On 8/4/22 10:24 AM, 0root via groups.io wrote:
What do you guys make of these calibration boards?
Accurate enough to give a result with <3% differences?
As noted elsewhere, the basic measurement uncertainty is around 1%, so if your cal standard uncertainty is 1%, then you're in the right ballpark.
But it *really* depends on what you're measuring.
Consider measuring a 40 dB attenuator, and you care about S21 accuracy. The NanoVNA can easily do <1% measurement uncertainty of a 0dB attenuator, but the uncertainty of a signal that is 1/100th the amplitude will be bigger. The signal is smaller, but the noise is the same.
If you look at uncertainty curves for Keysight VNAs, you can see how the uncertainty varies as a function of the return loss (S11) or attenuation (S21)
See page 15
You can see that for S21, at lower frequencies, the uncertainty in magnitude is 0.1 dB (about 1% for linear magnitude) for "low loss" components, but as the loss gets bigger (-70dB) the uncertainty starts to climb. That's what the NanoVNA can do, but it will start to climb at lower attenuations.
It's even more obvious in the S11 (Reflection Uncertainty). For a big mismatch (== lots of reflected power), the uncertainty is low (fractions of a dB), but as the load gets to be higher quality, the uncertainty climbs. A load with a -40dB S11 has an uncertainty of 3-4 dB (40%).
That's with -15dBm stimulus (comparable to the NanoVNA) 10 Hz IF bandwidth (500 times narrower than NanoVNA - that improves the SNR by 27 dB). So if the FieldFox does pretty good to -50dB S21, then the NanoVNA can do almost as well up to -25 dB, but then it will start to degrade.
The FieldFox also probably has a *better* calibration algorithm (more terms, etc), so that helps too.
keysight even offers a calculator for determining the uncertainties - it might be usable with other VNAs
