I am designing a control for a commercial micro distillery and I was talked into using an Arduino instead of a PLC. The still uses a number of temperature sensors and the critical temperature range where accuracy is needed is 70 C to 100 C. The still will never go over 100C. I looked first at analog sensors like RTDs, but the cost of the supporting circuitry to do it well is high. I also looked at semiconductor devices like the LM34 and the LM235, which are accurate in the range I need and are simple to use. However, they are analog devices and I am concerned about corruption of these very low level analog signals in a distillery environment. So, most recently I have been considering the DS18B20, which is fairly easy and inexpensive to use and has the accuracy needed in my temperature range. It is digital and it does send a CRC byte so I assume that signal integrity can be checked. Physically, it will be well protected in a thermal well mounted in a still wall.
Since I have no experience with these devices, I would appreciate any input regarding reliability or suitability of these devices.
You mention most of them, and each one has advantages and disadvantages, which you also mention.
DS18B20 : accurate, very accurate. Up to 125 degrees Celsius, above 125 degrees it is broke.
TMP36 and the like : analog
NTC : analog, versatile
RTD : for medium high temperature.
ThermoCouple : large range
The still only operates up to the boiling point of water and then it shuts down. So I'm not concerned about the 125C limitation.
If I use the DSB1820, I will purchase one like the Adafruit probe. However, it is inserted from the outside into a metal thermal well fastened to the still so it never touches liquid and it is totally surrounded by grounded metal. We will probably push the probe well into the well (a foot or more) and fill the well with heat conductive grease or the like. The leads will be terminated within a shielded box fastened to the thermal well on the outside of the still and a shielded cable will run from there back to the controller. Worst case, that cable is 15 feet.
A cable of 15 feet (4.5 meters) is no problem. But some cables cause problems.
The DS18B20 can use parasitic power with 2 wires, if you avoid that and use 3 wires it is more reliable. Is that a shielded cable with two or three wires ?
Sometimes temperatures are requested too soon after each other, or the checksum fails. So you always have to check if the temperature was valid.
The DS18B20 is so useful and so easy to use it is an obvious choice. The fastest read frequency is about 1Hz, which is surely fast enough for your purposes.
The cable length is OK as long as it is decent shielded stuff. All my DS18B20s are the water proof eBay versions and come complete with 5m cable and plug. I don't know why you need such a deep thermowell, but they would be fine in one.
Thanks for the very helpful information. The still is 350 gallons, so one of the sensors will be in a deep well. The others, not so deep.
I am definitely going to use the 3 wire method. If anyone has a suggestion for a shielded cable I should use, it would be appreciated. Since one of the 3 wires is ground, I imagine I only need 2 shielded wires?
oldradio:
I am definitely going to use the 3 wire method. If anyone has a suggestion for a shielded cable I should use, it would be appreciated. Since one of the 3 wires is ground, I imagine I only need 2 shielded wires?
All my probes came with plugs but I am using them in groups of three. If you are using more on one board, it it may be better not to bother with plugs and use a terminal board. I'm not sure all my cables are shielded. They all perform OK but I had trouble when a made some extensions with cheapo multi-conductor cable.
When you use 2-wire shielded cable and use the shield as a ground, the shielding is connected to the ground of the Arduino.
That means that high electric noise could lift the ground level near the sensor.
Since the communication over the cable is digital, this is not a problem.
A cheap thin shielded cable introduces more capacitance between the wires and the shield, reducing the maximum length.
And if the shield is connected to the ground of the Arduino, a shortcut to the shield could damage the Arduino.
It is all about how far you want to go, and how much money you want to spend.
Since your distance is not very long, you could even use Ethernet cable or USB cable.
The most important thing is to buy a known and good quality cable. Some cables from Ebay look thick and sturdy, but have very poor quality plastic and only a few thin hairs of copper inside.
Thanks for the suggestions!
The ethernet cable sounds interesting. After reading more I have decided to operate each of the 3 sensors on a separate wire and Arduino pin. It seems that most problems of a flakey nature happen when you add too many devices to long wires. I have plenty of I/O points and I can write the software to be used with each wire.
I have contacted Maxim support about operating the devices in the 80 C to 100 C range, and so far they have responded positively. They believe their life will not be degraded and whatever error they have will remain fixed. Since that error might be as much as 2 C, my plan is to measure the error at 80 C and 100 C and make corrections for each reading in software.
I would love to protect the Arduino from spurious signals that might get in through this wire. Is there such a thing as a bidirectional opto-coupler?
I wish I could provide a picture of the still, but it hasn't been built yet. The production still is many months away.
Either the DS18B20 or the LM34 should be suitable for your application. The DS18B20 provides potentially greater accuracy and resolution when interfaced directly to an Arduino, but can be problematic with long cables. Used properly, the LM34 should be usable over longer cable lengths, since there is only an analog voltage being transmitted. I suggest you use the DS18B20 unless you think the cables will be too long for it.
The accuracy of the DS18B20 is +/-2oC but the precision is up to .0625oC (12 bit) with a drift of +/-.2oC*.
What this means is that assuming you have a separate, more accurate temperature reference then you can calibrate each device and it will stay accurate to +/-.2oC for the next 1000 hours*. In your shoes I would believe that the drift specification is what you really care about, because once you get your still set and running right it will continue to run stably enough for the next month before you need to recheck the calibration.
Since my cables are short, I am going with the DS18B20. It appears I will have to periodically recalibrate to account for drift over 1000 hours use. This means I will have to write my software to make this an easy task.
