The measurements of highest precision in a standards lab are the 1:1 comparisons of similar impedance standards, particularly the comparison between the standards calibrated at the National Institute of Standards and Technology (NIST) and similar reference standards that remain in a lab. This application requires measurement resolution and repeatability to detect parts-per-million (ppm) differences but does not require extreme, direct reading accuracy. Two standards of very nearly equal value are compared using "direct substitution" they are measured sequentially and only the difference between them is determined. When reading directly in value, the resolution of the QuadTech 1689 is between 10 and 100 ppm. However it has a ∆ % mode where the difference between an entered nominal value and the measured value is displayed in percent with resolution of 0,0001 % which is 1 ppm. Single measurements made in this ∆ % mode at a one/second rate have a standard deviation of about 2 ppm at 1 kHz. The use of the instruments AVERAGE mode reduces this by  where N is the number of measurements averaged. Thus, an average of 5 measurements or more reduces this deviation to under 1 ppm. Using averaging, it is possible to get the difference between two impedance’s to 2 ppm or better. It should be noted that although averaging many measurements takes time, an automatic bridge like the 1689 can take a lot of measurements in the time it takes to balance a high resolution, manual bridge.
A favorite technique is to record 5 averaged measurements and take the median of these, which is very easy to do. This gives a record of the spread as well as increasing the precision, and is independent of a large error caused by a line spike, lightning or other non Gaussian noise source. It should be noted that the 1689 has a MEDIAN mode, that takes the middle value of three measurements, these median values can be averaged automatically to give one final result.
The 1 ppm resolution of the 1689's ∆ % mode is not limited to values near full scale as it is on six digit, manual bridge readouts. For these, the resolution of a six digit reading of 111111 is 9 ppm. The 1689 does not discriminate against such values; it has the same 1 ppm resolution at all values. It also has 1 ppm resolution in D and Q (tangent of phase angle), a useful capability but more important for dielectric measurements than for RLC calibrations.
Scaling a calibration of one impedance value to another value is another precision measurement required in the standards lab. There are many techniques used for this process which differ for different types of impedance standards; resistance, capacitance or inductance. One method common to all is to simply measure each value with a bridge, assume the bridge ratio is perfect and apply the ppm correction of the known standard to the unit being measured. This method gives high accuracy when using a transformer ratio arm bridge such as the QuadTech 1615 or 1616 Capacitance Bridges, but such bridges aren't available for inductance or high capacitance. The 0,02 % accuracy of the 1689 compares favorably with available manual bridges that make these measurements. Its ∆ % mode can be used to give ppm resolution for both measurements by entering a different NOMINAL VALUE for each measurement. This improves the instrument accuracy as well because it is not limited by resolution of the display. Moreover, ratio measurements on the same range are independent of a calibration error or drift in the internal standard. Measurements on different ranges can be improved by a recalibration using special standards, such as the calibration kit available for use with the 1689.
Another important feature not usually available in precision lab bridges is the multi-connection capability which allows both 4 terminal, Kelvin connections and 3 terminal, guarded measurements. This adds up to a 5 terminal capability (not 7) which is rarely found in lab bridges and which is particularly important if measurements must be made on both very high value and very low value impedance’s. The automatic open  and short circuit zero correction capability of the 1689 augments the multi-terminal capability by subtracting out the effects of unguarded stray capacitance and mutual inductance between connecting wires.
Obviously, automatic instruments such as the QuadTech 1689 have the advantage of speed because a balancing procedure is not required, but this is not a particularly important advantage when only a few calibrations need to be made. However, there are situations where balancing an ac bridge can be tiresome, especially when low Q components are being measured as they require an annoying series of alternate balances of the two adjustments needed to null an ac bridge. Speed itself can be important when checking multi dial decade boxes. For this task, a suggested trick is to use the MEDIAN mode which will reject the erroneous measurements made while the dial settings are being changed.
A final advantage of an automatic instrument with IEEE 488 bus interface capability is the opportunity of having the result printed out, thus avoiding the opportunity of making a mistake in recording the result, as well as making the record more legible. Moreover, with a computer in the system, any required correction calculations can be made without the chance of more errors, especially the all to common problems with + and - signs. Of course a computer, suitably programmed, also can lead the measurement technician through a complicated calibration process with prompts and procedures that ensure proper measurement techniques and precise data manipulation.

