How to Achieve Electrical Isolation for Thermocouples in Automated Fuse Testing Systems

I am working on a project to automate fuse testing. Per the test procedure, thermocouples are attached directly to the fuse furrules. A current is passed through the fuse until thermal stability is achieved. The TCs are going to be connected to an amplifier then to an A/D converter and finally to a microcontroller for analysis.

Here's the issues I am having that I do not know how to solve.

The power supply we use to drive the current will drive it at whatever voltage is conveient to achieve the curent required. So, if the fuse should open, there is the possiblity of imposing severeal hundred volts onto the thermocouple for a brief time. I say brief, but the few milliseconds will be sufficient to damage/destroy the electronics downstream of the TCs. What I need is a way of electrically islotaing the TCs while permitting the 0-15mV signal from tem to reach the amplifier undisturbed.

I do not want to mechanically isolate the TCs from the fuse as the introduction of any medium between the fuse and TC will introduce errors in temperature measurement. While that is not necessarily a chritical attribute, it will make calibration traceable top NIST more difficult.

Do any of you brilliant guys out there know of a solution that is not off the shelf technology (in other words I need discrete component type stuff) or bulky to the point of needing it's own desk (I need it all to fit in a box the sixe of a deck of cads)?

There are calibrated thermocouple meters that provide wireless connectivity for the data (I think some of the new Fluke meters do this).

The low voltage signal from the thermocouple is going to be difficult to isolate in the analog domain and I think you would sacrifice NIST traceability of the measurement system if you went that way. Isolating the data stream after the ADC (or current loop conversion if you want to go that way) is a better bet.

If you want to roll your own at the component level I don't see how you will end up with a traceable measurement system. If that isn't really an issue, you can use a battery powered ADC with a digital isolator as long as the worst case voltage is less than a couple thousand volts. if you want this type of solution post a response and I'll provide specific part numbers (also specify what kind of accuracy you need and over what temperature range).

The max voltage seen at the TC would be in the hundreds of volts. We use a Doble power supply to drive the current through the fuse, but in the event the fuse opens, the Doble will attempt to drive the current by increasing the voltage. Once the voltage has reached a threshold (based solely on the VA rating of the power amp at whatever current settign is entered) the unit trips off and power is removed from the circuit. Al told, the time is in the tens of milliseconds the high voltage will be applied, but I have not found a viable solution to it without introducing some sort of physical isolation between the fuse and the TC.

If you are aware of an ADC (single chip solution) that provides the kind of isolation I am looking for I will be forever endebted.

The ambient temp range is 18-25°C. The TC temp range will be from 18-250°C. I have designed the amplifier, ADC and microcontroller systems already, but redesign is an option if needed to accomodate the isolation.

I can't utilize the "off the shelf" systems like the Fluke. I need an embedded system that will be stand alone.

One possible solution:

Use a separate microcontroller, amplifier and ADC for each thermocouple.

Isolate the power using a ROE-0505S or similar. Add a filter cap on the output and follow that with a high PSRR/low noise LDO to get 3.3V (assuming your microcontroller and ADC are 3.3V).

Use a 3.3V serial link (async, SPI) to communicate with the domain outside the isolated circuit using either an opto-isolator (typically low speed) or a digital isolator like the Si88620ED-IS (Maxximm TI and Analog Devices also make digital isolators). You can get different isolators depending on the number of isolated lines you need in either direction.

Your thermocouple can now float beyond 1000V from the ground on the non-isolated side.

Be careful about unintentionally creating a capacitive coupling (through the PCB, etc) between anything on the isolated side and the non-isolated side. If you have fast transients on the power on the isolated side, the smallest capacitance (pF) can couple that fast edge over to the non-isolated side and cause all sorts of havoc.

I understand what you mean about using optos to communicate between isolated and non-isolated sides, but I don't follow the reason for the separate microcontrollers and amplifiers.

If I am using optos to electrically isolate the HV portion from the microcontroller, I only need one. I also don't follow the idea of the amplifier being in the HV zone. I know that the HV would be common mode, but is there an amp that can tolerate a common mode voltage of a few hundred amps without producing shrapnel?

re: isolation on the thermocouple side

I don't know of any amplifiers that will tolerate several hundred volts of common mode signal and still resolve a few mV accurately. Since this is a DC signal, you can't use a transformer.

Isolating the power, microcontroller and ADC is probably the easiest and cheapest way to go if you want to preserve accuracy on the temperature sensing. The microcontroller and ADC (assuming you use a dedicated thermocouple ADC) should be around $5 for the pair. If there is a solution to isolating the thermocouple analog signal I suspect it will cost a lot more than that and accuracy will suffer.

BTW, if you are using a bridge and standard ADC you should look at the dedicated thermocouple amplifiers made by Maxxim and others. If you are looking for really high accuracy you might want to look at platinum RTDs instead of thermocouples.

You didn't ever say what kind of accuracy you actually need. If it is in the 1 deg C range, you might try using a non-contact IR thermometer sensor and eliminate the thermocouple and voltage coupling problem completely.