I AM THAT I AM wrote:
On Sat, 04 Dec 2010 14:06:43 -0500, Phil Hobbs
wrote:

I AM THAT I AM wrote:
On Wed, 01 Dec 2010 20:24:09 -0500, Phil Hobbs
wrote:

Bob Masta wrote:
On Tue, 30 Nov 2010 13:11:18 -0500, Phil Hobbs
wrote:

Glen Walpert wrote:
On Mon, 29 Nov 2010 16:36:56 +0000, ratullloch_delthis wrote:

Greetings All

I'm looking for a recommendation for a low cost thermocouple DAQ that
works with ubuntu linux. Can someone recommend one. I'm trying to
record temperture measurements over time. tia sal22

http://www.mccdaq.com/daq-software/Linux-Support.aspx

Single channel thermocouple to USB with linux support $99, multiple
channels more $.

Thermocouples are really really horrible temperature sensors--almost as
bad as ICs. Their advantages are small size, relatively low cost, and
(potentially) high speed, but their disadvantage is that it's really
hard to get good measurements.

One reason for this is thermal conduction down the leads (which is a big
problem for most temperature sensors). There's a worse one, though:
due to their very low sensitivity, thermocouples are extremely
vulnerable to errors caused by offset drift in the circuitry. See
Figure 20.3 on P. 803 at http://electrooptical.net/www/book/draftthermal.pdf

Not to mention that they involve a reference junction for
which you must know the temperature. Unless you have an ice
bath handy, this involves an independent non-thermocouple
sensor like a thermistor, diode, or IC. So just use that
instead! (Unless you need really high temperatures.)

One good thing about TCs is that they don't need typically
calibration... they are supposed to conform to a standard
for the TC type (J, K, R, S etc). You just measure the
output voltage, correct for the reference junction
temperature, and look up the temperature for that voltage.
Accuracy is typically +/-2 degrees C. But the voltages are
really small: 1 mv or less at room temperature (0.10 mV
for R or S types).

I'd say, save the thermocouples for the really hot stuff.
For "normal" temperatures (say, freezing to boiling water
ranges), you can get much better precision and accuracy, not
to mention convenience, from an IC.

If all you need is a degree or two's accuracy near room temperature,
almost any method will work. IC temperature sensors are generally
fairly putrid--slow, inaccurate, and noisy.

Cheers

Phil Hobbs


Resistor bolometer 2 mm sq, 2 mm behind probably a Ge window (in a
little to-39 pkg) at whatever needed distance behind a half inch diameter
plastic Fresnel lens with some nice read circuitry and LCD display with
read and hold mode, etc., and a little laser focal point spotter.

Pretty damned good accuracy from every test I could put it through,
from new batteries, all the way down to both cells being dead... the
damned thing reads. (obviously the data cell is not completely dead at
that point)

$20 at Harbor Freight.

http://www.harborfreight.com/infrare...ter-93984.html

Some of those things are pretty useful--microbolometers have come a
really long way. Consistency is not the same as accuracy, though, and
all sensors relying on radiation are (a) vulnerable to emissivity
variations, and (2) slow, at least compared to an RTD or thermistor (or
thermocouple).

Temperature control lives and dies by loop bandwidth, just like every
other control system. Slow sensors == poor control. Inaccurate
sensors ==poor control. Fast, accurate sensors plus intelligent sensor
placement, insulation to reduce thermal forcing, thermal grounding of
leads, .... ==good control.

Cheers

Phil Hobbs


IR sensors are faster, and that includes a bolometer. There is zero
settling time, and emissivity will not be a factor, because just like
your sensor, this would be "placed" the same every time. Accuracy is
dead on, if it was calibrated right in the first place as the circuitry
is usually VERY linear and very accurate if any linearization corrections
are needed, they are usually hard wired in. I'll bet that it even has
ambient compensation built into a single custom MCU/do-it-all chip.
Things we had to engineer in with discreet components back when the
finished product was $500.

Well, it was precision lab instrumentation outputs. All the Harbor
Freight item has is a readout.

I'll bet that it can be relied on to plus or minus 0.3 degrees though.

They range from about 1.5 us to about half a second in response time.

It probably takes a typical TC junction a tenth of a second to settle
through with a 'bead' size of about .75 to 1 mm.


0.3 degrees is nowhere near close enough for instruments. For
industrial control, that would often be just fine. However, you
massively underestimate the contribution of emissivity error.

Radiation coupling is very poor at room temperature--a vacuum gap
between two surfaces of unit emissivity is equivalent to the thermal
conductivity of about 5 mm of air. (I once had occasion to calculate
that for a sensor design.) That means that the same sensor in intimate
contact with the given surface would be at least an order of magnitude
faster, and probably two orders.

Optical pyrometry also doesn't work too well through insulation.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net