-
Temperature Logging - What are some good entry points?
Posted by TheUnnamedNewbie on 13 Mar, 2018 08:37 -
So, I am a young and broke student (broke being relative of course, I can't complain compared to fellow students, but metrology isn't exactly the cheapest of hobbies...).
I am currently in the process of designing my own voltage source. For me this was a good project since when repairing multimeters that need calibration, I often require something like a 1 mV voltage, and generating that with a bench supply is tedious at best.
It has also been a learning experience for me to discover computer controlled instrumentation. My core meter is a Keithley 2000. I have no real references to compare this too, so for now it will have to do.
One of the things I want to try is measure the temperature coefficient of my board. I can see that if I make it very warm, it goes up by about 1/2 LSB (meter measuring it is not being heated up). However, I don't have any way to measure how warm I am making the system, so the next thing I need to get a real idea of the temperature coefficient is a way to measure said temperature.
This brings me to my question: What are some good entry points to do reasonably accurate temperature measurements? I don't need crazy precision or accuracy, as it will mainly be relative measurements, and I can't control the temperature that precisely anyways (the way I currently do the heating is by just having two big power resistors in the same box as the DUT, and then pumping 5/10 W into those). My requirement is that I can easily log this with my matlab code. USB would be nice, but I am willing to spend the extra few bucks on another USB-to-Serial converter. I don't have GPIB yet, though I am looking at it in the long run. Most of the logging devices I find on ebay etc are for use with cards, and ofcourse don't come with the cards (so the cost of entry is quite high). -
You can achieve temperature measurent using a readily and cheap available thermocouple sensor. You K2000 is able to convert the thermocouple voltage to a direct temperature reading. Excpect an achievable accuracy around 2 ... 3 °C.
Next stop would be a simple but precision NTC temperature sensor, you can get them at about 1 ... 2 EUR, at your favourite source like Reichelt or whatever, measure the resistance with any half decent Multimeter and do the conversion by yourself (or some simple scripting)
Next one: use that said NTC sensor and some arduino like board, you'd just have to connect a resistor to Analog Ref voltage and do the simple calculation in Arduino code. IMO the cheapest thermometer with logging capability you can build - An Arduino or similar, a NTC sensor, a resistor (of the same value as the 25°C value of the NTC), a 10nF cap for decoupling the analog input - and all the rest costs your labour but no money. A typical "precision" NTC temperatur sensor does not need any calibration to achieve +/- 1...2°C accuracy with this setup (done that many times before).
-
Maybe the simplest way is to use a LM35.
Give it 4V to 20V and it delivers an output voltage linearly proportional to centigrade temperature
10 mV/°C (2 °C to 150 °C) with an accuracy of ±0.5°C (at 25°C).
(VOUT= 10 mv/°C × T).
http://www.ti.com/lit/ds/symlink/lm35.pdf
-
[/quote]You can achieve temperature measurent using a readily and cheap available thermocouple sensor. You K2000 is able to convert the thermocouple voltage to a direct temperature reading. Excpect an achievable accuracy around 2 ... 3 °C.
Am I right in assuming that the error would be due to no cold-junction compensation?
I like the idea of the arduino. It seems so simple and obvious - I have all these arduinos laying round, I might as well use them. Perhaps I could build on top of this and look at getting a temperature sensor module I can read.
Since I have only 1 K2000, I'll have to put together a switching board. But I have a bunch of old Coto reed relays I might use, so I could do that as a next project.Maybe the simplest way is to use a LM35.
Give it 4V to 20V and it delivers an output voltage linearly proportional to centigrade temperature
10 mV/°C (2 °C to 150 °C) with an accuracy of ±0.5°C (at 25°C).
(VOUT= 10 mv/°C × T).
http://www.ti.com/lit/ds/symlink/lm35.pdf
This I like. I'll have to look at building a board to power a number of these. Thanks for the suggestion. -
You can achieve temperature measurent using a readily and cheap available thermocouple sensor. You K2000 is able to convert the thermocouple voltage to a direct temperature reading. Excpect an achievable accuracy around 2 ... 3 °C.
