Temperature probe with a DS18B20

[MikeK]

I got promising results from using my infrared thermometer to locate a current pig on a circuit board.  I want to confirm the results with a temperature probe touching the suspected faulty IC, since the IR thermometer is big and bulky.  I have a DS18B20 temp sensor that I can use.  I figure I'll sand the top of the sensor smooth (get rid of the seam) to get more and better contact with the overheating part?  I have no heatsink compound that I could find.  Any way I can improve my temperature measuring?  I have a small lathe...I could make an aluminum contact pad that I hot glue to the sensor top?

[MikeK]

Nevermind; I think I was overly concerned.  Any temp sensor up close to a part will be good enough.

[Ian.M]

Why not put the DS18B20's flat face with the part no. against the surface you want to measure for minimum thickness of case plastic between the surface and the die?  Also beware of heat loss via the DS18B20 leads.  You can minimize that by using very thin tracks near the DS18B20, or for a scratch-built probe, a couple of cm of hair thin magnet wire inside the rigid part of your probe, between the (shortened) leads and the cable back to your MCU.
A closed cell foam sleeve (or block) for thermal insulation over the body of the probe, ending flush with the sensor face would also be beneficial.

[MikeK]

Good tips, thanks Ian.

[Ian.M]

Note that *ANY* filler, applied in a thin enough layer between two approximately flat surfaces in contact is better than nothing as very thin air gaps are notoriously poor thermal conductors.  If you don't have heatsink compound, a small dab of any high temperature grease, or even toothpaste will do, but in your application its probably a good idea to avoid ones containing electrically conductive particles due to the future trouble stray traces are likely to cause. e.g. avoid graphite grease, copper and aluminum antiseize, and Noalox and similar zinc loaded electrical jointing compounds.

[Doctorandus_P]

The plastic of the DS18B20 does not conduct heat very well (In the TO-92 package). At least a lot less then the metal leads going into it and the leads act both as a thermal mass and a heatsink.

Any PN junction (from a diode or BE- transistor junction) can be used as a temperature sensor (with a temperature coefficient of approximately -2mV per degree centigrade. Transistors in a TO-218 package or the old metal can packages such as TO-18 seem particularly suited to be used as temperature sensors.

[Ian.M]

True, which is why I emphasized using very very thin connections to the sensor for the last few cm.  It may even be worth using fine enameled resistance wire rather than magnet wire, as e.g. Constantan alloy has a thermal conductivity of approx. 5% of that of Copper, but with care + good flux is still solderable.

One doesn't choose a DS18B20 over other sensors for its package's thermal performance, one chooses it for its factory +/-0.5°C calibration and  minimal interface circuit. 

BEWARE: There are a lot of counterfeit DS18B20 sensors out there, of dubious accuracy - if you didn't buy direct from Maxim or an authorized distributor, odds are you've got a fake.  See https://github.com/cpetrich/counterfeit_DS18B20 for details.

[MikeK]

BEWARE: There are a lot of counterfeit DS18B20 sensors out there, of dubious accuracy - if you didn't buy direct from Maxim or an authorized distributor, odds are you've got a fake.  See https://github.com/cpetrich/counterfeit_DS18B20 for details.

Indeed!  I've already be fooled by this.  That Arduino sketch had helped me ID the fakes.

[MikeK]

Interesting link below on shaving down the case.  Apparently there's a copper slab inside, for conducting heat.  Fake sensors probably use cheaper metal or do without.

https://hackaday.io/project/158004-trimming-down-dallas-ds18b20s

[magic]

In TO92, TO220 and similar packages the die is placed on a copper plate which is an extension of one of the leads, usually the central one but it can also be one of the side leads in TO92. Hence it is electrically and thermally connected with the lead, which is usually collector/drain or ground/VCC in case of ICs.

You could remove almost all the plastic material from the front side of TO92 to gain access to this copper - there is nothing in there. Just keep enough plastic that the package doesn't fall apart. The copper plate will still be one part with one of the leads and conduct heat away.

The rear side contains the die, as that guy found out, and also bond wires which connect it with the other two pins. You will damage the wires if you file too much of the rear, a millimeter or two before reaching the die itself.