Thermistors

Try the manufacturer's web site. If there isn't any, then try another manufacturer. Lots of info out there.

Best regards, Spehro Pefhany

I think Epcos may have some app notes - also try Vishay BC.

Thermistors *are* pretty vague btw. Considered a semiconductor solution ?

Graham

Are they ? Not really. There are some with 1% matching or less. Then you measue temperatures, you have some problems anyway. How are you going to calibrate ? A PT100 ? With a precision current of voltage source ? You're good when you reach 6 digits for a current or voltage source. Then apply the PT100 to the location isn't trivial either. On the other hand, calibration with a temperature reference. Water at the tripple point, outgaz it with vacuum, apply vaccum to reach the tripple point and hope you have some good contact. Good. Then you may need a second point. A temperature normal. Freezing mercury, non-alloyed. Yuck. Another idea....

I found NTCs good for repeatable 10mK accuracy around room temperature. Knowing where you are is a different matter but perhaps not really necessary.

Rene

In message , Rene Tschaggelar writes

Thermometrics Taunton Devon UK used to be STC they still make precision NTC and as described they can provide the best temperature information for many applications. Note I used to employ thermistors with good long term aging but un adjusted resistance to compensate crystal oscillators individually matching crystal/resistor/thermistor for best performance. Ignore demon address try snipped-for-privacy@XXXhotmail.com

Thanks to all who made datasheet/app note suggestions. After some reading I'm getting the impression that all thermistors have a non-linear curve? Am I limited to temp sensor chips (lm35 etc) if I need a linear response?

Thanks

There are compound thermistor devices that are linear, but with about

Maybe you could linearize the response over the region of interest, using analog or digital means.

Best regards, Spehro Pefhany

Precious-metal RTDs are one of the best temperature sensors if you need an accurate linear response and don't mind a little cost. They have only a tiny bit of nonlinearity over a narrow range, and most of that is easily eliminated with a taste of +ve feedback. They are also made very accurately (initial tolerance) and can be very much free of hysteresis and drift, depending on the construction you choose. Base metal types are popular with the comfort heating and A/C crowd, and for monitoring heavy motors.

There are plenty of other types of temperature sensors.

Best regards, Spehro Pefhany

Hello Dave,

No. There are also plain diodes and thermocouples. You have to deal with rather small signals here but with clever offset nulling (clamping, chopper amp etc.) it can be done.

Regards, Joerg

Except they don't use Vbe these days, they use delta Vbe at two different currents, which gives better interchangability. On the minus side, the signal level is an order of magnitude lower (more like

Best regards, Spehro Pefhany

I have in the past simply used the temp coefficient of Vbe with a simple transistor sensor.

It's linear with temp, you just need to select a device that has a reasonably consistent Vbe @ a given temp. Depends how accurately you're measuring.

I understand this method has been used for measuring die temp in modern CPUs.

Graham

All thermistors seem to conform to the Steenhart-Hart fitting function, which is pretty non-linear.

Over a sufficiently small temperature range, a thermistor-based sesnsing circuit can look reasonably linear.

As Spehro Pefhany has pointed out, there are a number of options available if you need to measure temperature over a wider range.

A thermistor plus Microchip Technology PIC microcontroller with a built-in A/D is probably the cheapest way to go if you need linear temperature measurements over a wider range.

A platinum resistance thermometer is certainly the best way of getting a good temperature measurement, but you have to dissipate quite a lot more power in the sensor, and even then the sensitivity in volts per degree is quite a lot less, to the point where you find yourself stuck with building an AC-excited bridge circuit to get reliable resistance measurement. That is the way to go if you want to build a microdegree temperature controller, but it tends to be an overkill if you are less ambitious.

----------- Bill Sloman, Nijmegen

In article , Pooh Bear wrote: [...]

Two 1N914s work quite will to get the difference in temperature. I used this fact in a controller for the pump in a solar hot water system many many years ago.

The important questions with temperature sensors are "how many", and "do you already possess some means of temperature calibration".

If the OP just wants one (or even just a few), then he goes for an already calibrated sensor, Pt100 or precision thermistor, at the appropriate price.

If there is a production quantity then he can go for a cheap homebrew sensor, (of whatever sort is most convenient), and invest some money in a temperature calibration test setup.

I like metallic resistive temperature sensors, (Pt, Ni, or even Cu in some apps), because the Beta-term is reasonably known and the linearising is simple.

Afair there is even a combination of Pt+Ni where the correct proportions make the Beta-terms cancel, but have never seen a proprietry implementation of this.

If you want to control at a set temperature thermistors are good, YSI? Used to do a pair on thermistors that gave a linear voultage over a linited temperature range.