thoughts about getting more precision without ref-voltage

StefanL38

Hi,

I want to use a PTC KTY-81-210 for measuring temperatures.

Now I have the following idea. The PTC is part of a voltage-divider.
+Vss

---

8500 Ohm----M

---

PTC(1300-4000 Ohm)---GND

I guess usually an external voltage-reference would be used to supply a stable voltage
as the reference for an ADC. (MCP3208 12 bit)

Now I played around with LibreOffice-calc to see what resolution I would get out of this setup.
Somehow I hat the idea to connect the reference-input of the ADC to a lower voltage than Vss.
If this is done through a voltage-divider 1 to 1 then if the supply-voltage would go up and down a bit

the voltages of the 8500 Ohm-PTC-divider and the reference input go up and down
with the same ratio and the digit-value-range stays the same.

Example connecting the 8500 Ohm-PTC-divider just to 3.3V powersupply of the propellerchip

For temperatures between -20°C and 100°C (range 120K) the resistance of the PTC varies
from 1367 Ohm to +100°C 3392 Ohm.

At "M" the voltage changes from 0,456V to 0,940V. Difference 0,484V
Through the the 1 : 1 voltagedivider the reference-voltage of the ADC is 1.65V

1.65V / 4096 digits = 0,4028 mV/Digit. Related to the 484 mV this means 1201 digits
which means aprox. 0,1 K / Digit resolution which would be pretty good.

As long as I disregard the non-lineraity of the PTC are my thoughts right?
or did I oversee something important?

As for the non-linearity of the PTC I'm thinking about a calibrating algorithm that would work this way:

A high-precision temperature-sensor is used as reference. Then connecting the PTC to the ADC.

Putting the reference sensor and the PTC near to each other into hot water while the propeller-chip aquires data from both sensors. As the water cools down the calibrating code stores a table of values temp (from Ref-Sensor) Digits from PTC.
Some hours later I put some ice into the water to cool it down to 0°C.
Or taking methanol out of the superfridge (-18°C) and let it warm up slowly to ambient temperature and then the calibrating is done.

What do you think about this ideas?

best regards
Stefan

kwinn

Your method of calibration should work well, but I am not so sure about the voltage divider idea. Since the PTC resistance varies with temperature the divider ratio for the resistor/PTC will vary quite a bit ( resistor/1367 to resistor/3392 ). Without a stable reference I don't see how you could compensate for that.

Beau Schwabe

One way to use the Vref is to create a differential signal and track the upper leg with Vref and sample on the lower leg. ... Or keep the Vref at 3.3V and look at the difference in both legs differentially through an instrumentation amplifier before an ADC.

                                 o--> V2
            3.9k           T     |    
+ 3.3V >----/\/\----o----/\/\----o----/\/\----< GND
                    |                 3.9k
                    o--> V1

V1 - To Vref
V2 - To ADC

or, send V1 and V2 to an instrumentation Amplifier to reject any common mode noise prior to sending the output to an ADC.
Using the Vref as V1 however, also has an effect of rejecting common mode noise without an instrumentation Amplifier.


Note: Notice what happens to the ADC output on the attached spreadsheet when the Voltage supply changes (only change the 'green' fields)