF , if use max 0,25 mm2 cable, max 50cm for sensot , will be cable capacitance 300pF, input/outuput capacitance for master MCU and 4 slave will be 5x5pF =25pF This gives us 75pF of margin per connector on the PCB. I thing on 100kHz with 2,2k pull up resistot it will work.
...but I'm still nervous about the I2C in the wires

 

[KlausST]

Hi,

it would be much easier if you give clear informations:

I need to heat and measure the temperature of several components within one device.

--> what does "several" mean? In my eyes 6 addresses are "several".
--> what does "within one device" mean? .. in the meaning of total cable length.

I2C bus length: It does not matter much if it is 20cm of PCB trace or 20cm of cable.

Klaus