Immersed Dallas One-Wire Water Temperature Sensors

[rubidium]

I like the Dallas DS18B20 one-wire (DOW) temperature sensors. They’re inexpensive, responsive, accurate, and interface nicely with the Crystalfontz series of 5-1/4” bay displays (I’m using the CF635). They’re also supported by Matrix Orbital displays, I believe. I have yet to see what I consider to be an acceptable solution to using these sensors immersed in a liquid-cooled system to measure coolant temperature. There are plenty of solutions that attempt to seal the sensors into the cooling circuit with silicone RTV compounds, particularly the so-called “aquarium grade” ones, but many of these eventually leak (just like most aquariums eventually leak) and almost all that I’ve seen are simply “sloppy” looking (largely caused by the poor manageability due to the high viscosity of the RTV). (Check out the entire thread: http://www.crystalfontz.com/forum/showthread.php?t=2142&highlight=dallas+water and you’ll see a classic example of what I mean - including the eventual failure.:eek:) There are also lots of attempts to place these sensors inside of sealed stainless steel or copper tubes (so-called thermowells), but I expect this approach significantly reduces the responsiveness of the naked sensor, and real thermowells are very expensive. The intention of this post is to outline the recipe for making the definitive liquid-immersed DOW temperature sensor that lives up to the performance, reliability and appearance standards that we're expecting in our builds.:up:

The Dallas sensors available from Crystalfontz come prepared with color-coded wire leads and a cast rubber “boot” that is molded at the base of the TO-92 sensor case to encapsulate the connection points (http://www.crystalfontz.com/product/WRDOWY17.html). These are relatively inexpensive, and are tidy, nicely-made, assemblies. For months, I've been doing a lot of research and testing of sealant compounds for this application. One approach that kept my interest for a while was to dip the DOW assemblies into Oatey Rain-or-Shine PVC cement, which completely encased the sensor in a very thin PVC film. Unfortunately that film wouldn’t adhere very well to metallic plumbing fittings, which are my preference in water cooling, so I wasn’t too confident with that. I also tried the time-tested and water-proof PC-7 epoxy that adheres extremely well to metal, but is too thick to work with in tight places, and really doesn’t adhere to the plastic sensor case very well. To make a long story short, in the end I selected electronics-grade 3M DP-270 clear conformal epoxy – a favorite of RC boat enthusiasts for potting circuit boards in submerged applications. This is available from McMaster-Carr (catalog page 3317), among other places.

The approach uses small G1/4 BSPP threaded/o-ringed nickel-plated brass compression fittings - the kind we’re all familiar with. I am using the Innovatek variety, designed for 6mm ID / 8mm OD tubing (http://www.frozencpu.com/products/2314/ex-tub-62/Innovatek_G14_thread_14_Screw-In_Type_Fitting_500179.html?tl=g30c101s176), because they are conveniently sized for this application, and I had some laying around. Others should work too. The basic fitting is shown here:
http://farm4.static.flickr.com/3028/3095313439_187a0b047a_o.jpg

To ensure high shear adhesion of the DP-270, the nickel plating of the interior surfaces of the otherwise solid brass fittings that will be placed in contact with the epoxy is first removed and the brass is given some "tooth". This is easily done with a bronze or beryllium-copper brush that is typically used for cleaning gun barrels:
http://farm4.static.flickr.com/3129/3095316803_3a572ed376_o.jpg
and repeatedly stroked through the bores while spinning with an electric drill. The base of the fitting is also slightly chamfered, and sanded with 400-grit aluminum oxide paper. Then the entire fitting is repeatedly wiped with methyl ethyl ketone (MEK), using cotton swabs. The result is seen here:
http://farm4.static.flickr.com/3146/3095322957_b03eb8d863_o.jpg
and here:
http://farm4.static.flickr.com/3173/3096169672_5bf01241b0_o.jpg

[Note: I also experimented with etching the brass with a 50/50 mixture of household ammonia (clear / non-sudsing) and household hydrogen peroxide (3%). While this is effective, it is no more effective than using the bronze brush and sandpaper, so I abandoned this more complicated process.]

