pt100 design project

Discussion in 'The Projects Forum' started by jiwhun1989, Apr 8, 2014.

  1. jiwhun1989

    Thread Starter New Member

    Jun 18, 2013
    8
    0
    Hey everyone for my project I'm designing a circuit board using pt100 (temp. Rtd sensor). I have to pick out an appropriate sensor , adc , and uc. I also need a SCA design to amplify my voltage from my sensor.

    Design must meet following specification:

    measurement of instrument will be between -40 c to +85c.
    Total error must be less than 0.1 c
    resolution of reading must be at least 24 bit
    Instrument must be capable of operation with supply voltage with supply range of 3-6 VDC
    Total power consumption of device must be less than 75mW

    For my power module it must be based on LP4915C LDOs

    Im not quite sure how to approach this project . I sort of picked out my sensor but I need some guidance to continue on with this project.

    Thank you
     
  2. THE_RB

    AAC Fanatic!

    Feb 11, 2008
    5,435
    1,305
    Who gave you those weird design specs? :eek:

    (A) Not many off the shelf temperature sensors will read better than 0.1'C. Most are good for approx 0.5'C accuracy.

    (B) "at least 24 bit" over 85'C?
    24 bits gives you a resolution of 16.7 million counts, over 85'C that is a resolution of 0.000005'C per bit. That's just plain stupid when you will be already struggling to find a sensor that can match your 0.1'C accuracy spec!

    Good information would be to tell us what this is measuring, what it connects to, and why it needs to be super-low power consumption.
     
    #12 likes this.
  3. #12

    Expert

    Nov 30, 2010
    16,248
    6,745
    I thought of that, but the number is so outrageous that I believed I must have done it wrong.
     
  4. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    2,673
    2,712
    If this were my project, using this, I'd use something similar to a TI ADS1242.

    Drive the RTD with 1ma and use the built-in PGA for front-end gain.

    Having 24 bits to play with is useful, and relatively inexpensive.

    Edit: Oh, and drive the RTD in series with a (precision) resistor to generate the Vref. This will eliminate the need for a precision reference source and calibration.
     
    Last edited: Apr 9, 2014
  5. #12

    Expert

    Nov 30, 2010
    16,248
    6,745
    This still boggles my mind. Counting to 16 million with a thermocouple is like measuring tree growth with a stop watch.
     
  6. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    2,673
    2,712
    A) it's not a thermocouple...it's an RTD.

    B) you are in possession of an instrument that measures temperature via a thermistor and 24 bit A/D.

    One of the benefits of 24 bits is that you can measure small signal changes buried in large offsets without requiring a drift-prone analog offset on the front end.

    The RTD he wants to use varies between 100 and 138 ohms from 0C to 100C. Applying a constant current, say 1mA, results in a 38mv change (from 0 to 100) riding on top of 100mV. You could just offset the 100mv prior to the A/D, but it's easier and more stable just to use more bits.

    Besides, 24 bits just means you have 24 bits worth of possible codes that come out of the device. All 24 will not be significant. With a very small signal, lots of PGA gain, and *perfect* noise free design, you'd be lucky to get 16 to 18 significant bits out of a 24 bit converter.

    Further, it appears that 24 bits is the assignment. Are you going to argue with the professor? Or just do it? It's possible the professor simply wants his students to gain experience with some of the more leading edge hardware available by building something without a lot of complexity.
     
  7. #12

    Expert

    Nov 30, 2010
    16,248
    6,745
    And you could still see the change in temperature when somebody opens the bathroom door down the hall. :D
     
  8. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    2,673
    2,712
    Ok. Let me put it this way. The minimum number of bits he requires, while still being able to avoid analog offsetting, is 13. This will give 0.05C resolution. But he'll have no headroom to do any additional noise processing. I'd go with a minimum 16 bits on this.

    But 24 bits is far more fun.
     
  9. MrChips

    Moderator

    Oct 2, 2009
    12,415
    3,354
    Let's say 24-bit is the specs in the assignment. Maybe the professor just wants you to understand and appreciate what 24 bits resolution means and the meaning of ENOB. Off the top you are going to lose 4 bits. I wouldn't be surprised if you lose another 4 bits owing to noise. Then you're down to 16 bits.

