On 3/16/2016 3:19 PM, Joseph Gray wrote:

Because that's not really making it colder or warmer? :-)

Sounds like you want a moving average, so you don't see the sudden
changes, even when they exist. Add some thermal mass to the sensor. Easy
way: waterproof it (epoxy, Plasti-Dip, etc.) and stick it in a bottle of
water.

You should use oil so that you do not get evaporative cooling.  I have
worked with temp rules that talk about oil filled 1 quart jars spaced
around the point you want to measure.

On Wed, Mar 16, 2016 at 1:22 PM, Mike S mikes@flatsurface.com wrote:

--

John Phillips

These questions come up regularly. It is much much much easier to average
it in software after the fact, then you can apply the filtering you want or
try different ones.

Look at this temperature graph (it takes a while to load, there is a lot of
data) (yes, it may even take longer than that):

http://www.eds-fl.com/test/plot.php?file=5-68.1.37.121.csv&length=full

The number in the bottom left corner of the graph is the moving average,
change it to smaller or bigger to see how much filtering you want.

On Wed, Mar 16, 2016 at 3:22 PM, Mike S mikes@flatsurface.com wrote:

On 3/16/2016 4:47 PM, John Phillips wrote:

Better still, but more potential to be messy. Another advantage is that
there's no need to waterproof the sensor. I was thinking along the lines
of sticking the wire through a small hole in a bottle cap, so there's
next to no evaporation or spill potential. Even using an open glass of
water, evaporation isn't likely to make enough of a difference given the
accuracy/resolution of a DS18S20.

Averaging in software is preferable for temperature measurement.
If you shield a sensor from air currents you leave it far more
sensitive to radiation errors and errors from conduction along the leads.
These problems only increase if you try to place a brick on the sensor to
average the response, you simply decrease the dependence on air temperature
and increase radiation and conduction problems.
cheers,
Neville Michie