The Dallas sensor assemblies supplied by Crystalfontz are prepared next. The plastic TO-92 sensor case is prepared for bonding by lightly - but thoroughly - sanding all around with 400-grit aluminum oxide paper, then wiping with MEK using a cotton swab. The rubber boot is also thoroughly wiped with MEK as well. Below, the sensor on the right is in original condition, while the one on the left has been sanded:
http://farm4.static.flickr.com/3025/3095334017_9da1c8f270_o.jpg

Without touching the areas just cleaned with MEK, the sensor assembly leads are passed through two small rubber grommets:
http://farm4.static.flickr.com/3037/3096198828_14abea7ca3_o.jpg

http://farm4.static.flickr.com/3104/3095357991_6f0ea5b38b_o.jpg
that will temporarily be used to support the sensor assembly in the bore of the compression fitting – properly positioning the TO-92 sensor case at the G1/4 threaded end of the fitting such that the sensor case protrudes slightly less than half-way (0.1”) beyond the end of the fitting:
http://farm4.static.flickr.com/3143/3096199608_4a6d2d095e_o.jpg

The fitting is then suitably supported so that its bore is oriented vertically, with the G1/4 threaded end (and TO-92 sensor case) up:
http://farm4.static.flickr.com/3126/3095358681_54f95124f0_o.jpg

The 2-component DP-270 is thoroughly mixed up in a beaker:
http://farm4.static.flickr.com/3278/3095359075_a88ea3cc43_o.jpg
and then poured into a disposable syringe that is equipped with a disposable Luer-Lok blunt cannula needle:
http://farm4.static.flickr.com/3181/3095359535_4371f05d97_o.jpg

The latter can be found at McMaster-Carr (catalog page 3342) among other places. The needle is then gently inserted down into the compression fitting bore alongside the thermal sensor assembly:
http://farm4.static.flickr.com/3183/3095361375_6d657e3d52_o.jpg
until the top-most rubber grommet is felt. Then, the epoxy is injected from the syringe until it rises to the level of, or ever-so-slightly above, the top-most surface of the compression fitting – being sure that the lower-half of the TO-92 sensor case is wet by the epoxy in the process:
http://farm4.static.flickr.com/3079/3095361887_d0b83489b6_o.jpg

After the epoxy is allowed to set (approximately 3 hours), the two rubber grommets are carefully removed, using suitable tweezers, from the tube barb end of the compression fitting. Now the fitting is suitably supported so that its bore is again oriented vertically, but with the G1/4 threaded end (and TO-92 sensor case) down this time. Another batch of DP-270 epoxy is mixed up, another needle-equipped syringe loaded with it, and the balance of the compression fitting bore is completely backfilled to encapsulate the sensor wires leaving the fitting:
http://farm4.static.flickr.com/3135/3095362327_619d24bc7b_o.jpg

After this second application of epoxy is allowed to fully cure (approximately 48 hours), the completed sensor assembly:
http://farm4.static.flickr.com/3290/3095362707_a39118859a_o.jpg
is ready for sleeving, installation and service. Economy of scale dictates that several sensors be prepared at one time:
http://farm4.static.flickr.com/3091/3095363319_424ec3080c_o.jpg
so as to limit the number of epoxy batches that need to be made up and the number of syringes consumed.

I'm putting mine in Bitspower Rotary-Q fittings:
http://farm4.static.flickr.com/3015/3096205006_925f53ee22_o.jpg
and these will be used to monitor the inlet and outlet water temperatures at radiators in my rig:
http://farm4.static.flickr.com/3057/3096205950_03a15d6b40_o.jpg

http://farm4.static.flickr.com/3184/3095364273_6dd2724917_o.jpg

http://farm4.static.flickr.com/3195/3095365165_64b0dfb510_o.jpg


A final note on using the Dallas sensors:
There's a little known, but noteworthy, accuracy improvement technique for the DS18B20 temperature sensors that I want to point out to anyone interested in using these. The DS18B20's are factory calibrated to have accuracy better than or equal to ±0.5°C, which is sufficient for most of our watercooling applications. However, if you want to push performance another order of magnitude, there is a simple procedure that compensates for the offset and curvature of the device error characteristic. Details are in a white paper found at http://www.maxim-ic.com/appnotes.cfm/an_pk/208. Note that the technique presented in the paper can be applied to any IC temperature sensor having a bandgap-based thermal circuit (such as the DS18B20), but generally not other sensors based on different technology.

enjoy,
rubidium

[Eddie3dfx]

Thank you for posting this.. yes, I have about 7-8 of these suckers lying around, so these innovatek quarter inch would work wonders..
I do believe they are aluminum though.. :(

I'm thinking of using plugs and drilling them.

