A few decades ago Hewlett-Packard began selling a fairly high performance impedance analyzer, the 4192A. This analyzer could measure component impedance over a frequency range of 5 Hz to 13 MHz, with a maximum resolution of 1 mHz--that's millihertz!
This analyzer used the standard technique of applying a sine wave of voltage (or current) to the component being analyzed and measuring the amplitude and phase of the resulting current (or voltage). From this the real and imaginary parts of the impedance can be determined. The ESR of a capacitor is just the real part of the impedance. To get a good measurement it's necessary to use a proper fixture with the analyzer. The fixture which was most convenient for measuring loose capacitors was the 16047A. Here's a picture of this fixture:
The 16047C uses the 4 terminal technique. There are 4 stainless steel spring loaded "blades" that grip a capacitor's leads for the measurement. The spring pressure exerted on the "blades" is not great, and one of the irritating problems with the 16047A is that dirt, solder from the cap leads, and oxidation products would build up on the "blades", leading to erratic measurements. The user had to constantly clean the fixture "blades".
Some years later, HP introduced another impedance analzyer, the 4194A. This analyzer could make measurements up to 40 MHz. At the same time, HP also made available a new fixture, the 16047C. This fixture was much improved; it featured thick gold plated copper bars that grip the DUT (device under test). There were two screw knobs that could be tightened to apply a large pressure to the leads of the capacitor being measured. This was a great improvement over the 16047A:
The years rolled by and once again HP/Agilent introduced another impedance analyzer, the 4294A. This analyzer could make measurements up to 110 MHz and also had better sensitivity and accuracy. The fixtures on all three of these analyzers are connected to the analyzer by 4 BNC connectors. The mechanics of this connection are not very rigid, and the 16047A and 16047C fixtures could be moved by just inserting a component (and in the case of the 16047C, by the act of loosening and tightening the two knobs). This movement would result in poor repeatablity of measurements.
When the 4294A came out, the new fixture that it came with (the 16047E) was very similar to the 16047C, but with brackets on each side that screw into the front of the 4294A and hold the fixture tightly to the 4 BNCs. This gave greatly improved repeatability:
The new specifications for the 4294A included better performance at milliohm impedances. When using these analyzers, to get the most accurate measurements it's necessary to perform an open/short calibration after the instrument warms up and just before making measurments. The "short" cal is performed by inserting a "shorting bar" into the fixture. To get milliohm accuracy, the shorting bar must really be almost zero ohms. It can't really be zero ohms, but for milliohm accuracy of measurements, the shorting bar should have a resistance substantially less than 1 milliohm.
The 16047E comes with a gold plated "shorting bar"; the bar can be seen in the picture above in the top middle of the fixture; it is held in place by a screw knob. Here's a view with the shorting bar removed:
This shorting bar is much smaller than the bar that came with the 16047A and 16047C. HP realized that to get it to function as a "short" of less that 1 milliohm, it would be necessary for the contact resistance in the fixture to be very low. Thus a small area of contact with the gold plated measuring contacts in the fixture would be required, and a high pressure must be exerted by tightening the two screw knobs (until it hurts your fingers when you've done dozens of times). The shorting bar that came with the 4192A and 4194A was much larger than the little gold plated one that came with the 16047E, and it isn't possible to get a contact resistance as low as can be obtained with the 16047E gold plated one.
When I was first using the 4192A and 4194A at work, it was inevitable that the shorting bar that came with the instruments would get lost. When this happened, we made a DIY replacement that was the same size and configuration as the official one that had come with the instrument. We no longer have the official bar that came with the 16047A, but here's a picture of the DIY replacement 16047A shorting bar and also the small one that came with the 16047E: