How to understand the "zero source impedance" in LISN insertion loss calibration

Question

Figure 1: 50 uH LISN Diagram

Figure 2: Input impedance of the LISN varies with frequency from 5Ω at 9 kHz to 50Ω at 30 MHz

In Page 27 of Calibration and use of artificial mains networks and absorbed clamps from EMC Standards, author Tim Williams, it says:

... The insertion loss of LISN is measured as a voltage ratio between the EUT port and the receiver port for all lines. It is normal practice to use a 50Ω system for the measurement, but as the input impedance of the AMN/LISN varies with frequency from 5Ω at 9 kHz to 50Ω at 30 MHz, driving the EUT port from a 50Ω source would give an input voltage that changes with frequency. The correct way to make the measurement is to feed the AMN/LISN and measuring system in parallel through a Tee adaptor, providing effectively a zero source impedance. The receiver port is terminated with a 50Ω load and readings taken on the attenuation measuring system over the required frequency range. The measurement system is then transferred to the receiver port and the 50Ω load transferred to the Tee adaptor. The insertion loss of the AMN/LISN is the difference between these two measurements.

My questions are

1. Why I called “the wrong way” get a different insertion loss from ANSI or CISPR “correct way”?

Figure 3: the wrong way to calibrate the LISN insertion loss

2. How can the so called correct way can create a “zero source impedance”, by 2 times measurement and parallel through AMN/LISN and measuring system? Can you make a drawing to explain it?

Figure 4: the correct way to calibrate LISN insertion loss

Answer 1

They're taking the ratio of measurements, so that the result is always in ratio to, or normalized again, the input-side voltage. Which is equivalent to having a constant voltage at the input, without having to wire up some impossibly powerful and wideband amplifier.

Doing a once-through measurement assuming the input voltage is constant, is erroneous, because as you know from the simulation, the input impedance and therefore input voltage is not constant.

This measurement can also be done in a single pass, with a pair of 10x probes and a dual-channel oscilloscope for example, but the standards aren't written assuming any particular kind of test equipment, and the most likely means one would have on hand are 50Ω sources and receivers.

You can measure the voltage gain in simulation by setting the source to 50Ω ESR, and viewing V(to_EMI_Receiver)/V(DUT). Note that R3 represents the EMI receiver as such, so a further connection "to" a receiver is kind of weird to be hanging off to the right. But in the sense that schematic placement doesn't have any strict semantic meaning, only the netlist, that's excusable.

(No relation, by the way. :) )