What temperature accuracy can be achieved without large investments in meters?

[Teza]

I would like to know if it is 19, 20 or 21 deg.C inside/outside my house.
Do I have to invest in a very expensive tool to get this correct or should normal thermometers be able to do this?

I might be a bit crazy when it comes to measurements of basically anything...    ...I want accuracy! 
Also I do not have much knowledge when it comes to the subject...

I have 5 electrical temperature sensors from different producers placed next to each other, included a ST-612 thermometer of unknown brand.
None of them will agree on the temperature (all within 3 deg.C from each other)

ST-612 Accuracy:
-50 to 0 deg.C is +- 2 deg.C
0 to 1000 deg.C is +-0,05% rdg. +1 deg.C

If the correct temperature would be 20 deg.C then the ST-612 will show between 18.9 and 21,1 deg.C

[Nusa]

Most consumer-level products don't even try for 1 degree accuracy, although they're good enough for ballpark purposes.

The ST-612 may be quite accurate, but you won't know for sure unless you have another accurate source. You could test it against easily reproduced temperatures, such as 100C or 0C.

This is an area where old-school is often superior. I'd suggest a glass laboratory thermometer, which typically come with a certificate of accuracy. The cheaper ones with 1C markings are under $15.

[kosine]

Most cheap temperature sensors will be thermocouple based, and they tend to be a little inaccurate, especially over time. If your meters are a few years old, it's unlikely they'll agree to within a degree or two, and as mentioned by Nusa, you need regular calibration to be sure.

Your best option for long term accuracy is to use a resistance temperature device (RTD), such as a PT100.

http://www.temperature.com.au/Support/RTDSensors/RTDaccuracyClassAClassB13DIN110DIN.aspx

Even the cheap ones on ebay will likely be good to within a degree. If you get some good 0.1% resistors (or better) and a similarly precise voltage reference, you can easily rig up a reasonably accurate thermometer with a cheap microcontroller and a display. Plenty of arduino based projects out there already.

https://www.instructables.com/id/Reading-Temperature-From-PT100-Using-Arduino/

[David Hess]

Kosine responded with basically what I would have said.

A thermocouple measurement stacks its cold junction measurement which is usually made with a thermister or PN junction.

Thermisters and PN junctions can do better than 1 degree with calibration but are usually worse.

An RTD is the way to go for reliably higher accuracy at low cost.  If you do not trust the instrument, then the analog circuit to convert an RTD output to a voltage is not difficult.

[2N3055]

Lowest cost without having reference to calibrate with is to use one of factory calibrated solid state digital sensors. They are getting better and better all the time.

[Teza]

I did not know about PT100 sensors, I can now do some more accurate google searches to learn more.
Will try to learn more about 2, 3 and 4 wire sensors and circuits as well as class A and class AA.

Thanks!

[David Hess]

Lowest cost without having reference to calibrate with is to use one of factory calibrated solid state digital sensors. They are getting better and better all the time.

An ice and boiling water bath with distilled water allows two point self calibration if a barometer is also used.

A water triple point cell is also feasible:

https://www.scientificamerican.com/article/tackling-the-triple-point/

Shawn Carlson also suggests a platinum RTD:

http://makerstutorial.blogspot.com/2008/07/by-shawn-carlson-last-month-i-described.html

[Gregg]

A quality mercury glass thermometer is hard to beat for a reference thermometer in the range posted.  It may be hard to find one these days.

[Nusa]

A quality mercury glass thermometer is hard to beat for a reference thermometer in the range posted.  It may be hard to find one these days.

For oral/rectal thermometers, that's pretty much the case. Mercury is no longer used.

For laboratory use, quite a selection of mercury thermometers are available, assuming you live in a jurisdiction that allows their sale. This is the message posted at one of the US sellers:

At least one product below contains mercury. Please be aware that mercury-bearing products CANNOT be shipped to the following states: California, Connecticut, Illinois, Louisiana, Maine, Massachusetts, Michigan, Minnesota, New Hampshire, New York, Ohio, Rhode Island, Vermont, Washington, and Wisconsin.

I've no idea how easy they are to access internationally.

Alcohol-based ones can also be extremely accurate for various ranges, and can be had marked in 0.1C increments. The OP's 1C requirements are much cheaper to meet, however, and I suggested accordingly in my first response.

[Dundarave]

I'm like you in terms of wanting to know "absolute and accurate" temperatures inside and out, but I've found that the greatest challenge in determining household temperatures accurately is that generally speaking, the temperature is constantly changing at that one particular point in 3-space where the sensor is sitting, and there is also hysteresis of some kind at play as well as the sensor tries to settle, whether it's a thermocouple or a glass mercury bulb, and it'll be different for every device.