    -40 to +85°C is as span of 125°C.
    0.1°C resolution is 1 part in 1250. Just over 10 bits. Hence the temperature readout needs about 11-12 bits. Since the RTD circuitry is going to provide a non-linear output, a 16-bit ADC ought to do the job.
     
  10. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    2,673
    2,712
    No sir. You must account for the fact that the the change in resistance (over the temperature range) is smaller than the resistance at the lowest temperature.

    For my estimation, I used 0C and 100C (it was easier...), not -40C and +85C. 0C is 100 ohms, 100C is 138 ohms, giving roughly 0.38 ohms per C or 0.038 ohms per 0.1 C. If you discount the 100 ohm offset, yes, 10 bits will suffice. But you need a total number of counts of 3631 to account for the offset, giving 12 bits. But, to keep quantization error to less than 0.1C, you need another bit, giving 13.

    I then like to throw in a few more normally distributed noise bits for additional digital filtering.
     
  11. MrChips

    Moderator

    Oct 2, 2009
    12,415
    3,354
    You didn't read my post carefully.
     
  12. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    2,673
    2,712
    Apologies if necessary. I re-read it but still don't see where you account for the offset.
     
  13. THE_RB

    AAC Fanatic!

    Feb 11, 2008
    5,435
    1,305
    I've done a lot of temperature control projects with ADC, including keeping a xtal regulated to approx 0.01'C temperature using a 10 bit ADC.

    Voltage noise from the temperature sensor and wiring alone can be >1 ADC count on a 10bit ADC. So that equates to >14bits of voltage noise on the 24bit ADC and maybe 10bits of signal left over. Sounds like a stupid spec to me.
     
    #12 likes this.
  14. jiwhun1989

    Thread Starter New Member

    Jun 18, 2013
    8
    0
    Sorry guys i haven't' been replying and thank you everyone for your comments.

    I am still new to this design thing so bare with me guys , so my instructor told me that he will narrow down list of ADC 's we can pick next week in the lectures so i will get back on that for you guys later.

    But for this week i just have to get the power module done with Low drop out regulator (How do i know how much to voltage to output using LP4951C) and also get my SCA part of design done. This may sound like a stupid question but here it goes ... I know the Pt100 RTD sensor reads the temperatures and converts them into resistance so how do i convert this resistance into voltage ( some of you were suggesting me to drive current source of 1ma?) and how do i know how much voltage i need in the output of SCA? Once again sorry for asking too many stupid question. I am still new to this whole thing

    List of Op amps instructor suggested (For SCA) are: OPA336, 3 Instrumentation op amps( ina155,ina327,ina333)
     
  15. jiwhun1989

    Thread Starter New Member

    Jun 18, 2013
    8
    0
  16. #12

    Expert

    Nov 30, 2010
    16,248
    6,745
    That file won't download for me.
    I hope you can download it and post it here.
     
  17. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    2,673
    2,712
    Technically, you don't need an analog front end for this project. The ADS1242 that I mentioned above has a digital PGA (programmable gain amp) that allows you to measure very small voltages.

    Additionally, if you use a ratiometric divider (RTD plus fixed percision resistor), you don't need a precision current source.

    For example, if your Vdd is, say nominally 3V, and you desire a 1mA RTD current and a ~2.5V voltage reference, you need a total of 3K in the divider (3V/1ma = 3Kohms). The precision resistor needs to drop 2.5V (for the reference), so a 2.5K 0.1% would be chosen. The RTD is nominally 100 ohms, so you just need an additional (non-precision) 400 ohms to make up the difference.

    Now, if your prof insists on an analog front end, then I would say he is now officially wasting your time...

    Does this help?
     
  18. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    2,673
    2,712
    BTW, what the heck's a SCA? You don't really mean CSA (chopper stabilized amp), do you?
     
  19. jiwhun1989

    Thread Starter New Member

    Jun 18, 2013
    8
    0
    signal conditioning Amplifier design (to Amplify the small voltage signal coming from my RTD) to like 3V using op Amps
     
  20. jiwhun1989

    Thread Starter New Member

    Jun 18, 2013
    8
    0
    List of requirements i have to complete :

    schematics for power module based on LP495x
    and tps791xx LDOs

    Detailed Design Analog module based on INA155/327/333
    Instrumentation amps or OPAx336 op-amp

    Schematics for the Digital module (A/D selection)
    outline the tasks need to be performed by the
    microcontroller
     
Loading...