[General] LM35 precision problem?

[Ma3ix]

Hi everyone,

I have precision problem with my temperature sensor LM35DZ who gives me error of about +/- 3 C degrees comparing with mercury room thermometer. I tested it with multimeter and also separately from my breadboard and always gives me error (for example my fingers/body temperature is 32 C :roll.
Due to datasheet it should be +/- 0.5 C at room temperature.

any suggestions?:idea:

 

[barry]

How are you measuring the voltage?

Measuring the temperature of your fingers is NOT the same as measuring your body temperature.

If you can, putting the sensor in an ice-bath would be a good measurement point.

Is your mercury thermometer measuring the same point as your sensor? Maybe hot components on your PCB?

There are a lot of elements of the design here that you haven't mentioned.

 

[schmitt trigger]

And..........how do you know the mercury thermometer is actually accurate?

Unless you have access to a calibrated thermometer traceable to a national standard, your best option would be to:
-ice/water bath to set 0C
-slowly boiling water to set 100C.

On the last measurement, you would have to adjust the temperature for the height above sea level from where you live. The higher elevation, the lower the temperature water will boil.
https://www.engineeringtoolbox.com/boiling-points-water-altitude-d_1344.html

Also, make sure the water is as free as possible of dissolved salts. Purchase some distilled water to perform your experiments.

 

[Ma3ix]

And..........how do you know the mercury thermometer is actually accurate?

Unless you have access to a calibrated thermometer traceable to a national standard, your best option would be to:
-ice/water bath to set 0C
-slowly boiling water to set 100C.

On the last measurement, you would have to adjust the temperature for the height above sea level from where you live. The higher elevation, the lower the temperature water will boil.
https://www.engineeringtoolbox.com/boiling-points-water-altitude-d_1344.html

Also, make sure the water is as free as possible of dissolved salts. Purchase some distilled water to perform your experiments.

Thats the good point , i dont know accuracy of mercury thermometer but +/- 3-4 C seems to me big difference :/

How can i do test with boiling water?Make some extension wire and put it in water, but is sensor (TO-92 package) waterproof?

Thanks;-)

- - - Updated - - -

How are you measuring the voltage?

Measuring the temperature of your fingers is NOT the same as measuring your body temperature.

If you can, putting the sensor in an ice-bath would be a good measurement point.

Is your mercury thermometer measuring the same point as your sensor? Maybe hot components on your PCB?

There are a lot of elements of the design here that you haven't mentioned.

Thanks for answer,

no others elements can cause extra heating because for testing im connecting sensor alone on pcb and measure voltage with digital multimeter . I tried also with putting 100nF capacitor between Vcc and ground, also RC (100 Ohm , 1uF) on Vout and nothing gets better .

 

[c_mitra]

Put it inside a ice bucket and see the reading. One point measurement will be sufficient because you cannot correct for span.

+/-0.5C is quite liberal margin. If it reads beyond that, there is something seriously wrong with the sensor (or the electronics). Perhaps 100C will be too much for a band gap based silicon sensor (it may show some non-linearity but that can perhaps be taken care of by +/-0.5C margin!)

 

[KlausST]

Hi,

How can i do test with boiling water?Make some extension wire and put it in water, but is sensor (TO-92 package) waterproof?

Maybe you have an empty ballon by hand....or something similar

Klaus

 

[Ma3ix]

I compared LM35DZ with digital multimeter (temperature measurement) and LM35 has +2~2.5 C at room temperature. The smallest difference gives me in the water, about +1.5 C at LM35.
Multimeter
range: 20°C ÷ 400°C
resolution: 0.1°C
accuracy: ±(3.0%rdg + 5°C)

For example at 20 C, the accuracy of multimeter should be ±(3% of 20 + 5?) or?

Is this temperature difference LM35DZ vs Multimeter okay?

