LibreVNA performance

[LogicalDave]

Can someone who owns a LibreVNA and has the lab-grade gear to evaluate its performance post a review?  I'm hoping to understand things like:

• Signal generator - calibrated frequency and amplitude accuracy (i.e. repeatability)
• Signal generator - harmonic and other spurious suppression.
• Spectrum analyzer - calibrated frequency and amplitude resolution, accuracy
• Spectrum analyzer - resolution bandwidth, max span/points, sweep rate
• VNA - dynamic range, calibrated repeatability, etc.
• General - how does performance degrade with frequency

I have a lab-grade SigGen and SpecAn as well as several field-grade VNAs; what I'm interested in is whether the LibreVNA is good enough to serve as a general-purpose RF bench tool (when bench space is limited).  There have been lots of prosumer RF tools coming out of recently: the TinySA is great for doing sub-GHz site surveys, the TPI-1002 serves many good purposes on the bench, and the NanoVNA V2 is serviceable for antenna tuning and such, but I have yet to find a good general purpose RF bench tool (SigGen, SpecAn, VNA) that is reasonably accurate, repeatable, and has good resolution and adequate filtering.  My application is mainly sub-GHz (860-930MHz) narrow-band FM.  Is the LibreVNA is the real deal?  Any comments on the strengths and weaknesses would be much appreciated.  Thanks!

[joeqsmith]

The designer as well as a few users are members here but I would join their mail list and ask there.     You may also want to include your minimum requirements.  I would also compare those with the published specs before you ask.  You may find a deal breaker.   

[jankae]

Well, I am obviously a bit biased but I'll try to answer these as honestly as I can:

Signal generator - calibrated frequency and amplitude accuracy (i.e. repeatability)

You (as the user) can calibrate the frequency as well as the amplitude. Frequency calibration might not be necessary, the TCXO has a tolerance of 1ppm. Amplitude calibration is definitely necessary if you require accurate output levels (without calibration they can be off several dB). You can calibrate the amplitude at up to 64 frequencies, it will be interpolated between the calibration points). Repeatability on my device after several months is better than 0.7dB (amplitude is still spot on compared to my SSA3021x which has an amplitude accuracy spec of 0.7dB)

Signal generator - harmonic and other spurious suppression.

Not good. There is a bit of filtering but you can expect a lot of harmonics. It is okay if you just need "any" signal at a specified frequency but it certainly is not a clean signal source.

Spectrum analyzer - calibrated frequency and amplitude resolution, accuracy

Same as with the signal generator. Both amplitude and frequency calibration is available, amplitude calibration is necessary to get good results.

Spectrum analyzer - resolution bandwidth, max span/points, sweep rate

RBW from 13 Hz to 112 kHz, maximum span is 6 GHz (but unreasonable slow at that point), number of points is fixed at 1001 (but this could be changed in the source code). Sweep rate is generally too slow for wide spans (e.g. 8.5 seconds/100 MHz at highest RBW). At low RBWs a DFT acquisition mode speeds up the sweep significantly:
10 kHz span with 100 Hz RBW: with DFT 1.7 seconds, without DFT 130 seconds.

The SA also includes a tracking generator option with a configurable frequency offset (useful to measure mixers).

VNA - dynamic range, calibrated repeatability, etc.

Fairly good between 10 MHz to 3 GHz. Minimum specified dynamic range is 90 dB for these frequencies but with narrow IF BW and averaging, most users actually report something between 110-120 dB on S21. Dynamic range decreases sharply at around 3 GHz, you can expect 50 dB dynamic range up to 6 GHz (better at most frequencies but there are some bad spots in there, so this is what I specified).

You have to keep in mind that the LibreVNA was primarily designed as a VNA. The signal generator/spectrum analyzer modes are sort of an afterthought. The hardware is already there so I thought it would be nice to include these functions in the software but it isn't optimized for them so you can't expect the best performance.

[LogicalDave]

Thank you so much for the clear, comprehensive, and quite unbiased response! I think I have a better understanding of its capabilities now.  I am very grateful to you and the others who are bringing lab-grade VNAs into the hobby price range; congratulations on designing such an impressive device.

One of the things I think is desperately needed in the hobby community is a *good*, affordable USB spectrum analyzer.  I've been working from home for much of the pandemic and it has also made me acutely aware of the importance of bench/desk space  

Although it sounds like LibreVNA is not really suitable for my spectrum analysis needs (no attenuator or image rejection), it still sounds far better than most other inexpensive options (most low-cost spectrum analyzers have very high RBW).  The signal ID feature is nifty too!

I use a VNA once in a while, but I use a spectrum analyzer very frequently, maybe one of these days you'll follow up the LibreVNA with the LibreSpecAn?

Either way, thank you again for your response and for the impressive product!

[bjornsh67]

Hi Jankae,

This seems to be an interesting option for doing VNA labwork. I have two old HP model VNA and these work good. But they are not so easy to move around.

Can you share more on how you have implemented 2 port calibration and if you support various test kits?

Bjørn

[jankae]

Hi Bjørn,

sorry for the late reply, I don't check this forum regularly.

Can you share more on how you have implemented 2 port calibration

Different calibration types are available with the LibreVNA but I assume you mean the standard 12-term SOLT calibration. I used the formulas from here: https://www.rfmentor.com/sites/default/files/NA_Error_Models_and_Cal_Methods.pdf
Other calibrations you could use are port 1/2 reflection only (requires only 3 measurements instead of the 7 or 8 for the SOLT) or a simple normalization (useful for quick and not so accurate through measurements).
I have also implemented a basic TRL calibration but since the LibreVNA only has 3 receivers (instead of the usual 4 the professional VNAs often have), only 8 of the 12 calibration terms are actually independent of each other in this case.
If you really want to know the details and understand C++ code, you can always check the source: https://github.com/jankae/LibreVNA/blob/master/Software/PC_Application/Calibration/calibration.cpp

and if you support various test kits?

Yes, different calibration kits are supported. I suggest you download the LibreVNA-GUI and have a look yourself. Without the hardware, you won't be able to actually apply a calibration (since you can't take the calibration measurements) but you should be able to see which calibration kit parameters are supported. I can also extend the software if a specific feature is required (as long as it takes a reasonable amount of work).

- Jan

[Feynman]

Does the LibreVNA support logarithmic frequency sweep?

[jankae]

Yes, it does. You can chose between linear and logarithmic for both the actual sweep and the representation of the data in the graphs (both settings are independent of each other, but most useful if the same is selected for both). Feel free to download the LibreVNA-GUI and look around.