And have you noticed how car speedometers vary? My work truck is bang on according to my GPS, but my personal car reads about 5Km/h low at 100Km/h. (at least that makes me less liable to get caught speeding....)
"Do I look like the sort of man...." No Max, I didn't really think you were.....just mentioned it in passing.
I'd agree though, with all our technological know-how you'd think we could get better than +/-5% accuracy in temperature measurement.
## ...but you have to calibrate them using ice baths and boiling water...
Do I look like the sort of man who wants to calibrate things with ice baths and boiling water (taking into account air pressure, of course, because water boils at a lower temperature in reduced air pressure).
I'm too young for all of this :-)
Also I'm too young to have lost my faith in technology...
Years ago, at a Tandy in France, I bought some temperature control modules with digital display. They were on a (really good) special so I bought 10 of them. I was as sceptical as you Max, so I fitted all 10 of them with batteries and tied all the sensors (which are on a 3 foot cable) together and found they read the same temp to within about a degree - mostly half a degree. I used one of them on a project (they have outputs that you can program to signal when the temp is below or above a certain level). I doubt if Tandy would still do them (or even if Tandy is still going there - they are not in France...) but you could try them Max - the code was 277-123. If you can't get one, let me know and I'll send you one (since I have 9 of them left after 15 years or so, I doubt I'll ever need them all....)
I recently got a type K Thermocouple for next to nothing on a sale, I also got some AD595 ICs that connect to a type K and spit out a voltage of 10 mv per degree C to within 1 degree C. Have not got to playing with it yet, but I'd like to see if its as easy as they say.
You can also get 3-terminal ICs that do much the same thing and are easy to connect to a digital voltmeter module, but you ahve to calibrate them using ice baths and boiling water.
The other thing you could do Max is head to your loacal Meteorological office - they'd have pretty precise thermometers that you could compare yours to?
Oh, and have a go at fixing the old one - usually the LCD connects to the PCB using rubbery strips called elastomeric connectors. You can usually put in a strip of paper between the housing and the LCD to push it harder onto the connector and sort out the missing segments or pixels. The old Nokia 5110 phones used to have this problem a lot, fixed many of them with this trick.
It's the beginning of the 21st Century -- we can create silicon chips with structures measured in a few tens of billionths of a meter -- and yet we can't create something for home use that can measure the temperature closer than 5 degrees...
Give me strength! (sob sob)
My thermostat at home has an external sensor as well. It, my car thermometer and the reported temperature on the radio can vary by 5 degrees C. I think (and I am not trying to justify it here) that the philosophy is that it should be repeatable only. Find the temperature you like irrespective of what it says it is and you will always be comfortable. If the thermometer says it is -5 degC outside, take a coat even if it is actually only zero.
The thing is that even if the sensors themselves vary in response, there's got to be some way to calibrate them in-circuit -- and you expect something like a $100 thermostat (which can't cost more than $15 to make) to be accurate to within say 0.5 degrees...
I have had several similar issues at home trying to test out an airconditioner and our fancy stove that my wife reckoned wasn't working properly. Through work I have access to calibrated Fluke thermocouple themrmometers and RTD/ thermocouple calibrators along with DVMs with different temperature sensors. All I can tell you is that there is a very wide variation in the accuracy of all the sensors. Obviously the more you pay the more accurate they become.
The stove said it was keeping a constant temperature, but the meters indicated variations of 10 degrees C above and below. I assume that the software algorithm of the stove averaged the two sensors and decided on a temperature. Obviously the baked goods were not agreeing with that average, which was the complaint.
In some of the work we have done, we found that there can be a self heating element (pun unintended) to the design. In the case of your thermostat, do you have the ability to reduce the backlight- that could be causing your offset.
It also depends on the technology. Thermocouples can be +/-5 degrees C inaccurate. RTDs tend to be better, but are much more expensive. Thermistors are also rather vague. Semiconductors can be good, but pricey. And then there is the circuitry around them.
Which brings us to the problem of calibration-how do you know what the temperature is when you calibrate it, when you don't have any accuracy in your instruments? My experience of the cheap end of Chinese manufacture is that the factories have no climate control and are cold in winter and steaming in summer. And as to accurate calibration- I watched in much amusement as a passive infra-red detector was checked. They hang the DUT on a cubicle divider set in a large empty factory space. Then a flunky (in a lab coat) walks across its field of view at a distance of 20 feet or so. If it "sees" him, the unit passes.
NASA's Orion Flight Software Production Systems Manager Darrel G. Raines joins Planet Analog Editor Steve Taranovich and Embedded.com Editor Max Maxfield to talk about embedded flight software used on the Mars on EE Times Radio. Live radio show and live chat. Get your questions ready.