[bigslappy]

that's an Great Idea ! PIX plz ..

[rubidium]

Thank you for posting this.. yes, I have about 7-8 of these suckers lying around, so these innovatek quarter inch would work wonders..
I do believe they are aluminum though.. :(

I'm thinking of using plugs and drilling them.

Check to see that the fittings are not nickel-plated brass. A touch of a file will quickly expose the brass if they are. Anyway, although mine were brass, that isn't to say aluminum fittings won't work in the procedure, per se. It's just that you don't want aluminum in your cooling loop with other copper components (like water blocks and rads), due to the corrosive action you'll eventually get between two metals so widely spaced in the electromotive series. But you know this...

The surface preparation is key - spare no time or attentiveness.

rubidium

[rubidium]

that's an Great Idea ! PIX plz ..

Just hit the hyperlinks to the photos I stashed on Flickr.

[bigslappy]

Just hit the hyperlinks to the photos I stashed on Flickr.

i'm an idiot agin !

[RedRaider]

That is some amazing work...:slobber:

[skinnee]

The syringe method is pure genius, as my sensor assemblies are somewhat messy on the top side.

To house the sensors I used the Bitspower LED stop fittings and drilled through them with a 5mm bit. This gives just enough clearance for the sensor to get through. To seal the sensor end and keep it clean of epoxy, I used a piece of a latex glove and heatshrink to seal it up. Applied the epoxy, used coffee stir straws to get the epoxy all the way to the bottom and fully surrounding the sensor. Done.

Put the finished stop fittings with sensors in a Delrin T, and looped them up.

[NAekuh]

dude definitely smart move!

awesome design!

[bigslappy]

can any of ya do us up a D.I.Y. article for us mental midgets in Da crowd !

PLZ

;>}

[ballz0r]

good use for those alphacool compression fittings... ive got a few of those layin around

[KaptCrunch]

i like them plumbing washers for dougnuts

great post ribidum

[Baleful]

Wow, pure genious. I love it. Thank you for sharing.

[Eternalightwith]

would you make some available for sale?

[rubidium]

would you make some available for sale?

I appreciate your interest in what I've done and I'd like to accomodate you, but I'm not really set up to make selling these sensors seem practical from either of our perspectives. After ordering the parts in the small quantities I needed and having them shipped from several different sources, factoring in the cost of the expendables used for prep and assembly, and my time, I'd hate to estimate what I ended up "paying" for each sensor.

rubidium

[Eternalightwith]

Yeah, you're right. I can do it, just have so much other case modding to do. I asked for a sample of the sensor from digikey, but I think they bkew me off. Cause I wanted like 10 :P

I appreciate your interest in what I've done and I'd like to accomodate you, but I'm not really set up to make selling these sensors seem practical from either of our perspectives. After ordering the parts in the small quantities I needed and having them shipped from several different sources, factoring in the cost of the expendables used for prep and assembly, and my time, I'd hate to estimate what I ended up "paying" for each sensor.

rubidium

[mcoffey]

Yeah, you're right. I can do it, just have so much other case modding to do. I asked for a sample of the sensor from digikey, but I think they bkew me off. Cause I wanted like 10 :P

You can get the probes from CrystalFontz if you want. They sell from one to one hundred.

@ OP. Great job and very well done. I just used an inverted G-1/4 1/2" fitting for mine with 5 minute epoxy with a tapped bottom to let the epoxy set. But your method is much better. I never though of the O ring as a stopper. Very clever and I'll use that method on my next batch.

andyc

[rubidium]

Yeah, you're right. I can do it, just have so much other case modding to do. I asked for a sample of the sensor from digikey, but I think they bkew me off. Cause I wanted like 10 :P

Crystalfontz sells them (complete with leads and connectors) for $5 each for quantities of 4 or more - which is a pretty good deal. These are what I used. In comparison, I just checked the Digikey web site and they want $5.04 each for the bare sensor in the TO-92 case, unless you want ginormous quantities (like thousands). If you still want the bare sensor at a lower price, you can buy them direct from the manufacturer, Maxim (https://shop.maxim-ic.com/storefront/viewpriceavailable.do;jsessionid=c0a80b0130d8f4da6 7b19a054b7abf87f0997cbd8128.e38Kb34Ta34PbO0SbhmPbh qOci1ynknvrkLOlQzNp65In0, enter DS18B20 in the search window) at $2.57 each.

rubidium

[Langer]

I bit of a silly question... but what do you have on the other end of these temp probes? and what hardware are you using to monitor them temps? do you think there wold be any reason why these couldn't be used with an AquaComputer Aquaero setup?

