Temperature sensor LM35 and long leads

[Pandora]

I know this topic has been discussed previously but I wondered if
anyone had personal experience of an application similar to mine.

I wish to monitor the air temperature (-5 to +30 deg C) at various
points in a void below the office floor. The void is normally
inaccessible, but the floor is being re-laid at the moment so I have a
"one-shot" opportunity to put the temperature sensors in place.

For each monitoring point, I plan to use National Semiconductor's LM35
three-lead temperature sensor with a 2K2 resistor in series with the
output lead. This sensor/resistor will be soldered to one end of a
two-pair screened twisted pair cable (Belden 8723) such that the output
and ground signals are carried on one pair, and the power supply will
be carried on the other pair (the two grounds will be commoned at the
sensor).

The max. length of the cable will be 5 metres (16 feet). I will power
the IC with either a 9V battery or floating 12V d.c. supply and use a
1M-impedance digital multimeter to measure the output voltage. When the
output is not being monitored, the LM35 supply will be switched off.

I require an accuracy of about ±2 deg C

Has anyone achieved good results in similar applications using the
above approach? My particular concerns are whether the capacitance of
the twisted pair cable causing an error in reading and whether the LM35
will latch up.

Do I require any bypass capacitors at the sensor end?

 

[John Larkin]

I know this topic has been discussed previously but I wondered if
anyone had personal experience of an application similar to mine.

I wish to monitor the air temperature (-5 to +30 deg C) at various
points in a void below the office floor. The void is normally
inaccessible, but the floor is being re-laid at the moment so I have a
"one-shot" opportunity to put the temperature sensors in place.

For each monitoring point, I plan to use National Semiconductor's LM35
three-lead temperature sensor with a 2K2 resistor in series with the
output lead. This sensor/resistor will be soldered to one end of a
two-pair screened twisted pair cable (Belden 8723) such that the output
and ground signals are carried on one pair, and the power supply will
be carried on the other pair (the two grounds will be commoned at the
sensor).

The max. length of the cable will be 5 metres (16 feet). I will power
the IC with either a 9V battery or floating 12V d.c. supply and use a
1M-impedance digital multimeter to measure the output voltage. When the
output is not being monitored, the LM35 supply will be switched off.

I require an accuracy of about ±2 deg C

Has anyone achieved good results in similar applications using the
above approach? My particular concerns are whether the capacitance of
the twisted pair cable causing an error in reading and whether the LM35
will latch up.

Do I require any bypass capacitors at the sensor end?

Noise spikes can latch an LM35, especially if the supply voltage is
above +5. Current-limit the supply, bypass Vcc at the LM35, and use a
series resistor between the LM35 output and the cable, as you
suggested.

The LM35 will work below 0 C if you pull its output below ground, in
which case the likelihood of latchup goes way up. You could use an
LM34, which works to 0 F with no pulldown. Better yet, use an RTD.

The LM35 has lots of personality.

John

 

[David Lesher]

For each monitoring point, I plan to use National Semiconductor's LM35
three-lead temperature sensor with a 2K2 resistor in series with the
output lead. This sensor/resistor will be soldered to one end of a
two-pair screened twisted pair cable (Belden 8723) such that the output
and ground signals are carried on one pair, and the power supply will
be carried on the other pair (the two grounds will be commoned at the
sensor).

I used about a dozen LM35's in a hyronic heating project. The App
Note had a current-loop design such that it ran on 2-wires, period.

I require an accuracy of about = B12 deg C

The current loop did introduce a fixed error of 1C, as I recall. I was
looking at the differential & didn't much care.

 

[Robert Baer]

I know this topic has been discussed previously but I wondered if
anyone had personal experience of an application similar to mine.

I wish to monitor the air temperature (-5 to +30 deg C) at various
points in a void below the office floor. The void is normally
inaccessible, but the floor is being re-laid at the moment so I have a
"one-shot" opportunity to put the temperature sensors in place.

For each monitoring point, I plan to use National Semiconductor's LM35
three-lead temperature sensor with a 2K2 resistor in series with the
output lead. This sensor/resistor will be soldered to one end of a
two-pair screened twisted pair cable (Belden 8723) such that the output
and ground signals are carried on one pair, and the power supply will
be carried on the other pair (the two grounds will be commoned at the
sensor).

The max. length of the cable will be 5 metres (16 feet). I will power
the IC with either a 9V battery or floating 12V d.c. supply and use a
1M-impedance digital multimeter to measure the output voltage. When the
output is not being monitored, the LM35 supply will be switched off.

I require an accuracy of about ±2 deg C

Has anyone achieved good results in similar applications using the
above approach? My particular concerns are whether the capacitance of
the twisted pair cable causing an error in reading and whether the LM35
will latch up.

Do I require any bypass capacitors at the sensor end?

1) A shielded pair will do the job: common, power and signal.
2) AFAIK the LM35 is not sensitive to capacitance, but a series 1K or
10K from the LM35 output and the cable would give more than sufficent
isolation.
3) You will have to calibrate each and every LM35 used to get +/- 2C
accuracy; finding a standard and getting a constant temperature stirred
bath that can be varied over that range will both be challenging, unless
you are willing to spend a fair amount of $.
4) OTH, they are not bad to 210C!

 

[Robert Baer]

Noise spikes can latch an LM35, especially if the supply voltage is
above +5. Current-limit the supply, bypass Vcc at the LM35, and use a
series resistor between the LM35 output and the cable, as you
suggested.

