# favorite temp sensors

D

#### David Lesher

Jan 1, 1970
0
So we're looking to sense temps in the ~~0-40C range, and
wondered what to use. In the past I've used LM35's and would
again, but all the cool kids seem to using this digital stuph,
so I was wondering what's a roughly parallel sensor that has
some ilk of digital output?

They'll feed a 32bit CPU so I'm sure we can scare up so IO lines...

B

#### Bill Sloman

Jan 1, 1970
0
So we're looking to sense temps in the ~~0-40C range, and
wondered what to use. In the past I've used LM35's and would
again, but all the cool kids seem to using this digital stuph,
so I was wondering what's a roughly parallel sensor that has
some ilk of digital output?

They'll feed a 32bit CPU so I'm sure we can scare up so IO lines...

Depends whether you want an accurate sensor.

At the moment interchangeable glass-encapsulated thermistors leave everything else for dead at +/-0.2C (or better, if you are prepared to pay more).

You've got to put together a bridge and an A/D to turn the resistance into a digital number, which is going to be a non-linear function of temperature

http://en.wikipedia.org/wiki/Thermistor

but sorting out the Steinhart-Hart relationship won't overload a 32-bit CPU\.

D

#### David Lesher

Jan 1, 1970
0
Depends whether you want an accurate sensor.

A few degrees won't matter....

M

#### [email protected]

Jan 1, 1970
0
David Lesher said:
So we're looking to sense temps in the ~~0-40C range, and wondered
what to use. In the past I've used LM35's and would again, but all the
cool kids seem to using this digital stuph, so I was wondering what's
a roughly parallel sensor that has some ilk of digital output?

The hobby 8-bit microcontroller people seem to like the Dallas DS18B20/
DS18S20 and similar parts. They speak Dallas' "1-Wire" protocol, so
you can string several of them on one I/O pin. The TO-92 case seems
to be popular but you can also get them as SOIC. They claim +/- 0.5 C
accuracy, and somewhere between 0.1 to 0.75 seconds per reading
depending on how many (configurable) bits of resolution you want.

I think those parts are popular in that area because they are always
libraries that will handle bit-banging the protocol for you.

These seem to be reasonably available 1 or 2 at a time (Digi-Key has
them in minimum quantity 1), but they now come from Maxim. For other
kinds of parts, I have heard varying things about Maxim's willingness
to sell parts in small quantities, where "small" is defined as "less
than 100,000 at a time".

When PC motherboards started sprouting temperature sensors, a lot of
them seemed to be National (now TI) LM75 or similar. The LM75 speaks
I2C, claims +/- 2 C and comes in a couple of surface-mount packages.
There are similar parts all the way from LM63 to LM94; some of the
options are SPI, different resolutions, programmable alarm points,
lower power consumption, etc. You need more than one I/O pin to make
these work.

These also seem to be reasonably available 1 or 2 at a time (Digi-Key).

Standard disclaimers apply; I don't get money or other consideration
from any companies mentioned.

Matt Roberds

D

#### David Lesher

Jan 1, 1970
0
There are simpler ways, methods of which I have posted multiple times
on this newsgroup.

I'm all ears. TO92 case is nice, fanatical accuracy is not required,
Price is part of the decision....we can always use just LM35's.....

B

#### Bill Sloman

Jan 1, 1970
0
There are simpler ways, methods of which I have posted multiple times
on this newsgroup.

Sure. Make the thermistor the frequency determining element in an astable, but then you have to cope with the temperature dependence of the reactive element.

I was recommending an accurate solution, rather than a simple one.
Slowman, You're a has-been... if you ever WERE in the first place. Go
away.

I may be a has-been, but I provided real - and potentially useful - information. You've just claimed to know a "better" way (probably incorrectly), without giving any indication of what it might be, and followed it up with gratuitous personal abuse. Definitely an exhibition of psychopathology, though you've skipped the criminal aspect on this particular occasion.

B

Jan 1, 1970
0
B

#### Bill Sloman

Jan 1, 1970
0
There are lots of simpler ways, they're just less accurate. Bill's
right about thermistors, especially if you're interested in working near
a single temperature. Junctions aren't in the same class, because
parameters like doping density and beta aren't as well controlled. IC
temperature sensors have crappy thermal time constants, their
temperatures are dominated by conduction through the leads (which is
sometimes what you want, but often not), and for accurate temperature
control, they're strictly in the "just keep banging the rocks together,
guys" class. (*)

Mixed technology wins again.