[Eternalightwith]

2nd that. That's what I want to do. :D

I bit of a silly question... but what do you have on the other end of these temp probes? and what hardware are you using to monitor them temps? do you think there wold be any reason why these couldn't be used with an AquaComputer Aquaero setup?

[rubidium]

I bit of a silly question... but what do you have on the other end of these temp probes? and what hardware are you using to monitor them temps? do you think there wold be any reason why these couldn't be used with an AquaComputer Aquaero setup?

Not a silly question, but the abswer to your question is unfortunately no. The temperature probes that support the Aquaero and similar units are basically thermistors - devices whose electrical resistance is a function of temperature. These are two wire devices that, simplistically, can indicate temperature by applying a known, regulated, voltage across them and measuring the resulting current flow through them to imply the resistance and hence the temperature. (If you have an ohmmeter, you can watch the resistance change while you heat the probe between your fingers.) More realistically, the thermistor is connected into a so-called bridge circuit that includes three other precision fixed-value resistors in a 4-member circular loop. A regulated voltage is applied across one diagonal of the loop and the voltage across the other diagonal is passed to an operational amplifier, where it is multiplied by an appropriate calibration scale factor and added to an appropriate calibration offset value, and then on to an analog-to-digital converter for eventual readout. The probes are strictly analog devices, and once again you can identify them by noting that they have two wires. One issue with these simple analog probes is that, to be accurate, the wire lengths should remain fixed to retain the calibration. Significant lengthening or shortening of the leads changes the total resistance of the probe and thus violates the calibration. Also, the probes must remain electrically independent of each other if there are several of them. Each must have it's own bridge circuit and op amp.

The Dallas (Maxim) DS18B20 are much more sophisticated solid state devices that completely measure the temperature and provide digital readout, provide device status flags including device serial number for identification, and accept software commands all via a single wire digital data bus. Hence the reference to it being a "one-wire" device. In actuality, there are three wires emerging from the device: power, ground, and the data bus. Being digital devices, the length of the leads is essentially irrelevant - so long as there's not so much signal loss as to turn a binary "1" into a "0" in a bit stream (but it would take several hundred feet of very thin gauge wire for that to happen). Many of the Crystalfontz displays (I'm using the CFA635) have a suitable microcontroller that is programmed via firmware to command the Dallas sensors and read them out. Because each sensor has a unique 64-bit ID burned into it during production, the sensors can simply be connected in parallel. The microcontroller can command them to identify themselves, command them to make a temperature reading, and a host of other functions.

rubidium

[Snyxxx]

Wow, excellent response rubidium.:coolup:

I always wondered about the differences between the Dallas sensors and those like the Bitspower that simply screw into the Tee or reservoir.

[Eternalightwith]

I wonder if it's possible to make a converter where as the output of the digital device is converted to a corresponding analog signal (voltage) that can be read by analog controls ie. Aquaero, etc.

Not a silly question, but the abswer to your question is unfortunately no. The temperature probes that support the Aquaero and similar units are basically thermistors - devices whose electrical resistance is a function of temperature. These are two wire devices that, simplistically, can indicate temperature by applying a known, regulated, voltage across them and measuring the resulting current flow through them to imply the resistance and hence the temperature. (If you have an ohmmeter, you can watch the resistance change while you heat the probe between your fingers.) More realistically, the thermistor is connected into a so-called bridge circuit that includes three other precision fixed-value resistors in a 4-member circular loop. A regulated voltage is applied across one diagonal of the loop and the voltage across the other diagonal is passed to an operational amplifier, where it is multiplied by an appropriate calibration scale factor and added to an appropriate calibration offset value, and then on to an analog-to-digital converter for eventual readout. The probes are strictly analog devices, and once again you can identify them by noting that they have two wires. One issue with these simple analog probes is that, to be accurate, the wire lengths should remain fixed to retain the calibration. Significant lengthening or shortening of the leads changes the total resistance of the probe and thus violates the calibration. Also, the probes must remain electrically independent of each other if there are several of them. Each must have it's own bridge circuit and op amp.

