Spectrometers for Temperature Measurements

For a list of suppliers and specialists, I'd start by looking here.

Thank you!

Look at www.omega.com

you can get temperature sensors with single wavelength and dual wavelength versions.

This sounds a fancy name for the century old method of temperature measurement using the disappearing filament technique!?

You all know, don't you? viewing the furnace light through a window and adjusting the filament of a bulb until the light from the filament equals the light from the furnace, then read off the temperature from the graduated scale...

In expert hands the temperature of molten metal could be measured to within 1 degree Celcius...

John.

Dear John

I agree that hot wire is cheap and near fail proof technology. It is like old coal rail engines.

It takes time to get used to those old devices and they require some skill and perfact vision. What if a person has problem with RED color identification or has poor focussing problem. I think the best way is IR now. Some cheap device use a lense and collect IR on a thermocouple and these are just about US$200 range. Omega equipments are actual spetral ratio type and are US$4000 range. I think very soon other cheal alternative in optical version will come at reasonable cost.

We should use advanced technology to support R&D and our better tomorrow. Perhaps for those who need to read only once a while that hot wire thing may be OK. It was so odd when I used some 30 years ago and never used that stuff again. I prefer fastest information method as this is an era of doing everything robotic way except for few personal stuffs.

Its all very well to use and talk about how we should all be using the latest and greatest instruments...

But I've come across some really weird cases where old is best... The easiest one is of course the need for portability and so a small battery handheld device is okay, so why not make it small using all the latest technology?

Because the instrument is only of very limited use, essential, but not required in any large quantities, so a small cheap meter is ideal...

Quite often I've found that for me to design a specialist instrument that will do exactly what the company wants, is cheaper than buying an off-the-shelf instrument that has all the bells and whistles together with functions and accuracy which in most cases will never be used!!?

Also, I design instruments that are easily repaired! On another thread people are talking about never repairing PCB assemblies, just throw them away and use a new one... What about if you haven't a new one?? What if it takes a week to get hold of one? The companies I've done work for may have to stop production waiting for the replacement board - Totally unacceptable! So the design should be easy to repair, calibrate etc... to be of any use...

I've had people in companies say to me, "okay it does the job but its a bit old fashioned looking isn't it?",

That to me is a compliment! There are some wonderful 'old' instruments around that do their job with excellent results, they are easy to repair and so keep on going decade after decade...

I've also been asked to repair some instruments that were the latest thing 12 years ago, but as they're surface mounted PCBs and use components that have long since been made obsolete as well as the fact that the software code burnt onto the chip isn't available, means that only 12 years from new it is scrap! and the customer has to be told to buy a new one from somewhere for as you say maybe $ 8000 an expense they usually aren't happy about!!

Sometimes the old designs are better...... John.

Dear John

I agree on those points and you have well said about the way new technology is done in careless way. People used to build electronics to survive for life and now we have no idea if it will last few months. Nothing to say about the hate of repair job. I love talking to you because you are from early know everything and do everything by hand era.

I needed Ir sensor for monitoring the rubber sheet temperature and this stuffs moves at high speed on conveyer. All contact thermocouple were no where near and old hot wire stuff just could do nothing. I needed <10ms response time. I hit on special IR devices and could handle it down to 1ms using specially designed lens and sensor.

I now need 10ms temperature response for the air temperature sensor but find nothing close to it. I need 0.1C accuracy. I am looking again at the same technology or perhaps fiber optic version. My sensors will be placed on a balloon so I expect some air flow around and that may help in heating-cooling the sensor fast. I will use an array of sensors as I am afraid that few will break by air shock.

Vaishala sensor perhaps in 1s response time. They ask huge money for their sensors.

Dear Shyam,

Hmmmm that's a fast response time and accuracy.

Have you looked into something like a platinum thin film sensor deposited on a heat insulating substrate? I'm thinking that with a very fine thin platinum trace it should give a fast response and being platinum would have a predictable temperature coefficient...

Either that or a fine platinum wire suspended in the air flow, similar to a bulb filament without the glass envelope?

Other wise, as you say, the only way I can think of measuring with that speed is with an infrared thermometer, but then the accuracy would be a problem...

John.

Dear John

Those platinum resistance film sensor PT1000 or PT2000 on alumina (Al2O3) substrate, temperature sensing for air have response time of 6s and perhaps are fast only when dipped in oil.

I have collected almost each and every type of temperature sensor in the world. I am now waiting for development of quantum dot temperature sensor that can convert about 10000nm into perhaps 1000nm using 10 Phonon into one photon and Quantum dot coupled with IR diode will give us fast signal. Perhaps that may take few years to have a new sensor.

I have this thermoluminescence temperature sensor also. This sensor require charging of thermoluminescent material by gamma irrediation. This is fast enough to microseconds level responses and can be less than micron size grain powder. It is nice to illuminate heated material like meteroids. Lots of material makes it slow. This sensor is not for regular use but only for research data and once a white extra-ordinary experiment.

