How do you Choose the Best Vector Network Analyzer?

There are many choices when you look for a Vector Network Analyzer (VNA) to purchase. What factors should influence this important purchase decision? Engineers tend to focus on electrical performance alone, but many other factors must also be considered. Let’s examine them first.

Physical Factors
At one time, a VNA was large and heavy and took up most of the space on the lab bench. Years ago, it made sense to build a VNA in this way. The technology to miniaturize the directional bridge, the signal sources and the digital processing didn’t yet exist. After the first analog VNAs, second-generation VNAs utilized advanced microprocessors of the time such as the MC68000 from Motorola to perform digital signal processing and provide rudimentary calibration. With the advent of the PC, it eventually became economical to embed a single board computer with a Windows operating system (OS) in the VNA and write the operating software in C, C++ or C#.

This progression made sense, the VNA evolved with technology. Today, the simplest laptop PC is powerful enough to run the processing to apply calibration and display the results. A USB connection has plenty of available bandwidth to transport data from the VNA measurement device. There is no longer any need to embed a Windows operating system with an integrated display in the VNA; In fact, it is a liability. Operating systems evolve and must be updated. Failure to do so can result in an unacceptable security risk. Unfortunately, updating the embedded operating system usually results in a non-functional VNA requiring expensive service.

A modern VNA can have miniaturized components. Integrated signal sources replace expensive multi-loop phase locked loop designs. Up to 18 or 20 GHz, a miniaturized Wheatstone Bridge replaces the large coaxial bridge with ferrites first inspired by Wiltron and others in the 1950s. These innovations have resulted in a robust VNA design with a much smaller form factor. The modern VNA with a USB link to a Windows laptop can be as small as 6.3” W x 9” L x 1.75” H and weigh as little as 3 ¾ pounds. This makes it simple to move the device from one bench to another or carry it in a briefcase or backpack. One such device from Copper Mountain Technologies (CMT) is used by a scientist at CERN who uses his laptop connected to a VNA powered by a lithium-ion battery to perform measurements along the 27 km collider ring.

One other factor to consider which is commonly overlooked. What kind of connector is used on the front panel of the VNA? Is it a simple 3.5 mm female connector or is it something more robust? If a 3.5 mm male to 3.5 mm male test cable is connected between the VNA and Device Under Test (DUT) there will be issues with keeping the front panel connection properly tightened when the attachment is made to the DUT. Torquing the connector to the DUT—with the proper torque wrench—applies an almost equal negative torque to the cable connector at the VNA. If both were torqued equally, there is a significant danger that the front panel connection will be slightly loosened, enough to invalidate all subsequent measurements. This is a disaster and, in some cases, can cause more economic damage to the project than the cost of the VNA itself. A VNA with an “N” connector, and N to SMA or 3.5 mm test cables solves this problem. If a 3.5 or 2.4 mm connector must be used on the front panel, it is best to utilize female NMD versions. This large connector will not un-torque itself.

Figure 1 - SC5090, 9 GHz VNA with N Connectors

Software Considerations
If the operating software is running on the VNA itself, there are potential security risks. Measurements and potentially measurement result files might be stored on the internal drive. Defense contractors are obligated to purge all stored data under certain circumstances and penalties are possible for failure to do so. It is very helpful that the USB VNAs provided by CMT do not store information. All settings and data are erased when the unit is powered down. The attached laptop saves the settings and data and secure procedures for laptops are well established and easy to follow.

How do you get your data from the VNA to your laptop? With legacy VNAs, the user would have to use a floppy disk or a portable USB drive to store the measurement data and use a “sneaker net” to transport it to the computer, where further analysis is done. Some later models sported an ethernet port to share data over a local network. This is exceedingly dangerous if the OS is out of date. Older VNAs running Windows 7 are still relatively common and notoriously vulnerable to even the simplest malware exploits. A USB VNA is the obvious choice for security and simplicity.

Analytical functions such as Time Domain analysis, Time Domain gating, embedding, and de-embedding are often not included with the VNA software and are only available after a significant upcharge. CMT provides these at no extra charge on all VNA models except the lower-cost M model VNAs. Offset frequency measurement, necessary for mixer characterization, is also a standard feature of all CMT VNAs (except for “M” again).

