A thermocouple circuit is by its nature a differential measuring device, producing an emf which is a function of the temperatures of its two junctions. One of these junctions is at the temperature which is to be measured and is referred to as the measuring junction. The other junction is maintained at a known temperature and is referred to as the reference junction. (In a practical thermocouple circuit (see Section 2.3) copper wires are often connected to the thermocouple alloy conductors at the reference junction. The term reference junctions will be used to refer to this situation.) If these junctions are both at the same temperature, the presence of the copper “intermediate metal” introduces no change in the thermocouple's emf. If they are not at the same temperature the analysis of the circuit is complicated. Moffat  gives a helpful analytical method. The Seebeck coefficient (thermoelectric power) of many common thermocouples is approximately constant from the ice point to the upper temperature limit of the materials (see Section 3.1.1). For such thermocouples, an uncertainty in the temperature of the reference junction will reflect a similar uncertainty in the deduced temperature of the measuring junction. However this situation does not exist for all thermocouple pairs. Notable exceptions occur in the case of the high rhodium-in-platinum alloy thermocouples [35, 36]. In particular, if the reference junction of a platinum-30 percent rhodium versus platinum-6 percent rhodium (Type B) thermocouple lies within the range 0 to 50 C (32 to 122 F), a 0 C (32 F) reference junction may be assumed, and the error will not exceed 3 mV. This represents about 0.3 C (0.5 F) error in high-temperature measurements.