You should add a heatsink to the exterior of the LibreVNA. It gets pretty hot after a few minutes and you'll get some pretty nasty temperature drift due to that. I personally measured some TekBox attenuators and the accuracy was very good up to 6GHz with a heatsink. Without a heatsink you get a drift of about 5dB.
Oh I noticed, we were using it for awhile during filming and I could feel it getting very warm. I've never tried doing a frequency test but I think I need to now!
It should perhaps be added that for most applications (and probably virtually all hobbyist applications), simply terminating the third port at the characteristic impedance and then measuring the S-parameters with a 2-port VNA and filling in the S-parameter matrix, should yield sufficient accuracy. There should be no real need to recalculate the S-parameters for the error. The National Physical Laboratory did a study that compared the accuracy of this method against the more rigorous mathematical-correction method. The EVM error they observed was under about 0.02 for frequencies below 50 GHz and under about 0.05 for frequencies above 50 GHz. You can get "close enough" results by simply terminating the ports using termination loads that are closely matched, and then filling in the matrix with the observed 2-port results. Here is NPL's analysis: projects.npl.co.uk/hf-circuits/docs/presentations/20150603-anamet/03.pdf
To be honest I have never used the NanoVNA so I can't really comment on it. But based on reviews of these tools I have seen on youtube, I'm glad I put down extra for the Libre, even if it is not a game changer for client projects. Other thing I really appreciate is the software interface for the Libre, sometimes with lower cost hardware the capabilities of the hardware is not the problem, it's the quality of the software that comes with it. My little 200 MHz Hantek scope is just as capable as the equivalent 150 MHz Tektronix scope for about the same price, but the Hantek software is garbage.
@@Zachariah-PetersonHi Zach. Did you get the chance to check drift etc? I’ve used very high end vna’s and also nanovna but my main issue with cheaper ones is reliability/repeatability especially when you power off the unit and power it back on next day and load file then does the cal hold good for same cal file when you look at smith chart or you see a drift on market positions between low and highest freq point? Thanks
@@jholworld I only use this for demos or test boards, but you're right the cal might be off if you come in the next day. What I do when I need to use it is I first turn it on, let the unit come up to its operating temp, recalibrate it, and then start using it. I've had to do the same thing in the past with spectrometers.
In this situation where the 3rd port shares the same reference (ground) as the other two ports, you do not need to terminate directly, you just need to know the reflection coefficient at that port. Obviously you could terminate to 50 Ohms, but you could also leave open or short it. I calibrated with the open circuit standard for the Libre, which should closely mimic the open circuit I have with a physically open port. It might be more accurate to have terminated the 3rd port with the open calibration standard, but I did not check this so it is still an open question.
I realy, realy love your videos. If the 50R load has a BW larger than 5GHz, would not be better to terminate the unused ports? How about the filters. How are they physically implemented? A schematic of the splitter would be nice. Finally, I believe that NanoVNA is way more popular than LibreVNA (due to the price tag I guess). For this reason, probably more people will relate to the video if the former VNA was used.
Thank you, Yes NanoVNA is much more popular just because of the price, but when I purchased the Libre it had the higher bandwidth so it was worth the price. I've used it on some test boards since buying it with good results. In a real 3-port or 4-port measurement it would be best to terminate with a calibration standard if you are planning to terminate to a specific impedance value (not an open or close) because of the BW requirement as you mention. In my case, I left it open because it's just a demo, but I did check that the open on that connector actually behaves close enough to the open circuit calibration standard terminator so that there is little error. But in general if you are not sure or have not measured that an open on your connector really behaves like your open circuit calibration standard, then you should just terminate with the calibration standard because that standard will (or should) be good up to some frequency limit beyond the measurement instrument's bandwidth.
You should add a heatsink to the exterior of the LibreVNA. It gets pretty hot after a few minutes and you'll get some pretty nasty temperature drift due to that.
I personally measured some TekBox attenuators and the accuracy was very good up to 6GHz with a heatsink.
Without a heatsink you get a drift of about 5dB.
You can also boost the frequency up to 8GHz by using harmonic mixing.
