?
Just to keep things in perspective, the input VSWR will be a function of the design of the 478A sensor and will also be affected by the performance of the 432A power meter to some degree. However, if a healthy 432A power meter is used and a typical 478A sensor is used that has the same blocking caps shown in the service manual then I think that the input VSWR at 50 MHz is typically going to be about 1.07:1 at 50 MHz.
?
Mine has been 1.07:1 for over 30 years for example. The source VSWR of a typical HP 1mW reference (50 MHz) is spec'd to be better than 1.06:1 and I would expect it to be a bit better than this in reality. The mismatch uncertainty between 1.07:1 and 1.06:1 is about +/- 0.01dB. I've done load pull testing on my Anritsu 1mW source that suggest it has a VSWR much lower than 1.06:1 and in this case the mismatch uncertainty falls below +/- 0.01dB.
?
However, if lots of 1mW references were being measured and they had a spread in source VSWR that covered all phase angles of a 1.06:1 VSWR then this could cause some spread in results for someone trying to get repeatable results below about +/- 0.01dB using DC substitution when comparing a H72 sensor against a standard 478A sensor. But I'd argue that the drift caused by handling the 478A sensor would cause similar spreads unless some time and care was taken when doing the comparisons.
?
In theory at least, adding the matching section would minimise the spread seen with the standard sensor but I'm not sure it is really that worthwhile to chase down errors in the ballpark of +/- 0.01dB unless this is being done at a decent calibration/metrology lab.
?
?
--
Regards
Jeremy
