1
\$\begingroup\$

I'm designing a temperature measuring circuit using a PT100 on a wheatstone bridge. In the temperature range I'm measuring (0-45C) I should get 0-25.5mV from the bridge and then amplify the signal to 0-5V using an instrumentation amplifier (I was thinking on something like the AD8221) so I could read it using a PIC.

Then here comes two problems:

1- Amplifier common mode:

I cannot afford any common-mode voltage. My main supply is +12V so I was thinking on using a charge-pump in order to provide +-12V supply to the in-amp and avoid output clipping.

2- Analog input protection The +-12V supply solution forces me to protect the PIC's analog input from voltages below 0V and above 5V using clamp diodes at the analog input but I don't actually know if this is valid solution or it affects the readings in any way.

I don't know if I'm in the right way with this approach so any insight or tip is appreciated. Thanks

\$\endgroup\$

2 Answers 2

2
\$\begingroup\$

Putting a single element variable resistor into a wheatstone bridge inherently produces non-linearities and to reduce these you need to have the three other fixed resistors a lot higher in value than the 100 ohms of the PT100 at 0 degC.

This of course reduces yield i.e. the output voltage variation with resistance is significantly diminished. Another problem is keeping the excitation voltage stable. Given that you are using an ADC I would consider making a ratiometric measurement where Vcc both generates excitation and is used as the ADC reference voltage.

Another consideration is that the ADC input to your PIC will not be perfectly matched to produce digital zero at precisely 0 volts and digital full scale at precisely 5 volts. It may also not have great linearity (look up the INL figure of the PIC's ADC in the data sheet) - I have seen INL errors of over 10 LSBs in some PICs and, if it is a 10 bit ADC, this represents a 1% full-scale error.

Rounding up, I would use a 1k resistor to feed the PT100 from local Vcc and this will generate a voltage range of 454.5 mV at 0 degC to 525.6 mV at 45 degC. I would then use an op-amp to amplify this by 8 and this produces a voltage range of 3.636 volts to 4.205 volts at the ADC input.

Sure, it only uses about 11% of the ADC range but that is still 116 LSBs and for a 45 degC range that is a resolution of 0.39 degrees.

If that appears too "klunky" then you might also consider that the PT100 accuracy may be worse. With some care you could re-scale the range by using a bias resistor fed into the inverting input of the op-amp: -

enter image description here

The above circuit (LH side for 117.47 ohms and RH side for 100.00 ohms) produces an output range from 455 mV to 4.561 volts i.e. less "klunky".

Non-linearities from the simple 1k/PT100 potential divider can be easily compensated by the PIC.

\$\endgroup\$
16
  • \$\begingroup\$ This looks simple and nice but I don't see how this solves the common-mode voltage problem, 0.45 to 4.56 V are still hard to get with a single supply op-amp at 0-5V, let me know if I'm wrong. \$\endgroup\$
    – Bizcochito
    May 9, 2018 at 10:17
  • \$\begingroup\$ @Bizcochito there are plenty of op-amps that are rail-to-rail input and output devices. All the leading manufacturers make them. Microchip and TI are probably the cheapest if cost is a big deal. You also need to be looking at devices with decent input offset voltages sub 1 mV. \$\endgroup\$
    – Andy aka
    May 9, 2018 at 10:23
  • \$\begingroup\$ If you don't mind I could suggest using the Vref- and a smaller refference to get rid of the dc offset and use the ADC more efficient. I'm still try to figure the solution to be VDD independent using also some amplification \$\endgroup\$
    – Dorian
    May 9, 2018 at 11:19
  • \$\begingroup\$ The minimum Vref+ - Vref- is 1.8V for PIC ADC Vref-> GND Vref+<VCC \$\endgroup\$
    – Dorian
    May 9, 2018 at 11:21
  • \$\begingroup\$ @Bizcochito Something like an MCP6021 springs to mind. \$\endgroup\$
    – Andy aka
    May 9, 2018 at 11:22
1
\$\begingroup\$

A variation to your approach is to use a high resolution Delta-Sigma A/D with fully differential inputs, precluding the need for a separate In-Amp. I have used this before to measure the output of a Wheatstone bridge containing a strain gauge.

The reference I'm including is the one I have used; others are available that can work equally well. Here is the Link

As shown, the A/D input connects directly to the bridge output. Interface to your PIC is SPI. Power requirement for the A/D is compatible with your PIC. The only other voltage required is bridge excitation.

enter image description here

\$\endgroup\$
1
  • \$\begingroup\$ This looks interesting, I have to check if I can drive both SPI and UART modules in the PIC at the same time since I need to output the data via RS485 \$\endgroup\$
    – Bizcochito
    May 9, 2018 at 9:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.