The accuracy is decent between 6GHz and 8GHz.
Oh I noticed, we were using it for awhile during filming and I could feel it getting very warm. I've never tried doing a frequency test but I think I need to now!
What would be required to achieve measurement up to 8GHz@@DaSmik101
It should perhaps be added that for most applications (and probably virtually all hobbyist applications), simply terminating the third port at the characteristic impedance and then measuring the S-parameters with a 2-port VNA and filling in the S-parameter matrix, should yield sufficient accuracy. There should be no real need to recalculate the S-parameters for the error. The National Physical Laboratory did a study that compared the accuracy of this method against the more rigorous mathematical-correction method. The EVM error they observed was under about 0.02 for frequencies below 50 GHz and under about 0.05 for frequencies above 50 GHz. You can get "close enough" results by simply terminating the ports using termination loads that are closely matched, and then filling in the matrix with the observed 2-port results. Here is NPL's analysis: projects.npl.co.uk/hf-circuits/docs/presentations/20150603-anamet/03.pdf
I was not aware of the study but will definitely take a look.
So do you feel the about $600 difference in price from the nanoVNA is worth it? (The Libra is now right at $700)
To be honest I have never used the NanoVNA so I can't really comment on it. But based on reviews of these tools I have seen on youtube, I'm glad I put down extra for the Libre, even if it is not a game changer for client projects. Other thing I really appreciate is the software interface for the Libre, sometimes with lower cost hardware the capabilities of the hardware is not the problem, it's the quality of the software that comes with it. My little 200 MHz Hantek scope is just as capable as the equivalent 150 MHz Tektronix scope for about the same price, but the Hantek software is garbage.
@@Zachariah-PetersonHi Zach. Did you get the chance to check drift etc? I’ve used very high end vna’s and also nanovna but my main issue with cheaper ones is reliability/repeatability especially when you power off the unit and power it back on next day and load file then does the cal hold good for same cal file when you look at smith chart or you see a drift on market positions between low and highest freq point? Thanks
@@jholworld I only use this for demos or test boards, but you're right the cal might be off if you come in the next day. What I do when I need to use it is I first turn it on, let the unit come up to its operating temp, recalibrate it, and then start using it. I've had to do the same thing in the past with spectrometers.
Can you make a presentation for VNA calibration based on v1.5.1?
No promises but I can add it to our list, we have a very long list of topics so it might take awhile to get to this.
Why are you not terminating third port?
In this situation where the 3rd port shares the same reference (ground) as the other two ports, you do not need to terminate directly, you just need to know the reflection coefficient at that port. Obviously you could terminate to 50 Ohms, but you could also leave open or short it. I calibrated with the open circuit standard for the Libre, which should closely mimic the open circuit I have with a physically open port. It might be more accurate to have terminated the 3rd port with the open calibration standard, but I did not check this so it is still an open question.
Great Tutor!
Excelent video, it heps me a lot to understand. Regards.
Hello, where can i fin the file of calibration "SOLT1" beacause it wasn't in my release folder ?
I realy, realy love your videos. If the 50R load has a BW larger than 5GHz, would not be better to terminate the unused ports? How about the filters. How are they physically implemented? A schematic of the splitter would be nice. Finally, I believe that NanoVNA is way more popular than LibreVNA (due to the price tag I guess). For this reason, probably more people will relate to the video if the former VNA was used.
Thank you, Yes NanoVNA is much more popular just because of the price, but when I purchased the Libre it had the higher bandwidth so it was worth the price. I've used it on some test boards since buying it with good results. In a real 3-port or 4-port measurement it would be best to terminate with a calibration standard if you are planning to terminate to a specific impedance value (not an open or close) because of the BW requirement as you mention. In my case, I left it open because it's just a demo, but I did check that the open on that connector actually behaves close enough to the open circuit calibration standard terminator so that there is little error. But in general if you are not sure or have not measured that an open on your connector really behaves like your open circuit calibration standard, then you should just terminate with the calibration standard because that standard will (or should) be good up to some frequency limit beyond the measurement instrument's bandwidth.