Initially, the VNA consisted only of some breakout boards of various RF chips that I designed and used to get some experience with RF design. Once the result seemed reasonable good, I decided to built everything into one PCB and create an enclosure for it, this device was the result.
I didn't plan to do anything with this beyond this initial prototype but several people got interested and, thanks to Hugen, development continued. However, this is still just a hobby project of mine. I'll keep improving it as best as I can (and find the time for it) but please do not expect professional level support.
Yes. No matter what happens this design and all future improvements by me will be published on my GitHub account (same as all my other projects).
Yes, an initial batch is available at Aliexpress. Production and sales is handled by Hugen, I am not directly involved in the sales process.
Absolutely. All required files are available and if you are not afraid to solder a lot of small components you can certainly build your own. The total BOM cost (without the aluminium shielding) is about 200€ when only buying parts for one PCB.
There are some basic instructions to help get started on the software side.
You are probably right. Development of the VNA consists of several different parts (schematic, RF layout, embedded firmware, FPGA configuration, GUI application, digital signal processing,...) and I am certainly not an expert in all of them. Please get in touch if you have ideas for improvements (write a mail or create an issue).
There are some advantages to having the second stage mixers (mostly to compensate for limited fractional dividers in the MAX2871) but for most operations, a single stage conversion would probably work almost as well. However, compared to other parts the cost for the second stage conversion is not that high and without it certain functionality would not be available (e.g. spectrum analyzer mode with low RBW). There is a bit longer explanation here.
The actually are RC lowpass filters in front of the ADC but they are not very sharp and even for signals way above the nyquist frequency they only provide a small amount of attenuation. With my combination of final IF and ADC samplerate, proper alias filters are difficult to implement. Thankfully, they are not absolutely required in a VNA (same reason why image reject filters reject filters after the mixers are not required: the device already knows which frequency to look for, because it generates the stimulus signal itself).
However, in spectrum analyzer mode, all these missing filters are a problem and lead to a lot of images that have to be removed by signal identification.
I am using 250kHz as the final IF frequency. Compared to some other VNA projects this is rather high and requires a fast (more expensive) ADC. However, it also makes the sweep rate faster: each measurement point takes at least one IF period (more if noise should be reduced to practical levels).
Why not sample the ADCs directly with (a more powerful) STM32? Or even use the ADC on the µC itself?
- I wanted more than the 12bits provided with by internal STM32 ADCs (and also better SNR)
- Sampling the three 16bit 1MSamples/s ADCs with an STM might be possible but certainly would be a challenge
- More flexibility with the FPGA controlling all critical RF components (e.g. setting up all PLLs concurrently for each measurement point, further increasing sweep rate)
- Enough power for future features (e.g. spectrum analyzer mode with FFT)
- Sorry, thats just what I had experience with (also, the toolchain is free, just not open-source)