Ok, I admit it - this one doesn't really have much to do with arduinos, but some of the people on this forum are pretty smart so maybe someone can help me.
I've built an arduino temperature logger to record what's happening at various points around my central heating system. It's based around the TMP36 temperature probe.
I've checked the probes are accurate by calibrating against melting ice and steam. But I'm having trouble getting believable values when the probes are clamped to 22mm copper pipes. the problem is that the pipes are convex and the probe has a flat end - so I suspect that I'm not getting a good contact between probe and pipe. I'm clamping the probes to the pipe using various springs, and don't believe the values simply because the probe is telling me a pipe is at 45C when it's too hot for me to touch!
I've tried a bit of thermal grease, but that seems to run off as soon as the pipe warms!
Hi,
Have you tried wrapping the assembly with cloth or spray foam to seal the immediate area around the pipe and sensor.
As you would gather its all to do with heat loss, so minimise the heat loss and the environment around the sensor will approach that of the pipe surface.
Another way is to make a small bracket that has curve of pipe on one edge and a hole, the hole is a snug fit for the sensor and add with thermal grease.
You make it as small as possible and then wrap it again to minimise heat loss.
Exposed sensors like that on surfaces in the open will not give accurate temperature readings.
Have you tried wrapping the assembly with cloth or spray foam to seal the immediate area around the pipe and sensor.
Not sure this is practical in this situation. I'm working in a very confined space.
Another way is to make a small bracket that has curve of pipe on one edge and a hole, the hole is a snug fit for the sensor and add with thermal grease.
this is a possibility - I'll have a look at what I've got to make the bracket from.
Fulliautomatix:
But I'm having trouble getting believable values when the probes are clamped to 22mm copper pipes.
I've tried a bit of thermal grease, but that seems to run off as soon as the pipe warms!
It doesn't sound like you are using thermal paste. The picture shows temp probes under a short length of pipe insulation. They were embedded in a dab of paste and the insulation is sufficient to keep them against the pipe. In the bottom left corner, another sensor is in a properly made thermowell. It doesn't appear to work any better. I understand silicon goo also works well. Anything is better than air. 
Hi,
I have used thermal paste and also metal-filled epoxy. The TMP36 has very little thermal mass, Also run the first 3+ inches of wire along the pipe and insulate with pipe insulation so there is not a wire conduction heat loss.
Seems to work well for me..
And I have used DS18B20 sensors on plastic pipe in a similar way with good results. They may be better because they are digital so wire length is not important, and you can connect multiple sensors on the same circuit / wires. See examples HERE.
And if you need to measure temperatures up to 1000F see the examples HERE. (My woodstove flue gas is 533 F at the moment).
Fulliautomatix:
Says it is on the tube! it's the same stuff I use on heat sinks - isn't that what I should be using?
It sounds right but not doing right. Perhaps it is outside its specification. Silicon 781 is good for 150C but should be applied at room temperature, and I think it should still go under a foam sleeve.
I see a pipe plug at the top of the copper pipe can you put a thermal well in there thats the best place other than directly in the actual fluid there are also thermistors that thread in and measure actual fluid temp take a look at or for a Balboa M7 thermistor
rogertee:
I see a pipe plug at the top of the copper pipe can you put a thermal well in there thats the best place other than directly in the actual fluid there are also thermistors that thread in and measure actual fluid temp take a look at or for a Balboa M7 thermistor
The plug is where a matching dial type came out. The probe is 250 or so long and just misses the turbine. A wet PT 1000 could also go in there. It is 100 long. The little digital suffices, but they are far from weatherproof. There are PT1000s at the bottom in a typical installation. They have never been used as DS18B20s do it all. That position at the top is not ideal for thermowells. They are normally transverse to the flow.
I have a thermowell for the second DS18B20, but the current arrangement is fine.