NanoVNA – how accurate does the LOAD need to be – part 1?

A reader of EFHW transformer measurement – how accurate does the load need to be? asked whether the discussion applies more generally, in particular to the loads used for calibration and measurement with a VNA.

In this article, unless stated otherwise, reference to |s11| and ReturnLoss are to those quantities expressed in dB. Note that |s11|=-40dB is less than |s11|=-20dB. ReturnLoss and |s11| are related, ReturnLoss=‑|s11|.

Measurement 1

As a basis for discussion, let me offer an example measurement.

Above is a scan of a certain DUT after SOLT calibration of the NanoVNA.

Note:

  • the extremely low value of |s11|, quite jittery but averaging around -85dB;
  • reported impedance is very very close to 50+j0Ω; and
  • the reported |s11| and Z at the marker reconcile (making allowance for rounding errors) when Zref is taken as 50+j0Ω.

Experienced practitioners will recognise that the most likely way to measure extremely low |s11| like this is that the DUT is the calibration LOAD component, or one that is of same value within a tolerance of 1 part in 10,000 or 0.01%. In fact, the DUT is the LOAD, let’s call it LOAD1 for reference.

But… there is a serious problem.

Measurement 2

Let’s calibrate again using another LOAD, call it LOAD2 for reference.

Above, a new measurement of the same DUT being LOAD1.

Note:

  • the low value of |s11|, averaging around -29dB;
  • reported impedance is very very close to 47+j1.6Ω; and
  • the reported |s11| and Z at the marker reconcile (making allowance for rounding errors) when Zref is taken as 50+j0Ω.

Analysis of the problem

Stand alone, both displays look consistent, there is not an obvious problem. The reason I could say there was a problem is information that has not yet been presented.

DC resistance of the loads (measured using an instrument with uncertainty of 25mΩ):

  • LOAD1=47.05; and
  • LOAD2=49.81.

For most good loads, the DC resistance is a very good indicator of VSWR at low frequencies (the exception is loads with a DC block).

The implied value of VSWR wrt Zref=50+j0Ω is:

  • LOAD1: VSWR=1.07 (ReturnLoss=29.24dB); and
  • LOAD2: VSWR=1.004 (ReturnLoss=54.41dB).

So, LOAD1 is a bit shabby, whereas LOAD2 looks quite good.

The insidious thing is that for measurement of many DUT, it might not be apparent that calibration based on LOAD1 is poor. Of course for DUT with high ReturnLoss50, the error will be relatively small and will probably not matter, but for lower ReturnLoss50, the error may be significant.

The next part will offer a practical method of approximately quantifying the uncertainty in ReturnLoss.

Conclusions

Without knowledge of the characteristics of the LOAD used for calibration, the measurements of themselves are of little value.

Even if the SHORT, OPEN, LOAD calibration parts are known in theory, poor application (eg dirty connectors, connectors not adequately tightened, unusual temperature of equipment, aging, damage) can result in degraded uncertainty.