Vector network analyzer (VNA) are used to measure scattering parameters of high frequency circuits. When frequency is high enough the reflections of the waves start to matter and distributed effects need to be taken into account. VNA can be used to analyze reflection and transmission coefficients of circuits at high frequencies.
For example ideally antenna would radiate all the energy it gets, but all antennas reflect some of the energy back to the source and only radiate energy at certain frequencies. With VNA amount of energy reflected as function of frequency can be measured. Amplifiers also reflect some energy from both input and output and have some amount of gain. All of which can be measured using VNA.
Unfortunately VNAs are often very expensive and way out of my budget. Newest cutting edge VNAs with very wide frequency band can have insanely high cost. For example starting price of Anritsu’s 110 GHz VectorStart ME7838A VNA is $575,850. Even used VNAs for lower frequencies are often several thousand dollars. At ebay cheapest used 6 GHz two port VNAs seem to sell for about 2,000€, still way more than I’m willing to pay.
Since I can’t afford even a used VNA I decided to make one myself with a budget of 200€, tenth of what they cost used and about 1/100 of what they cost new. Of course it isn’t going to be as accurate as commercial VNAs, but I don’t need that high accuracy and it’s a good learning experience anyway.
So how does VNA measure reflection and transmission of signals? Operating principle is simple, but implementation is more challenging. Theoretically VNA consists of signal source that is used to excite the device under test (DUT), two directional coupler per port that measure transmitted and reflected waves and detectors at the end of the couplers that can measure both amplitude and phase of the signals.
Signal source generates a test signal which is routed to one of the ports. Part of the signal is coupled by the receiver directional coupler and its phase and amplitude are measured. Rest of the signal goes out of the VNA port and into the device under test. Some of the signal is reflected back to the source port and it is measured by another directional coupler. Ratio of reflected power to transmitted power is used to calculate the reflection coefficient of the DUT.
Non-reflected part of the signal goes through the DUT and can either be attenuated or amplified after passing through the device. When the test signal comes out of the DUT, part of the signal is coupled by the directional coupler on the second port and its phase and amplitude are measured. Rest of the signal passes to the termination where it is absorbed. Transmission coefficient is calculated as ratio of received power to transmitter power.
When measurement is repeated with the source switch connected the other way, reflection and transmission coefficients of the DUT can be measured from the other direction.
However in practice measurement isn’t so simple. Biggest difference is length of the transmission lines inside the VNA and cables connecting the DUT causing loss and affection the measured phase. At 6 GHz wavelength on PCB is about 3 cm. For the phase difference between receivers to negligible distances from source, couplers and DUT should be much smaller than that. Especially cables connecting the VNA to the DUT need to be much longer than that so that device can be connected. There is also losses on cables, couplers and transmission lines inside the VNA. Directional couplers aren’t perfectly directional and they couple also some signal coming from the other direction. Source and load matching aren’t going to be perfect and will also reflect some signal back. There are also reflections from the internal components of the VNA. All of the errors also have some frequency dependence.
However situation isn’t hopeless as all of the error terms can be solved from measurements of devices with known reflection and transmission coefficients. When error terms are known, real reflection and transmission coefficients can be solved from the measurements. Usually very accurately characterized short, open and load standards are measured on both ports and through line is used to calibrate the transmission from port to port.
Four receiver VNA
If you want to see a more detailed block diagram of VNA, take a look at for example PNA-X Service Manual N5242-90001. On page 119 there is a very detailed block diagram of the RF parts. Above is a simplified version of the diagram.
Source is implemented using a phase locked loop and often frequency multipliers are used to reach the higher frequencies. To keep the output power level constant as a function of frequency, output power after the output amplifier is measured and attenuator before the amplifier is adjusted until the sensed power is correct.
Power coupled into the directional couplers is high frequency and it needs to be mixed down before it can be detected. Super heterodyne receiver with one intermediate frequency is often used receiver architecture that avoids complications with mixing straight to the DC. In this case the signal exists only at one frequency and this allows setting the intermediate frequency very low, about few MHz, and doing the final mixing digitally. Digital mixing has advantage over analog implementation in that while no analog component can be perfect, digital mixing can be made as accurately as needed. Analog mixers add noise, phases of the LO signals of two mixers aren’t perfectly equal, performance varies as a function of temperature and operating voltage and so on. None of these errors exist with digital mixing and measured result is much more accurate.
While this is a good architecture for making a VNA, it has a drawback of needing many expensive components. 30 MHz – 6 GHz mixer costs about 10€, high accuracy ADCs about 10 – 20 €, fast microcontroller, or better, FPGA is needed to interface to the ADCs, control switches, toggle other signals and communicate with computer. Just these components cost at least 100 € and many more components are still needed such as PLLs, oscillators, filters, PCBs, power converters and so on. Whole board would be way too expensive so something has to be removed to save money.
Single receiver VNA
Most radical way to simplify the block diagram is replacing the receivers with single receiver and SP4T switch. This removes three ADCs, mixers and filters while adding a single switch. Signal processing is also simplified since now we must only measure one ADC instead of four. This change does have some drawbacks. Firstly the SP4T switch isn’t perfect and it will have some leakage between the receiver channels. In theory it can be calibrated out, but it will reduce dynamic range of the measurements. Secondly previously all of the four channels could be measured at the same time, but now only one channel can be measured at once. This increases the time required to measure a single frequency sweep by four times.