1Wire how often and how much

[MN]

Anyone polling a 1 wire circuit of DS18B20s.  How often do you poll them?
And what accuracy do you store them to.

I was just thinking saving to 3dp is a waste of time really and maybe storing to nearest 0.5oC would be good enough.

MN

[KenB]

MN,

The DS18B20 is accurate to +/-0.5C across its operating range and has a 9bit to 12bit temperature output.

It takes approximately 750mS to produce a reading.

If you consider the thermal mass of the fluid or pipe that you are measuring then clearly it cannot exhibit  very rapid changes in temperature.

Taking a reading every 6 seconds and then averaging 10 readings to give a log of temperature every minute might well be sufficient for your needs. Saving this data to one-tenth degree should be perfectly good enough to spot trends.

As an aside - I am currently working with a thermistor sensor which I have linearised by approximating its output to several straight line graphs and using y=mx+c to get the temperature in degrees C. This is easy enough to do on the PICaxe chip working in BASIC. It works surprisingly well and follows a commercial thermocouple digital thermometer to within 1/10th degree.

I average 64 readings to eliminate the noise and then read the average every 6 seconds.  I'm currently measuring outdoor temperatures and have noticed that these seldom change by much more than 0.1C per minute.




Ken

[wookey]

I'm polling every 30 seconds at the moment. I used to do it more often but decided that was excessive.

One day I'll change things so that I can read some sensors more frequently than others.

[sleepybubble]

I'm polling ad-hoc, must get that slug writing to the database tonight!

[wyleu]

I go round the event loop a certain number of times and then package up the data with a start and stop time event and fire that off.
It's an area of debate that is probably chaotic, because the reason your monitoring is your trying to gauge performance to allow you to make decisions based on how frequently you make measurements.