DS18B20's or K-Type thermocouples

Homebrew Talk - Beer, Wine, Mead, & Cider Brewing Discussion Forum

Help Support Homebrew Talk - Beer, Wine, Mead, & Cider Brewing Discussion Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Rbeckett

Well-Known Member
Joined
Dec 23, 2011
Messages
993
Reaction score
72
Location
Bronson
Which temp sensor is going to give the best performance. The DS's have a much narrower range and output directly in centrigrade. simple + and - 5 VDC and a digital out. The K-Type thermocouples are are 2 wire through the sensor wiring set up with a much larger range. Is the large range going to make reporting accuracy at the lower end of mid level mushy? Or maybe a combination of the two. K-type in the tuns and DS's in or on valves and fittings. That would add a lot of data colection points for way cheap. I got 10 DS's for 9 dollars and 6 K-Type for about 35 dollars. any ideas on placing them in the most effective pattern? Or better yet any place to avoid putting them to prevent damage or leakage. Or even a different temp sensor that works better all together? Comments or opinions?
Thanks
WCB
 
I use the DS18B20s (as does the Brewtroller) and I guess I haven't come across a reason to really look at anything else. I have mine in the stainless probe ends from "Brewers Hardware" along with the compression fittings and connected via some M12 cables.

I think kladue has had some concerns regarding the conversion time with this sensor so hopefully he'll chime in.

In regards to conversion time as it relates to resolution, see Table 1 here: http://www.maxim-ic.com/app-notes/index.mvp/id/4377
 
I have used 100 ohm RTD's and type J thermocouples with the opto22 hardware, but these take additional interface circuits to render useful outputs for anything else. There is a voltage output sensor with 0-5Vdc output that looks interesting http://www.analog.com/en/other-products/militaryaerospace/ad22100/products/product.html. This would play well with a 5V reference analog input board, and 0-5V output pressure sensors for level measurement.
 
digital IC temp sensors and thermocouples are very different, and picking one over the other really depends on what hardware you are reading it with or what you are doing with the data. each of them have their accuracy, range and response time in the spec sheets. its up to you to pick whats best for your application.

if i were using any type of microcontroller or computer (not for just a digital panel meter), i would go with a DS18B20. a digital signal is much less prone to interference from near by electrical loads and variations (heaters, pumps, sensor wiring, legnth of wires). with a resistance or voltage (analog) measurement of a thermocouple, you might have some interference that skews the readings only slightly, or non-linearly, which can be hard to spot or correct. with a digital signal, you either get the correct sequence of 1s and 0s, or you get a corrupted string which immediately tells you something is wrong.
 
The DS sensors are sooo slooow in reading values out that more than a couple and a fast changing process PID loop is in trouble. With the I2C analog system I can read and decode about 16 - 12 bit points by the time one DS responds to a command and returns a 12 bit response.
 
The DS sensors are sooo slooow in reading values out that more than a couple and a fast changing process PID loop is in trouble. With the I2C analog system I can read and decode about 16 - 12 bit points by the time one DS responds to a command and returns a 12 bit response.

In reading some of the Brewtroller code and documentation, it seems like the 750 ms for the DS18B20 is only in parasitic mode and the time is more like 10 ms when powered; they're also only reading the first two bytes of data to speed up the read.

The newer Brewtrollers are using the DS2482; not sure if this helps the conversion time or is just a bridge to I2C.

Any idea what the read time is on the sensors you're using? I'm kind of curious to check these out and see if it makes sense to switch.
 
The AD22100 sensors are voltage output devices, they are connected to a 8 input 12 bit I2C analog chip. the measure and read commands take 25 Ms each to scan an input channel, 400 ms for all 8. The analog chip is using an external 5V reference so it can handle 0-5v output pressure, flow, temperature devices with industry standard outputs. The hex values returned on the read response are passed to an array for use in another loop for conversion which is running with about 5 Ms intervals for each conversion and transfer to the output array.
 