In addition, inside the house, there's usually a multi-degree temperature gradient floor to ceiling and window-to-wall, with moving convective air, the heating system coming off and on, etc. leaving the exact "room" temperature in some doubt.   For outside measurements, there are heat gradients as you move farther away from the building structure, and of course the sun/cloud changes that also add to the constantly changing "actual" temperature, not to mention the continual radiant heat changes from the ground and structure as things cool down at night. 

What really made me I realize that I was essentially chasing an impractical level of accuracy was that regardless of what any guage or thermometer said, my wife insists on defining what the acceptable room temperature is!  And that then involves comfort and thus humidity levels, etc...  So I've given up. 

[Teza]

@David Hess
No building instructions tho.

@Gregg
True, but in 2017 mercury was forbidden to use in products, this goes for most of the world.
There is other good solutions tho.

@Dundarave
That is very true, but when your homeautomation control turns the heat on and off, it is a better chance to get the desired comfort temperature around the house if the temperature input is more accurate than 3 deg.C
I am measuring about 150 cm from the floor, but I want to know if the temperature is 19, 20 or 21 deg.C  (Even if it does not really matter elsewhere than in my head    )

-----

I do not have any skills in electronics or software like KiCad, but with the help so far, I think I have a starting point.
What I have learned so far is that the best option may be to use a 4-wire PT100 RTD sensor and convert the signal with a RTD to digital converter.
Took me some hours to find information, but I think I have found the red line to follow.



I think I have found usable components from Farnell, but time will show.




Believe I will buy a circuit from https://www.adafruit.com/product/3328, learn how it works and learn how to communicate with it, before I go further with trying to recreating it.

[beanflying]

A Quick and dirty option using off the shelf blocks and KISS.

Arduino Pro Micro or even something WiFi if that floats your boat.
Max 31865 Board eBay auction: #163401382594
4 wire RTD capable of over 100C
Display of your choice ?

The code if you haven't used an Arduino before might take a bit but their is Libraries for the MAX for a start.

Calibrate it in an ice slush bath and at boiling point as both can be achieved with decent accuracy. And even cross check it with a 4 wire Multimeter if you have one and the NIST tables.

Apart from that you are on the right sort of track 

EDIT Quick check plenty of youtube videos you might like to take a look at too https://www.youtube.com/results?search_query=max31865+pt100

[2N3055]

SHT35 ... You want humidity measurements anyway..

[Teza]

The building blocks seems like a good place to start, but I would like to learn more about the circuit and it's components.
How to troubleshoot it and change it to my needs.

Question:
The Max 31865 Board eBay auction: #163401382594, you linked to, it looks like a rip off, even the text seems to have been ripped off.
How do you know that they have used the 0,1% accurate reference transistor in the circuit?
Without this, the circuit and sensor will not be accurate, as far as I understand.

[beanflying]

It is fairly common for the Chinese to make clones of Adafruit etc. boards (generally they are open source for a start) I have used a few of them over the last few years and I haven't had a dud so far. If the IC on the boards are genuine or not is another whole subject but you really won't know until it arrives and even then sometimes you can't tell.

The block option isn't as much of an electronics challenge but it does lend itself to dropping on a breadboard, trying different sensors/rtd's and experimenting with WiFi and Wired Arduino type Micros. You will learn more about programming and calibrating going down this path and then roll what you have learned into a final circuit/product. There is also some combined Temp/Humidity serial output sensors around the AM2305 is affordable and I think under 0.5C? Accurate.

The only way to test accuracy in a home sense with any of your options is Iceslush and boiling water and that would be true for any sensor genuine or not. The Spec sheet and RTD you have really needs checking when you go chasing more accuracy. Or just jump straight in the Deep end  https://www.eevblog.com/forum/testgear/sika-18600e-dryblock-temperature-calibrator-teardown-pictures-testing/msg2249781/#msg2249781

[David Hess]

@David Hess
No building instructions tho.

No building instructions for the triple point cell or the RTD assembly?

The instructions are pretty detailed and I think the original Scientific American articles had photographs if you want to buy the back-articles.  More is available online if you search.

Given the practical limitations Dundarave identifies, I would be inclined to use a metal packaged silicon sensor or a metal packaged transistor in a delta-Vbe circuit and custom calibrate them using ice and boiling water baths.  When they were reasonably priced, we got fine results with the AD590.

[NivagSwerdna]

As soon as you have more than one you are doomed.

You could always use something like a DS18B20 which is meant to be reasonably calibrated which would probably meet your requirements otherwise you are down the road of thermistors and thermocouples and that way madness lies.

I had a requirement to keep two fridges at as close to 5C as possible once... Not easy

A first step towards madness perhaps....

https://www.kandrsmith.org/RJS/Misc/Thermometers/absolute_ds18b20.html

[coppercone2]

4 wire RTD (the more expensive the better) and a drywell that is calibrated.