The millimeter-wave (mmwave) frequencies of 30-300 GHz offer incredible opportunities for innovation. As we reach the bandwidth limits of lower frequencies, engineers are looking towards mmwave frequencies to help accommodate the explosive growth of wireless devices. One of the goals of 5G is to decentralize mobile phone transmission from big cellular towers to numerous small hot spots, also known as cells. A single cellular tower can only support so many devices, so increasing the number of cells will relieve mobile traffic as more devices demand bandwidth. Millimeter-wave signals attenuate quickly in the air so multiple cells can use the same frequency without interfering with each other.
This is just one example of how the industry is moving towards higher frequencies. Gigabit WiFi devices, automotive radar, and secure military and aerospace radar will all depend on mmwave frequencies. Vector network analyzers typically have maximum frequencies of only 67 GHz, so frequency extenders are required to test at these higher mmwave frequencies.
Frequency extender modules allow vector network analyzers to characterize devices at frequencies up into the hundreds of GHz. But there are more considerations than just your instrument’s frequency capability when making mmwave measurements. How do you validate that your measurements are accurate and reliable? You need to minimize uncertainty by:
- Minimizing cable loss
- Stabilizing temperature
- DC biasing
Minimizing Cable Loss
Cable loss increases significantly with frequency, as seen in Figure 1. Even a very good cable will lose more than 1.1 dB over 8 cm at 110 GHz and higher, which has a strong impact on measurements. To put that in perspective, a standard 0.5 m cable could lose over 9 dB.
Figure 1: Cable Loss and Frequency
This amount of loss is unacceptable when testing low-power devices. External frequency extenders can be placed right next to the DUT to minimize cable loss at high frequencies. This means that you only need to account for high frequency loss between the extenders and the DUT rather than the entire length between the DUT and the network analyzer.
Figure 2: Frequency extenders placed right next to the DUT, a wafer
Frequency extension with poor temperature stability will bring down the performance of your entire system, regardless of how good your network analyzer is. When measuring wafers with thousands of devices, your measurement equipment will heat up. Smaller instruments like frequency extenders are more susceptible to ambient temperature changes than full-sized vector network analyzers. Higher temperatures agitate charge carriers and create thermal noise. We can see temperature’s direct contribution to the noise power in the following formula:
Where k is the Boltzmann constant in J/K, T is temperature in K, and B is the measurement bandwidth in Hz. In addition to the thermal noise, mechanical effects of heat can introduce measurement errors. Metal connectors expand and can shift when heated. This can lead to impedance mismatch and phase errors, especially at high frequencies. At a wavelength of 2.7 mm it only takes 1.35 mm of movement in the measurement plane for a 180° phase shift.
A measurement setup is only as strong as its weakest link, so frequency extenders ideally have temperature stability on par with high-end VNAs. Rugged, high quality connectors and convection cooling help mitigate thermal errors. Figure 3 shows how a measurement setup featuring frequency extenders with good temperature stability (blue) compares to a setup with more temperature drift (red) after 8 hours of measurement.
Figure 3: Drift Impact on Calibrated Measurements
Many RF devices are active, meaning they require a DC bias to operate. Some frequency extenders have an internal bias tee which adds a DC signal to the mmwave test signal. The equivalent circuit in Figure 4 shows us how this works. The inductor blocks AC and the capacitor blocks DC so the inputs cannot interfere with each other. The output contains both a DC bias and an RF test signal.
Figure 4: Bias Tee Equivalent Circuit
Variations in bias current can lead to variations in measurements. The DC bias is susceptible to small leakage currents and the farther the bias is from the DUT, the more current will leak. Placing the bias tee within the frequency extender brings it as close to the DUT as possible, limiting current leaks to provide a consistent bias.
In Figure 5 we see several measurements of drain-source current vs gate-source voltage on a FET. Zooming in on the measurement reveals that there are actually different I-V curves. This is due to ground current leakage varying between measurements. Keeping the bias close to the DUT helps minimize errors like this.
Figure 5: FET I-V Curves
As we’ve seen, the frequency extension modules have a few ways of helping your mmwave vector network analysis.
The small size of the modules allows you to bring the measurement to the device. This reduces cable loss between your DUT and your instrument by minimizing the cable length the mmwave signals need to travel. The modules also minimize temperature drift errors with precision hardware and temperature regulation. Finally, the modules are able to bias active devices with a built-in bias tee.
The N5295AX03 frequency extender module uses all of these advantages to make accurate continuous sweeps from 900 Hz to 120 GHz.