The applications given below fall into two categories. In some of them, the QuadTech 1689 fills a definite need and can make better measurements than any available manual bridge. In others, the 1689 can make the required measurements, but not as well as specialized, expensive lab instruments. This latter category extends to all ac impedance measurements at or below 100 kHz, so for many standards labs that don't require the highest accuracy, the 1689 becomes a "standards lab in a box."

A made-to-order application that uses most of the features mentioned above is the inter‑comparison and scaling of inductance standards. This application is particularly important because there has bean no inductance bridge available with ppm resolution except, maybe the GenRad 1632 which was discontinued several years ago. The 1632 was a six‑digit, two‑terminal bridge (one grounded) with only 0,1 % direct‑reading accuracy. It was very good for comparing two, similar decade‑valued (1, 10, 100 etc) inductors. It wasn't particularly good by today's standards because of the 0,1 % direct reading accuracy. Balancing a 1632 to high precision was a slow procedure especially when measuring standards like the QuadTech 1482 Inductance Standards which have low Q values, especially at 100 Hz , where many measurements must be made. Moreover, the effective ac resistance of these standards is primarily the resistance of copper wire which has a temperature coefficient of almost 4000 ppm/degree C. Even a 1/100th of a degree change in the temperature of the wire due to ambient changes or applied power causes an annoying bridge unbalance which makes the inductance measurement difficult. An automatic bridge like the QuadTech 1689 has no such problem; the resistance can change at any (reasonable) rate without affecting the inductance reading (if you measure equivalent series inductance). Because of this, and the 1689 ease of use, the instrument is ideal for making 1:1 comparison measurements on inductance standards.

A five‑terminal (guarded and Kelvin) capability would have been an advantage in calibrating these inductance standards, but because they were being calibrated long before either 3‑terminal or 4‑terminal capability was available on an inductance bridge, NIST uses a grounded, 2‑tenninal connection. With early precision meters, ground was guard, not one of the main connections, but this did not cause a problem. The only thing to remember, when measuring something like the QuadTech 1482 Standard Inductor is to tie the case to its LOW terminal (with the link provided) and to insulate the case from ground. This keeps the internal stray capacitance’s that effect high‑inductance measurements the same as the 2‑terminal calibration by NIST.

Standard inductance’s are particularly hard to scale in value by combining two or more units in series or parallel (as is done with resistors) because of their size, their low Q values and the stray capacitance’s involved when two are connected together. Transformer‑ratio‑arm bridges, capable of precise scaling, are not available for inductance measurements. They can be made from commercially available parts, but are difficult to construct and use. Fortunately, extreme accuracy is not required because the best NIST calibrations have an estimated uncertainty of only 0,02 %. However, ratio measurements used to scale these calibrations should be much tighter to avoid adding errors. Ratio measurements of 2:1, made on a 1689, on the same range and using the ∆ % readout, typically have errors totaling less than 20 ppm. This allows comparisons of inductors of intermediate value (those values starting with a 2 or 5) to be compared against the even decade values with negligible added error. Scaling calibrations over a 10:1 range is less accurate because a range change is often required, or if on the same range, there is apt to be increased non‑linearity error. They should be good to 50 ppm at 1 kHz over the basic overall range of the instrument.

Precision capacitance bridges such as the QuadTech 1615 and 1616 referred to earlier have ranges up to 1 μF at high accuracy. This range can be extended somewhat by using external standards, but the accuracy deteriorates rapidly as the capacitance is increased because of the inductance of the wiring and the leakage inductance of the ratio transformer used in the bridge. These bridges are three‑terminal; there is a guard but single connections are made to each end of the capacitor being measured. For good accuracy at higher values, four‑terminal (Kelvin) connections are needed to remove the effects of self inductance. Automatic short‑circuit zero corrections are a good way to remove the remaining effects of mutual inductance between leads.

