>> Anybody with experience with DS18B20 at 30m ?

> Meeemeeemeee ;-))).
>
> Even at 100 metres, but if you want it reliable you cannot
> connect it "one pin only", it is necessary to use "active
> pull-up" instead of pull-up resistor and sometimes even "control
> slew rate"... with this length the line falls easily into oscillations...


Me to, with rather large networks (tens of sensors, altogether hundreds of metres of cabling), wired carelessly ("hala-bala"), simply as it happened.

The most important is to have a well designed driver and correct timing. Then it works very well. I used output of processor (three-state driving) + passive pull-up, once, for a couple of tens of metres, when there was no other option, and it worked - but as I wrote above, it is necessary to have a well written timing. If it is out of bounds, it works for a metre or two, but as soon as there is a longer cable, it starts making errors. I also put a 5V8 unipolar transil next to the driver (both supply and data), the same next to the sensors on exposed locations (e.g. sensor on the roof).

Sometimes it happened (say 5 times altogether in 30 applications a 20 sensors in 5 years), that some sensor went to latch-up and that took down the whole net, it helped to toggle the supply and it went on merrily. Since some time I preventively switch off the supply for 100ms and switch it on again - applications that have this don't have the mentioned problem, but anyway from time to time some of the Dallas fails definitively (e.g. draws a big current from the supply wire, if it would be in latch-up, but it won't clear by interrupting the supply), the reason I suspect to be something of physical nature, maybe improper packaging of the sensors or something like that, but I am not competent in this, so it's only a guess. Last time this happened when two sensors failed shortly after each other, when pool water got into them. A better insulated sensor was put there and it appears it is OK.

Lukas Grepl