A thermometer.

Referring to answer given by Dr Andy mere thermometer (say glass thermometer) can not be used with conductivity monitoring instrument for temperature compensation. Only one can read temp of solution from thermometer and provide correction factor for manual correction of conductivity. But this is not asked by me in my original question.

"Hence how one can blindly believe those temp sensors?".  Short answer: one cannot.

I would hesitate to "blindly believe" any temperature sensor, and regular calibration is strongly advised if operation of your instrument depends on accurate temperature measurement. 

What kind of temperature sensor are you using, and how often is it calibrated?  What accuracy do you need?  What does the instrument manufacturer recommend?

4-terminal platinum resistance thermometers are very stable, and can be accurate to 0.001 K.  Is this good enough for your application?  Industrial Pt 100 sensors are rugged, stable and inexpensive, but less accurate than standard or secondary standard platinum resistance thermometers.

Thermocouple thermometers are widely used, and are reasonably accurate if employed correctly. Euromet publish a guide to calibration, which you may find useful.  For temperatures below 500 C, Platinum resistance sensors are probably more appropriate.

UK's National Physical Laboratory offer an on-line good practice guide which you may find helpful.  The Wikipedia entry on resistance thermometry has more background information.

http://www.euramet.org/fileadmin/docs/Publications/calguides/EURAMET_cg-8__v_2.1_Calibration_of_Thermocouples.pdf

http://www.npl.co.uk/publications/good-practice-online-modules/temperature/types-of-thermometer/

http://en.wikipedia.org/wiki/Resistance_thermometer

Hello Shri Shaooji!

I do agree with Alan Robinson! You need to decide between the performance (accuracy and stability) and more importantly the cost.

Primary or secondary laboratory standards are costlier to be used with a pH meter.

Pt100 or Pt100 grade RTDs or thermocouples are best for your day-to-day use of pH meter but it demands more frequent calibration. In my experience, I have seen the pH probe looses its calibration at certain amount of time depending on your uses. The best way is to use new probe. Unfortunately I did wanted to investigate the real cause of the probe failure but could not carry out in detail.

To summarize, a best quality Pt100 or thermocouple can be a solution with proper electronics/instrumentation.

Thanks!

We also tried with Pt100 and Pt1000 RTD along with our indigenously developed pulsating sensor based instrumentation. It gives quite accurate measurement with good precision. But the calibration does not hold for a long time. Only in some cases it holds about 1 month. Moreover now a days whatever temperature probe is used in commercially available conductivity meter or pH meter the solution temperature as read by the glass thermometer differs from the reading as obtained from the instrument. 

Can you give us some idea how far the calibration drifts?  0.01oC or a few degrees? What is the temperature range of interest?  Is it a drift in zero (ice point), or a change in scale factor?

For your Pt100 sensors, is the drift outside the limits specified in IEC 751:2008?  I believe this is +/-0.15 oC +/-0.002 x T(oC) for grade A, although tighter tolerances are available.  Are you using a 2, 3 or 4 wire configuration?

You appear to be using liquid in glass thermometers as your reference.  How often are these calibrated?  NIST recommendations are in the link below.

http://achelpers.marketlabinc.com/ProductTechData/Forms/Manual/9040-155%20NIST%20In-House%20Cal.%20Thermometers.pdf

We are using glass thermometer with Hg as sensing material. It is well certified by NIST. Moreover we use to check with ice point in every one month interval. Hence there is no error in primary calibration source. But when we use Pt-100 or Pt-1000 with our own instrumentation it works initially with good precision and accuracy. But when it is used with conductivity meter for temperature normalisation first few days it works satisfactorily. Then the temperature differs even about 2 deg C after one month. A few units work satisfactorily about 4 months. We also realised that the conductivity calibration holds for a long period even upto one year with our pulsating sensor based instrumentation. But due to change in temperature the conductivity reading deviates.

A few more suggestions.

If you connect a new Pt100 sensor, and get accurate results, it is likely that the metering electronics are OK, and the problem lies with the sensor or the connection leads.  A drift of 2 deg C per month is very high.  Manufacturers typically quote drift closer to 0.05 oC over 5 years.

How is the sensor element encapsulated?  Is it possible that the packaging is not hermetic, and that your conductive medium percolates into the housing and forms a parallel electrically conducting path?  Could there be chemical attack of the sensor, the substrate or the connecting leads?  Could corrosion of the encapsulation induce strains in the sensor or its substrate?

Have you examined a failed sensor to check for corrosion or other damage?  Is the resistance stable after removal from the test environment?

Do you routinely use a 4-wire sensor configuration to reduce sensitivity to changes in the connection leads?

If your mercury in glass reference is stable, then your measurement environment probably doesn't attack glass very quickly.  Is it possible to slide a platinum resistance sensor into a long glass tube, fused shut at one end, and immerse this in your solution to protect the sensor and the connection leads.

Could the sensor be damaged by ionising radiation or something else you have not mentioned?

http://www.temperature.ie/ptacc.htm#

http://scitation.aip.org/content/aip/proceeding/aipcp/10.1063/1.4819578

http://heraeus-sensor-technology.com/media/webmedia_local/media/pdfsen/hstproduktbroschuere.pdf

Kind attention to Dr. Robinson. Thank you very much for possessing so much interest to solve my problem. Thanks for providing so much information by giving suitable links. We normally use Pt-1000 and Pt-100 RTD elements for temperature normalisation in conductivity measurement. Both types of RTD elements are well encapsulated with SS sheath. There is no question of entrapment of solution to the probe housing and corrosion. Moreover the solution used is mainly water of different types. Of course some probes are working for about 4-5 months without change in calibration. Some specimens do not perform satisfactorily. Initially precision and accuracy are maintained. But after 1 month there is variation in temperature reading. Our instrumentation is pulsating type. The instrumentation is designed in such a way that the primary signal output from the sensor head is in direct digital form which does not need any further signal conditioning for digitalisation. This concept is utilised in many of our instruments and such instruments are used for various applications which are well documented. Of course it is true that the manufacturers claim 0.05 deg. temp shift after 4-5 years. But we do not believe in it.

The 0.05 deg C I quoted is the temperature stability of the platinum resistance sensor element.  To realize this performance you must use appropriate instrumentation electronics.  A well-designed AC bridge with stabilised drive, lock-in detection or precision rectification of the error signal, combined with a high quality analogue to digital converter referenced to the bridge drive voltage, should be stable.  A simple DC bridge with a low cost differential amplifier could easily exhibit the drift you report.

What performance do your suppliers claim for the modules you use?  If your operating conditions comply with their recommendations, and the drift exceeds their performance specification, surely you have a claim against them?  Either the equipment is faulty, or the design is unable to meet the performance claimed.

If you can show that two sensor units from the same supplier diverge by more than twice the specified drift, then you don't even need to demonstrate reference calibrations which are traceable to international standards.