PT100 and Thermocouple choice

[MicroMaster]

I need to do a temperature control. resolution needed is 0.5 Degree Celsius. Accuracy is not that much great critical.

The operating temperature range is normally only 100 to 200 Degree. Normally need to kept somewhere near 150 Degree Celsius.
The client proposed a thermocouple, but considering the complications of compensation and linearizing issues associated with thermocouples, I suggested for PT100. I am using this sensors for the first time.

Similar devices in market seems to be using both these sensors but most of them are on thermocouples.

May I Know the pros and cons of making the selection between these two? I have searched and read the difference and got a general idea, but looking for input from practical experience.

Thank you

Roy

[leptro]

hi,
i have used a thermocouple with the special circuit AD595, you have nothing to do, all compensation and hard work ar done by the circuit itself. You will have an analog value at the output, just connect it to your controller adc.

have also a look at the MAX6675, it has a digital output (spi) , an do the same job.

Regards.

[MicroMaster]

Thanks for the info.
AD595 is locally available but costly (550/ Indian Rupees).
But at the same time thermocouple cheaper than PT100 here.
Thermocouple 150/INR and PT100 is 250/INR.
But PT100 needs constant current and some conditioning circuits, I think. (Right?)
Thermocouple  + AD595 can go to ADC directly and then process as I did for LM35 output!

Now I have to decide which way to go!

Thanks
Roy

[MicroMaster]

The final cost is important, but not more than the professionalism and quality of the product. I am planning to implement a PID based temperature controller in this system.
To be frank, the idea to use a very costly chip for that Thermocouple conditioning seems a inability on my part. (May be a bad idea to think that way!)

If the final cost is important and you have a micro controller solution with some spare FLASH and running time, you can supply the PT100 by a simple resistance and linearize the data with the micro.
Regards

You are suggesting that even a constant current source is not needed. Can you give some more details?
This is how I understood the PT100 usage so far..
PT100 gives a change of 0.385 Ohms/Degree Celsius.
So to be handled by a 10 bit ADC channel of PIC to give 1 Degree resolution, we must get 4.9mVolts of change. So a constant current around 13mA can effect that volatge change in PT100.for that I can use a regulator in constant current mode. But if I want .5 Degree resolution I may have  to double the current. But that increase may not be OK because of other factors.
So is my understanding of the situation, workable?
How this 'linearize' ? You mean any look up table or calculation?

This can be avoided (by a bridge configuration) but complexity and cost increase.
Regards

Some more info please.

[ktek]

Simple,
the   noise always present on the pt100 comes in aid ,
in your case, in order to increase the resolution of the pic of 1 bit ,
you make 32 averages reads and you divide for 16
and therefore you obtain the resolution of  .5 Degree
With this system you can increase the resolution of 1 or two digit, beyond it becomes inaccurate
the conversion slows down but,  however temperature has a slow course

 

[engamor]

Hello! This is a subject on which I can say I have a good experience. In the 100 to 200 deg. Centigrade range a PT100 RTD is the most professional choice. BUT you shall consider the length of wire to connect it in circuit. Professional PT100 sensors are 3 wire or 4 wire (best) the extra wires will allow compensation.
You may want to consider using a PT1000 , that is a less accurate sensor, but usually does not need the 3 or 4 wire complication, and it is anyhow much better than a thermocuple.
Definitively you cannot drive a PT100 ( or PT1000) with 13 mA! Forget it! That will increase the temperature of the sensor by a BIG amount. Maximum drive is normally 1mA.
Yes, with PT100/PT1000 you do have to offset your reading and make some calculations (linearization etc.) That is true for the thermocuple as well.
The Thermocouple is a low cost sensor choice but you will need a secondary temperature sensor for ambient temp. correction, quite often a thermistor, and some compensation scheme in digital or analog form as well. You can find a good deal of design projects to implement both solutions. To get a good idea have  a look to the following application note from TI :slaa216 (search for it in www.ti.com), it is complete with source files. 
Finally: if your interest is mainly in the controller part of the design, why not buy a complete 4-20 mA temperature transmitter? That is not so much costly and you will get a really professional solution complete with fixation hardware and calibration. Start from there and add your controller ...
If you only need a couple of instruments this is the best way to go. If you are planning in the 100s then ok, I would work on the PT100/1000. 

