Precision thermometer: question / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

Precision thermometer: question

1 2 next

Move Thread

LAN_403

IamSmooth

Thu Feb 06 2014, 08:11PM

Registered Member #190 Joined: Fri Feb 17 2006, 12:00AM
Location:
Posts: 1567

I am building a precision LCD thermometer and would like to make sure I am going about it correctly. I am using an RTD probe that will have a resistance that varies from 20R to 100R (approximately). I can use an Arduino to measure the voltage to get the temperature. The arduino can measure from 0 to 5v. This is my full range which I want to use. The question is what is the correct and simplest way to do this. I am planning to do this:

I plan on using a resistive divider to get a reference voltage equal to the lowest voltage I expect to measure on the RTD. I will put both voltages into a non-inverting opamp. Ov = lowest voltage. I will also have a 10k resistance in series with the RTD so changes in RTD resistance are very close to linear changes in voltage. Since this voltage across the RTD is small, my opamp will have a gain such that the maximum voltage is 5v.

Is there a better approach? This method allows me to use trim-pots so I can calibrate my low and high values.

Sulaiman

Thu Feb 06 2014, 08:58PM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

At work we repair many types of temperature controllers,
quick answer; the time/effort/cost involved vs. ready made unit.
it's equivalent to designing and building a D.M.M. vs buying one.

Usually compensation for offset, gain, linearity and ambient temperature.

Have a look at Analog Devices website for too much information.
Microchip also have some good stuff as do Texas Instruments.

If you want to build a PRECISION anything, lots of research required.
( assuming PRECISION to mean better than commonly available)

The FIRST thing to check is that the sensor(s) meet your 'precision' requirements (e.g. gain, offset, repeatability, drift .....)
Some probes have a 2-wire rom inside with calibration mapping.

Wastrel

Thu Feb 06 2014, 11:38PM

Registered Member #4095 Joined: Thu Sept 15 2011, 03:19PM
Location: England.
Posts: 122

I think that is good advice. If you want to try anyway, I would avoid the bridge and go with an op amp controlling a transistor as a constant current source and use a 4 point probe. I'd put the op amps and anything sensitive to thermal variations (not the current source transistor) in a thermally controlled box.

The other thing is that repeated exposure to cryogenic temperatures may turn the sensor to dust after a few thermal cycles.

BigBad

Fri Feb 07 2014, 03:26AM

Registered Member #2529 Joined: Thu Dec 10 2009, 02:43AM
Location:
Posts: 600

fwiw I built an arduino based sous vide controller using a thermistor.

All I did was put it in series with a resistor put the 5v/0v from the arduino across them, and plugged the middle straight into the arduino's analogue input. I think the resolution was about 0.2 degrees.

I just took the data from the manufacturer and interpolated it; it seems to be within a degree accuracy without calibration. I'd advise you to ditch the pots; you can always calibrate it with software if you have some other way to measure temperature more accurately.

Shrad

Fri Feb 07 2014, 08:26AM

Registered Member #3215 Joined: Sun Sept 19 2010, 08:42PM
Location:
Posts: 780

you have to use a precision voltage reference and override the one inside the ÂµC of your arduino (there are config registers for doing so)

I advise you to power your design with 5V and use a precision low drift DC/DC converter to feed 3,3V to your ÂµC (so whatever your charge may be you're sure to always have a precision voltage to compare against your precision reference)

you also have to use a 4 wire RTD and use the shortest cable length you can, along with a precision instrumentation amplifier (linear appnotes deliver clear and precise examples)

another idea would be to use a wheatstone bridge which achieves a ggood precision if you have it temperature stabilized and use precision carbon resistors (1% laser trimmed)

but be aware that using a cheap arduino would render most of those precautions useless due to the design and ÂµC not being designed for precision (unless the design is checked to comply with best design practice and the ÂµC is a good one)

Sulaiman

Fri Feb 07 2014, 09:24AM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

+1 for calibration in software
... that's how most 'modern' temperature controllers do it
potentiometers have their own problems, and cost

Have you decided on required temperature range, precision and sensor required ?

So, you've chosen a suitable sensor,
next is to choose the precision for A/D conversion
8-bit = 0.4% resolution
10-bit = 0.1 % resolution
16-bit = optimistic (0.0015%)
To allow for unused ranges above and below required range etc.
aim for twice the precision you actually require,
e.g. 8-bit = 1%, 10-bit =0.2%

but, before you start,
My Maplin multimeter with thermocouple input and RS232 output cost GBP9.99
(-50C to +200C type-k thermocouple in this case)
.... ALL RANGES (V, I, R, C, TEMP etc.) displayed and/or recorded on P.C.
ELECTRICALLY ISOLATED, with nice enclosure, terminals and leads
(A USB I/F is better, but more expensive)
... do you need to D.I.Y. ?

P.S. many semiconductor manufacturer's application notes have done all the work for you, are pretty much guaranteed to work, and evaluation pcb/kit available.

Bjørn

Fri Feb 07 2014, 09:42AM

Registered Member #27 Joined: Fri Feb 03 2006, 02:20AM
Location: Hyperborea
Posts: 2058

Don't forget the noise you will get on the readings. Some Arduino boards are not made well enough to give precision readings.

...

Fri Feb 07 2014, 03:05PM

Registered Member #56 Joined: Thu Feb 09 2006, 05:02AM
Location: Southern Califorina, USA
Posts: 2445

In an application like this, where you are feeding a 10bit A/D converter, the normal precautions associated with 'precision' do not really apply.

I can say though that with a 100ohm RTD you will definitely want to use a 4-wire setup, with a current source wired across 2 wires, and a voltage sense across the other two. Ideally, you would derive the current from the arduino internal voltage reference (doing it this way removes any possible drift from the arduino reference), but that starts to get complicated. Although I bet if you thought about it for a while it would be possible to build the voltage->current adapter and instrumentation amplifier to read from the sensor output in a single quad opamp.

Alternatively if you switch to a 1kohm RTD the error effects from the contact resistance will be a lot smaller, and you might be able to get away with just making sure to use low resistance connections and still keep your 10bit resolution as long as you are not scaling the 10bit window over too small a temperature range (for a 1k RTD changes a few ohms per degree depending on the temperature)

Shrad

Fri Feb 07 2014, 05:39PM

Registered Member #3215 Joined: Sun Sept 19 2010, 08:42PM
Location:
Posts: 780

there are fast 12bit A/D converters in SPI... 12bit usually fills the gap cost between 10 and 16 bits and you gain a nice precision factor

there are also interpolation techniques you could use for gaining precision, and there are at least three or four different types of A/D conversion schemes used in ICs each with their advantages

Avalanche

Fri Feb 07 2014, 09:44PM

Registered Member #103 Joined: Thu Feb 09 2006, 08:16PM
Location: Derby, UK
Posts: 845

You can get a lot more than 10 bits of resolution from a 10 bit ADC if you oversample.

The basic method is to add a small amount of random noise to your signal, then (for example) 16 samples averaged will give you a 12 bit result, or 256 samples averaged will give you a 14 bit ADC result.

Obviously the drawback is that your effective sampling rate goes down rapidly with an increase in bit depth

1 2 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob