are circuits that are driven by a low frequency pulse, which has very high harmonic content. The example circuit shown in Figure 2.21 is typical of older VNA sampling receivers such as found in the HP8753 or HP8510 VNAs. In the circuit, the diode pair acts like a switch, which is driven with a very short pulse from a pulse generator driven by a voltage‐controlled oscillator (VCO), operating at relatively low frequencies. The short conduction angle (the amount of a cycle that the diodes conduct) of the pulse means that the frequency content is very high, sometimes referred to as a harmonic‐comb, and the sampler can convert frequencies much higher than the VCO drive.
Figure 2.21 Schematic of a sampler.
It is not unusual to use harmonics of up to 200 times the VCO frequency. Because the conduction angle is so short, the effective input impedance is high, essentially multiplied by the maximum harmonic number, and this means the effective noise figure of the sampler is high as well. Since the conduction angle does not depend upon the frequency being measured, the noise figure does not depend upon the frequency being measured. The conduction angle and conversion efficiency can be adjusted by adjusting the diode bias so that they are not quite turned on, and the pulse provided by the VCO gives the extra current to turn on the diodes fully.
Some advantages of samplers are that they do not require a high frequency local oscillator, as a mixer would, and they can simultaneously down‐convert a signal and all its harmonics to the intermediate frequency (IF). This capability is exploited in a sampling oscilloscope, and now in some of the modern versions of sampling VNAs used for non‐linear measurements. However, for the most part samplers have been largely abandoned in VNAs due to a number of difficult problems that they present.
The foremost problem is the degraded noise floor in the sampler‐based VNA. The effective noise floor is further reduced as the conversion efficiency of higher‐order harmonics typically degrades near the top of the sampler frequency range. Almost all modern VNAs use some form of mathematical response correction on the sampler response so that frequency response of the VNA receiver to a constant input power appears flat over its entire frequency range. This response correction, which removes the effect of roll‐off in real conversion loss, has the consequence of increasing the apparent noise floor of the sampler at higher frequencies.
A secondary problem with samplers is that the rich harmonic‐content of the VCO dictates that the sampling receiver has many regions where it is sensitive to other signals, such as harmonics or spurious signals of the VNA source or DUT, and spurious signals present at the DUT output. This makes sampling receivers particularly poor at measuring mixers or frequency converters, where the sampling comb‐tooth can cross mixer output signals at many different frequencies. Figure 2.22 shows an example of spurs from a source generating responses in a sampling receiver. In this case, the source signal is generated by mixing a 3.8 GHz fixed RF signal with a 3.8–6.8 GHz swept yttrium‐iron‐garnet (YIG) oscillator LO. The mixed product provides the desired 0–3 GHz source output, but spurious signals at 2 · RF − LO and 3 · RF − 2 · LO do show up in the 0–3 GHz VNA measurement receiver band. While small, these spurious responses do degrade the S21 accuracy.
Figure 2.22 Spurs from a source crossing a harmonic of the VCO.
Another difficulty with sampling mixers becomes apparent when measuring filters in the stopband and is caused by remixing signals reflected off the DUT back into the input reflection receiver such as the b1 receiver. Because this effect has the appearance of a signal bouncing off the input reflection of the filter stopband and then bouncing off the b1 mixer (at a different frequency), it is sometimes called sampler bounce (or mixer bounce in the case of mixers). Designs of these components must be carefully considered to avoid these bounce signals, and the basic design of samplers make them especially susceptible to this particular type of crosstalk.
For these reasons, and the fact that creating wideband mixers with full‐band RF‐frequency LOs has become much more cost effective, the use of samplers in VNAs has been phased out.
2.2.5.2 Mixers
Almost exclusively, modern VNAs use mixers as VNA receivers. The mixers are driven by a fundamental or low‐harmonic‐order LO. This provides a much larger conduction angle for frequency conversion and as such provides a lower noise floor. The trade‐off, of course, is the cost of the LO, which may be due to its increased complexity.
2.2.5.3 Noise Floor
With mixers used as VNA receivers, the critical performance attributes are noise floor and input compression point. A fundamental LO provides best performance; using a third‐order harmonic of the LO to obtain higher RF frequency response theoretically degrades the conversion loss by 9 dB, and fifth harmonic degrades it by 14 dB, based the idea that the LO drives the mixer into square‐wave conversion. This degradation in conversion loss is represented as an increase in noise floor for most VNAs, due to the internal factory‐based response correction.
2.2.5.4 Spurious Responses
Mixers have much lower unwanted spurious responses compared to sampler‐based receivers. The primary spurious (or unintended) conversion occurs at the IF image of the desired RF signal. If the RF is above the LO (low‐side mixing), the image will be one IF below the LO (hi‐side mixing). Because mixers used as VNA receivers have lower noise floor and lower spurious responses, a much wider class of measurements that can be performed including noise figure, two‐tone IMD, and even some modulated measurements.
2.2.5.5 Phase Noise
The LO distribution attributes are critical in the performance of mixer‐based systems. The phase noise of the LO contributes directly to errors (primarily trace noise) in the phase of the measured signal. Even the trace noise of the amplitude response can be degraded by phase noise as the signal moves across the IF bandwidth filter. Providing a common and coherent LO to all the mixers improves the trace noise of the measurement by allowing the effects of LO phase noise to be reduced when the measured parameter is a ratio, such as gain or return loss.
2.2.5.6 Isolation and Crosstalk
Isolation between VNA receivers is important when measuring high‐dynamic‐range devices such as filters. In almost all measurements, the reference path has a large signal (as it measures the incident wave) and so is a constant source of leakage signals. Partly for this reason, it is common to provide additional loss in front of the reference channel mixer (5 or 10 dB) to lower the incident signal level and provide greater reverse isolation for mixer bounce. This also helps to keep the reference channel mixer operating in its linear region for higher source power signals, avoiding compression in the reference mixer.
In fact, there are four primary crosstalk paths in VNAs, listed here:
1 RF signal from the internal source or reference receiver to the transmission test port. This crosstalk signal is independent of the DUT characteristics, and its level remains constant regardless of the DUT properties but will change with test frequency. Leakage of the reference mixer IF to the test mixer IF path has similar characteristics, but its value doesn't change with frequency.
2 RF signal from the reflection test receiver (b1) that leaks to the transmission test receiver. This signal depends upon the input reflection of the