IMHO, 12 bit temperature resolution is silly for what we're doing. If the sensor is returning results in Celsius, that's 0.06 of a degree.
At least for my system, 10 bit (0.25 of a degree), or 11 bit (0.125 of a degree) are more than accurate enough, and when using DS18B20 in powered mode provide max read times of 180 and 375 ms, respectively.
These numbers are not cumulative unless you're polling each sensor individually, which is foolish. Poll the entire I2C collection, get all the results (at close to the 180 and 375 numbers plus 10 ms or so for transmit time), and apply your logic to the array.
 
The DS sensors are sooo slooow in reading values out

what is your definition of "slow"... because a few hundred miliseconds isnt slow to me. a thermocouple in a thermowell is going to take entire seconds to fully respond to changes in temperature (not just begin to register the change). either way- i dont think an extra second here or there makes any difference.
 
If this was just a single temperature point, the DS would be adequate, but when trying to measure flows, levels, temperatures, and feed the information to multiple PID loops, time becomes important. The constraint of the analog input chip is that it requires individual channel polling, group data requests are not possible, upside is you can poll the channels in any order you want.
Having built and operated a 2 vessel brewing system with a flash boiler, precision control and fast response was required to make it operate. The current project is working toward a lower hardware cost control system with greater accuracy and flexibility than currently available.
 
I am using the DS18B20's throughout, made some tri-clamp thermowells for easy removal/cleaning:

TC_temp_probes.jpg


The cost of monitoring additional temperatures throughout the brewing room becomes very low when using these digital sensors. Even at 750ms polling for the highest resolution you can use basic filtering to provide more timely inputs to control algorithms if needed.
 
Nice looking work there,
for your application the DS sensors would make sense as the process temperatures are not changing very fast with electric heating and the volumes involved. With the 10 bit resolution of the analog inputs on the Arduino platform the DS sensors make a lot of sense with the wide temperature range needed and lower resolution analog inputs available. It is when you get into higher heat inputs to small volumes that speed of response is needed, and increased resolution makes loop tuning easier with smaller steps in the input providing finer control of output.
 
It is when you get into higher heat inputs to small volumes that speed of response is needed, and increased resolution makes loop tuning easier with smaller steps in the input providing finer control of output.

Couldn't agree with you more. I would think that the thermowell or mechanical sensor interface is critical in high frequency measurements as any additional thermal mass acts as a low pass filter/delay.

As always, that is a lot of pretty equipment you are putting together. Seems as if you have enough extruded aluminum to build a roller coaster.
 
The automated system is 4 years old now, and was just a make work R&D exercise to work out the design for an automated gas burning indoor brewing system. Software applications were another make work exercise that combined recipe with process control features to see what could be included and made to work. First application was built in Java, which sucks at native memory management, second was a conversion to .Net which eliminated the problems when displaying process graphics that Java has.
Now the next make work exercise is to come up with a comparable touch screen modular control system with analog outputs for proportional control, and a flexible application with run time setup that eliminates the need for hard code changes.
 
When considering a thermocouple's accuracy keep in mind that you also need to take into consideration the cold junction compensation temp sensors accuracy. TC's work by creating a voltage across a junction between 2 dissimilar metals. When the 2 thermocouple wires reach the electronics you get a second junction between the wires and the connector going into the PCB. To compensate for this second (cold) junction you need another temp sensor (thermistor, RTD, thermal diode, etc) located close to the second junction to get the temp out at the end of the probe. In practice this works by reading the CJC sensor and converting that into volts based off of a look up table. The you subtract that voltage off of the voltage measured by the TC. Finally you use another look up table to convert from volts to temperature. Also thermocouples produce an extremely small voltage that needs to be amplified before it can be read by any ADC. Any inaccuracies in you amplifier circuit (op-amp and resistor tolerances) will affect your overall accuracy. IMHO TC's are not nearly as accurate as RTD's or thermistors and are a lot more difficult to implement properly.
 
Given that the PIC and Atmel chips only offer 10 bit analog inputs, direct connected RTD's or thermocouples are out of the question. With a suitable IC built for thermocouple or RTD use, they can render reasonably accurate readings via a voltage or digital output, and boards are available for this purpose. Thermisters are not a linear device and require formulas to correct the response across a typical brew system temperature range, but with appropriate resistance selection will yield an output suitable for the low resolution analog inputs on the popular microprocessors.
 
