next topic »

[Reply #15 on:]

It is possible to log up to 40,000 samples per second to a SD card with an Arduino.  See the binaryLogger example in fastLoggerBeta20110802.zip  here http://code.google.com/p/beta-lib/downloads/list.

I used a 12-bit MCP3201 ADC.

[Reply #16 on:]

I'm constructing a 'thermometer' using a TI ADS1248.

It's a 24bit ADC with current generator, bias voltage, reference voltage, 7 input channels (or 3 differentials), PGA and can sample up to 2kSPS.

I've got a working breadboard prototype (a real mess) which measures an RTD (4-wire measurement) for cold-junction compensation and measures a thermocouple. At 1SPS, I've got a stDev of 0.005C.

I've build a small library which communicates with the chip over SPI. I'm planning to open-source the lib if there's interest but I'll have to clean it up a little bit first.

[Reply #17 on:]

I've got a stDev of 0.005C.

That seems like a hellaciasly small standard deviation... what kind of temp sensor are you using, and what is the possible temperature range?

[Reply #18 on:]

I just re-ran a small test. It's a Pt100 thinfilm RTD

I got these numbers for my room, it's in degrees celcius:
24.60506
24.60647
24.61212
24.60990
24.61151
24.60405
24.61272
24.61796
24.61050
24.62159
24.63429
24.63066
24.63429
24.63288
24.63308
24.63530
24.63207
24.63026
24.61454
24.62824
24.60869
24.61333
24.61292
24.61353
24.62018
24.62441
24.63611
24.63167
24.62623
24.62663
24.63046
24.62119
24.61957
24.62340
24.61917
24.62401
24.62361
24.62925
24.61796

that gives you a stDev of 0.0093 but you can see it fluctuating, the temperature in my room just isn't that stable.

If I measure a resistor with very low temperature coefficient I get a stDev of 0.0046 so the measurement error is probably closer to 0.005 and my room fluctuates another 0.005

[Reply #19 on:]

The specs look like the pt100 is good to 600C. How high have you gone above room temp with your measurement?

[Reply #20 on:]

not much, up to 35C, the temp of my fingers I'm using the RTD to do the cold junction compensation for a thermocouple so it only measures room temperatures in my setup.

Check out omega.com for a huge range of sensors. High temp is usually measured with a thermocouple, but those are less accurate and more susceptible to noise.

What you want seems a bit unrealistic though. Even a fast response sensor has a delay of 1sec and the thermal mass of the sensor will act like a low-pass filter. What you'll measure will be a moving average of the actual temperatures.
http://en.wikipedia.org/wiki/File:MovingAverage.GIF this is how a moving average looks like. It's from the Dow Jones stock market but the principle is the same. Suppose the green values are the actual temperatures, then what you'll measure will be the yellow MA line. A small short spike in temperature will cause only a small bump in the MA.

Noise and samplerate are also inversely correlated. The 0.005C stDev is at 1SPS. You get an n bit gain with 4ˆn oversampling.

(disclaimer: I'm not an EE, I just spend to much time on wikipedia )