Support Considerations
What is the total cost of ownership for the VNA? Are there critical features that come at additional cost? What is the cost of the annual or bi-annual calibration? Where is the VNA to be serviced if necessary? Is service for this model even available? How long is a typical calibration or service turn-around? These are all critical questions to consider before purchasing a VNA. 

Calibration and repair for CMT VNAs can be performed at either the Indianapolis office in the US or the CMT EU office in Cyprus. Both have ISO-17025 certified laboratories traceable to NIST and all models are supported metrologically. Turnaround time is 5-10 business days. Repair, if needed, can also be done at either office at very reasonable rates.

Some older model VNAs from vendors other than CMT are no longer supported and cannot be repaired at any cost. The up-front cost of a used VNA like that might be low, but if it fails, it is relegated to the recycle bin.

What about customer support? Do I have to pay a recurring subscription fee to receive ongoing support for my equipment? This is common with many VNA manufacturers. Some vendors offer tiered subscription levels based on the number of units owned and monthly payment. CMT offers full support for all models at no additional cost. There is no recurring subscription payment.

Hardware
The ultimate question: what problem are we trying to solve?

A VNA is needed to perform specific laboratory measurements. If one isn’t available, the obvious solution is to purchase one. But what if a noise-figure meter and a spectrum analyzer are also needed? Should I get them all in one box? You could and some VNA manufacturers offer that, but you need to consider whether the additional cost adds significant value. A stand-alone power meter can be moved around easily, but not so with a large form-factor VNA. Does the Spectrum Analyzer function degrade the VNA performance? Yes, usually. Again, isn’t a stand-alone spectrum analyzer more convenient? An engineer using the gear as a spectrum analyzer is also tying up an expensive VNA. If we want to make robust VNA measurements, why not just purchase a VNA?

Copper Mountain Technologies produces only VNAs and VNA calibration accessories. This is the company focus. 

Electrical Performance
Now that we understand the true cost of ownership of a VNA, we won’t be blindsided by extra charges for necessary features and the cost of maintenance and support. We can move on to examining the VNA's RF performance and measurement capabilities.

Dynamic Range
For those who need to measure high-performance filters, the insertion loss (S21) in the passband near 0 dB and deep stopbands as low as -130 dB must be seen simultaneously. This requires a VNA with a dynamic range of 130 dB or more. The setting of the IF Bandwidth (IFBW) directly affects the noise floor and the measurement time for each frequency point. The receiver noise floor goes down with lower IFBW, and the measurement time goes up. Measurement time is approximately 1.3/IFBW. By tradition, all VNA manufacturers specify dynamic range in a 10 Hz IF Bandwidth. Each point will take about 130 mS and a 201-point sweep will require about 52 seconds to complete. (two sweeps for a full 2-port calibration, one in each direction).

Due to DSP fundamentals, the measurement timing for a 10 Hz IF Bandwidth does not vary significantly from one VNA manufacturer to another. However, the dynamic range is due to the design of the receiver and its intrinsic noise floor. For comparison, the 9 GHz Cobalt series has a guaranteed dynamic range of 148 dB. For reference, here are the typical dynamic range for a few CMT VNA models:

Figure 2 - S5243 - 44 GHz Compact VNA

Measurement Speed
As previously mentioned, measurement speed is related to IF bandwidth. A VNA with a 1 or 2 MHz IF bandwidth capability can make high-speed measurements. The 1.3/IFBW estimated measurement time per point does not hold true at these high bandwidths as other factors, such as Phase Locked Loop settling times and band-switching times, come into play. For reference, here are some typical measurement times per point for a few CMT VNA models:

Calibration
Calibration is extremely important. You can own a VNA capable of very precise measurements, but without the proper calibration kit, that precision may never be realized. After calibration, the remaining residual errors determine the effectiveness of the calibration process, and the residual errors are directly related to the uncertainties of the calibration method. A mechanical calibration kit will provide calibration with a moderate level of uncertainty. The absolute floor for reflection uncertainty will be set by the return loss of the wideband load, perhaps 30 to 35 dB. A mechanical kit with a sliding load is much better, followed by a fully data-based kit with all standards characterized by a Touchstone file. Finally, the best calibration is provided by an electronic calibration module (ACM). CMT provides 4-Port ACMs to 20 GHz and 2-Port versions to 44 GHz.