[DTiziano]

Hi MicroMaster

The basic circuit is just a resistance in series to PT100.
To increase the ADC resolution multiple samples can help but a better solution is to change the
ADC reference (Vref- and Vref+). Take care not violate the micro specification for this function.
Attached a simple excel simulation of this ideas.
If the power dissipated by the PT100 is hight you can consider an ON/OFF solution.
While ago I made i similar system, the power supply comes from a LM317, by switching to ground the ADJ pin the PT100 current goes very low, so the power. In my application the PT100 was very tiny and under vacuum, so even o low power generate error.
For you application you have can use the micro power supply equal to the Vref+.
Some verification also have to be made how to calibrate the unit and the error from component tolerances and drift.
Of course the conversion from voltage to temperature may require some linearization, but this should not be difficult to do with the micro. I think that a linear interpolation on a two or three segment should be enough.

The line connection also is a source of error that must be evaluated.
If not under control and it is too high the suggested solution solution may not be good.


Regards

P.S.
The back ground colored cells are the ones to play with, the others are calculated.

[MicroMaster]

Thank you all. Now I got a substantial stock of information to start with plus documents. As for now my feel is to go for PT100 sensor.
The little confusion remains is about the normal current used for driving PT100. Is it around 1mA as mentioned above? If so my resolution will be greatly reduced.
Constant current or not, what is the optimum current for PT100? (The spreadsheet default value is 20mA)
Also if we need to drive in more currents, we may energize PT100 only when about to read.
Correct me if I am wrong.

Thanks a lot

In the spreadsheet as we change some values the cell back color changes,- some sort of a indication! How is it to be interpreted ?

[engamor]

May  I insist that:
1) You cannot drive a PT100 at 10mA, not even for a short time, if you want to discriminate 0.5 deg.centigrade -
2) You cannot have even a short run of cable when using a PT100 as a sensor , 1 meter is too much already ! You must have a 3 wire or 4 wire arrangement.
3) A 10 bit ADC cannot do the job in my opinion.
4) Some analog conditioning circuit is REQUIRED to solve the above problems.
OR you can use a microprocessor with a better resolution ADC and that includes a PGA. The 4wire correction can be done in firmware. The simplest ones that I use succesfully are  the MSP430F4270 or MSP430F479 with 16 bit sigma-delta converter. I can guarantee that no less than that is necessary.

[Old_but_Alive]

at the risk of being flamed for "advertising"

have a look at

http://www.parallax.com/Store/Sensors/TemperatureHumidity/tabid/174/CategoryID/49/List/0/SortField/0/Level/a/ProductID/96/Default.aspx

it uses a thermocouple, has a Dallas 1wire output, so is easy to use with a micro.

there are loads of good info, and code to help you.

p.s. I dont work for parallax

Mike

[engamor]

Sorry but  Isuspect that the parallax design is not good for the application. First of all it seems to work only up to 127 centigrades, second ... have a look at how they suggest to "build" and connect the thermocouple! Speaking of professional ...

[Old_but_Alive]

the +127 is just the cold compensation temperature max, not the thermocouple sensor.

as for the thermocouple connection, i dont disagree, it should use proper thermocouple plug connections.

I was just giving the guy a good starting point

[engamor]

May  insist, it appears that the 127 limit is for the whole measurement ... and the problem on the thermocouple "build" is making the contact of the two wires on the sensing termination ... by "twisting" the thermocouple wire, this is a bad example... of course the info is interesting anyhow. What I am warning against is a certain "amateur style" of the company.

[Old_but_Alive]

from the link I gave :-

    * 1-Wire interface allows multiple devices with just one BASIC Stamp I/O pin
    * Cold Junction measurement: 0 to +127 °C (0.125 °C resolution)
    * Low power consumption: Active current = 90 µA max, Sleep current: 2 µA max

Key Specifications:

    * Power Requirements: 2.5 V to 5.5 V @ 90 µA (max)
    * Communication: Dallas 1-Wire® Interface with unique 64-bit device address
    * Dimensions: 1.08 x 0.43 x 0.3 in (27.5 x 11 x 7 mm)
    * Operating Temperature: -4 to +158 °F (-20°C to +70°C)

Kit Contents:

    * Thermocouple Module
    * K-Type (Chromel/Alumel) Element Temperature Range: +32 to +1873 °F (0 to +1023 °C)
    * J-Type (Iron/Constantan) Element Temperature Range: +32 to +1873 °F (0 to +1023 °C)
    * T-Type (Copper/Constantan) Element Temperature Range: +32 to +752 °F (0 to +400 °C)


cant you read properly ?, the cold compensation temperature max is +127C

I agree their method of thermocouple mechanicals is bad.

