LM92 temp problems - again

Hi to all. I have 2 lm92's connected to a controller. One reads 22 deg C and the other reads 24.5 deg C. How can they read such different temps. They are within 1cm of each other in free air. There are just 4 small wires from each to the controller board. Is there a problem with these sensors that make them read such different temperatures. During soldering none of the legs were heated for more than about 1s at a time. cheers Rob

The fundamental problem with most temperature sensors is self-heating.

The LM92 draws 0.35mA typical, 0.625mA worse case. You haven't specified the voltage you are applying across your sensors - it has to be at least two volts. and a controller might apply 24V.

The LM92 data sheet specifies a package thermal resistance of 200C/W for a device mounted on a printed circuit board made wth 2 ounce copper (one ounce copper is more usual).

24V by 0.325mA is 7.8mW, which corresponds to a temperature rise of 1.56C with a 200C/W thermal resistance. The worst case 0.625mA corresponds to 15mW or 3C.

If you've got the parts suspended on wires in free air, the thermal resistance to ambient will probably be higher than 200C/W, and the temperature rise correspondingly higher.

My guess would be that the difference in measured temperature that you see corresponds to a difference in current consumption (which you can measure). Adding clip-on or glued-on heat-sink to the part should help a lot.

Farnell does a thermally conducting rubber-based adhesive (part code

130-485) which might be useful. I don't know whether it is worth of the order of $50 per tube.

--
Bill Sloman, Nijmegen

Hi there. Both LM92's are running from 5V. That's about 1.75mW or 0.35 deg C rise in temp. The difference I'm seeing is 2.5 deg C. That's a big difference. Still stumped:0( Cheers Rob

Tape them together with heat sink grease to thermally couple them and see what happens.

Hmm, the LM92 accuracy spec at +5v is -0.13/+.53 degrees C, so 2.5 degrees difference _is_ more than the maximum 0.53-(-0.13)=0.66 deg. C expected.

1) Are the parts really measuring the same temperature? Try making sure by clamping both to one big bar of copper or aluminum, the whole covered in bubble-wrap or some such to protect from air currents. 2) Digital output loading: if high, could cause self-heating. 3) Are both active and inactive for equal amounts of time? That could produce differential heating. While active the LM92 draws 0.35-0.63mA; inactive the parts draw

1.75mW is 0.35 deg. C for a thermal resistance of 200 deg. C per watt. The 200C/W figure was for a part mounted on a printed circuit board made with 2 ounce copper tracks.

Having the hanging on thin wires in mid-air is going to raise the thermal resistance by quite a lot - maybe enough to get it to 2000 degrees C per watt.

The package is small, and the temperature rise relatively low so the Rayleigh number could well be below 600 - which would imply that there wouldn't be any significant convection cooling, which might push the thermal resistance well above the 150C/W figure usually quoted for discrete transistors in TO-92 packages.

--
Bill Sloman, Nijmegen

Hi there. Something to think about. I'll try some of the suggestions to get things the ic's at the same temp.

put the 92's on sockets....grab a a handful, move them about, see what follows what, the devcie or the postion? how are the two hook ed pu? sepratley or intermixed? any differences in the "ground return" path?

Good luck