The Dallas (Maxim) DS18B20 are much more sophisticated solid state devices that completely measure the temperature and provide digital readout, provide device status flags including device serial number for identification, and accept software commands all via a single wire digital data bus. Hence the reference to it being a "one-wire" device. In actuality, there are three wires emerging from the device: power, ground, and the data bus. Being digital devices, the length of the leads is essentially irrelevant - so long as there's not so much signal loss as to turn a binary "1" into a "0" in a bit stream (but it would take several hundred feet of very thin gauge wire for that to happen). Many of the Crystalfontz displays (I'm using the CFA635) have a suitable microcontroller that is programmed via firmware to command the Dallas sensors and read them out. Because each sensor has a unique 64-bit ID burned into it during production, the sensors can simply be connected in parallel. The microcontroller can command them to identify themselves, command them to make a temperature reading, and a host of other functions.

rubidium

[Langer]

Thank you very much for clarifying that in such immense detail mate.

[rubidium]

I wonder if it's possible to make a converter where as the output of the digital device is converted to a corresponding analog signal (voltage) that can be read by analog controls ie. Aquaero, etc.

Of course you can, but why would you want to? Essentially you would be sequentially making an A-to-D conversion (internal to the Dallas), followed by a D-to-A conversion in some custom circuit, followed by an A-to-D conversion within the Aquaero. In addition to the expense of that custom circuit, you will experience the loss of accuracy inherent in three rather than one conversion steps (e.g. due to quantization, non-linearity, and other sources of error). For $8 why not just consider the Bitspower analog sensors for the Aquaero? As far as I know, they should be compatible, and from what I've seen, they look pretty sweet.

BTW, I am doing fan control based on water temperature measurements with the CFA635 and Dallas sensors, in case I didn't make that clear earlier. In my Project Erinyes build, I have 3 cooling loops. Two of these loops each cool a pair of processors in a 4-socket system, while the third cools a PSU. The Dallas sensors that I assembled as described in this thread are placed in the inlet and outlet stream of each of the 3 PA120.3 radiators in the rig. An example of the temperature readouts for the two CPU cooling loops, via the CFA635 can be seen here:

http://farm4.static.flickr.com/3364/3176947902_9ef99a44b0_o.jpg

The fans on those respective radiators are also being controlled by the CFA635, with fan power being a function (that I specify) of the measured temperatures:

http://farm4.static.flickr.com/3367/3176947918_5d5719e44a_o.jpg

I spent all of 10 minutes crudely setting up this readout and control on the CFA635. Once I get the rig fully up and running, I can spend more time and do fancier things.

rubidium

[rubidium]

Thank you very much for clarifying that in such immense detail mate.

And that's one hot job you're doing on your rig, mate.:coolup: I am lurking ...

rubidium

[link1896]

for people not wanting to use a Crystalfontz or other type of off the shelf i/o box, be it for cost, asthetics, requirements, whatever, the dallas one wire temp sensor can be used with the picaxe family of microprocessors. For example, the 08M picaxe is about 4 bucks, and can communicate with a pc via RS232.


See Peter H Anderson's Temperature Measurement using the Dallas DS18B20 - PICAXE-08M page for more details (http://www.phanderson.com/picaxe/ds18b20_08m.html)






.

[rubidium]

for people not wanting to use a Crystalfontz or other type of off the shelf i/o box, be it for cost, asthetics, requirements, whatever, the dallas one wire temp sensor can be used with the picaxe family of microprocessors. For example, the 08M picaxe is about 4 bucks, and can communicate with a pc via RS232.


See Peter H Anderson's Temperature Measurement using the Dallas DS18B20 - PICAXE-08M page for more details (http://www.phanderson.com/picaxe/ds18b20_08m.html)

.


Indeed. Thanks for bringing this up. Picaxe is cool. I don't have it, but I'd like to get one to play with.

rubidium

[Eternalightwith]

Wouldn't you have to write your own software for it though?

[rubidium]

Wouldn't you have to write your own software for it though?

Sure, but there are some code examples if you scroll down the page provided by link1896. It's certainly not an "out of the box" solution like Crystalfontz or others offer, but the picaxe kits look like they can be fun if you're willing to invest some time.

rubidium

[Nickel020]

Great guide, lots of useful information in this thread!

I have a question regarding the accuracy of the probe: If it is off by say the maximum of 0.5C, will it always be off that same amount or can it fluctuate by +/-0,5C?

I got me a thermometer for 50€ that's guaranteed to be accurate to 0.1C, and when I tested normal temperature sensors on my Aquaero, they weren't always off by the same amount, so I'm looking for something good and accurate enough for testing.

