This blog post describes a DIY tool for VNAs that can allow users to get going with easily attaching components or circuits to the VNA and getting usable measurements of the frequency-dependent behavior of them, for the most popular range of things that a VNA does (known as S11 measurements, which can be used for viewing VSWR charts, return loss and so on). Ordinarily this would require 'calibration' tools and techniques to attach the component or circuit without affecting the calibration, which can be quite difficult (at least for me!). It also makes the process more reliable, using a button-press instead of a procedure involving swapping out connections at least four times.
The 11-minute video here describes the project:
The project consists of two boards assembled in this awkward way, unfortunately!
What's a VNA, and What's "Calibration" in this Context?
Vector Network Analyzers (VNAs) allows users to see how components, component networks, and other circuits work at a particular frequency or a range of frequencies.
VNAs connect to the device-under-test (DUT) and then the VNA sends a signal, at a known frequency, and also measures what comes out of the DUT (either reflected off the device, or whatever passes through and egresses from a different port).
The VNA compares the amplitude and phase of the measured signal with the originally sent signal and then represents the information in several ways on charts. The information allows the user to understand many things, such as how efficient a circuit is working, or if a circuit isn't 'tuned' as well to a frequency as might be desired, and so on.
As can be imagined, for such measurements to be accurate, the phase and amplitude that is measured shouldn't have significant errors. However, at high frequencies, even the length of an interconnecting cable between the VNA and the DUT will cause a difference in phase, and that will cause a large error in the measurement. It is clear that connected cables need to be compensated for so that they do not play a role in the measurement of the actual DUT. The procedure is known as a 'calibration'.
There's another issue, in that at the impedances and frequencies usually involved, amplitude and phase measurements need to be very precise. A fraction of a millimeter will cause an error. Furthermore, at high frequencies, a few pF of stray capacitance could cause an error too. The stray capacitance can never be eliminated practically, so it follows that for accurate measurements, that capacitance needs to be compensated for too.
It can be seen that the VNA needs to perform quite a few calculations! Unlike instruments such as oscilloscopes or multimeters which can just dump acquired measurement values to the display (that's of course not strictly true anymore, but it used to be; especially with the more analog 'scopes and multimeters), in contrast, a VNA needs to perform a lot of calculations before anything can be represented on the displayed chart(s). In the past, humans were the calculators, taking measurements from a multitude of instruments before a single result could be calculated (these were really network analyzers for electrical mains power networks across towns/cities, operating at very low frequencies of course).
The VNA is blind without being able to 'probe' the attached cables and connectors, and then performing some internal calculations so that actual device-under-test measurements eliminate everything else (as much as possible) from the final displayed results. It's a very significant part of using a VNA because there can be a dozen or more sources of error that the VNA needs to correct for, based on the probing!
There are many ways to do this, but one basic method is called 'OSL' or 'SOLT' which stands for 'Open-Short-Load' or 'Short-Open-Load-Thru'. It relies on attaching devices on the end of cables that resemble (closely but not exactly) an open circuit, a short circuit, and a 50-ohm load. It is impossible to make it exact, so parameters also need to be entered into the VNA to let it know just how much the devices differ from perfect open and perfect short, to correct for these errors. The parameters consist of coefficients for a function and some electrical length values.
The photo below shows a traditional OSL tool being used with a VNA, to perform the calibration on the end of a cable, so that the device-under-test (DUT) can be attached to that cable afterward.
That traditional tool comes with a file containing the parameters mentioned above, and they need to be uploaded into the VNA.
Modern VNAs such as the FPC1500 now accept a fully automated calibration technique, known as Electronic Calibration. It is a lot more sophisticated. An electronic calibrator contains in-built open/short/load and possibly more references, all electronically switched under control of the VNA (the FPC1500 has a USB port to connect the electronic calibrator). The parameters are automatically shared over the control connection between the VNA and the calibrator.
What can the DIY tool do, and what can't it do?
This DIY tool allows the user to not use any other calibrator, and to be able to directly connect components or circuits to the solderable component location, and get usable results under certain scenarios (in particular, the scenario must not be for very high frequencies).
Currently, with the tool, I am able to connect up to any arbitrary component or network of components, or any circuit, directly to solder pads labeled 'DUT' and get quite usable results up to 1 GHz (I have not tested further).
The photo below shows an arbitrary network (it is a resistor and inductor in series) acting as the Device-Under-Test.