Its a kit for amateurs, but will work well providing a pcb is designed with proper connectors, and proper thermocouples from say omega

Its a good starting point for a solution though.

[MicroMaster]

You must have a 3 wire or 4 wire arrangement.
Some analog conditioning circuit is REQUIRED to solve the above problems.

Thanks friends for the input.
If I have to go for higher bit ADC I need to use MCP3304 or something. Needs external reference etc. I have to try weather it can be accommodated with PIC ADC.

As you suggested I have a three wire PT100.  white and 1 blue.
I can use some analog conditioning circuit also.
Can you provide some suggestions in this context?

Thanks

[engamor]

The first example circuit at page 136 (pdf page= 74) of the book referenced above is clear enough on the priciple of measurement of a PT100, but does not provide a 3wire compensation. I can search for something in 3 wire but that will take time. It's a long time I left such "analog" approaches and use a 16(or more) bit ADC. In my opinion that is the best thing to do anyhow.

[DTiziano]

In the spreadsheet as we change some values the cell back color changes,- some sort of a indication! How is it to be interpreted ?

The Vref- and Vref+ must be lower and greater respectively to the min - max input voltage.
The back ground color indicate that this condition is not meet.

For the PT100 driving current problem, I do not agree, at 20 mA and 200 °C you have 70 mW power dissipated by the sensor.
Yes it can be self heated, but that depend also on the final sensor mass (sensor holding if exist).
This self heat can be reduced by supplying the PT100 just for the few ms required by the reading.
Low current can be used of course, at low current the self heater is lower, but the noise contribution increase and a better ADC is required.
Here as  everywhere the project is a balance of performances , complexity and costs, and only the designer may know which solution is better  tailored.
We can give suggestions and ideas to evaluate.

If the PT100 supply line can be keep under control or is included in the calibration for example, than also the two wire solution can work.
Take care also that resolution and precision are two different thinks.
Yes typically hight resolution is married with high precision, but again there are applications were the resolution can be a little better than precision just because the final result is trimmed (no hight precision required), but good resolution help to keep it on track.

Regards

PS
For three wire sensor see microchip AN682 pag.7 for example.

[engamor]

Oh la! I found a nice circuit and explanation in the following : http://ww1.microchip.com/downloads/en/AppNotes/00687b.pdf
In it you get a costant current source at 1mA AND a 3 wire compensation circuit. The ADC is 12 bit anyhow, you may still want to make your own considerations and try to use a 10 bit, but in my opinion this will not give any good. Believe me, I do know the matter. To day you spend less money and get better results using a 24bit sigma-delta and make the differential job in firmware.

[engamor]

Nice example! So, now, you can understand the reason you cannot really do the job with a 10 bit ADC!  And, by the way, there certainly are thermistors that can be used in the range from 100 to 200 celsius, but they are neither cheap nor accurate. A Pt100 or PT1000 << IS >> the choice in that range.

[oldvan]

I need to do a temperature control. resolution needed is 0.5 Degree Celsius.
Accuracy is not that much great critical.

The operating temperature range is normally only 100 to 200 Degree. ...

Nice example! So, now, you can understand the reason you cannot really do the job with a 10 bit ADC!

How would one be so silly as to conclude that?

The problem is defined to cover a 100 degree range in 0.5 degree increments, requiring 200 data points.

A 10-bit ADC offers 1024 data points, over 5 times as many as minimally necessary.

Scale and offset sensor output voltage with op-amps such that 100 degrees gives 0 volts, 200 degrees gives
5 Volts, feed this voltage to a 10-bit ADC and get on with it.  No need to make a mountain out of a mole hill.

[MicroMaster]

Thanks a lot, friends.
I am going to use PT100, three wire configuration, which is available.
For the processing I think I narrowed it down to two options.
One is to scale and offset the PT100 output to accommodate in the PIC10 bit ADC- less work
Next is to use that Bonnie Baker Microchip App note circuit with 12 bit ADC- more adaptable for future needs.
Thanks again

[engamor]

How would one be so silly as to conclude that?

