Author Topic: Precision pt100 frontend  (Read 12302 times)

0 Members and 1 Guest are viewing this topic.

Offline awallinTopic starter

  • Frequent Contributor
  • **
  • Posts: 694
Precision pt100 frontend
« on: June 17, 2013, 10:50:34 am »
Hi all,
I'm designing/simulating a precision pt100 frontend for temperature measurement between +22C and +42C:
http://www.anderswallin.net/wp-content/uploads/2013/06/pt100_piiri.png

The main ideas are:
- 4-wire connection for obvious reasons
- ratiometric output, i.e. measure both sensing resistor voltage and reference resistor voltage - the ratio of these two should be insensitive to drift in the sensing current
- switchable +/-500uA "AC"-excitation. Using the sum signal observed with +500uA and - 500uA should get rid of asymmetric error sources (thermovoltages etc.)
- output centered around +22C by way of the 112 Ohm reference resistor

Some of the details:
- By simulation I found that an integrated instrumentation amp has quite high input offset voltage (which results in asymmetric output when we switch the sensing current direction) as well as input bias current (which sources or sinks a variable amount of current into the amp inputs). Hence I've assembled the instrumentation amps from zero-drift op-amps (AD8551) with <10pA input bias currents and low input offset voltage.

Any ideas for how to switch the +5V and -5V voltages to change the direction of the current source? An analog switch with low resistance?

I have not run a sensitivity or temperature-drift simulation yet, but the design goal is stability of 1 mK (0.001 K or C) over long time periods (weeks).

The +/-10V output will be AD-converted at a rate of about 1 SPS, so the analog bandwidth can be < 1Hz. I should probably add low pass-filters into the instr-amps or between them.

Any other improvement ideas or suggestions?

Anders
 

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13694
  • Country: gb
    • Mike's Electric Stuff
Re: Precision pt100 frontend
« Reply #1 on: June 17, 2013, 11:06:42 am »
That seems ridiculously complicated.
Use a delta/sigma ADC that has a differential reference input.
You can then measure the ratio of PT100 resistance to a reference resistor directly, and you don't need any amplifiers or a precise current or voltage source - just one with low noise. 
Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Offline chickenHeadKnob

  • Super Contributor
  • ***
  • Posts: 1054
  • Country: ca
Re: Precision pt100 frontend
« Reply #2 on: June 17, 2013, 11:32:20 am »


was just looking at this. Analog Devices has a 24 bit adc ad7195 with differential input Programable gain and AC or DC excitation.:

http://www.analog.com/en/analog-to-digital-converters/ad-converters/ad7195/products/product.html
 

Offline awallinTopic starter

  • Frequent Contributor
  • **
  • Posts: 694
Re: Precision pt100 frontend
« Reply #3 on: June 17, 2013, 11:44:41 am »
That seems ridiculously complicated.
Use a delta/sigma ADC that has a differential reference input.
You can then measure the ratio of PT100 resistance to a reference resistor directly, and you don't need any amplifiers or a precise current or voltage source - just one with low noise.

If I understand correctly this simple approach will always put the signal "on top of" the basic 112 ohm (or similar) resistance of the sensor at ambient temperature.
The pt100 resistance will vary between 108 and 116 ohms in my application. In relative terms that's 8/116 or about 7%.
Using only 7% of an ADC's dynamic range will waste about 3 to 4 bits of resolution.

A possible solution is to use a bridge-connection to center the signal around 112 Ohms, but that will have worse nonlinearity than my suggested circuit.
A variation I've seen is to use a single chip containing two precisely matched current sources, one is put through the sensing-resistor, the other through a reference resistor, and the voltage difference detected.

Anders
 

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13694
  • Country: gb
    • Mike's Electric Stuff
Re: Precision pt100 frontend
« Reply #4 on: June 17, 2013, 12:27:48 pm »
That seems ridiculously complicated.
Use a delta/sigma ADC that has a differential reference input.
You can then measure the ratio of PT100 resistance to a reference resistor directly, and you don't need any amplifiers or a precise current or voltage source - just one with low noise.

If I understand correctly this simple approach will always put the signal "on top of" the basic 112 ohm (or similar) resistance of the sensor at ambient temperature.
The pt100 resistance will vary between 108 and 116 ohms in my application. In relative terms that's 8/116 or about 7%.
Using only 7% of an ADC's dynamic range will waste about 3 to 4 bits of resolution.


Maybe, but bear in mind you can easily be starting from 20-24 bits so that may not be an issue. Some ADCs also have programmable gain to optimise the input range. You can also optimise the value of the reference resistor and drive current to maximise resolution in the area of interest.
Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Offline poorchava

  • Super Contributor
  • ***
  • Posts: 1672
  • Country: pl
  • Troll Cave Electronics!
Re: Precision pt100 frontend
« Reply #5 on: June 20, 2013, 07:34:14 am »
I'd go for National Semiconductor's (yeye, I know ... "TI") LMP90100 sreies of chips. I have been using them for different projects for RTDs, thermocouples and semiconductor sensors (and stuff like force transducer).

For RTD you can find quite a detailed description in the datasheet, and there's also an application note.

The circuit itself consists or 2 8-to-1 muxes then differentially amplified by programmable gain amplifier and then either 16- or 24-bit sigma-delta ADC. As a bonus up to 2 matched programmable current sources.

