Designing a rapid reacting 'real' air temperature sensor

[MikeSharp01]

As I was driving around Kent today I was musing on the car's exterior air temperature sensor and how fast it was, or was not, reacting to actual air temp changes  outside the car. I know where it is in the car, down under / behind the front fairing and it must have some signal processing to smooth out the delta T's - observations seemed to say that the reading would update every 20s or so.

 

I then started to wonder about the challenges of designing a very fast reacting air temperature sensor and logger that would display the curve of temperature as you drove into and out of valleys, across hill tops and through frost hollows. I then mused about 'wind chill' (Don't panic I know its not a problem for things without 'internal heat') and how the humidity / density of the air might cause a problem as it passed across the sensor as heat transfer, in theory anyway  I think, will vary depending upon the density - If you think about it a bucket of water around the sensor will have more effect that a single warm molecule otherwise in a vacuum.

 

I also dimly recalled my thermo dynamics studies (mostly asleep I think I was) and, way back at the turn of the century, noticing that side impact in cars was being measured by temperature changes, driven by the old PVT (Pressure Volume & Temperature) gas laws, in the door air, with only a fractional (1ms) lag behind what a pressure sensor could achieve. (I found a ref HERE).

 

In then end, many of these effects and constraints demand a sensor with very low mass (I did not say thermal) in order that the work required to raise and low its temperature is very small. So I am now on the look out for a very low mass, and hence fast reacting sensor - any ideas?

 

 

[Onoff]

I can't help, g'night!

[Jeremy Harris]

A plain DS18B20 takes around 750ms to measure and output a 12 bit temperature, already calibrated in deg C.  Easy to read and display the data, too.

 

https://datasheets.maximintegrated.com/en/ds/DS18B20.pdf

Edited December 12, 2017 by JSHarris

[MikeSharp01]

Just now, JSHarris said:

A plain DS18B20 takes around 750ms to measure and output a 12 bit temperature, already calibrated in deg C.  Easy to read and display the data, too.

Thanks Jeremy that feels a bit slow and I wonder how long it takes to react / rate of reaction to a temperature change in the air around it?

[dpmiller]

car thermometers have massive amounts of processing, to attempt to negate heat rising from the road surface . The thermistors used actually have a pretty good response so you might try tapping into the signal at the current one using a nice high-impedance ADC.

[SteamyTea]

Isn't it the plastic case of the DS18B20 that slows the reaction time down.

Car fuel injection uses a hot wire and a Wheatstone Bridge, they work very quick.

Or how about sensing the pressure inside a sealed vessel.  A mate of mine who paraglides has an altimeter that can sense less than half a meter of movement and issue a warning i.e. up or down.  Only cost a few quid.  (I bought some temp/humidity/pressure sensors for my RPi, but not got around to playing with them yet.

[Jeremy Harris]

11 hours ago, MikeSharp01 said:

Thanks Jeremy that feels a bit slow and I wonder how long it takes to react / rate of reaction to a temperature change in the air around it?

 

Probably best to go back to basics and assess the boundary parameters of the thing you are trying to measure. 

 

In this case, you're looking to measure changes in the temperature of atmospheric air masses at a particular height.  Because of natural air movement, the scale of the terrain etc, these changes will only happen over large areas, certainly many tens of metres across, most probably hundreds of metres across.

 

Additionally, you need to separate out still air temperature (the thing you are trying to measure) from the cooling effect of air motion on the sensor,  plus the cooling effect of evaporation if the sensor gets wet.  That means the sensor needs to be in a shielded location where it won't be affected by either the air flow around it or moisture. 

 

Decide how accurately, and how quickly, you need to measure the air temperature, then look at what is realistically achievable with any form of sensor.  We used to measure outside air temperature using a sensor tucked inside the static tube of a pitot, inside the head but not directly affected by the air flowing around the head, because it was buffered via an air pocket.  That generally worked well, but I can say with certainty that atmospheric air temperature changes are both small and gradual, over a wide area.  We used a mix of different sensors, from fast reacting thermistors and diode junctions (both a complete pig to characterise and linearise if you want accuracy) to thermocouples and hot wire sensors.  The latter can react very quickly but are more sensitive to accelerated convective heat loss, so are better for measuring air flow rate than they are for measuring temperature when placed in a shielded container.

