low frequency / 9 kHz - 30 MHz Conducted Emissions measurements in spectrum analyzer

[exitrip]

Hello!

I just got a LibreVNA kit to do some WiFi tuning, and I'm generally very impressed with the project!

I've also got some other work doing Conducted Emissions pre-screening and am thinking to characterize the LF to HF spectrum analyzer inputs. I've got the other necessary hardware for these tests, etc... Some quick playing with the VNA suggests it doesn't do anything particular to resolve extra frequency bins at LF (but I really have only used the unit+software once for about an hour so far....) Maybe there's a combination of existing settings that are suitable for LF measurements already?

If not, this post is to ask for quick guidance/help/interest about the possibility of contributing the LibreVNA project to optionally improve measurements (or add a separate mode even) at "low frequencies". I haven't reviewed the schematic or source code yet, so could a kindly neighbor link to parts of the code that implement the sampling control and SA functionality?

Are there any hardware blockers to a higher resolution, low frequency capture mode? I've got lots of experience with firmware, FPGA, and RF circuit design, but not an excess of free time, so any pointers, info, gotchas, or help would be very much appreciated before I dive into the repo.

Cheers,

[jankae]

Hi, I'll try to answer these points but in generally I can not recommend the LibreVNA for such low frequencies.

Some quick playing with the VNA suggests it doesn't do anything particular to resolve extra frequency bins at LF (but I really have only used the unit+software once for about an hour so far....) Maybe there's a combination of existing settings that are suitable for LF measurements already?

There is nothing special going on for LF. In spectrum analyzer mode, the LibreVNA just tunes its LO frequencies to match the RF frequency it is looking for and then samples the ADCs. There are several limitations to the spectrum analyzer:

• low frequency roll off due to 100 nF coupling capacitors
• low frequency roll off due to limited working range of the mixer
• aliasing of other frequencies because it does not use a high 1. IF (see section 2.3.2 in here) and also does not have any preselector filters. I am trying to compensate for this by implementing signal ID but especially for low frequencies the implemented algorithm lets quite a few aliases through
• Specifically for EMC measurements: there is no quasi peak detector (although that could be changed in the firmware)

All in all, the LibreVNA simply was not designed as a spectrum analyzer. If you have nothing else it can be used for some simple measurements but I would always recommend a dedicated spectrum analyzer.

I haven't reviewed the schematic or source code yet, so could a kindly neighbor link to parts of the code that implement the sampling control and SA functionality?

Most of this is a hardware limitation. For very low frequencies, the signal is heavily attenuated along the way to the ADC. There is not much you can do about that unless you change key RF components.

The only point where I see room for improvement is the firmware. Two things are probably not ideal:

• my very simply signal ID algorithm does not work very well for frequencies in the range of 37 MHz to 87 MHz. See this part of the code:
  

  LibreVNA/Software/VNA_embedded/Application/SpectrumAnalyzer.cpp

  Lines 167 to 180 in 3483b7b

  	// depending on the measurement frequency this is not possible or additive mixing has to be used
  	if(freq >= HW::getIF1() + HW::LO1_minFreq) {
  	// frequency is high enough to shift 1.LO below measurement frequency
  	LO1freq = freq - HW::getIF1();
  	LO1inverted = true;
  	break;
  	} else if(freq <= HW::getIF1() - HW::LO1_minFreq) {
  	// frequency is low enough to add 1.LO to measurement frequency
  	LO1freq = HW::getIF1() - freq;
  	break;
  	}
  	// unable to reach required frequency with 1.LO, skip this signal ID step
  	signalIDstep++;
  	/* no break */

• Maybe you could implement the required quasi peak detector but I believe that might need additional processing in the FPGA

Are there any hardware blockers to a higher resolution, low frequency capture mode?

Nothing is blocking high resolution at any frequency (down to 1 Hz resolution because all frequency variables are integers in the firmware). I believe that is already possible with the firmware as it is now. Can you give an example measurement that shows the limitation you run in to?

I've got lots of experience with firmware, FPGA, and RF circuit design, but not an excess of free time, so any pointers, info, gotchas, or help would be very much appreciated before I dive into the repo.

For the spectrum analyzer function, you mainly need to look at these files: SpectrumAnalyzer.cpp and SpectrumAnalyzer.
Part of the DFT processing is also happening in the FPGA and that is implemented in DFT.vhd, Sampling.vhd and Sweep.vhd.

I am afraid that the VHDL code is not exactly high quality. It was my first serious FPGA project and it was mostly learning by doing (I am still not an expert in FPGAs but would probably do some things different now).

To get a high level overview of what the FPGA does, its SPI protocol description is probably a good starting point.

I appreciate any input and improvements here but personally I would not spend the time on improving the LibreVNA for EMC measurements at such low frequencies. The hardware was just not designed for that and I believe the performance will never be really good for this specific application.

[exitrip]

Thank you so much for your quick response and thorough introduction! The hardware limitations you go detail were exactly the information I was looking for.

I look for to getting to know the project better regardless of the fact that it's not a good fit for CE testing.