That's actually the most versatile sensor interfacing IC i know (let alone the OP07 :D).
I love the smell of FR4 in the morning!
 

Online Andy Watson

  • Super Contributor
  • ***
  • Posts: 2074
Re: Precision pt100 frontend
« Reply #6 on: June 20, 2013, 09:34:10 am »
- ratiometric output, i.e. measure both sensing resistor voltage and reference resistor voltage - the ratio of these two should be insensitive to drift in the sensing current

I'm struggling to see this. If I've read your circuit correctly you appear to have offset the pt measurement, this is not the same as making a ratio. The scaling of the pt resistance is still dependent on the excitation current.
 

Offline babysitter

  • Frequent Contributor
  • **
  • Posts: 893
  • Country: de
  • pushing silicon at work
Re: Precision pt100 frontend
« Reply #7 on: June 20, 2013, 08:38:26 pm »
22 to 42 degrees centigrade ? Hey that sounds like you go for body temperature!

Just do as almost everybody else in the patient monitoring industry and go for a Thermistor which is compatible to YSI4000 :)
The deltaR vs. R is much higher in that range, and they are well-tolerated and the resistance is big enough to get rid of the 3/4 wire requirement.

I'm not a feature, I'm a bug! ARC DG3HDA
 

Offline ftransform

  • Frequent Contributor
  • **
  • Posts: 728
  • Country: 00
Re: Precision pt100 frontend
« Reply #8 on: June 20, 2013, 09:10:14 pm »
It is easier to get greater resolution at that range using a thermistor. I think you might find good quality thermistors which may beat out a typical RTD at this range, but don't quote me on it.
 

Offline babysitter

  • Frequent Contributor
  • **
  • Posts: 893
  • Country: de
  • pushing silicon at work
Re: Precision pt100 frontend
« Reply #9 on: June 21, 2013, 08:25:40 am »
I am back, trying to get you off the Pt100 trip :)

For convenience, the attached picture shows a teardown of a temperature measurement module circa-1988 from a Siemens patient monitor, accepting such YSI400-compatible Thermistors. (Besides 2x temperature probe input with big sockets, it also has optical communication facilities and a wireless power supply with the pot-core on the other end.)

The Thermistors are usually specified for +/-0.1°C accuracy (not resolution) in body temp range, R25 is 2252K. Big "headphone style" sockets but with way bigger diameter accept the temp. probes.

This page can be used to give you a nice lookup table, I bet you find the steinhart-hart-coefficients in the web too to set up a online calculation.

http://www.advindsys.com/ApNotes/YSI400SeriesProbesRvsT.htm

You might find cheap disposable  probes of that stype.

Greetings
Hendrik
I'm not a feature, I'm a bug! ARC DG3HDA
 

Offline poorchava

  • Super Contributor
  • ***
  • Posts: 1672
  • Country: pl
  • Troll Cave Electronics!
Re: Precision pt100 frontend
« Reply #10 on: June 21, 2013, 06:49:45 pm »
With pt100 the voltage reference is actually no needed.  What you need is a reference resistor.  You can then measure voltage drop across it and and compare to voltage across the pt100.  On the other hand really low tempo and high precision resistor costs quite a lot :-(
I love the smell of FR4 in the morning!
 

Offline awallinTopic starter

  • Frequent Contributor
  • **
  • Posts: 694
Re: Precision pt100 frontend
« Reply #11 on: June 24, 2013, 12:22:57 pm »
- ratiometric output, i.e. measure both sensing resistor voltage and reference resistor voltage - the ratio of these two should be insensitive to drift in the sensing current

I'm struggling to see this. If I've read your circuit correctly you appear to have offset the pt measurement, this is not the same as making a ratio. The scaling of the pt resistance is still dependent on the excitation current.

Multiplication or division in analog electronics is not nice, so the idea would be to do the division in the digital controller.
The measured signal voltage will be: (R_pt100 - R_ref)*I
measured reference voltage: R_ref*I
The ratio of these should be insensitive to (small, slow) variations in the sensing-current I.

Thanks also to everyone else for their comments!
 

Offline poorchava

  • Super Contributor
  • ***
  • Posts: 1672
  • Country: pl
  • Troll Cave Electronics!
Re: Precision pt100 frontend
« Reply #12 on: June 25, 2013, 07:35:22 am »

- ratiometric output, i.e. measure both sensing resistor voltage and reference resistor voltage - the ratio of these two should be insensitive to drift in the sensing current

I'm struggling to see this. If I've read your circuit correctly you appear to have offset the pt measurement, this is not the same as making a ratio. The scaling of the pt resistance is still dependent on the excitation current.

Multiplication or division in analog electronics is not nice, so the idea would be to do the division in the digital controller.
The measured signal voltage will be: (R_pt100 - R_ref)*I
measured reference voltage: R_ref*I
The ratio of these should be insensitive to (small, slow) variations in the sensing-current I.

Thanks also to everyone else for their comments!

That's how I usually so it. make a current source, it doesn't have to be high precision, a transistor, 2 resistors and 2 diodes will do just fine. Connect RTD and reference resistor in series. Measure the differential voltage across the reference resistor and the RTD. Vrtd/Vref=Rrtd/Rref. Done.
I love the smell of FR4 in the morning!
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf