06-05-2012 05:00 AM
Solved! Go to Solution.
06-05-2012 08:19 AM - edited 06-05-2012 08:23 AM
The resolution of your final measurement will depend on exactly what combination of thermistors and resistors you use. Any National Instruments board which can supply both output and input will work for you, provided you pick the right set of resistors/thermistors. The 6009 certainly qualifies.
That said, however, if you are concerned about cost, you can do things much more cheaply, but pay attention to stability and resolution. This is an interaction between your hardware and software. Two possibilities:
Good luck. Let us know if you have further questions.
06-06-2012 02:23 AM
Thanks for reply!
Another motivation to use thermistor to measure temperature, the higher resolution. We have some NI USB-TC01 cheap devices, but their resolution (also accuracy) is quite low: about 0.1 Celsius.
We have several USB-6009 and we could try to build with this a thermistor sensor. Lets consider we have a thermistor 3 kOhm (at 25 Celsius) serially connected to a 3 kOhm precision resistor. If we want to measure a temperature change of 0.01 Kelvin, it will result (using the Steinhart-Hart equation with the coefficients mentioned in wikipedia) in about 0.28 mV voltage change output of the voltage divider. The specification of the USB-6009 states that the A/D resolution is 14 bits, which means 0.15 mV nominal resolution (0-2.5 Volt range). So we are already very close to the theoretical resolution of the USB-6009. Also, using thermistors it is possible to reach resolution under 1 mKelvin, and certainly this level is not reachable by the USB-6009...
What do you think, is our rough calculation above right?
thanks!
06-06-2012 02:46 AM
Also, I guess if we use a Wheatstone bridge configuration, two precision resistors and two thermistors (all of the same nominal resistivity), we can get a better sensitivity? (so the output voltage is zero at the nominal temperature, and changing polarity above and under it...)
06-06-2012 06:25 AM
Going to 10mK resolution with high repeatability the cheap way is tricky. (absolute deviations ignored 😉 )
Keep in mint the drift of your source..
I just have seen a solution using a REF200 (a dual 200µA precision current source) a 1k reference resistor and a Pt1000. R and Pt have their own current source, the voltage difference is amplified by a instrumentation amplifier to enable a SE measurement in an optimal range.
But I like the soundcard solution. Using a precision resistor outside the NTC range (of interest) and a DPDT relay to swap the channels (driven by an astable 555) and using tone detection on the four measurements I would bet one can beat the 14bits ... leaving the uncertainty of the resistor and the nonlinearity of the soundcard AD at a modulus frequency.
06-06-2012 08:10 AM
I would like to second Henrik's comments about repeatability at high resolution. Temperature measurements with thermistors are especially tricky. You have to worry about:
In general, if you want anything more than about 1% repeatability, things start to get interesting. Of course, that is what makes it fun to do.
If you go the sound card route, don't forget that although most modern computers have fairly high resolution ADCs (some up to 24-bit at 96kHz), you need to look at the noise figures to get what you can actually resolve. Off-board solutions (e.g. M-Audio Delta 44) will give you better noise figures, since the ADCs are isolated from the noisy interior of your computer.
Good luck and have fun.
06-07-2012 11:16 AM
thanks everyone for the good advices!