 

[andre_luis]

dont know accuracy of mercury thermometer but +/- 3-4 C seems to me big difference

I compared LM35DZ with digital multimeter (temperature measurement) and LM35 has +2~2.5 C at room temperature.

At least for me, now it is not too clear if you're concerned on accuracy or fluctuation on measured values. Anyway you should consider that mercury thermometers are somehow built with intrinsic "low pass filters" due to inertia of the fluid, the glass mass, aswell the capillarity force exerted by the pipe, so any instantaneous change in room temperature would be not perceived as it is forthwith sensed by the LM35 sensor.

The water is a nice material to compare performance of both thermometers because of its high specific heat capacity which tends to keep itself at a steady temperature insensitive to sudden changes in room temperatures, therefore you should find some way to totally submerge them in order to avoid effects of the dynamics of the air on the LM35 thermometer. I'm assuming you had not encapsulate it within a bulky thermally conductive housing.

 

[betwixt]

I have to calibrate batches of a similar device with analog output. What I do is attach long wires, bundle them all together and put them in the fridge overnight with a calibrated reference thermometer. The wires pass out of the fridge door to a PSU and a DVM. In the morning I read all the voltages (they always differ sightly) and mark the cables with the measurement. Then I open the fridge and immediately note the reference temperature. From the temperature and individual voltages I can produce a calibration offset for each device which is used in the final equipment build.

I have a very forgiving wife who kindly lets me comandeer the fridge occasionally!

Brian.

 

[c_mitra]

accuracy: ±(3.0%rdg + 5°C)

For example at 20 C, the accuracy of multimeter should be ±(3% of 20 + 5?) or?

Is this temperature difference LM35DZ vs Multimeter okay?

If the real temp (really not known) is x, and the display on the multimeter says y, then you can safely expect x to be somewhere in the interval of y+(0.03*y+5) and y-(0.03*y+5)

Note the interesting additive term 5; this suggests that the measured value is always on the higher side. Most likely your multimeter uses a thermocouple based sensor.

LM35 devices report temperatures in degrees Kelvin (absolute temp) and there are subtle differences.

 

[Ma3ix]

If the real temp (really not known) is x, and the display on the multimeter says y, then you can safely expect x to be somewhere in the interval of y+(0.03*y+5) and y-(0.03*y+5)

Note the interesting additive term 5; this suggests that the measured value is always on the higher side. Most likely your multimeter uses a thermocouple based sensor.

LM35 devices report temperatures in degrees Kelvin (absolute temp) and there are subtle differences.

So, the conclusion is that multimeter can not be good reference point for comparison with LM35 cause it also have its own accuracy error, and difference of 2, 2.5 C on LM and multimeter can be expected?

 

[c_mitra]

All measurements are subject to errors; some errors are random and we can account of them in mathematical terms. Temperature fixed points are also having their own sources of errors.

- - - Updated - - -

I have to calibrate batches of a similar device with analog output. What I do is attach long wires, bundle them all together and put them in the fridge overnight with a calibrated reference thermometer. The wires pass out of the fridge door to a PSU and a DVM. In the morning I read all the voltages (they always differ sightly) and mark the cables with the measurement.

That is not the recommended way; You should get a big bucket and fill it with clean crushed ice to about half full. Also put some liberal amount of water (about 20% so that it fills the gap between the ice). Use this primary standard (fixed point) to check for 0C - you can use this bucket as a standard. (actually the fixed point is the triple point and not the melting point).

 

[andre_luis]

So, the conclusion is that multimeter can not be good reference point for comparison with LM35 cause it also have its own accuracy error, and difference of 2, 2.5 C on LM and multimeter can be expected?

All sensors have their own displacement on average, and once inserted at the probe this adds further variables on its thermal characteristics, so that it is expected that the circuit where they are placed on have the capability of adjusting either its offset and slope, therefore accuracy of the LM35 sensor itself is not an issue.

So far it is still not clear whether the exact issue refers to accuracy or methodology for taking a reliable reading.