The major contributor to calibration kit uncertainty is the load standard. If the load's return loss is 30 dB (at the highest frequency), then there will be ±3.3 dB uncertainty for 20 dB reflection measurements and ±1 dB for 10 dB reflections, which is not stellar performance. A sliding load can be effectively 40 dB, and an ACM about 47 dB. Reflection measurements such as S11 are limited by the residual directivity, not the receiver noise floor.

Uncorrected VNA Parameters
It is instructive to examine the uncorrected parameters of the VNA. These would be:

• Source Match
• Load Match
• Directivity

All three of these parameters are corrected by user calibration, so they don’t contribute greatly to overall measurement uncertainty in practice. However, more drift over temperature could be expected if excessive correction is required to reach reasonable values.

The uncorrected and corrected values of these three parameters for the model C1209 are:

Noise Floor
The dynamic range of a VNA is the difference between the maximum output power and the receiver noise floor in a 10 Hz bandwidth. For the C1209, with +15 dBm output power and 148 dB dynamic range, the noise floor is -133 dBm in a 10 Hz bandwidth or –143 dBm/Hz. Transmission measurements such as S21 are noise-limited. Therefore, the dynamic range specification is a derivative parameter, with noise floor and maximum output power being primary.

Direct Receiver Access
If power amplifiers are to be measured by the VNA, then Direct Receiver Access (DRA) is a very useful feature. DRA allows the user to move the directional measurement bridge outside the VNA, perhaps between a pre-amplifier and a power amplifier. A direction coupler can serve as a bridge as shown in Figure 3.

Figure 3 - DRA on PA input

Another directional coupler might be used to reduce the signal level seen by the VNA on port 2 as shown in Figure 4. S11 and S21 of the Amplifier may be measured in this way. Other DRA configurations can perform full 2-port measurements.

Figure 4 - DRA Connections for PA Measurement

Millimeter-Wave Capability
Millimeter-wave measurements (above 30 GHz) may be necessary. The S5243 model mentioned earlier covers up to 44 GHz, but some engineers need to measure automotive radar at 81 GHz, and there is activity to 330 GHz and beyond. These high frequencies are attractive for extremely broadband data applications.

It is possible to purchase a VNA that operates from low RF frequencies up to 110 GHz and higher. However, a VNA like this is extremely expensive, and unless there is a pressing need to sweep the entire range in one measurement, it is difficult to justify the expense. Calibration over the entire range is more art than science, and the connectors are delicate and expensive and have a short operational lifetime. If the needed measurement occupies one or more waveguide bands such as WR15, 12, 10, 8, 6.5, 5.1, 4.3, or 3.4, it is possible to extend the range of a CMT 9 or 20 GHz VNA to cover the range. Copper Mountain Technologies is the only company offering millimeter wave extension based on a 9 GHz VNA. This greatly reduces the overall cost of the system.

VNAs with extenders are ideal for engineers designing automotive radar, working in the 60-90 GHz, WR12 band and the solution is much more affordable than a “DC to daylight” broadband VNA.

Automation
A good VNA will respond to Standard Commands for Programmable Instruments (SCPI). For repetitive tasks and measurements performed in conjunction with other test equipment, it is ideal to script the test in a programming language such as Python, LabView, C++, C#, or VBA. A USB-based VNA makes this simple. The UI software runs on the host PC and responds to SCPI commands sent to a TCPIP socket from a test script running on the same or some other PC on the network.

Manufacturing test routines use automation to control the VNA, make measurements, and record the results. CMT offers a “Manufacturing Test” plug-in to simplify this process.

Embedded VNAs
It is entirely possible to embed a VNA into a larger system. Smaller versions of standard VNA products meant to be “bolted in” have found their way into industrial, medical, and agricultural applications.

Conclusion
An engineer has many choices when selecting a metrology-grade VNA for their application. The information given here should help with the decision process. Careful consideration of the total cost of ownership, along with a clear understanding of how the VNA will be used, is essential.