Am I right in assuming that the error would be due to no cold-junction compensation?
Yes, that's an error source, and you'd have to take the limited accuracy of the TC itself into account, so my guess is quite a rough guess. Both errors will add up, the offset caused by the Cold Junction, and the linearity and whatsoever error from the TC itself.
-
I like the idea of the arduino. It seems so simple and obvious - I have all these arduinos laying round, I might as well use them. Perhaps I could build on top of this and look at getting a temperature sensor module I can read.
What kind of sensor you use, depends on your SW skills ...
I don't like these LM35 (and similar others, there are sensors with I2C so you don't have to mess with analog) not too much, because they need at least 3 wires (more HW). A precision NTC uses 2 wires and doesn't care about polarity, but requires a bit more effort in SW - for calculating the resistance and then doing some lookup / interpolation to convert the resistance to temperature. Usually I trade less HW effort to a bit more SW for this kind of application. Results are quite similar.
Another advantage of the NTC over the LM35 is the ratiometric measurement, you don't need an accurate reference voltage (but stable). To achieve the accuracy using a LM35 type sensor, you'll need an accurate and stable reference voltage.
-
Quote
so you don't have to mess with analog
which mess?QuoteTo achieve the accuracy using a LM35 type sensor, you'll need an accurate and stable reference voltage.
nope -
Hello,
I've made two such temperature-controlled boxes so far, and have been logging temperature a lot lately.
https://www.eevblog.com/forum/projects/mini-tempco-characterization-rig/
https://www.eevblog.com/forum/metrology/temperature-stabilised-enclosures-heating-and-or-cooling/msg1410031/#msg1410031
Some of my go-to temperature sensing solutions:
Analog:
NTC 10K, plus a 10K resistor to linearize it. Cheap, 2 wires, hard to beat. The glass-encapsulated thermistors are supposedly very stable over time.
https://www.digikey.com/products/en/sensors-transducers/temperature-sensors-ntc-thermistors/508?FV=1140050%2Cd880007%2Cffe001fc%2C402b81&quantity=&ColumnSort=1000011&page=1&stock=1&k=ntc&pageSize=25&pkeyword=ntc
MCP9700A / MCP9701A: a cheaper competitor to the LM35. 19.5mV/C in the MCP9701A, so a bit more sensitive.
https://www.digikey.com/product-detail/en/microchip-technology/MCP9701A-E-TO/MCP9701A-E-TO-ND/1212511
For the Analog solutions, if you are interested in only a narrow temperature range, you'll want an op-amp circuit to do some gain + offset to "zoom in" on the signal. EEVblog forum user T3sl4co1l showed me a trick where you can do this with 1 opamp and 3 resistors, rather than a full instrumentation amp.
second screenshot here shows how to model this for the MCP9701A in LTSpice:
https://www.eevblog.com/forum/beginners/using-diodes-for-subtraction/msg456677/#msg456677
first screenshot here shows how to do this for an NTC 10k (the .asc file is also attached there if you want to play with the resistor values in LTSpice):
https://www.eevblog.com/forum/metrology/temperature-stabilised-enclosures-heating-and-or-cooling/msg1412203/#msg1412203
Digital:
The Si7021 seems to be the good solution here: easy-to-use (I2C) assembled boards are cheap on ebay, and this guy did a pretty deep analysis which showed that they are pretty good:
http://www.kandrsmith.org/RJS/Misc/Hygrometers/calib_many.html
https://www.ebay.com/sch/i.html?_nkw=si7021
I should note also that my temperature control methods are typically digital / PID, but there are also purely analog ways of doing this, some of them being pretty simple!
