RTD help

Hi, Joseph. I guess the right answer would be, "It depends".

First off, if you've got a 1.1V(?!) ADC reference voltage, you might have a little trouble with a Wheatstone bridge whose output is centered around 2.5V.

Second, a straight Wheatstone bridge output isn't linear. Since you suggest you're bringing the output to a computer or microcontroller (ADC), I guess you'd be linearizing the signal in software, but still...

If you've got a computer or microcontroller at the receiving end, things can be a lot simpler than that. Maxim/Dallas Semi and several other manufacturers make semiconductor temperature sensors that will give you a straight serial input to your uC accurate to within 2 degrees C. That's as simple as you can get.

If you're concerned with tenths of a degree, though, an RTD might be the way to go. For better accuracy and linearity, it might be better to cobble up a precision current source for your RTD, and then use a diff amp with an offset voltage for the sense terminals as your input to the ADC.

Again, it all depends on your application, and your project requirements. If you want more help, you might want to give some more information, like:

• You've said long-term accuracy and drift aren't important. What is your minimum precision specification (degrees, tenths of a degree, C or F or what?)

• Exactly where is your ADC signal going?

• Is this a school project or a real problem? If it's real, what is the application for? Describe what you're trying to do. If it's a school project, what level (trade school, college class project, college senior project)? Also, when's it due?

There are so many different ways to do this problem.

Cheers Chris

Hi, Joseph. I guess the right answer would be, "It depends".

First off, if you've got a 1.1V(?!) ADC reference voltage, you might have a little trouble with a Wheatstone bridge whose output is centered around 2.5V.

Second, a straight Wheatstone bridge output isn't linear. Since you suggest you're bringing the output to a computer or microcontroller (ADC), I guess you'd be linearizing the signal in software, but still...

If you've got a computer or microcontroller at the receiving end, things can be a lot simpler than that. Maxim/Dallas Semi and several other manufacturers make semiconductor temperature sensors that will give you a straight serial input to your uC accurate to within 2 degrees C. That's as simple as you can get.

If you're concerned with tenths of a degree, though, an RTD might be the way to go. For better accuracy and linearity, it might be better to cobble up a precision current source for your RTD, and then use a diff amp with an offset voltage for the sense terminals as your input to the ADC.

Again, it all depends on your application, and your project requirements. If you want more help, you might want to give some more information, like:

• You've said long-term accuracy and drift aren't important. What is your minimum precision specification (degrees, tenths of a degree, C or F or what?)

• Exactly where is your ADC signal going?

• Is this a school project or a real problem? If it's real, what is the application for? Describe what you're trying to do. If it's a school project, what level (trade school, college class project, college senior project)? Also, when's it due?

There are so many different ways to do this problem.

Cheers Chris

Or even simpler, use an LM35 (or related IC). Just put your 5 V in, and connect the output to your ADC for a linear 10 mV per degree C. But, like the other poster in this thread said, the best solution is going to depend on your application.

--
Steve

Thermocouple twisted ends and ohmmeter?

How would you amplify the T/C so I can read it wth a 10-bit ADC? thx

...plus I have to limit the current to 1mA! Oh well.

I will be linearizing via software.

I need faster response than they can offer.

Ok...how about a LM4040 4.096V ref diode and a couple resistors to give me

1ma throught the RTD. 1mA across the PT100 RTD will output roughly .1V. A change of 1C will only give .1004V...not much of a delta! I was looking at the INA126 instrumentation amp...would you suggest a diff one (no pun intended)?

1 degree C (repeatable to +/- .1 in a 15s time frame.).

It is the ADC in a uC.

This is a real world app due asap.

thx, jj

To me, this sounds like a good application of a small bead thermistor with a series resistor. Connect this series pair across the full scale reference voltage for the ADC. Pick a series resistor that has a resistance about equal to the thermistor's resistance at the center of the measurement range, and you get the largest sensitivity. No amplification needed.

As usual, Mr. Popelish cuts to the chase for the best response. A thermistor can change 3% or more in value for a 1C change in temp.

A couple of notes, though:

• If you need any long-term accuracy in addition to your precision, be careful about the cheapie thermistors. Omega makes a series of thermistors that are well-characterized over a good range of temps (say, 25C to 100C).

• The series resistance method is best for changes in temp over a small range. For larger ranges, you may be better off with multiple current sources or a fixed current source and a programmable gain amplifier.

• If you're going to be running a constant power across the thermistor, be aware that self-heating may be a bit of an issue. If so, allow time for the temp to stabilize before beginning precision measurements.

• You've said you need faster speed than you can get from the units with a serial output. Be aware that your main issue may actually be the thermal constant of the sensor itself (the time required for a sensor to indicate 63% of delta t), particularly if it's in free air instead of a liquid. This could easily be 1 to 10 seconds, which makes the delay of the digital end look less significant.

You might want to look around the Omega.com website for some helpful information on thermistors, particularly these:

Good luck Chris

Ain't that the truth!

$$15 cough. Decided to browse Digikey and do some calculations: US Sensor, 10K, $.80. DO-35 pkg, 5 Second time constant in still air. I'll use 6.2K as my series resitor, decided to use 5V (Vcc) as V ref.

Calculated my max heating is 1.0mW which is below the 2mW for my part. This gave an error .25 degreeC, which is OK since I'm only reading temperature deltas.

LM34 is 1 minute.

Sincere thanks to all, jj

I'd like to differ with all the other posts: You say that actual temperature is not needed. I'd like to advocate using a simple diode fed from a constant current source. Essentially, you are measuring the resistance drop of the diode with temperature.

This can be incredibly linear over a huge temperature range, basically from liquid helium to junction melting temperatures. The only drawback is that (by itself, anyway) it is uncalibrated. But if that doesn't matter, it is certainly simple, and way more linear than a thermistor. If even linearity is not critical, it is often made even simpler by cheating on the current source, since the resulting nonlinearity can be dealt after digitization.

Incidentally, this is essentially the method used in precision temperature chips. They get around the calibration issue by running the diode at 2 different currents whose ratio is accurately known, which allows absolute temperature to be computed.

Best regards,

Bob Masta dqatechATdaqartaDOTcom D A Q A R T A Data AcQuisition And Real-Time Analysis

Home of DaqGen, the FREEWARE signal generator

OK. Try to remember, though, that many times, the hardest part of any engineering project is knowing when not to cut corners.

I like to get the job done, and I know I value my time, even if (and especially if) my employer is getting uncompensated overtime because I'm still playing around with it.

A few extra bucks for a characterized thermistor can make all the difference if it turns out you might need some accuracy as well as good short-term precision.

Also, it means you can just drop in a replacement without recalibrating your project if the sensor is damaged.

Looking back, I think many of the "do-overs" I'd like to have started from trying to cut one small corner to achieve a more elegant or minimum-cost solution. Many times, just splurging a little on that one component would have kept me out of problems. If you're just trying a "one-off" control circuit, you should approach the problem a lot differently than if you are going to be making them by the hundreds or thousands.

Something to consider, for what it's worth. That and $3.49 will get you a grande cappucino at Starbucks.

Cheers Chris