DS18B20 Heads Up

rq3

Senior Member
Last year I built a controller for my fridge/freezer based on a Picaxe 20M2 and three DS18B20 temp sensors. It's been working great BUT:

One of the sensors monitors the temp of the fridge evaporator, so it is exposed to below freezing temps, and then thaws to above freezing temps (to ensure that the evaporator defrosts). This is how the original fridge controller worked, with two sensors, one to monitor the defrost of the evaporator, and one to monitor the actual temp of the fridge contents. If both were warm enough, the compressor would turn ON.

The evaporator temp sensor has now failed twice! The second attempt involved potting the DS18B20 in a small stainless steel tube with urethane having a glass transition of -72C. This material is specifically designed to pot components exposed to extreme temperature variations, as it stays relatively flexible to prevent stress to the potted components. I use it extensively in a commercially available product with no problems over many years.

I can confirm that the DS18B20 cannot tolerate being exposed to freeze/thaw cycles for very long, having experienced failure of this component twice in those conditions. The failure seems to be failure of the internal wire-bonds within the TO-92 package. Whether this is due to moisture (not likely, given the urethane potting) or thermal expansion (likely), is unknown.

The DS18B20 units that remain frozen (freezer control), or remain above freezing (refrigerator control) seem fine. I "fixed" the fridge by shutting off the evaporator sensor in software (thanks to the picaxe!) and just rely on the remaining "fridge contents" sensor to control the fridge compressor.

Just thought this might be of use to anyone using the DS18B20 sensors.
 

westaust55

Moderator
Thanks for the information.
Good for potential future reference when someone provides such feedback.

Have you contacted Maxim IC in relation to your findings to see if they can confirm the problem is as you suggest?
 
Last edited:

PhilHornby

Senior Member
Strange...

Last year I built a controller for my fridge/freezer based on a Picaxe 20M2 and three DS18B20 temp sensors. It's been working great BUT:
...
...
I can confirm that the DS18B20 cannot tolerate being exposed to freeze/thaw cycles for very long, having experienced failure of this component twice in those conditions. The failure seems to be failure of the internal wire-bonds within the TO-92 package. Whether this is due to moisture (not likely, given the urethane potting) or thermal expansion (likely), is unknown.
I built a very similar project - which has been running since June 2014, without any problems (touch wood!). It's in a small under-counter fridge of circa 2002 vintage which originally used a mechanical 'whirly-gig' to sense the temperature of the cooling plate at the back. When it hit about -20ºc, it powered-off the compressor and didn't re-trigger until the cooling plate temperature climbed to some arbitrary, above zero value. The air temperature in the fridge, could be anywhere between 0ºc and +12ºc ... seemingly at random.

This fridge must have been out-of-date when purchased, because modern ones have the cooling 'plate' arranged so that you can't see it from inside the fridge - presumably, just the other side of the internal plastic casing. Cheap fridges in the UK, still use this rubbishy whirly-gig arrangement though.

I used an 08M2 and 2 x DS18B20 - one attached to the underneath of the top shelf, and the other attached to the cooling plate (I drilled a small hole in it :eek:). A single LED conveys status information (including flashing S-O-S in morse, if the temperature exceeds 8ºC!)

The control algorithm is simple: Compressor ON when temperature of top shelf > 4ºC and OFF when < 2ºC. Every 8th cycle, is a defrost operation. The compressor remains OFF, until either the cooling plate >0.5ºC or the top shelf reaches 5.5ºC. This seems to give adequate defrosting and average temperatures of between 0º and 5ºC everywhere within the fridge. Due to inadequate suppression?, the program occasionally resets when the compressor is turned ON - and necessitated another addition to the algorithm: it cannot be switched ON within two minutes of it being switched OFF. If you don't observe this rule, the compressor suffers some kind of internal 'fit' and trips-out ... but not before convulsing violently enough to make the whole fridge move! :eek:

The DS18B20 variants I used, were the 'waterproof' ones - which appear the same as these: https://www.adafruit.com/product/381
 
Last edited:

srnet

Senior Member
I can confirm that the DS18B20 cannot tolerate being exposed to freeze/thaw cycles for very long, having experienced failure of this component twice in those conditions. The failure seems to be failure of the internal wire-bonds within the TO-92 package. Whether this is due to moisture (not likely, given the urethane potting) or thermal expansion (likely), is unknown.
What was the range of temperatures it went through, low to high and how many times did this freeze thaw occur before failure ?
 