The 1689 has these capabilities of four‑terminal connection and auto zeroing as well as extreme range and good accuracy. The normal range of the 1689 extends to 0,099999 F, but using its RATIO mode the display range can be extended (nominally) to 10,000 Farads! However, don't look for a capacitor of that value to measure, for even if one could be found, the 1689 would not be able to measure it. However it can measure 1 F with fairly good accuracy at 100 Hz even though its specifications might not indicate so. This is because the 1689 specifications assume that the zeroing calibrations, open and short, are made at 1 kHz only. If these are made at the frequency of measurement, the accuracy of extreme values depends mainly on the repeatability, which can be improved by averaging. For example, at 100 Hz the accuracy specification at 1 F is 120 %, but with a short‑circuit calibration at 100 Hz, accuracy is about 5 % for one measurement and 2 % if 10 measurements are averaged. And yes, there are standards of capacitance at such values, for example the QuadTech 1417 Capacitance Standard. There are special fixture considerations that improve the measurement accuracy of the 1689 by 5 to 1 at such values. More important are measurements on standards of lower values, between 1 μF and 1 F, again such as the 1417. The 1689 can measure these with good accuracy, usually better than required since these standards are less stable than lower valued ones, such as the QuadTech 1404 or 1409 Standard Capacitors.

There are bridges, such as the QuadTech 1615 mentioned earlier that measure capacitance from 1 μF or less with better precision and accuracy than the 1689. The 1689 is not necessarily recommended for inter comparisons of reference standards in high‑level labs, it is however, very adequate for calibrating the reference standards of lower‑level labs and all working standards and decade boxes.

Generally, the 1689 is more sensitive than the 1615 for capacitance of 1 μF or higher at low frequencies. The 1689 applies more voltage and has much better repeatability if many measurements are averaged over the time it would take to balance the 1615. Further, the 1689 is four‑terminal and the 1615 is not, so there are connection errors, particularly those due to series inductance at higher frequencies. The repeatability of measuring a 1000 pF capacitor with the 1689 at 1 kHz is good. The standard deviation should be about  where N is the number of measurements averaged, so comparisons can be very good. The accuracy is less at lower values, so lower values should be compared at a higher frequency. At 10 kHz one can compare a 100 pF standard to better than 1 ppm or a 10 pF standard to better than 10 ppm.

It is interesting to note that the repeatability of the 1689 is comparable to that of the precision 1615 if measurements are made at the same level (1V rms.) and if averaging is used to make the overall measurement time the same. An automatic instrument is not necessarily less precise than a manual one. They both use the same laws of physics and the automatic instrument has the advantage of statistical data manipulation.

The QuadTech 1689 is probably unequaled for ac resistance measurements over its frequency range (12 Hz to 100 kHz). Unfortunately many precision resistance measurements call for dc instead of ac, even though ac measurements avoid thermal voltage errors, have lower noise and can use precise transformer‑ratio scaling techniques. Moreover, the size of the unit of resistance, the ohm, is determined from ac measurements and has to jump from ac to dc. Attempts have been made to use ac for the most precise resistance measurements, but the dc habit has been hard to break.

For most resistors, the ac‑dc difference is negligible at 100 Hz or even 1 kHz. For flat‑card wire‑wound resistors, the difference can be less than 1 ppm up to 1 Mohm if equivalent parallel resistance is used at high values to avoid errors due to lumped parallel capacitance and series resistance is used at low values to avoid errors due to series inductance. Lower measurement frequencies should also be used for very low values to avoid skin effect errors. There are significant differences for high‑value, coil‑wound resistors, because of capacitance not inductance, and for high‑value, multi‑resistor networks such as decade boxes and build‑up standards. The ac‑dc difference of resistance standards are generally very small and often can be easily determined by measuring it, and a small metal film resistor of similar value at both ac and dc. Here the assumption is made that the film unit has negligible ac‑dc difference (which it probably does) and that it was stable for the time required (which it usually will be if one doesn’t heat it up by applying too much power or touching it). Once such differences are determined, ac could be used for precision calibrations.

The QuadTech 1689 should be considered for use in the standards laboratory. It can make some calibrations more accurately than possible with traditional instruments and will make many other required measurements easier and faster. It can be used for even more measurements in lower-level labs and for almost all RLC measurements when ac resistance measurements are acceptable.