For slightly wider ranges, platinum RTDs are all the go, especially if
you pulse the excitation so you can use a higher voltage without
horrible self-heating.

As I pointed out back in 1978

Sloman, A.W. "On microdegree thermostats", Journal of Physics E: ScientificInstruments, 11, 967-968 (1978).

platinum resistance thermometers have a lower thermal resistance to ambientthan thermistors, which does compensate - to some extent - for the lower sensitivity. Any place I'd have wanted to sue them I'd have had to use AC excitation to get the sensitivity I wanted. Reversing DC will do - but a bifilar wound Blumlein transformer bridge is a very neat way of setting up a very stable bridge, and that usually takes a couple of kHz if you want to usea nice compact ferrite cored transformer.
But AFAIK nobody makes thermistors with built-in I2C or one-wire or
whatever the OP is looking for. So ICs are probably useful for
something after all.

Too true.

G

#### George Herold

Jan 1, 1970
0
LM71, 3 wire SPI.

For a bit more money I saw a project that had a whole bunch of
DS18B20's that did a temperature profile of metal plate. (Maxim
though...)

George H.

K

#### [email protected]

Jan 1, 1970
0
So we're looking to sense temps in the ~~0-40C range, and
wondered what to use. In the past I've used LM35's and would
again, but all the cool kids seem to using this digital stuph,
so I was wondering what's a roughly parallel sensor that has
some ilk of digital output?

They'll feed a 32bit CPU so I'm sure we can scare up so IO lines...

I've been using the AD7414. With appropriate wiring and board
stuffind, I can get eight on an I2C interface (used for thermal
testing only).

N

#### Nico Coesel

Jan 1, 1970
0
David Lesher said:
So we're looking to sense temps in the ~~0-40C range, and
wondered what to use. In the past I've used LM35's and would
again, but all the cool kids seem to using this digital stuph,
so I was wondering what's a roughly parallel sensor that has
some ilk of digital output?

They'll feed a 32bit CPU so I'm sure we can scare up so IO lines...

Microchip has some pretty cheap ones. Like everything from Microchip
they are a bit crappy but if accuracy isn't the prime goal...

G

#### George Herold

Jan 1, 1970
0
There are lots of simpler ways, they're just less accurate.  Bill's
right about thermistors, especially if you're interested in working near
a single temperature.  Junctions aren't in the same class, because
parameters like doping density and beta aren't as well controlled.  IC
temperature sensors have crappy thermal time constants, their
temperatures are dominated by conduction through the leads (which is
sometimes what you want, but often not), and for accurate temperature
control, they're strictly in the "just keep banging the rocks together,
guys" class. (*)

Mixed technology wins again.

For slightly wider ranges, platinum RTDs are all the go, especially if
you pulse the excitation so you can use a higher voltage without
horrible self-heating.

But AFAIK nobody makes thermistors with built-in I2C or one-wire or
whatever the OP is looking for.  So ICs are probably useful for
something after all.

Cheers

Phil Hobbs

(*) "So a big hello to all intelligent life-forms everywhere, and to
anybody else, the secret is just keep banging the rocks together, guys."
-- Douglas Adams, "The Hitchhiker's Guide to the Galaxy"

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

Briarcliff Manor NY 10510 USA
+1 845 480 2058

hobbs at electrooptical dot nethttp://electrooptical.net- Hide quoted text -

- Show quoted text -

Not a digital solution, But I've been using diode connected
transistors (in TO-220 packs) as temp sensors. I like that it comes
with a mounting hole :^)
I did a calibration run (versus a spendy lakeshore diode) and then a
single point calibration after that. (all transistors with the same
date code.) I've got to do another calibration in the near future,
and it will be interesting to see how different the next batch is.
How important are all the process variations for a diode connected
transistor?

George H.

K

#### [email protected]

Jan 1, 1970
0
Microchip has some pretty cheap ones. Like everything from Microchip
they are a bit crappy but if accuracy isn't the prime goal...

Seiko has some cheap ones, too.

J

#### John Devereux

Jan 1, 1970
0
Jim Thompson said:
(1) Connect a transistor as a diode (B-C short)

(2) Operate it at low enough current that bulk resistive effects and
dissipation are nil

(3) Observe basic equation: vBE = kT/q * ln(Ie/Is)

(4) Pulse transistor with squarewave of current, Imax = 10*Imin

(5) Observe voltage squarewave at B-C of value kT/q * ln(10) (P-P)

(6) Gain this up (AC-coupled) to desired amplitude

(7) Maybe do DC restore if you like ;-)

(8) Pretty much device independent, though large/power devices make it
easier to observe condition (2)

Is this accurate to +/- 0.2'C (without calibration)?