Jeremy Harris

Senior Member
There are a lot of fake DS18B20's around, especially on a certain well-known auction site. I've bought and used, or tested, a few dozen DS18B20's, from various sources, and found that around 2/3rds of them have markings that are inconsistent with the genuine Maxim parts. Some of the apparent fakes work OK, but some have a temperature offset error of up to around 1 deg C. All the ones that have what looks like a genuine Maxim product marking seem to work fine.

I now only buy these from an accredited re-seller, one of the big name component suppliers that will guarantee that they are genuine parts, even though this is a lot more expensive. Sadly, I've not found a definitive way to determine the suspected fakes from the genuine ones, as the markings seem to vary from batch to batch. I only started looking at them when I was testing a data logger on the bench and was getting different temperature readings from four DS18B20's that were all sitting at room temperature. I taped them all, flat side down, to a block of aluminium and checked their readings against a calibrated glass lab thermometer that was stuck in a hole filled with thermal grease in the same alloy block. The fakes seemed linear over a reasonable range, but had a consistent offset that was outside the Maxim spec.

So far, all the DS18B20's I've bought from non-accredited suppliers seem to be fakes, as far as I can tell. Some came from smaller UK component suppliers, and I'm guessing they may well have sourced them direct from the Far East. If they have temperature errors my guess is that they may well have other manufacturing defects that are not so obvious. It reminded me of the fake power mosfet's I came across a few years ago, they could only be detected by a variation in the product marking and slightly higher than spec Rdson, but would fail (usually explosively) in switched mode controllers because they weren't able to withstand anywhere close to the dV/dT of the genuine parts.
 

techElder

Well-known member
but would fail (usually explosively) in switched mode controllers because they weren't able to withstand anywhere close to the dV/dT of the genuine parts.
I learned my lesson about dV/dT faults from the manufacturer of a 10K amp output (low voltage) machine with counterfeit SCRs in a (high voltage) variable power supply. KABOOM! The first time it was loaded to 50%.

Months of waiting for delivery and all the downtime wasted.

It used to be that eBay suppliers were selling manufacturer's "off" products, but not anymore.
 

Jeremy Harris

Senior Member
It used to be that eBay suppliers were selling manufacturer's "off" products, but not anymore.
Very true, but I was still surprised to find that there were Chinese semiconductor manufacturers making counterfeit components on such a large scale. The MOSFETs I had the problem with were IRFB4110's, a 100V, 180A, 4.5mohm Rdson part. The fakes had dubious looking product markings and codes, that seemed to indicate they were made in IR's Mexican facility (this was before IR became a part of Infineon), which was odd as IR didn't make this particular device at that plant.
 

srnet

Senior Member
I would suggest that there is not a problem with geniune DS18B20s being exposed to freeze thaw cycles.

The DS18B20 on $50SAT went from -30C to +30C and back down to -30C on every orbit. There were just over 9,000 orbits before we stopped hearing from it, there was no indication of problems with the DS18B20.

This extreme temperature cycling was one of the discoveries from $50. Previous to this the larger Cubesats had stabilised at a few degrees below zero, with relativly small amounts of temperature variation, so we were not expecting the temperature cycling on $50SAT to be so severe.
 

rq3

Senior Member
What was the range of temperatures it went through, low to high and how many times did this freeze thaw occur before failure ?
28 degrees F to 37 degrees F. I'm guessing probably 24 cycles per day for 7 or 8 months. The sensors that don't go through the freeze/thaw cycle are fine (so far). These were purchased from Mouser, a major supplier here in the states, so I can only assume that these were genuine Maxim devices.

I bought half a dozen at the same time; maybe I just got a flakey lot?
 
Last edited:

PhilHornby

Senior Member
28 degrees F to 37 degrees F. I'm guessing probably 24 cycles per day for 7 or 8 months.
Only ±2°C ? ...

The one that's attached to the 'cooling plate' (for want of a better name), in my fridge is subjected to -20°C -> +5°C. (I had no idea that temperatures that low existed in the fridge section, until I embarked on this project).

Here's some pictures of mine:-

https://www.dropbox.com/s/aihl81oo4pqxc65/IMG_0070.JPG?dl=0 - Much abused DS18B20 attached to 'cooling plate' - caught in the middle of an abandoned defrost cycle (door opened?)

https://www.dropbox.com/s/i285c29jn4p5erm/IMG_0071.JPG?dl=0 - blurry close-up of the one attached to the underside of the shelf.

(The Forum software refused point-blank to upload these ... in any size, or any format by any means I can think of ... so they're on Dropbox :(. There might be a bit of clue from the attempted 'Website upload' - it said "error writing temporary file")

(See Hippy's post on next page - he managed to upload them.)
 