The problem is defined to cover a 100 degree range in 0.5 degree increments, requiring 200 data points.

A 10-bit ADC offers 1024 data points, over 5 times as many as minimally necessary.

Scale and offset sensor output voltage with op-amps such that 100 degrees gives 0 volts, 200 degrees gives
5 Volts, feed this voltage to a 10-bit ADC and get on with it.  No need to make a mountain out of a mole hill.

The problem is that every dummy can do the calculus above, BUT the real world is not as you think. I said I do this job all the time. You seem to search and try to insult other people, but maybe can't really know the matter. The 0.5 °C is approx. 0.2 Ohm difference ,over approx. 150 Ohm base resistance. Now the points to resolve are 750 at bare minimum, which doesn't work anyhow with a 10 bit ADC in the real world.
Secondly it is NOT that easy to scale and offset the PT100 plus compensate the 3wire . I can only repeat that the real way to go is use a high resolution converter. But at this point, as most other people DOES KNOW EVERYTHING I will only describe the real circuit to the interested guy, so that he will really gain his job. Please write me in private.

[Ichan]

The problem is that every dummy can do the calculus above

I use such a dumb calculus on my 20 - 50 degree range 0.1 degree resolution (300 points) thermometer with 10 Bit ATMega internal ADC. Sensor is 10K thermistor with lookup table linearization feed to the ADC through an analog front end which do what Oldvan said about scaling and offsetting plus lowpass filtering.

I can get a stable, consistent, and responsive reading with 0.1 degree resolution over the full range of 20 - 50 degree. Analog front end, noise handling, and stable References (both for ADC and sensor) are the keys.

-ichan

[bbarney]

Please write me in private.

 
First off we don't condone private share's like this as it doesn't benefit the community as a whole
So far this thread has been quite informative and we want to keep it open for the everyone but if it is going to turn into a flame war over who's right or wrong than it will be closed and so will a few accounts
from what I've read here everything has been covered to use this right from a Hobby level to professional level so everyones input has been of value.
Only the Op knows what degree of accuracy he really needs and is capable of getting away with, even though this may not be the best way in your opion it will work .
do not give up expressing your views because you don't agree with poster "A" or "B"  as there are people who still need information regarding more professional applications and would be interested to see the circuit and at the same time we have members who only want to experiment with these at a hobby level and 24 bit adc would be overkill for them not to mention some of the terms and math involved
the bottom line is we have members ranging from  amature to professional and the professional people tend to get ticked off when posting thinking we are all professional's here when in fact were not so try and remember that

[zuisti]

In the early 80's years I worked a lot  with this issue.
I was looking for a long time, this drawing is one of my old things.

It is a simple analog PT100 linearizer circuit, using only one OPAMP.
I prepared the circuit in Proteus (attached), it's working as well.
The output is 10 mV/Celsius, working in the range -100 ...+400 degree
Its linearity error is max. +- 0.1 degree (+- 1 mV).
Please be patient, the spice simulation is too slow ...
  
Unfortunately, I still cannot find my corresponding mathematical derivations,
but the values of the items were on the drawing.
 
Well, I know, these extreme values, but is required for the correct operation.
I remember glory days of the resistance wire was prepared by supplementing the necessary
values. It needed a high-precision resistance measurement too :-)

I prepared also a variation of this circuit to compensate the resistances of the measuring wires,
but I have not found yet...

However, at my humble opinion, this doesn't necessary in your case, MicroMaster.
Please have a look at this simple calculation:

You wrote:
"resolution needed is 0.5 Degree Celsius. Accuracy is not that much great critical"

Well, if measuring cable's lenght 5 m and cross-section 0.5 mm2 is assumed:

- the sum resistance of this wires is appr. 0.36 Ohm, corresponds to 1 Celsius in case
     of a PT100.
- the temperature coefficient of resistance of copper is
     alpha = 3.92 ... (4.041) ... 4.3 x 10-3/kelvin  
- to cause an error 1 celsius the delta T (change of temperature of the wires) must be  
     min. 232 Celsius (worst case), this is not realistic I think : -)

Look at this link for example:
http://www.allaboutcircuits.com/vol_1/chpt_12/6.html.

I hope I could help.
Welcome
zuisti

Edited:
The link for the circuit:
http://hotfile.com/dl/80867570/53fd6e2/circ.rar.html
pw:zuisti