to the port 1 output and drive the DUT directly from the booster amplifier. This generally results in poor measurements of DUT due to mismatch and gain errors in the booster amplifier. In this approach, it is common to add a booster amplifier, normalize the S21 trace, and then add the DUT and measure the resulting gain relative to the normalized booster amplifier response. However, the normalization has errors due to mismatch between the booster amplifier and the load port. And the measurement has errors due to mismatch between the booster amplifier the high‐power DUT. Further, the input match or S11 of the DUT cannot be reliably measured because booster amplifier isolation eliminates the ability to measure a signal reflected from the DUT. A second error in gain measurements often occurs with this direct approach due to gain drift or gain compression of the booster amplifier.
A better systematic approach is to add the booster amplifier behind the test port coupler, and use a second coupler as the reference‐channel tap to generate a signal proportional to the booster amplifier output signal that can be routed to the reference channel. In this scenario, a directional‐coupler rather than a power splitter is typically used to provide lower loss after the booster amplifier. The output of the reference channel is directed through the test port coupler of port 1 so that the S11 of the DUT can be accurately measured. In almost all cases, an accurate measurement of high‐drive‐power devices requires a booster amplifier followed by a reference coupler. Chapter 6 provides a detailed discussion of high‐power amplifier measurements, including several alternative block diagram configurations to support various power levels.
2.2.9 VNA with mm‐Wave Extenders
Frequencies above 30 GHz are technically millimeter wave (as in they literally have a wavelength less than a cm), and it is common to consider mm‐wave above 50 GHz to over 120 GHz. At these high frequencies the losses in the test leads become so great than it often makes testing quite difficult because of the power loss from the VNA port to the DUT reference plan. This is particularly true in on‐wafer applications where the large size of the VNA means it must be placed some distance from the probes, and that means long test cables.
Thus, at these high frequencies many equipment manufacturers make mm‐wave extenders or remote heads. These heads typically have built‐in frequency multipliers for the source and LO, built‐in mixers, and built‐in directional couplers. Figure 2.30 shows a representative block diagram of a mm‐wave extender. The actual implementation can vary a lot, but modern mm‐wave heads all have this similar style.
Figure 2.30 mm‐Wave Head block diagram with broadband capability.
The mm‐head shown represents a broadband head and will have a coaxial test port (typically 1 mm or smaller) that can pass DC. The low‐band path takes the RF input signal directly and passes it through to the test port. There are a couple of basic methods to provide the low‐band measurement capability: some manufacturers use the directional‐couplers in the base box (port 1, for example) to drive the RF input and will measure S‐parameters from low frequency to the first multiplier band using the normal VNA receivers in the base box. This reduces the need for broadband couplers in the head but may add instability due to the long test port cables from the base box to the mm‐input. One manufacturer uses broadband receivers in the mm‐receiver portion of the head, as well as broadband coupler. The RF signal is only a source, and even at low frequency the mm‐head couplers are used. This provides a quite stable result, but it is a difficult design to achieve such a broadband coupler. The LO multiplier either can be a discrete multiplier or can rely on higher‐order harmonic conversion in the mixer. The reference and test IF signals are returned to the base box where they are processed to produce the desired detection.
Older mm‐wave broadband systems (not shown) placed the combiner in front of the mm‐port. This allowed the use of a full waveguide coupler in the head (which of course cannot pass low frequencies), but the loss of the combiner at the output leads to system instabilities, and this style of broadband head has largely been obsoleted.
One version of mm‐heads, so‐called banded‐heads, typically use waveguide‐based couplers and cover only the waveguide band. The test port is a waveguide as well. These heads have only a single source and LO multiplier, and of course they have no low frequency path at all. Banded heads operate over nearly all the waveguide bands from V band (50–75) up to THz frequencies of 1.5 THz.
The reference and test IF paths, just like in the broadband head, are returned to the base box for signal conditioning and detection in the digital IF.
Figure 2.31 shows a picture of a broadband, 4‐port mm‐wave system. This particular unit allows measurements from 900 Hz to 130 GHz, with full S‐parameter calibration as well as other features of modern VNA such as broadband spectrum analysis (see Chapter 8) and noise figure capability. The long test umbilical contains a composite cable that has all the RF and IF cables plus DC bias, DC power, and switch control. The switch control is needed to switch between filter paths in the head to maintain a good level of spurious performance. The length is needed as these systems are often used in on‐wafer applications. Fortunately, because the test port couplers are inside the head, there is good stability even with the long interface cables.
Figure 2.31 A 4‐port, 900 Hz to 130 GHz VNA system using mm‐wave extenders.
Source: Photo courtesy of Keysight Technologies.
2.3 VNA Measurement of Linear Microwave Parameters
In this section the fundamentals of making microwave measurements are discussed for a variety of parameters, along with the consequences of the practical limitations of the RF hardware detailed in the previous section. This section discusses measurement methods and the sources of errors and other complications to making microwave measurements related to test equipment limitations.
2.3.1 Measurement Limitations of the VNA
The systematic error terms (described in Chapter 3) are well known, and many methods are used to essentially eliminate their effects on the S‐parameter measurement results. However, other hardware limitations in a VNA are cannot be so easily removed, and special care must be taken to diminish the effects of these limitations.
2.3.1.1 Noise Floor
The consideration of system noise floor is often not included in the specifications of measurement accuracy, under the assumption that noise effects can be eliminated with sufficient reduction in IF bandwidth or increase in averaging factor. While this is theoretically true, in many circumstances the increase in measurement time makes this impractical. In some situations, such as real‐time tuning of microwave filters, the IF bandwidth must be increased to achieve a real‐time update rate. The noise effects increase 10 dB for each 10× increase in IF BW, so it is a simple matter to compute the effective noise floor at any IF BW given a noise floor at some IF BW, typically 10 Hz.
There exist two distinct noise effects in S‐parameter measurements: noise floor and high‐level trace