Go Down

Topic: DS18B20 self heating or what? (Read 2181 times) previous topic - next topic


Some nasty spikes in the column reporting the difference between sensors 1 & 2. See attached.


OK, it looks like two things have happened to turn this into a bit of a comedy of errors.

First, I'm afraid it looks like my flippancy has muddied the water. Second, I think you have deleted a message that came through just as I was going out the door for a surf.

What I posted was no test, no experiment, just some DS18B20s going about their job in the normal manner and producing exactly what was expected of them. For the most part, they are just reading ambient temperature. The reason why sensor#3 is still in the drawer is that I'm not interested at the moment. Any variations from the bleeding obvious is just me stirring up the graph on the Android, which isn't often..

Despite my intervention, there should be enough data to demonstrate that there is nothing to get worried about. I submit that this thread is nonsense - an exercise in studiously looking for a problem that isn't there, instigated by a failure to read the data sheet.

I don't know what the hell's going on with that graph. It initially appears to be made from the first twenty readings, but bears no relationship whatsoever with the data in the file.  The readings in the file hardly vary and the difference column is equally unexciting. The values on the X scale are not even in order. 


I agree that we have beaten this non-topic to death. Wish I were surfing!


I think that if you want calibrated temperature readings, you need to use an offset value for each individual sensor. I think that the 'standard' sensor is not calibrated for an offset, so it will read consistently for itself within 0.5 degrees as per spec, but the reading is not guaranteed to be within a certain deviation from a standard temperature. So sensors from different manufacturing batches would maybe all read the same, but not the same as a different batch.

That is why I always talked about comparing two or more sensors with each other. That way the error is meaningless.


Maybe better late than never

Just been doing some tests withDS18B20 using the Maxim 1-Wire viewer software.
Bare device, 5-V supply, 1-second polling - self-heating up to 2-Celsius and rising.
Stuck an old Stanley knife blade on as a heat sink and self-heating at 30-second polling less than +0.5-Celsius.
Obvious really. Slugs it a bit. Someone's uncertainty principal about measurements changing the measurement; so un-powered DS18B20 very accurate, powered up, not so good - strap it to a copper hot water pipe with thermal paste, no problem. But any device, PT100 etc, will self-heat.


self-heating up to 2-Celsius and rising.
That sounds more like user-abuse, and I suspect that, if it was even faintly true, Maxim would have gone out of business long ago, and nobody would be talking about DS18B20s today.

While I am one of the guys who fixes his sensors to copper pipes full of cold water, I might well be in the minority, and I bet there are thousands out there who use the sensors out-of-the-box to measure air temperature. And it does a proper job.

I guess any device indeed can self-heat and it just might be that the 125C upper limit of the DS18B20 has something to do with reliable heat dissipation but, if there is a perceived problem, it might be a good time to look at the power input - the watts that are the original source of the heat. I note in the specs that some of the current is measured in nanoamps. I'm not sure I want to ask what a nanoamp is. Similarly, I believe PT100 work on extremely low power, hence their need to run through op-amps to be read by Arduino.


Feb 21, 2015, 04:48 am Last Edit: Feb 21, 2015, 04:51 am by jremington
I've noticed that DS18B20s will "self heat" quite remarkably when left out in the sun!

However, I've never observed a temperature rise when suspended in the air, in a reasonably dark room.


The gotcha with these sensors is that people use them in the black epoxy package and
expect them to measure air temperature, which they don't.

You need one inside a metal package if you want to measure air temperature, otherwise
they pick up radiant heat from nearby objects (people) and read upto several degrees out - air temperature measurement is confounded by the radiant heating.

Black bodies absorb heat radiation efficiently, metals don't.

Strong forced air blowing on a DS18B20 will make them measure air temperature much
more accurately - a batch of 12 under a blower like this I measured had a standard
deviation of < 0.1degC, so they are very good sensors.

If you are in the business of temperature measurement you need to allow for conduction,
convection and radiation and understand what you are actually measuring.  Exposed human
skin can radiate around 50 to 100W per square metre to surroundings indoors and much more towards
the sky (500W/m^2)
[ I won't respond to messages, use the forum please ]


That sounds more like user-abuse, and I suspect that, if it was even faintly true, Maxim would have gone out of business long ago, and nobody would be talking about DS18B20s today.
Not quite true. This was measured using Maxim's 1-Wire  DS9490R RJ11 adaptor (their own bit of kit) and 1-Wire Viewer software. A rough test, to be fair, but user abuse? No. This was the difference between 1-second polling with 5-V Vdd and 30-second polling with Vdd and a bit of heat-sinking. A minute die in the middle of a lump of plastic is going to react more to what's going on inside than the outer surface. There's a lot of thermodynamics going on. Eventually it will all equilibrate. Plastic is an insulator - usually - so it going to take a while for changes either way to settle. More likely the die will react to the device lead temperature. The chances are that these are tested in laboratory grade environmental chambers with plenty of circulation, not on the average DIY user's table. Nanoamp? 10 to the minus 9, Pico to minus 12, Femto to minus 15. I worked with electrometer amps at Femto levels, so hopefully learned by past mistakes.

The usefulness of 1-Wire far outweighs absolute accuracy and it does for me, give or take a degree or so.


I can't argue any of this in detail but, if you see a self-heating error of 2C and rising, it's time reconsider the procedure, reconsider the reputation of the company supplying the gear, or simply give it all away.

My problem is that I have never seen it, but I don't get too scientific about it either. All I have done is put a plain-vanilla Brannan -10>50 mercury-in-glass thermometer beside a DS18B20, and see what I expect to see. This is just as well, since the purpose of the DS18B20s was to replace the laboratory glass thermometers

I recognise MarkT's comments but all mine are in metal package.


Hello Nick

All points taken. I guess the message is always to consider and thoroughly think through the application. I spent quite a few decades doing nothing but measurements in some form or other from redox/pH electrodes - hence the electrometer amps - through to complex chemical analysers - temperature was always critical. This was pre-1-Wire, but the gold standard was always PT100 for temperature, and always 4-lead compensated. That was then, but now it's 1-Wire which I am learning, like Arduino. I am working on a novel heat recovery system for domestic dwellings, using the heat gain from concrete roof tiles and have a string of DS18B20's dotted around the roof-space. A few weeks ago, outside temperature 6-C, tile heat gain 30-C and summer-time tiles can be 60-C. If it works, it's a lot simpler than solar panels and grabs warm air or possibly hot water at relatively low cost. There is a load of stuff to learn and the next job is to design a servo operated duct valve to do the warm air diverting.


Very interesting. There is a guy in Geelong who was doing the same thing. His problems were no storage and a lot of air handling, as I recall it, but I can't find his card.

I'm sure the PT100 is still the gold standard, but the DS18B20 is more than adequate for ordinary folk. You commented before on some flying leads I had in a picture. They are from PT500s that have never been used and just serve as blank plugs in the pipework. They were installed because because I thought they would probably react faster, which may be true but probably irrelevant.

Go Up

Please enter a valid email to subscribe

Confirm your email address

We need to confirm your email address.
To complete the subscription, please click the link in the email we just sent you.

Thank you for subscribing!

via Egeo 16
Torino, 10131