Electronic "fever" thermometers - what is the temperature sensor?

[Tim Shoppa]

Digital fever temperature thermometers (wherever you stick them)
are ubiquitous these days, especially with mercury being a no-no to
the current generation. (Hey, we all played with it when we were
kids, and we're perfectly normal, right? ;-) )

What's the temperature sensor in these? Cheap LM35-style sensors
have a rated accuracy of 1 degree C, which is like 2 degrees F, and
I think for this application a thermometer would need much better
accuracy - better than half a degree F, or preferably a tenth of a
degree F (which is the display precision of most of the consumer devices).
There are better versions of these temp sensors but none rated to
be better than half a degree, as far as I can tell.

Maybe there's a special class of LM35-style sensors which can be
factory-calibrated to be known good to a tenth of a degree F near 98.6F?

I can imagine a thermistor based on some sort of phase-change near
98.6F which might do all this magic but I've never seen anything in this
range in any of the catalogs. (There are some indium and cesium alloys
with melting points in this region, that I've seen these used in certain
industrial temperature-measuring applications that require extreme
accuracy and repeatability, but I'm pretty sure that there's no such
stuff in a consumer fever thermometer).

Quartz crystal thermometers (e.g. HP 2801) are good to better than a tenth
of a degree F, but are also much more expensive (circa $1K) than the
consumer thermometers. I could imagine this technology being
mass produced though (a mass-produced crystal with a highly
temp-sensitive cut might be made for less than a $1 and the rest is
quote-unquote easy).

There are also IR-reading thermometers at the high end of the consumer
range (circa $100) and I haven't a clue how they can make measurements
so close to ambient with an accuracy of 10 degrees F, much less 0.1
degrees F! ???

Tim.

 

[Spehro Pefhany]

Digital fever temperature thermometers (wherever you stick them)
are ubiquitous these days, especially with mercury being a no-no to
the current generation. (Hey, we all played with it when we were
kids, and we're perfectly normal, right? ;-) )

What's the temperature sensor in these? Cheap LM35-style sensors
have a rated accuracy of 1 degree C, which is like 2 degrees F, and
I think for this application a thermometer would need much better
accuracy - better than half a degree F, or preferably a tenth of a
degree F (which is the display precision of most of the consumer devices).
There are better versions of these temp sensors but none rated to
be better than half a degree, as far as I can tell.

Maybe there's a special class of LM35-style sensors which can be
factory-calibrated to be known good to a tenth of a degree F near 98.6F?

They use an inexpensive precision thermistor. The technology was
developed in Japan, in conjuction with an MCU company which developed
the original processor. Cheaper and better than an LM35, and they
don't require power supply regulation.

Best regards,
Spehro Pefhany

 

[Tim Shoppa]

They use an inexpensive precision thermistor. The technology was
developed in Japan, in conjuction with an MCU company which developed
the original processor. Cheaper and better than an LM35, and they
don't require power supply regulation.

Is this thermistor available off-the-shelf or is the best way to get
one to buy a $5 digital thermometer and throw away the rest? (After
using "the rest" to get the calibration curve for the thermistor, of course!)

Tim.

 

[Spehro Pefhany]

Is this thermistor available off-the-shelf or is the best way to get
one to buy a $5 digital thermometer and throw away the rest? (After
using "the rest" to get the calibration curve for the thermistor, of course!)

Tim.

If you need the packaging, probably to take one apart would be
easiest. Radio Shack used to sell a dipped through-hole version of the
part for a reasonable price (but not 1% or 2% accuracy). I have some
around that are 5% accuracy (about 1°C) without adjustment, 10K @25°C
and Beta of 3977+/-0.75%. You can have one if you'd like.


Best regards,
Spehro Pefhany

 

[Hal Murray]

If you need the packaging, probably to take one apart would be

easiest. Radio Shack used to sell a dipped through-hole version of the
part for a reasonable price (but not 1% or 2% accuracy). I have some
around that are 5% accuracy (about 1°C) without adjustment, 10K @25°C
and Beta of 3977+/-0.75%. You can have one if you'd like.

If I start with a brand-X type thermister, how stable or
repeatable is it? If I go through a calibration dance,
what sort of accuracy can I get?

 

[Spehro Pefhany]

On Sat, 25 Sep 2004 15:09:05 -0500, the renowned

If I start with a brand-X type thermister, how stable or
repeatable is it?

Better read the specs. If you don't go too much above room
temperature/body temperature and don't get it wet, I'd think within a
tenth or two tenths degree C stability for an epoxy packaged unit.

If I go through a calibration dance, what sort of accuracy can I get?

You ought to be able to get very close near the points of calibration,
but in practice I think more than single-point calibration gets very
expensive. You can buy single high-precision thermistors from YSI and
others (for much higher prices than the consumer parts) or you could
use a platinum RTD, which has much better stability than a thermistor.


Best regards,
Spehro Pefhany

 

[Rene Tschaggelar]

If I start with a brand-X type thermister, how stable or
repeatable is it? If I go through a calibration dance,
what sort of accuracy can I get?

stability is one thing, accuracy another.
With a bridge setup and a 20bit ADC you can get whatever
accuracy. The values may differ from part to part, but
the accuracy is there. What affects stability and repeatability ?
Provided the electronics has stable conditions and provides
stable conditions to the thermistor, there could be aging.

Manufacturing peltier controllers working with NTCs, we set
the temperature to the process and it just stays there.
Some physical processes would require retuning when it did
drift, but that was never the case. Downto 10mK resolution
at least. So I can claim 10mK absolute stability.

Rene