https://www.eevblog.com/forum/metrology/temperature-stabilised-enclosures-heating-and-or-cooling/msg1408952/#msg1408952
https://www.eevblog.com/forum/metrology/poor-man's-reference/msg1224071/#msg1224071
More links:
https://www.eevblog.com/forum/metrology/usa-calibration-club/msg1231174/#msg1231174
Here's the simplest Arduino sketch I have which uses the Si7021. It uses software I2C, so you can use any two pins (and use more than one Si7021): https://github.com/cellularmitosis/logs/tree/master/20180304-1590b-isothermal/Arduino/regulated
I've spent a lot of time fooling with PID constants:
https://github.com/cellularmitosis/logs/tree/master/20180114-pid-oven-tuning
https://github.com/cellularmitosis/logs/tree/master/20180205-pid-oven-tuning
Feel free to have a look around in here, I keep all of my logging results and scripts / Arduino sketches in this repo: https://github.com/cellularmitosis/logs
-
The use of a LM35 lies precisely in the simplicity of its application. It has a linear coefficient of x10, and delivers directly a measurement in °C at a k2000 with it's math-function MX+B (x10), regardless of its supply voltage.
So it depends on what and how you want to measure.
-
The MCP9700 has a 10mV/C step as well.
Edit: note: the 19.5mV/C of the MCP9701 makes 1 LSB per C for an 8-bit ADC. -
The MCP9700 has a 10mV/C step as well.
Ah, ok. Good to know. -
I have to agree - if you have a meter already (and if you're temperature is within range), I would suggest the LM35/LM34 is a simple & quick solution because: they are relatively accurate (get the better grade for better accuracy, but the cheaper units might be enough for you), they require no calibration, require virtually no software and they're very forgiving of power supply. THE industry de-facto standard if you've got a meter on hand already. You're ready to go in the time it takes to hook up a few wires to your power source and meter...and then you get on with your testing.
-
and they're very forgiving of power supply.
I should mention that I've had problems with the MCP97xx's when using leads which are several feet long with no capacitor (it oscillated if I recall). If you are using them on leads which are several feet long, you'll need to solder a cap across the power supply leads. A 0.1uF ceramic does the job. -
Hi,
I have been using USB ttl converter and DS18b20. They are cheap and very accurate for that price. And you can put multiple sensors to one converter. There is good support for interfacing with PC e.g. digitemp.
USB ttl converter with CP2102 works well, something like these: https://www.ebay.com/itm/6Pin-USB-2-0-to-TTL-UART-Module-Converter-CP2102-STC-Replace-FT232-New/273107801014?epid=1438383738&hash=item3f967e53b6:g:zUUAAOSwvRpap25u
And for sensors something robust like these:
https://www.ebay.com/itm/Waterproof-Digital-Thermal-Probe-or-Sensor-DS18B20-Length-1M/201544352532?epid=1158405043&hash=item2eecfac314:g:Q88AAOSwXeJXehjS
Wired like in this example:
http://kernelreloaded.com/reading-temperature-with-wr703n-and-ds18b20-on-openwrt/
Wictor -
Precision Thermometry on a budget is *DIFFICULT*
Ordinary Thermometry, down to approx +/-0.5°C is pretty easy using a silicon temperature sensors e.g. the Dallas DS18B20 1-wire sensor quotes that accuracy from -10°C to +85°C, and logging multiple sensors on a 1-wire bus is easy, so its good when you want to monitor temperature gradients and don't need a very small low thermal mass probe. Other silicon sensors are available in smaller packages - an advantage when building low thermal mass probes.
However without the capability to check your temperature sensors calibration and long term drift, its all pretty futile. Just about the only standards within the reach of a competent amateur on a tight budget would be DIY triple point of water and Gallium melting point cells, which would at least give you a two point calibration near room temperature. -
Hi,
Here's what I use:-
http://www.dogratian.com/products/index.php/menu-sensors/menu-usb-tnh-sht10
Available from here:-
https://www.ebay.com/sch/dogratian/m.html?_nkw=&_armrs=1&_ipg=&_from=
The Ebay seller is the developer and has written a wee VB app that reads the sensor and displays on a graph. The source code is available of you want to extend it's functionality.