The Aquaero is great for controlling fans but I guess it was never designed to have the accuracy need for proper testing.

Also, using a file should be no problem when removing the nickel from the fitting I guess, right? I wouldn't know where to get something to clean a gun here, we don't have all that many gun shops here in Germany ;)

[mcoffey]

Great guide, lots of useful information in this thread!

I have a question regarding the accuracy of the probe: If it is off by say the maximum of 0.5C, will it always be off that same amount or can it fluctuate by +/-0,5C?

I got me a thermometer for 50€ that's guaranteed to be accurate to 0.1C, and when I tested normal temperature sensors on my Aquaero, they weren't always off by the same amount, so I'm looking for something good and accurate enough for testing.

The Aquaero is great for controlling fans but I guess it was never designed to have the accuracy need for proper testing.

Also, using a file should be no problem when removing the nickel from the fitting I guess, right? I wouldn't know where to get something to clean a gun here, we don't have all that many gun shops here in Germany ;)

Within the range we test the accuracy is more like .2c, and that's realitive to the probe itself. If it's off by .2c, then it will always be. In other words, it would read 20.0c as 20.2c, and 25c as 25.2c. There's not enough varience to worry about.

These are the best rigs commonly availible for testing, period. Unless you want to drop 5000-6000 for a lab grade setup that is. But there's a lot more to good testing than the probes you use.:D

This is the best guide and the most pro way to do it. But you can also just take a fitting as is, slip the probe in, tape the underside, fill with 5 minute epoxy and your done. The main thing is, just make sure the epoxy covers where the metal contacts the plastic housing on the probe.

andyc

[Snyxxx]

rubidium,

One of the hassles I always have is to apply a little bit of offset in software to get each of the thermistors (2-wire BP type) calibrated with each other. I usually use a bigNG digital sensor as the "base" and calibarte the others to it.

Are these Dallas sensors any different in this regard. If you were to stick say three of them in a glass of ice water, would they all read the same?

[rubidium]

rubidium,

One of the hassles I always have is to apply a little bit of offset in software to get each of the thermistors (2-wire BP type) calibrated with each other. I usually use a bigNG digital sensor as the "base" and calibarte the others to it.

Are these Dallas sensors any different in this regard. If you were to stick say three of them in a glass of ice water, would they all read the same?

Out of the box, the Dallas sensors have +/-0.5 degree C accuracy, which implies that any two of them chosen at random can give readings that possibly differ by as much as 1 degree C as a worst case. However, being bandgap-based sensors, they have a very repeatable second-order error characteristic, so that it is possible to calibrate them to have +/-0.05 degree C accuracy - a factor of 10 better - and have them stay that way thereafter. The procedure is outlined here (http://pdfserv.maxim-ic.com/en/an/app208.pdf). Although the procedure discussed in the paper uses a NIST-tracable reference thermometer, I believe that with a little work one can get a bunch of Dallas sensors and cal them as a lot, so long as lots of measurements are made across a wide temperature range, and at least one of those measurements is made at an absolute reference temperature - such as the boiling point of distilled water at a known ambient atmospheric pressure.

[CrazyJoe]

Out of the box, the Dallas sensors have +/-0.5 degree C accuracy, which implies that any two of them chosen at random can give readings that possibly differ by as much as 1 degree C as a worst case. However, being bandgap-based sensors, they have a very repeatable second-order error characteristic, so that it is possible to calibrate them to have +/-0.05 degree C accuracy - a factor of 10 better - and have them stay that way thereafter. The procedure is outlined here (http://pdfserv.maxim-ic.com/en/an/app208.pdf). Although the procedure discussed in the paper uses a NIST-tracable reference thermometer, I believe that with a little work one can get a bunch of Dallas sensors and cal them as a lot, so long as lots of measurements are made across a wide temperature range, and at least one of those measurements is made at an absolute reference temperature - such as the boiling point of distilled water at a known ambient atmospheric pressure.

AHA ! ... that's exactly what I was looking for Rubidium ! Thanks so much for all the insight ... very informative ! :up:

[rubidium]

Actually, to correct myself slightly, if one were to calibrate the DS18B20 probes according to the procedure outlined by Maxim/Dallas, one could hope to achieve +/-0.0625°C accuracy (and not 0.05°C). This corresponds to the least significant bit of the on-board 12-bit A-to-D converter within the device. Still, that's pretty damn good.