Thermopile in heat flux sensors is also very fast and can sense in microseconds. That is what NIST uses in heat flux sensors. These are also used in gas leakage detection in NDT applications. I was executive committee member of NDT Society of India.

Dear John.

Of course I know this method. But in this case "expert" also must add correction value to result which depends from Emissivity of the object and the temperature of molten metal could be measured to within 1 degree Celcius (resolution of measurement) but real error may by much more than 1 Celsius. By the way measuremet of the high temperature object (1000-1800 C) with type B Pt/Rh termocouple,its master gage instrument, have error conected with reading deviation of thermocouople, 3.5 - 5.1 Celsius..

In my case I need to reciving radiation from narrow spectral bands and than I can calculate multicolour temperature and object temperature using these radiation measurements. For example, OceanOptics S2000 spectrometer have 2024 points in 0.5-1.1 mkm spectral region, for my method I need at least 4 points in spectrum. But this device very delicate for foundry condition.

You can use multimode optical fiber bundle having quartz or fused silica as core. I use 10mm dia face with large number of fibers placed in steel SS316/L clading. What is the temperature you think your fiber will experience? These fiber buldles are expensive to about US$2000 for 1m length. You can also try conduit to get the light out. As temperature is high, you can use visible light zone and not IR. You can also use rising portion of the high energy spectrum. Small fiber means low light intensity and means more sensitive sensor.

Ocean Optics is a very good company. Ask them their 7-fiber bundle. I use 40000 fiber bundle. You can use a pin hole with lense or optical window material in the wall.

Thank you, Dr. Shyam.

In my application I work with 1200-1600 Celsium region. For radiation conduct I use Thorlabs fiber-optical cable FT-1.0 EMT with SMA 905 connectors (not so expansive devices). In principle in hot zone I install only Focus Unit with cooling jacket. Spectrometer installation place is 4-10 m from object.

I agree with you that Ocean Optics is a very good company. S2000 is old model of spectrometers. May be modern Ocean Optics devices more compartible with my application devices. I'll study them more detail.

Perhaps just using two narrow pass band filters with suitable photodiode should work in your case. You can also chop the light and change the filters mechanically using some small motor chopper. Using CCD monochromator is not the right choise as the well size of the CCD is small and dynamic range is limited. Photodiode current can change in several decades with high accuracy.

You are quite right, CCD linear detectors are not so good devices for pyrometer measurement like Photodiodes. But in some cases, for examle for temperature measurement of oxidable metalic surfaces with non stable spectral distribution of Emissivity, methodic errors for classic one-color or two-color pyrometers much more then acceptable level. But for SCMT-technoogies instrumental errors of CCD technologies is not so crytical and its using much more comfortable and gives good results (acceptable level of measurement error) than much more complicated Photodiode detectors application for SCMT-technologies.

For 1200C to 1600C you can also get UV tail. Did you try some of the UV range spectrum also? Fire generates a lots of UV and you can use metal to generate electrons using UV (Photoelectric effect). Look at Hammatsu UVTRON. You can get direct temperature response.

Perhaps your emission peak is in MID IR range. Hence, 1000nm and one more near band at about 700nm to 800nm will be ideal on rising edge. This you can detect using US$10 IR photodiodes. If you use chopper for IR beam light then room light leak can be rejected easily. Detection on tail portion will be costly for you. For 400C and above even read zone becomes intense due to T^4 emission. 1200-1600C I will prefer high side zone detection with cheap photodiodes. If fact it is good to kill the tail portion.

I just looked at more sources and found that people have already used Quantum well temperature sensors. People are real fast these days. They are leaving me and John far behind. Do you agree John?

There are few water molecule absorption bands in IR and MID IR. The are at 900nm, 1900nm and 2200nm or in near zones. I do not have axact wavelengths hence look into remote sensing waveform data as they use these bands to know the moisture on ground from above. Just avoid these bands in your measurement as moisture may be dynamic stuff in the furnace and your reading will also become full of dynamic oscillations. Those are not temperature fluctuations but absoption fluctuations. There are also absorption bands for CO2 and CO between 1000nm and 1200nm zone. If you look at molecular absorption spectrum of all possible compounds that are present in the gases then you can easily get the right bands that are not affected and black body spectrum is cleaner. Simple rule is to keep away from resonant IR absorption bands. These are trade secrets and are embedded in good designs which other often ignore and get poor results. Use IR transmission windows (zones of frequencies that have least absoption like 1550nm etc). There are three such windows which you can easily locate from literature by hunting. They will give you much better results.

Thank you, Dr. Shyam.

I know this problem. By the way, I did made some tests alowed to excluding of CO2 and H2O absorbtion bands on measurment results. My OceanOptics S2000 developed on SONY ILX511 CCD linear detector base. As you know it is Si-detector with 530-1100 nm spectral band. In these band I obtained some IR transmission windows (prviously studied from literature) and good measurement results on opened oxidable metal surface. But noise problem, as for me, the main prolem for such spectrometers in temperature measurements. Because of algorythms for noise resistance for CCD linear detectors is very slowly and delaying of measurment time.

Do you agree with me?