Specifications:
Operating Temperature (CPU): -20°C ~ +70°C
Connection to PC: USB (CDC Virtual COM port)
Power: USB Powered
Temperature Accuracy: +/-0.5°C at 25°C (max +/-3°C at 120°C)
Temperature Resolution: 14bits
Temperature Range: -40°C to 120°C
Humidity Accuracy: +/-4.5 %RH
Humidity Resolution: 12bits
Humidity Range: 5 %RH - 95 %RH
Ian. -
A boiling point bath is fairly imprecise - depending on the number of nucleation sites on immersed surfaces a significant degree of superheat can occur. The melting point of Gallium at 29.76 °C would be a better choice, but all exposed metal must be carefully lacquered as Gallium will form an amalgam with many metals, and any dissolved metals will compromise the purity of the bath, risking shifting the melting point. 99.99% Gallium is readily available and fairly affordable in small quantities.
-
A boiling point bath is fairly imprecise - depending on the number of nucleation sites on immersed surfaces a significant degree of superheat can occur. The melting point of Gallium at 29.76 °C would be a better choice, but all exposed metal must be carefully lacquered as Gallium will form an amalgam with many metals, and any dissolved metals will compromise the purity of the bath, risking shifting the melting point. 99.99% Gallium is readily available and fairly affordable in small quantities.
Thanks Ian, I've been hunting for a material to use for a point between 0C and 100C. The best option I had found so far was boiling acetone, but that's pretty dangerous, and I have no way of knowing how pure the acetone from the hardware store is, nor am I interested in distilling it myself.
If anyone else has any other boiling / melting point suggestions between 0C and 100C, please chime in!
This also brings up the issue of linearity between 0C and 100C. Everything I've come across online suggests calibrating at 0C and 100C for the device in question, but the linearity between those points seems to just be assumed. On that topic, what sort of device would be our best source of linearity? A PT100 perhaps? -
With LM35 type devices (in the TO93 package) the dominate temperature input to the chip is via the leads. Leave them long...
I have added hooks to Lady Heather to support an environmental sensor. I want one that does humidity and pressure and (at least) two temperature channels. The temp sensors should be remotely mountable (on leads). It should speak over RS-232 or a standard serial-looking USB device (i.e no proprietary drivers. Output rate of 1 Hz is adequate. Resolution is probably more important than absolute accuracy. I prefer RS-232 interfaces since you can almost always find a serial-RS232 dongle that will work with your system, but can have issues with getting a specific USB interface to work.
I'm leaning towards an BME280 for humidity and pressure (and maybe a temperature channel). I haven't decided on what to use for the remote temp sensors. I've thought about PT1000's, but interface chips tend to be a bit pricey. Thermocouples are probably out... not all that accurate, etc. Thermistors are still in the running... maybe using those 18 bit Microchip A/D's. I'd like the sensors to be small and fairly low thermal mass to get fast response times.
If I have to build one, it will probably be based on a ATMEGA-328.
-
texaxpyro,
You have the option of using the MAX31855 or MAX31865 and there are breakout boards available for a quick solution. The TC version isn't that accurate (~6C) and the RTD version is ~0.5C. -
If anyone else has any other boiling / melting point suggestions between 0C and 100C, please chime in!
Virgin coconut oils melt at 24 °C (76 °F)-- the actual value is imprecise, but it will remain constant for a given batch of oil. This can be measured by a PRT, and then used as a calibration tool for other thermometers. The more accurate the calibration, the better (but the more expensive it gets). There will be a thermal plateau just as with gallium. In a hermetically sealed container (with 1 atmosphere of Argon) this melting/freezing point will be quite stable. Pretty cool-- something you can buy at your local grocery store!
