I'm working on a PT100 based temperature probe sensor to read temperature difference between two points inside a pipe with flowing water, the requierements are:
- 1mK resolution over 0-45C range.
- dual PT100 setup to detect calibration discrepancies.
- 0.1C absolute accuracy would be optimal but what is actually important is getting reliable differential readings between matched probes.
- RS485 communication.
Restrictions I have:
The PCB can't be bigger than 100mm diameter to fit into the probe's head and since I need a 2xPT100 setup o don't have space for signal conditioning + ADC for both so I have to rely on high resolution ADC only.
What I have on the design table so far:
I'll be using a 2xPT100 DIN 1/10 for max accuracy (+-0.006 C error at 45C) with a 22 bit delta-sigma ADC (ENOB:21.9 but actually 20.9 since it uses two's complement output). This is the simplified circuit of one probe that show how both PT100 interface witch the ADCs and the MCU:
EDIT: Just to clarify, the PT100s inside each probe need to be compared against each other (in the MCU) so it can be detected when the probe needs calibration. Then temperature between probes must be compared to know the temperature difference between points.
For this circuit I get a signal that varies 12.6572mV from 0 to 45C (100-117.47 Ohm), then I get 2.27mK resolution and a 1.54mA current through the PT100, If I wanted to go down to 1mK I only see two solutions:
- Use a lower value for R1 to get more resolution and also lower Vref/Vex to adjust the current trough the RTD
e.g: R1=500Ohm for 0.975mK resolution and Vref/Vex= 1.024V for 1.65mA
- Keep the R1 and Vref/Vex values and get more resolution trough Oversampling and decimation method. The PT100 response time T90 typically is no less than 3 seconds so even with the 13SPS offered by the MCP3551 I think I should be able to get those extra 1.2 bits I need but I'm not sure.
Is there anything else I could do to improve the resolution?
I've selected the MCP3551 ADC because it's ENOB is pretty high 20.9 against all those 24bit sigma-delta ADCs out there that have no more than 19 ENOB, and besides, It's just an 8pin SOIC IC so if someone can suggest a non massive ADC that outperforms this one I'm open to change it.
If you have any question about the problem, please ask, if I've left any important information behind or think I could elaborate more, please ask.
EDIT: I add the whole system sketch
EDIT2: I've been suggested to use ADS124S06, It provides around 23 bit resolution without enabling the PGA,I could also enable the PGA and meet common-mode specs because this IC allows to connect the PT100s in parallel and use the voltage through Rref as level-shift. The downside is that it's a 32pin IC so routing it could be a nighmare...