The LM35 will work below 0 C if you pull its output below ground, in
which case the likelihood of latchup goes way up. You could use an
LM34, which works to 0 F with no pulldown. Better yet, use an RTD.

The LM35 has lots of personality.

John

I am using a 8V battery (dying 9V) in a noisy environment, and have
run the LM35 to at least 210C on a number of occasions and have seen
zero problems.

 

[Robert Baer]

I used about a dozen LM35's in a hyronic heating project. The App
Note had a current-loop design such that it ran on 2-wires, period.




The current loop did introduce a fixed error of 1C, as I recall. I was
looking at the differential & didn't much care.

....and, one can always subtract out that error.

 

[Pandora]

<snip>

3) You will have to calibrate each and every LM35 used to get +/- 2C
accuracy; finding a standard and getting a constant temperature stirred
bath that can be varied over that range will both be challenging, unless
you are willing to spend a fair amount of $.

Robert, why would I have to calibrate the LM35? Their worst case error
is ±2°C

 

[[email protected]]

1) A shielded pair will do the job: common, power and signal.
2) AFAIK the LM35 is not sensitive to capacitance, but a series 1K or
10K from the LM35 output and the cable would give more than sufficent
isolation.
3) You will have to calibrate each and every LM35 used to get +/- 2C
accuracy; finding a standard and getting a constant temperature stirred
bath that can be varied over that range will both be challenging, unless
you are willing to spend a fair amount of $.

A single point calibration is all you'd need to get rid of the bulk of
the error on an LM35 (or an LM34). An ice bath is the easiest fixed
point to set up, but it isn't much use for the LM35 though it would be
fine for the LM34 already recommended by John Larkin.

The NBS procedue for setting up an ice-bath can be found on the web.
I've followed it - more or less - and got stability to something like
0,001C, We used de-ionised water to make the ice and for the waer in
the bath, but I had to break up the ice-cubes with a hammer on the
tiled bench top, so I can't be absolutely confident about the actual
temperature

 

[Steve Kavanagh]

If the LM35 and resistor are just dangling at the end of the cable you
may find that self-heating eats up a fair bit of your error budget.

Steve

 

[Steve Kavanagh]

Or use an LM335 which reads out in Kelvin.

Steve

 

[[email protected]]

If the LM35 and resistor are just dangling at the end of the cable you
may find that self-heating eats up a fair bit of your error budget.

The LM35 is tolerably good as far as self-heating goes - around 300uW.
A thermistor dissipates less heat (10uW is typical, and they tend to
become unstable at higher dissipations) but offers rather higher
thermal resistance to ambient. It is a lot easier to stick a decent
heat-sink on a TO-220 packaged LM35 (Farnell order code 431-6721) than
it is to improve the thermal resistance to ambient of a thermistor.

 

[John Larkin]

2) AFAIK the LM35 is not sensitive to capacitance...

The datasheet suggests otherwise.

John

 

[David Lesher]

Robert Baer wrote:

Robert, why would I have to calibrate the LM35? Their worst case error
is = B12= B0C

LM35's came in a zoo's worth of species; some were much tighter specs
than others, some had a wider temp range, etc etc.

 

[Robert Baer]

Robert Baer wrote:




Robert, why would I have to calibrate the LM35? Their worst case error
is ±2°C

There is percision, accuracy, and error; three different things.

 

[Robert Baer]

If the LM35 and resistor are just dangling at the end of the cable you
may find that self-heating eats up a fair bit of your error budget.

Steve

Milliwatts???

 

[Robert Baer]

The datasheet suggests otherwise.

John

Suggestions and experience are totally different animals.

 

[John Larkin]

Suggestions and experience are totally different animals.

Do you know more about the LM35 than National does?

John

 

[Steve Kavanagh]

Milliwatts???

Yes....the OP suggested a 12V supply and 2.2k resistor...thats around
50 mW for the sensor and resistor together - that will warm up a TO-92
and small resistor quite noticeably. Using the TO-220 package rather
than the smaller options, a small heat sink and putting the resistor at
the other end of the cable will all help greatly of course.

I've just been playing with an LM335Z circuit. I saw around 0.5 degree
warmup drift while testing with the sensor soldered to a small PC board
(partly from sensor self heating and partly from heating from the 100
mW or so dissipation in other the components on the board). The
warm-up drift is insignificant now the sensor is mounted by itself on a
heatsink.

Steve

 

[Gary Peek]

I wish to monitor the air temperature (-5 to +30 deg C) at various
points in a void below the office floor. The void is normally
inaccessible, but the floor is being re-laid at the moment so I have a
"one-shot" opportunity to put the temperature sensors in place.

For each monitoring point, I plan to use National Semiconductor's LM35
three-lead temperature sensor with a 2K2 resistor in series with the
output lead.

If you can handle a bit wider tolerance, you might consider the LM50.

National must have figured out how much trouble capacitance loading
is with the LM34/35, because they made the LM50 with an output that
can drive much more.

The LM50 has an offset in the output so that you only need a positive
supply for both positive and negative temperatures.

Only disadvantage for me was the tiny package.

Gary Peek
Industrologic, Inc.

 

[John Larkin]

If the LM35 and resistor are just dangling at the end of the cable you
may find that self-heating eats up a fair bit of your error budget.

Steve

The resistor is in series with the output, to isolate the chip from
cable capacitance, so dissipates no power. But yes, self-heating needs
to be calculated, especially if the thing's not on a pc board.

The cable itself makes a decent thermal antenna!

John