This DIY tool can't replace a proper electronic calibrator, nor a decent manual OSL tool, primarily because I don't have the parameters to supply to the VNA. As mentioned above, the error terms cannot be computed by the VNA without that information for the open and short. However, if the frequency range of operation doesn't exceed 1 or 2 GHz, then some inaccuracy will have a smaller impact on the results than if the results were required at much higher frequencies.
It is possible to compute the parameters, which I hope to do, but that requires some modification to the OSL board that is being used currently, and it may even require rework to the RF Switch board, to improve the 50-ohm transmission lines, for better operation at high frequencies.
RF Switching Unit
This project consists of two circuit boards. The first is the RF Switching Unit, and it is already described here: Building an RF Switching Unit
The second board required for this project implements the Open, Short, and Load 'standards'. I copied a published Texas Instruments design for this. For the Device-under-Test location, I did a Gerber copy-paste of the Open standard and removed a rectangle of solder mask. This is wrong, really the DUT location should be extended a little further. At the time I created this OSL board a couple of years ago, I was only interested in operating to a few tens of MHz for a particular project, and I didn't put as much consideration into it as I should have.
The PCB files are attached to the blog post, ready for sending to any PCB factory.
The board consists of four RF connectors, and two 0402-sized 100 ohm resistors (1% tolerance; it is good to buy a reel of 100 1% resistors, and then use an accurate multimeter (null out the probe resistance first) to select two resistors closest to 100 ohms each).
The photo below shows some components soldered into the DUT location too.
Assembling the Boards
Screw the two boards connectors together : )
Using the Device
This is quite easy. First, copy across the ti102.ckit file (attached to this blog post) into the FPC1500 VNA (you can use the FPC1500 free InstrumentView software for this). The file contains the parameter values shown here (mainly guesses; please share any improvements that you find).
Next, solder on the component(s) or circuit into the DUT position, and then connect it to the VNA port.
In the FPC1500 VNA go to Calibrate->Calibration Kit and scroll down to select ti102. Then, go back one menu step and select Full 1-Port. The VNA will request confirmation to continue:
Next, just follow the prompts on the display : ) For instance in the screenshot below, the FPC1500 prompts the user to connect 'Open' to the VNA port. In the case of this project, the DIY tool has a single button and four LEDs. All that needs to be done is to press the button (hold it down to power on, if it isn't already powered up) until the 'Open' LED is lit, and then press Continue on the VNA.
Once the calibration is done, press the button on the DIY calibrator once more to select the DUT. Your VNA should now be successfully displaying the results. Select the desired view on the VNA, for example, Smith Chart, or VSWR.
Example Scenario #1: Inductor and Resistor in Series
For this scenario, I decided to find an inductor that came with a table of measurements for different frequencies (this is known as an S-Parameter list, or S11, or Touchstone file. S11 means a measurement that uses only one port on the VNA. S12 would be a measurement where the device-under-test is attached to ports 1 and 2).
By having such information, I could plot the expected/simulated measurements, and overlay the actual measurements obtained with this DIY project, to see if the project was working properly or not.
The Smith Chart below shows the simulated and measured results for the series inductor and resistor, for a frequency range of 2 MHz to 1 GHz. There are interval markers on the orange and blue lines at 100 MHz intervals. It can be seen there is a negligible difference up to around 600 MHz, and then there is a slight difference. Higher errors toward the edge of the Smith Chart are to be expected.
Example Scenario #2: Amplifier Input Port Measurement
For this scenario, I decided to use a circuit, in this case, an MMIC Amplifier. Again, this was so that I could take a look at the S-Parameters in the datasheet, to see what the expected behavior was, for comparison purposes.
The photo below shows how it was soldered on.
The Smith Chart below shows more of a discrepancy between simulated and measured, but the manufacturer would have measured the S-Parameters under ideal conditions whereas I used a scrap piece of circuit board from another project to build up the amplifier circuit. I think the difference is not too bad, and this is zoomed in on the Smith Chart (often users are only looking for the measurement to be close to the 50-ohm position, not precisely on that spot).
The DIY semi-automatic Electronic Calibrator is usable up to around 1 GHz but it could be improved further. I think it's fairly usable to at least 1 GHz, and it saves a lot of time compared to manual calibration. As well as that, it saves a lot of time being able to directly solder on components or circuits. It's also very low cost in comparison. The commercial Electronic Calibrator is the ideal tool of course, but a direct test fixture method may be handy from time to time, despite the limitations of frequency range and accuracy.
If you have any comments/suggestions or end up building it or have improvements, please do share them, it would be great to learn how to improve it all.
Thanks for reading!