 

If you want to experiment,  without having accuracy then you could try just using a thermistor probe, placing it in a well-shielded location, and measuring the variation of resistance.  It won't be linear (if you want to get the temperature from one you have to compensate for both offset and the inherent non-linearity) but you can get very low heat capacity, fast response time, miniature glass bead thermistors (I probably have a few somewhere).

 

My guess is that you will rediscover what others have found, that air temperature doesn't change quickly and that  the variation in temperature across small air masses (those with dimensions in tens or hundreds of metres) is very small indeed.  My experience of doing this with aeroplanes has been that by far the biggest factor is altitude, specifically the environmental adiabatic lapse rate, which varies with humidity, but is around 2 deg C per 1000ft in still air (more info here: https://en.wikipedia.org/wiki/Lapse_rate ).

 

Finally, you can get the DS18B20 to read to 9 bit resolution in around 94ms, if that sort of resolution is good enough for what you are looking to do. 

 

Edited December 13, 2017 by JSHarris
typos

[dpmiller]

Grant used to have a special probe for ambient air and coldrooms for the Squirrel datalogger which was basically a thermistor inside a ping-pong ball.

[MikeSharp01]

6 hours ago, JSHarris said:

air temperature (the thing you are trying to measure) from the cooling effect of air motion on the sensor,

Thanks Jeremy lots to think about there but specifically,  from above,  I was trying to understand the cooling effect due to motion of the air around the probe. I seem to recall, and as I said Thermo 1 & 2 were not my favourite classes when studying, that the cooling effect of wind is about breaking through the boundary layer of air around the sensor and that the sensor cools more quickly but I cannot, for the life of me, quite see how the temperature can go below the ambient temperature of the air around it no matter how fast it is travelling although a pressure drop caused by the air rushing by might explain it.

[SteamyTea]

If you have less air pressure, there are less molecules.
Temperature is the mean speed of those molecules (with some caveats).  So less molecules there are, the less likely they are to hit the probe, regardless of the absolute temperature.

Then there are radiative forcings, but that is a whole different thing for normal temperatures in normal air.

 

[Jeremy Harris]

There's also the effect that all practical sensors use some power, even if it's a tiny amount, and some are generally compensated for self-heating from that power to some degree, if they are "smart" sensors, like the DS18B20 or similar.  "Dumb" sensors, like thermistors,  have to have a current flowing through them to measure their resistance, and that current very slightly heats the sensor.  Airflow over it will then enhance the rate of convective cooling and change the shape of the resistance versus temperature curve.

[Temp]

I would think that a naked welded tip thermocouple would be about as fast as you can get.

[Jeremy Harris]

It's hard to beat these small (1.8mm diameter) glass bead thermistors for a fast response over the sort of air temperature ranges of interest:  http://www.farnell.com/datasheets/2137899.pdf

 

The only problem is that they need to be compensated for non-linearity and offset, but they respond to temperature changes very quickly.  Somewhere I have a stash of even smaller ones, around 0.5mm diameter, that were used in pairs in a half bridge inside a bottle variometer sensor in a glider.  They are even more sensitive to tiny changes, but they are a pig to handle, with hair fine wires,

[dpmiller]

Pretty sure a car MAF sensor as mentioned by @SteamyTea will be faster still,  I believe they can now see indivudual cylinder pulses. Many are now thin film rather than wire.

[Jeremy Harris]

6 minutes ago, dpmiller said:

Pretty sure a car MAF sensor as mentioned by @SteamyTea will be faster still,  I believe they can now see indivudual cylinder pulses. Many are now thin film rather than wire.

 

Ye, it would, but they are specifically designed to be operated at a contant current (with the occassional over-current cleaning burst) and to measure mass air flow, not temperature.  All the ones I've seen use a bead thermistor in the air chamber to compensate the hot wire for air temperature changes, so that the sensors reads mass flow rate reasonably accurately.

[Guest Alphonsox]

Most automotive temperature sensors are integrated with other functions. My company designs and manufactures things like this for automotive pressure/temperature

https://www.infineon.com/cms/en/product/sensor/integrated-automotive-pressure-sensor/manifold-air-pressure-sensor-map/kp275/

[MikeSharp01]

As I don't need hundreds I will get the Farnell jobie and have a play with it. I can use my pico scope and laptop to log it and wait for a frosty day.