In order to perform calibrations with a high degree of accuracy, reference standards must be employed at every range or decade of the measuring or calibration instrumentation.

Clearly, this can be difficult and costly since these standards must be highly stable and their precise values must be known with a high degree of certainty and with a sufficient resolution. To minimize the cost and difficulty, more practical means of performing such calibrations is to use transfer standards.

If one has a single standard that is calibrated by a national laboratory, one can then transfer the "certified" accuracy by comparing the "certified" standard to the transfer standard for as many as three decades.

The resulting accuracy of the transfer process can be much better (e.g. 1 ppm) than the accuracy of the transfer standard itself (e.g. 15 ppm). This may be understood as follows: a stable, but only moderately accurate, ruler could be used to accurately transfer measurement from one object of accurately known length to a second object of unknown length. This transfer is virtually limited only by the accuracy of the known length.

The IET HATS-LR Series of transfer standards consist of 12 matched equal value resistors of value R, designated as R1 through R12, which may be connected in series or parallel combinations to produce any number of values such as R/10, R and 10R. This permits the progressive transfers to higher or lower decades. For resistances above 1 MΩ, the HATS-Y Series of transfer standards may be used, and the same discussion applies.

Setting for Various Resistance Combinations

To obtain a resistance R of one step, any single resistor may be used, but it is clearly advantageous to use as many of them together as possible in combination. This not only allows the applied power to be divided among the set, but permits the use of a number of resistors in determining the net statistical resistance, always better for a larger number. In particular, 9 resistors are connected in a series-parallel combination. The best method to implement this circuit is to use the Model HATS-LR-SB set of shorting bars.

Similarly, the value of R/10 may be implemented by a parallel combination of 10 resistors. This again may be conveniently done with the shorting bars. This takes statistical advantage of 10 resistors in combination. Of course, using 10 resistors in a series combination will produce 10R with the same statistical and power advantage.

It is important to note that any series, parallel, or series-parallel configuration results in the net deviation being equal to the average deviation for that group of resistors no matter how they are connected, as long as the applied power is divided equally among the resistors. This is clearly the case with the R/10 and the 10R configurations, i.e. they have the same deviations. It is also true with the 9 resistor series-parallel configuration, since the effect of deviation of a single missing resistor may be safely neglected. This property is very useful since it permits making accurate transfers across three decades with one single unit.

Calibration Transfers

As an example, a 10 kΩ standard may be compared with a HATS-LR unit with 10 kΩ steps connected in a series-parallel configuration, as described above, to provide a net 10 kΩ resistance. Once a comparison is made, a net deviation of 10 resistors (approximately the same as for 9 resistors) is obtained.

This average or net deviation remains constant for a series combination, and therefore the standard is effectively "transferred" with the same deviation plus the transfer accuracy of the unit to another decade, 10R or 100 kΩ in this example.

This deviation is also transferrable to 1 kΩ by using the HATS-LR in the parallel mode.

This process may be continued with another transfer standard. 1 MΩ steps in this example could first be configured in the R/10 mode to produce 100 kΩ, which would be compared to the first standard set in the 10R mode. This now produces the additional values of 1 MΩ and 10 MΩ with known deviations close to the original standard. Only the transfer accuracy errors have to be added for each transfer.

Referring to the same example, a transfer may of course also be extended downwards. A standard with 100Ω steps would be set in a series for 1 kΩ and compared with the original standard and would subsequently provide a transfer at 100Ω and 10Ω.