Homebrew PCBs

We should probably move this to Electronics_101. ;')

A silicon diode can be used as a temp sensor for cold junction
compensation for thermocouples. Very linear around room temp,
approximately -2mV/C to -2.2mV/C. Practically free, add an OpAmp
before feeding it to your micro.

You can build a very simple and cheap wheatstone bridge thermometer
with a silicon diode and a DMM, without any Op Amps. A low dropout low
power linear regulator is the only active component needed, output can
be scaled to 1mV/C or 1mV/F.

It does change very slightly at the extreme ends of an Si diode's
useful range, but within the range of room temperature it is very
linear. You can get complicated and calibrate by bringing distilled
water to a slushy frozen and liquid state and then boil distilled
water to get 0 and 100C and calibrate the exact mV/C, or you can stick
it under your tongue and call that 37C and assume -2.1mV/C (split the
difference btwn 2.2 and 2). Note that I've labeled it as negative
because the voltage goes down as temp goes up.

Just looked this up on Google:
<http://www.physics.rutgers.edu/ugrad/387/material_phys_pc_I/silicon_diode_therm.pdf>

Excerpt:
**********
From a practical point of view, an important
aspect of silicon diodes used as temperature
sensors is the extent to which they may be
interchanged when a diode becomes damaged.
The voltage across ten IN4148 diodes was
measured with each diode first placed in ice/water
mixture and then in steam (diode current set to
625 A). At 0 C the mean voltage was 0.6422 V
with a standard deviation of 0.0031 V. At 100 C
the mean voltage was 0.4334 V with a standard
deviation of 0.0018 V. Care was taken to ensure
that the diodes were electrically insulated so that
no conduction was possible through the water in
which they were situated. Using the mean values
of voltage given above and assuming linearity
between temperature and voltage between 0 C
and 100 C, the equation relating temperature in
C and voltage is
T .C/ D \u2212478:9V C 307:6 (1)
where V is the voltage across the diode in volts.
Based on the variability between diodes as
observed in our sample, we can conclude that if a
diode is chosen at random (and assuming no other
sources of uncertainty) there is a probability of
about 0.7 that the temperature, as inferred by the
voltage across the diode, will be within 1 C of the
temperature inferred from substituting the mean
voltage of the ten diodes into equation (1). While
for many applications this would be acceptable,
there is of course no substitute for recalibration
when diodes are replaced.
**********

Plugging their numbers in works out to -2.088mV/C for a 1N4148.

Steve Greenfield

--- In Homebrew_PCBs@yahoogroups.com, "John Craddock"
<John.Craddock@...> wrote:
>
> Hi Philip,
> Yes, I read the article and have been doing some research of
thermocouple interfacing. My problem with the max6675 is that Maxim
lists the price at US$3.88 in 1K quantities. By the time I get a
single unit to OZ it costs me A$ 10 times that amount whereas cold
junction compensation with an LM335 costs me A$2.72 plus an
instrumentation op amp at about A$4.00 plus using a micro with ADC
channels gives me the digitising for free (no SPI programming). BTW
the AD595 (Analog Devices equivalent to the max6675) would cost me
$35.98 as well. Sooo, I am heading down the LM335 track with an 18F
series PIC (unless I can find a lower price on the 6675 or the 595.
BTW, using glass covered @K@ type thermocouple wire welded at the hot
junction end with a blow torch is a pretty cheap way of getting
reasonable thermocouples. Some of the code in the elector article is
useful as it is written in C and therefore reasonable portable.
> Regards
> John C
>