The resulting design specification called for a scanner consisting of a universal temperature reference junction (UTR), an interface board, a microprocessor board, and a power supply.

The scanner would have to function in ambient temperatures from –30°C to +50°C and withstand vibration and shock levels consistent with flight test applications. The specification required that the scanner pass the CE requirements for both light and heavy industrial environments.

The initial design would accept up to 16 different shielded thermocouple inputs. The input configuration would be adaptable. Brass lugs would be available for users to connect thermocouple wires directly to the UTR. The UTR could also be adapted to accept various types of thermocouple connectors.

The scanner communication would be Ethernet TCP/IP. For configuration only, the design included a secondary RS-232 connection. The target accuracy of the instrument was ±0.5°C over the normal usable range of the input thermocouple.

The two major sources of error in thermocouple measurements are secondary junctions and reference junction errors. Good instrumentation practices can eliminate errors in secondary junctions.

These practices include, but are not limited to the following:

Reference junction errors are more difficult to eliminate. An error in the measurement of the reference junction temperature will be a bias error that directly affects the final calculated temperature.

There are three accepted methods to correct for reference junction temperature:

There are, of course, variations to each of these techniques. But effectively, all of the variations are performing only one of these fundamental tasks. The errors introduced into the measurement vary with the method used.

This design team’s first decision was to use the electrical compensation method, #2, for the reference junction. The plan incorporated an existing sixteen-channel passive UTR. This product is an aluminum plate with three brass screw terminals (positive, negative, and shield), for each of the 16 inputs.

The brass screw terminals are electrically isolated yet thermally connected to the UTR plate. This UTR used a calibrated RTD to measure the temperature of the plate.

After initial testing, the team discovered that small temperature gradients existed across the UTR. The errors resulting from this were large enough to prevent the instrument from meeting the target accuracy.

The team redesigned the UTR plate to include a second RTD to provide a more accurate measurement of the UTR temperature. A second set of tests showed that the UTR temperature error maxed out at ±0.05°C by averaging the two RTDs.

Recall that a major source of errors in thermocouple measurement is secondary junctions with temperature gradients at the junction points.

Keeping all of the junction points at the same temperature can minimize this. Thus, the designers decided to house all of the critical components and connections to the signal board in a zone box.

The zone box and internal mounting plates are aluminum to improve heat conduction and prevent hot spots from forming inside the module. The external components are stainless steel to help isolate the module components from ambient temperature changes.

Initially, the designers insulated the zone box to prevent temperature gradients, but tests showed that the accuracy of the reference junction temperature measurement was better if they heated the zone box and the UTR above ambient temperature.

With insulation on the instrument top cover, the UTR temperature remained several degrees above ambient temperature. This created an effect similar to the effect an internal oven might have had on the instrument without adding the oven components and controls.

By keeping the UTR temperature above the ambient temperature, small changes in ambient temperature have no effect on the UTR temperature. Tests showed that the response time of the UTR and zone box combination was about four hours.

This very slow response prevented temperature gradients across the UTR, making the junction temperature insensitive to fast ambient temperature changes.

The signal board proved to be the most difficult part of the project. The initial design was an analog input section consisting of 16 isolation amplifiers, one for each thermocouple input, in order to achieve 1,000-volt DC isolation.

The output of each input amplifier multiplexed to a programmable gain instrument amplifier. The signal from this amplifier passed to a 16-bit analog-to-digital (A/D) converter and finally passed to the microprocessor board.

In theory, this appeared to be a good approach to prevent errors from grounded thermocouples and noise from DC voltage spikes. Although the board worked exactly as the team had planned, it turned out to be a calibration nightmare.

Calibrations required several hours, as each input amplifier had several trim pots to adjust. Small temperature changes within the zone box caused large output errors. These errors could not be controlled but were merely minimized by recalibrating the unit whenever the UTR temperature changed more than 0.5°C.

This was not practical for a unit that might install in an outdoor test facility, where temperature swings could exceed 40°C during a test.

After some experimentation, the design team incorporated a 22-bit A/D converter for each thermocouple input. This solved several problems.

First, the microvolt signals from the thermocouples converted to a digital signal almost immediately after leaving the reference junction. This minimizes the effect of noise and circuit drift.

Second, the signal-to-noise ratio increased from 100 to 160 decibels.

And finally, any drift in the A/D converters could easily be corrected by permitting periodic zero and span correction of the A/D converters via a software command.

When this command executes, the A/D converters switch off-line, the spans compare to a reference voltage, and the inputs short to get an updated zero. The process can require up to several minutes, depending on the setting of the average variable.