Last edited:

hippy

Technical Support
Staff member
(The Forum software refused point-blank to upload these ... in any size, or any format by any means I can think of ... so they're on Dropbox :(. There might be a bit of clue from the attempted 'website' upload - it said "error writing temporary file")
Not sure what's going on there. I visited both images, right-clicked and did a Save Image from Firefox, then uploaded here. Seem to have got the order swapped below, your first is on the right ...
 

Attachments

srnet

Senior Member
28 degrees F to 37 degrees F. I'm guessing probably 24 cycles per day for 7 or 8 months. The sensors that don't go through the freeze/thaw cycle are fine (so far). These were purchased from Mouser, a major supplier here in the states, so I can only assume that these were genuine Maxim devices.

I bought half a dozen at the same time; maybe I just got a flakey lot?
So your sensor had a very limited range of thermal cycling, -2C to +3C, a cyling range of 5C, and they failed in 7 or 8 months.

The $50SAT sensor went through -30C to +30C, a cycling range of 60C (12 times higher) for 20 months and did not fail.

Yours may have failed, but are there perhaps other reasons for the failure ?
 

rq3

Senior Member
So your sensor had a very limited range of thermal cycling, -2C to +3C, a cyling range of 5C, and they failed in 7 or 8 months.

The $50SAT sensor went through -30C to +30C, a cycling range of 60C (12 times higher) for 20 months and did not fail.

Yours may have failed, but are there perhaps other reasons for the failure ?
Perhaps, perhaps. It's just interesting that only the sensor that cycles through the freezing point failed, twice. The other two sensors (one is always well frozen, the other never is) are fine (so far). As I said, I suspect I got a bad lot, since they were all purchased together from a reputable supplier.

Just something to keep in mind if using these. It's unusual to see modern integrated circuits fail when used as intended. The failed units show a constant 0 degree Celsius output.
 

techElder

Well-known member
It's unusual to see modern integrated circuits fail when used as intended.
rq3, seems like you've made a conclusion before detecting the problem.

Actually, it's more logically likely that there is a circuit problem causing the failures rather than cycling a few degrees around freezing. Hard to say without seeing the circuit connections, but you've done some of the proof by detecting two failures in the same circuit.
 

westaust55

Moderator
If the "failed" DS18B20 is only showing 0 degC, assuming that the resolution is in whole degrees then open circuit on the data line beyond the data line pull up resistor is the likely case - whether in the chip as proposed or other reasons.

Open circuit in the data line at the temp sensor would mean that all bits are read as "1".
16 x "1" equates to -0.0625
 

rq3

Senior Member
If the "failed" DS18B20 is only showing 0 degC, assuming that the resolution is in whole degrees then open circuit on the data line beyond the data line pull up resistor is the likely case - whether in the chip as proposed or other reasons.

Open circuit in the data line at the temp sensor would mean that all bits are read as "1".
16 x "1" equates to -0.0625
Two sensors have failed. Replacement of only the sensor fixed the issue the first time. The second failure mode was identical to the first. The circuit uses three of these sensors, all running from the same well by-passed supply that is operating the picaxe itself. It is most definitely a DS18B20 failure, apparently due to a wire bond issue within the TO-92 package itself. I don't find this surprising given the package and the environment, nor do I care to do an etched package electron microscope analysis. Just a heads-up on my experience for others using this device.
 

premelec

Senior Member
@rq3 - if you get a 3d failure please let us know for sure... consequences of failure of this part can be pretty serious in many applications.. perhaps you would be willing to send your failed units to MAXIM for their expertise and comment.
 
Last edited:

srnet

Senior Member
Perhaps, perhaps. It's just interesting that only the sensor that cycles through the freezing point failed, twice. The other two sensors (one is always well frozen, the other never is) are fine (so far). As I said, I suspect I got a bad lot, since they were all purchased together from a reputable supplier.

Just something to keep in mind if using these. It's unusual to see modern integrated circuits fail when used as intended. The failed units show a constant 0 degree Celsius output.
Well, one possibility is that moisture ingress is the problem.

In Space its a vacuum and there is no water vapour around to freeze\thaw.
 

PhilHornby

Senior Member
Frozen...

It's just interesting that only the sensor that cycles through the freezing point failed, twice. The other two sensors (one is always well frozen, the other never is) are fine (so far).
But why would the freezing point of water have any significance?...

...unless :-

Well, one possibility is that moisture ingress is the problem.
Maybe I got lucky choosing the 'waterproof' variety for my very similar project. I only did so, because I couldn't think how on earth to mount a bare DS18B20 inside a fridge. Needless to say, I obtained them from eBay - provenance unknown...
 
Top