Hmmmmmm....
Edit: http://www.daliborfarny.com/2012/06/13/sealed-argon-tube/
edit2: https://youtu.be/GAMRHcbE3g0?t=14m4s
-
If anyone else has any other boiling / melting point suggestions between 0C and 100C, please chime in!
Virgin coconut oils melt at 24 °C (76 °F)-- the actual value is imprecise, but it will remain constant for a given batch of oil. This can be measured by a PRT, and then used as a calibration tool for other thermometers. The more accurate the calibration, the better (but the more expensive it gets). There will be a thermal plateau just as with gallium. In a hermetically sealed container (with 1 atmosphere of Argon) this melting/freezing point will be quite stable. Pretty cool-- something you can buy at your local grocery store!
Some years ago I needed something (some substance) that could provide reference temperatures - phase transitions, like a melting point - in the range 50-100 deg. C, with a resolution of 0.1 deg. C. The complication was that I had to detect the temperatures optically, from outside a heated chamber.
Highly purified synthetic waxes didn't work. The optical response curves were way too sloppy.
I ended up using temperature-sensitive liquid crystals, at 55, 75, and 90 deg. C, custom-made by Hallcrest UK (www.lcrhallcrest.com). The transitions were sharp enough to resolve 10 mK temperature differences, well within spec. The resolution was limited by the temperature control in the application, not by the liquid crystals themselves. For various reasons, I used the raw liquid crystal, not the microencapsulated form. I've always wanted to find out just how sharp those transitions really were. Could they resolve 1 mK? Maybe.
The US patent that resulted (https://patents.google.com/patent/US8345215B2/en) doesn't show my best data (which came later).
Getting back to the OP's application, I suggest an interchangeable thermistor, good for +/- 0.2 C or better, e.g.:
http://www.ussensor.com/products/thermistors/leaded-thermistors/interchangeable-thermistors
https://www.ametherm.com/thermistor/precision-interchangeable-thermistor
For linearity, maybe a PT1000, e.g. Honeywell HEL-776 or -777 (Digi-Key, Mouser), measured ratiometrically. Hard to beat the performance/price of a good thermistor, though. -
texaxpyro,
You have the option of using the MAX31855 or MAX31865 and there are breakout boards available for a quick solution. The TC version isn't that accurate (~6C) and the RTD version is ~0.5C.
Yes, I've been leaning towards the MAX31865. What would work better... PT100 or PT1000? My gut says PT1000, but that comes from when I was thinking of a discrete solution. -
Some useful links:
https://www.scientificamerican.com/article/tackling-the-triple-point/ - Water 0.01 °C
http://www.isotechna.com/v/vspfiles/pdf_articles/Fixed%20Points%20of%20the%20ITS-90%20and%20Confidence%20in%20the%20Metal%20Freezing%20Points%20of%20ITS-90.pdf - various fixed point cells including Gallium
http://makerstutorial.blogspot.co.uk/2008/07/by-shawn-carlson-last-month-i-described.html - a DIY precision platinum RTD thermometer using an off-the-shelf platinum film sensor.
https://thecavepearlproject.org/2015/03/30/using-ds18b20-one-wire-sensors-to-make-a-diy-thermistor-string-pt-2-calibration/ - Serious effort put into DIY calibration of silicon sensors -
Thanks for all the replies here. I'll post responses and follow-up questions throughout the day as I read up on some of the suggestions made here during my breaks.
As a first comment, I would like to point out that while I like the idea of using the very well behaved K2000 (at least to everything else that measures low-frequencies in my lab), and it has the added advantage of me already understanding how to get data from it onto my computer, it only shifts the problem - I go from having to design and build a temperature sensing rig, to having to build a switch matrix rig. After all, I would still want to use the K2000 to do whatever other measurements I want to do.
I will start by looking at the different sensors and complexity involved. I'll start by looking at user cellularmitosis' two links and take it from there!