Dallas ds18b20 dreaded 85.0 reading.

[dr pepper]

I've used these before and not had any problems with them, but my current project repeatedly comes back with 85.0 degree readings with nearly no aquisition delay.
I've tried a couple of sensors and had the same results, they are not the + versions that only operate on paraitic power, they are 3 pin ones.
The circuit has a 12v to 45v inverter to drive a vfd which does create a little noise, putting a 100u accross the 5v has improved things, now I occaisionally get an accurate reading amongst 85.0c ones.
Is this going to be a power supply issue?, I could maybe run the 18b20 off the 12v rail through a resistor, zener and cap arragement if absolutely necessary.

 

[picbits]

Bung a load more decoupling capacitors as close to the sensor as possible - prefereably a 100nF, 10nF and a 1uF ceramic - see if it makes much of a difference.

If you have an inverter on the power line, a 100uF capacitor won't do an awful lot to get rid of noise on the line although it may help a bit for voltage droops etc.

 

[dr pepper]

I temporarily held a 100n across the power supply pins on the ds, didnt work, I then soldered a 100n right up next to the connector on the board that goes out to the ds and changed the ds for another, that worked, I since tried the old ds that was reading 85.0 and it still does even with the extra 100n.
Seems like using one of these without good decoupling wrecks them.

 

[picbits]

Possibly worth banging a 5.1v zener across the pins as well if its spikes that could be killing them

Power supply noise can be a nightmare. Hope you get it sorted.

 

[dr pepper]

Its sorted, has been working all night, thanks for the input.

I wonder if I connected the ds the wrong way round, quite possible as the wires are all the same colour.

 

[tunedwolf]

Power it through a choke with decoupling caps on the far side of the supply, perhaps also fitting a tranzorb across it. Of course the best course of action would be to eliminate the noise at source if you can as it will also be affecting other circuitry, even if it's not immediately apparent

 

[dr pepper]

I didnt go quite to that extent, I used the same technique I use with infra red receiver modules which are sensitive to noise, 5v comes through a 100r and theres a 100u cap accross the ds, the ground going direct to the supply, all works well now.
I spose the analogue section of the sensor needs a reasonably quiet supply to be able to do its job.
There must be some mechanism that causes the chip to abort if the supply isnt good enough.