Given that the PIC and Atmel chips only offer 10 bit analog inputs, direct connected RTD's or thermocouples are out of the question. With a suitable IC built for thermocouple or RTD use, they can render reasonably accurate readings via a voltage or digital output, and boards are available for this purpose. Thermisters are not a linear device and require formulas to correct the response across a typical brew system temperature range, but with appropriate resistance selection will yield an output suitable for the low resolution analog inputs on the popular microprocessors.

Hi

Thermistors are also quite cheap and they are easy to cable (unlike thermocouples). Getting a bunch of cheap thermistors to directly interchange - not so easy. If you can calibrate them individually, then that's not an issue.

Bob
 
Thermisters are sensitive to source voltage fluctuations, one of the reasons they see limited limited use, and are found where accuracy and stability are not required.
 
When I use thermistors I measure the voltage across the thermistor and across the ref resistor to take voltage supply fluctuations into account. I have not found a need to individually calibrate each thermistor when I sample this way.
 
Thermisters are sensitive to source voltage fluctuations, one of the reasons they see limited limited use, and are found where accuracy and stability are not required.

Hi

As long as you do a ratiometric reading (same reference to the thermistor bridge as to the A/D) the voltage fluctuations drop out first order. Since the thermistor has a very high rate of change in resistance versus temperature, it's potentially a very high resolution sensor. They are indeed used in high accuracy applications when resolution is the key parameter.

One example is here:

http://us.flukecal.com/products/tem...nsors/5640-series-thermistor-standards-probes


Bob
 
Thermistors are not seen in much beyond the cheapo commercial controllers, the temperature range limitation and non linearity work against their wide spread use.
A better choice is a 0-5V output amplified temperature sensor that is precalibrated and delivers linear response to temperature change like the AD22100K from Linear, no curve correction formulas are needed, and no calibration when switching sensors.
With the low cost single chip solutions out there now for thermocouples, it would seem to be better to devote time and effort in that direction as thermocouples do not require external drive power, are readily available, and usually do not additional calibration when switching sensors.
 
Hi

Well, for some applications (like high precision single set point stuff) thermistors are very much the sensor of choice and have been for at least 50 years. For wide range work, indeed their rapid rate of change works against them.

Bob
 
Thermistors are much more accurate than thermocouples. Keep in mind that thermocouples need a second temp sensor for the cold junction reference. Therefore the thermocouple is no more accurate than its cold junction reference. I have seen thermistors used for cold junction compensation on almost all of the thermocouple circuits I have encountered. National instruments uses them on their $500 plus precision data acquisition instruments.

The non linearity comment is bogus. You can easily compensate for this by using a look up table.

Also thermistors are extremely easy to implement. All you need is a single resistor and a voltage source to bias them.

Thermocouples produce extremely small voltages that need to be amplified with a gain of at least 30 to be able to read them with an ADC. Any inaccuracies in your amplifier circuit will be multiplied by your gain value. You will have to use resistors with very high tolerances. In the end you will still have to calibrate your amplifier circuit to account for these errors.
 
Hi

With a thermistor both the nominal resistance and the beta (rate of change) vary from part to part. For enough money you get pretty tight tolerances on both numbers. You can also calibrate each one and eliminate the issue.

The gotcha comes when you want to do wide range (say -55 to +125C) measuremnet. Thermistors simply change to much over a range like that for a simple circuit to work well. Thermocouples or RTD's are a better choice there. Thermocouples are the bang for the buck kings when it comes to wide range measurement.

Bob
 
You weld these yourself or did you buy these somewhere?
Thanks,
-Chris

I am using the DS18B20's throughout, made some tri-clamp thermowells for easy removal/cleaning:

TC_temp_probes.jpg


The cost of monitoring additional temperatures throughout the brewing room becomes very low when using these digital sensors. Even at 750ms polling for the highest resolution you can use basic filtering to provide more timely inputs to control algorithms if needed.
 
Back
Top