G

#### George Herold

Jan 1, 1970
0
Is this accurate to +/- 0.2'C (without calibration)?

Not in my experience. (but my experience is fiarly limited... a few
transitors tested.)

I always got a number that was a bit off ~0.3%, so about 1 degree at
room temp. I always assumed the error was due to the transistor
beta... Since the current is Ic and Ib. (I think I got a temperature
that was always a bit high, but I'd have to check my notebook.) You
could add some beta 'fudge factor'.... but then beta changes with
temperature too.

It also depended a bit on the collector current. (1 uA to 10uA were
'nice' currents)

George H.

D

#### David Lesher

Jan 1, 1970
0
So midst the noise I've seen the following mentioned:

LM71
DS18B20
LM75

Of those, only a few seem to come in non-surface mount packages
and I'm not sure how we'd couple surface mount devices to the
copper cooling lines.

Given that we just want to measure the coolent temperature on
both sides of the liquid-cooled heatsink, and not reinvent
anything [1]; it looks like the DS18B20 may do the trick.

It's 2-4X the price of LM35's but still affordable.

Thanks for the suggestions.

1] Or rather "we are inventing enough on the product; we don't
need to do so on the test instrumentation too..."

W

#### whit3rd

Jan 1, 1970
0
Is this accurate to +/- 0.2'C (without calibration)?

It's an absolute-temperature measurement, so +/- 0.2 C implies better
than a part-per-thousand error band. Voltage references and resistor
values for the current pulse would have to be better than 0.1%, so probably
the answer is NO. The only way to get components with that accuracy
is with trimming, and that means ALL qualified sensors are dependent
on calibration or trimming. You can choose which, but you
must pay, either way.

Calibration after soldering is better than trimming before soldering.

W

#### whit3rd

Jan 1, 1970
0
On Friday, April 12, 2013 9:27:51 AM UTC-7, David Lesher wrote:
[about temperature measurement, with mention of 0.2C accuracy]
Given that we just want to measure the coolent temperature on
both sides of the liquid-cooled heatsink...

If you want the difference to be accurately measured, that's
actually an ideal place to put in a thermocouple, or several
thermocouples in series. If you use a single batch
of thermocouple wire, it'll have a good zero-temp-difference
characteristic guaranteed by design. Input pipe is hot
junction, output pipe is cold junction, and an autozero op amp
can condition the signal easily.

K

#### [email protected]

Jan 1, 1970
0
So midst the noise I've seen the following mentioned:

LM71
DS18B20
LM75

Of those, only a few seem to come in non-surface mount packages
and I'm not sure how we'd couple surface mount devices to the
copper cooling lines.

Which section of your specification was that in? ;-)
Given that we just want to measure the coolent temperature on
both sides of the liquid-cooled heatsink, and not reinvent
anything [1]; it looks like the DS18B20 may do the trick.

It's 2-4X the price of LM35's but still affordable.

Thanks for the suggestions.

1] Or rather "we are inventing enough on the product; we don't
need to do so on the test instrumentation too..."

J

#### Jon Kirwan

Jan 1, 1970
0
<snip>

(1) Connect a transistor as a diode (B-C short)

(2) Operate it at low enough current that bulk resistive effects and
dissipation are nil

(3) Observe basic equation: vBE = kT/q * ln(Ie/Is)

(4) Pulse transistor with squarewave of current, Imax = 10*Imin

(5) Observe voltage squarewave at B-C of value kT/q * ln(10) (P-P)

(6) Gain this up (AC-coupled) to desired amplitude

(7) Maybe do DC restore if you like ;-)

(8) Pretty much device independent, though large/power devices make it
easier to observe condition (2)

A published source on this is found here:
http://cds.linear.com/docs/en/application-note/an45f.pdf

Starting on top of page 7.

....

I take George's recent experiences using these seriously,
though. I'll add a short note to his writing on the topic.

Jon

Replies
1
Views
887
Replies
0
Views
2K
M
Replies
26
Views
2K
M
R
Replies
21
Views
2K
rickman
R
K
Replies
4
Views
1K
Klaus Kragelund
K