Temperature Sensors

It sound's like a very good way to get accurate temperature readings., as long as you are using calibrated instruments and thermocouples.

Thanks for your reply.

My concern is with the temperature accuracy. I am using a tape to attach the probe to the FPGA. Moreover, my other concern is I need to only measure the temperature on a particular part of FPGA and not the entire chip as such.

What type of a thermocouple are you using? I think you would need a very fine gauge probe to do this.

You can always calibrate your system using an ice bath and boiling water. I could send you a procedure for this if needed.

Pranav, what you are doing is fairly common and ok for quick and dirty temperature measurements. I've used the same method myself... the questions that usually come up are

1. was the tip of the thermocouple making good contact at the point of measurement?

2. As temperature increases... usually the tape mastic will soften, which could cause the thermocouple to shift or raise depending on how well it was placed... did the thermocouple stay exactly where you wanted it?

To eliminate these issues, I would suggest to use a good thermal-conducting epoxy (like O-bond 100/200/300 or so) check out Omega's website for more details.

It shouldn't be difficult at all to instrument your fpga at multiple points using this method. It will take a little more time for curing (if you have an oven, this can be reduced, just be careful not to exceed any of your components temp specs). You will have much more accurate measurements with this method. (Not knowing specifically what you are looking at, I would say that you could easily arrange 5 thermocouples in an "X" configuration... or even 9 in a tic-tac-toe board arrangement depending on the size of the chip).

Even for just one thermocouple it is better to bond thermocouples with thermal epoxy than to use tape for best results in my opinion.

As an alternative, you may consider using a Dallas 1-Wire sensor via serial or USB connection to PC, running Dallas software (free from http://www.maxim-ic.com/products/ibutton/software/1wire/OneWireViewer.cfm ) for PC monitoring or link it xl spreadsheet for logging. May also improve accuracy over t/c.

I checked that out, could be usefull, but can you calibrate it?

Generally no need to calibrate - they are accurate to a degree or so.

More info;

http://www.maxim-ic.com/appnotes.cfm/an_pk/3229

We use the DS18B20 Dallas sensors for multi-sensor (up to 100) single wire networks for food storage environment monitoring. The DS18B20 accuracy is +/- 0.5degC and ideal for these applications. Calibration is the same for any sensor, you can only measure the difference to a calibrated reference and take account of the appropriate correction for the process sensor. Apart from 0.1mm or 0.2mm +- thermocouple wire pairs, Pt100 and Dallas are more accurate and economical up to 125degC range.

A big thanks to everyone!!

I will try to arrange thermocouples in X configuration and try and see how it works. The main challenge I have in this project is I have an IP(Intellectual Property) thats embedded in a chip(FPGA), and the location is unknown. This IP is to be sensed using thermal sensors and the temperature changes happening inside this IP monitored. So , my challenges are

1) How do I locate the IP effectively? I thought of placing multiple thermocouples and thus get readings from all of them. But, I am worried that placing many would induce noise, cause interference and reduce the accuracy. I looked at infrared sensing of chip(FPGA)as an option but I would have to invest on having a complex Image Processing software to do this.

2) If I succeed in finding a solution to effectively locate the IP how do I get an efficient, robust and cost effective sensor to do the job of communicating with the PC.

I hope some of you can suggest me some ways.

Many thanks

The problem with using a thermocouple for this sort of measurement is that the entire thermocouple junction needs to be at the temperature you are attempting to measure in order to obtain an accurate reading. To understand this problem, imagine a metal sphere sitting on a hot surface in a cooler ambient. The sphere will never reach the temperature of the hot surface because heat is being removed from the sphere faster than the surface can supply heat. Although you can invent geometries to minimize this problem, you can never eliminate it. Think about the reasons for using thermowells. They were develpoed to eliminate this type of problem. Two excellent alternative methods are 1) use a temperature sensitive property of the FPGA, such as a diode junction to use for temperature measurement, 2) use an infrared microscope to measure surface temperature.