Should you trust your VSWR meter? asked an interesting question, and based on experience, including a relevant example, concluded:
The answer is no, like any measurement instrument, prove that it is trustworthy in the intended application.
It went on to ask:
If the VSWR meter is designed to fail, why does it fail?
This article contains an analysis of the analogue circuitry of the IC-7300 directional coupler to explain the likely cause of its poor behaviour.
IC-7300 directional coupler schematic
Above is an extract of the IC-7300 circuit in the area of the directional power coupler used for VSWR measurement. The circuit is a quite conventional Bruene coupler, and its response is similar to several types of directional couplers that produce a DC output voltage from a half wave detector.
The diodes used above are Schottky diodes (though the symbol doesn't indicate so).
Above is the measured response of a half wave detector using a common Germanium diode. Although Vout is almost linearly related to Vin at Vin above 1V or so, the graph shows the response is quite non-linear below 0.3V, especially towards 0V. The response for a Shottky detector is somewhat similar but poorer (contrary to popular ham opinion).
So if this type of detector was used in a coupler that delivered 10V RMS for Pfwd=100W, it would deliver close to 10V DC output from the forward detector. If the VSWR was 1.2, Pref would be 1W, the reflected detector input would be close of 1V RMS and the detector output close to 1V DC.
Now consider the case at Pfwd=10W and VSWR=1.2, Vdc for the forward detector is close to 1V DC, and at Pref=0.1W the RMS voltage will be close to 0.1V and from the curve, you can see that will yield about 0.02V DC from the reflected detector, much lower than the 0.1V needed for a linear response, and so the meter is going to show way too little reflected power and therefore grossly underestimate VSWR (it will show more like VSWR=1.04).
It is possible to linearise the response of these detectors, but the behaviour of the IC-7300 suggests no attempt is made at linearisation, there is none evident in the analogue circuitry and no evidence in the measured performance.
Now we can make inferences about this onset of non-linearity. If the IC-7300 is connected to a load with VSWR=1.5, the meter reads correctly from Pfwd=100W down to not less than 74W, at which point Pref=3.0W.
All bars below 1.9 are inaccurate, and indeed separate measurement of VSWR at 7.14MHz showed it to be 1.6 rather than 1.2 as displayed.
It is designed to fail by depending on the non-linear range of the reflected power detector, and wrapping it with a graphing application does not overcome the problem that GarbageIn -> GargageOut.
This dependence on the non-linear range of the reflected power detector is a common attribute of VSWR meters, and the user needs to validate that the VSWR indicated on a stable load is independent of power level so as to exclude this common trap.
Online advice tends towards advising hams to accept the lower indicated VSWR as the more accurate, and like most online advice, without sound explanation. It is simply a case of looking at the world through rose coloured glasses, believing what you want to believe and denying the less palatable.
The subject of linearisation will be picked up in a following article.