Does anyone have any suggestions on how to improve the Thermal Conductivity for DS18B20 sensors pushed into ‘pockets’ (small tubes embedded in pipes or tanks)?
I have a few pockets on the tanks with sensors in them but they are ‘lose’ and I wonder if there is a thick paste to assist conductivity?
Heat sink compound (the zinc-loaded white paste) or Silicone Grease. I suspect the latter is only good if the pocket points in the right direction - downwards.
Yes, that was part of my concern. These pockets are all horizontal so if the grease/paste does not stay thick at higher temperatures, I’m wasting my time and money!
I did read somewhere about using kitchen type aluminium foil loosely crumpled around the sensor and then pushed into the pocket with a thin stick. If you do it right, the sensor should be fairly tight in the pocket (but still easily removable) and the aluminium should sort out the thermal conductivity.
I remember thinking after I’d read it that maybe copper foil may be a better idea…
I don’t think the paste that I’ve used on transistors & heatsinks would flow of its own accord, though it can be spread. The consistency is almost like marzipan, it certainly wouldn’t drip. The 2.7 W/m·K syringe (on the page linked above) is good to 300 °C (it says).
Concur.
I’ve used quite a bit of the stuff on consumer and military hardware.
It has always stayed where it was applied.
Do you need to improve conductivity?
If the pockets are reasonably deep the internal air temperature of the pocket will be pretty much the same as the liquid temperature anyway.
Don’t know is the short answer.
It has been on my mind for a while so I’d like to see if it will make a difference 
My solar installation is using these tubed type of sensors by design.
They have addes some sort of spring to the enclosure which pushes the sensor to the inside of the tube. There is no mention anywhere of adding thermal grease for better conduction in the manuals. They do however suggest to close the pipe with some insulation material, so I added some rockwool after the sensor was mounted and didn’t really have any issues measuring the temperature inside the boiler at different layers.
This is the other solution they provide:
I bought this stuff for a try out.
You probably only need a very thin layer to help improve metal to metal contact for heat transfer from pocket to sensor. This most likely will help with sensor response speed but not with accuracy.
As somebody already said, its probably likely that the air in the pocket that is actually heating the sensor as that is where the most contact is. AIthough people do, I would avoid slopping on loads of thsi stuff as it will then slow down transfer of heat through the air giving you a worse response. Just a sliver, to help get some meaningful metal to metal contact like with CPUs and heatsinks.
Why not thermally conductive glue/epoxy? Are we talking solar collector high temps or something else?
I would want them to be removable
90°C for me.
I agree that conductive epoxy is the best solution. I have a similar problem with thermistors in metal sheaths on my HWC and they read up to 10C low, despite conducive paste. Unless the sheath is mounted side on with lots of paste there simply isn’t enough contact area. Mine are placed at say 30 degrees with only the tip in contact and it’s just not enough area.
I guess one solution would be to epoxy a short copper tube onto the tank and slide the sensor into that with paste filing all the gap between tube and sensor.
Yes - engineering wise that is the best solution - (a) physically forcing the sensor to touch the tube wall and (b) stopping draughts coming down the tube.
