🚀 go-pugleaf

smithju@motcid.UUCP (Justin A. Smith) writes:

>strong@tc.fluke.COM (Norm Strong) writes:

>>If the temperature range does not exceed normal environmental
>>temperatures, (0 - 100C) I would strongly advise a semiconductor sensor
>>instead of a thermistor.  The output is quite linear.  There are several
>>designs around using a transistor and an ordinary DVM chip, the whole
>>thing operating from a 9v battery.
>
>	What type of semiconductor sensor would i use, where can i get these??,
>i would also like to see some data sheets/designs using these sensors.
>Are they expensive?.

They are CHEAP!! They are easily available!! They are called TRANSISTORS!!
(that's what the man said :-)

For a forward biased base-emitter junction at constant current and a
sufficiently large forward voltage, an approximation is:

	VBE = (kT/q)ln(If/Isat)

where:
	VBE  = base-emitter voltage
	k    = Boltzmann's constant
	T    = junction temperature in degrees K
	q    = net charge in coulombs
	If   = forward current
	Isat = saturation current

This yields a temperature coefficient of approx. -2.5mv/deg C for a garden
variety, small-signal transistor (e.g., 2N2222A) forward biased at approx.
100uA. There are some temperature dependent terms in Isat which complicate 
matters a bit, but the net result is a temperature sensor with linearity
comparable to a platinum RTD (albeit with a more limited working temp range).

The problem is that this tc varies from device to device, so using off-the-
shelf parts in a product design requires matching devices for DC gain (hFE) 
and VBE at room temp. The solution: Motorola offers a family of parts, called
MTS102, MTS103 and MTS105 for which they do just that; they screen parts from
one of their standard, small-signal transistor lines for matched hFE and VBE.
They offer them in matched lots at very good prices (well under $1.00 in
reasonable quantities) and usually are happy to sample you.

I have used these parts with nothing more than a simple JFET constant-current
source and an 8-bit SA A/D and achieved < +-1 degree C from -40 C to +100 C
(Mot rates the MTS102 at +- 2 degree C from -40 C to +140 C). One down-side
is the relatively long thermal time constant (8 sec in flowing air) because
they're in TO-92 plastic packages, so don't count on these things for quick 
response. (This is perhaps the only area where thermistors shine; because
they can be fabricated in extremely small sizes, their thermal mass and 
time const. can be very low).

If you're just hacking and are willing to calibrate the tempco yourself, then
any good small signal transistor should do.

Good references on these parts and the use of transistors in general as temp.
sensors are:

    MTS102, 103, 105 Data Sheet, Motorola Semiconductors, DS2536 R1, 1981.

    O'Neil, P. and Derrington, C., "Transistors - a hot tip for accurate
    temperature sensing", Electronics, 11 Oct 1979.

    Ohte, A. and Yamagata, M., "A Precision Silicon Transistor Thermometer",
    IEEE Transactions on Instrumentation and Measurement, Vol. IM-26, Dec 1977

Bruce R. Koball
Motion West
2210 Sixth Street
Berkeley, CA  94710
415-540-7502
bkoball@cup.portal.com