The article proposes an improvement to the VK3AQZ RX1 Noisebridge.
The RX1 bridge uses inexpensive components and one of the shortcomings is the linearity of the potentiometer used for the variable resistance element of the measurement bridge. Fig 1 shows the relationship between measured resistance and the logging scale, which is essentially rotation of the shaft as a percentage of maximum.
The concerns are the data points at the very extreme, they depart from otherwise fairly good linearity over most of the track. In my experience this is common, but worse, the extent and slope of the non-linear region varies from pot to pot.
The departure at the low end is of greatest concern as one often wants to make measurements of impedances with very low R, eg stubs, inductors, capacitors.
Statistically, the Standard Error of R in the linear fit in Fig 1 is 8.2Ω. By fitting only over the 6% to 95% range, the Standard Error of R is reduced to 0.4, so R can be predicted linearly from the logging scale with much less error by using only the 6% to 95% range of the pot.
More importantly, the large error at the low end of the scale can be improved dramatically.
Fig 3 from the kit instructions shows the basic bridge circuit. If we want to offset the operating range of the potentiometer, we need only add a small resistance to the unknown leg of the bridge, between the 'Series C' and unknown socket in the above circuit.
Studying the measurement data suggests that adding 10Ω in series on the unknown leg of the bridge will offset the potentiometer to that it is on the substantially linear part of the track at 0Ω unknown load.
Fig 4 shows the improved fit to the data points at the low end by adding the 10Ω resistance to the unknown leg of the bridge. Zero ohms corresponds to 5.3 on the logging scale, and the resistance is given by the expression R(Ω)=-32+6.015*Dial .
The resistor was added to the noise bridge between the PCB and the unknown jack, see Fig 5. The scale was repositioned on the potentiometer shaft so that it properly indicated null on a 50Ω load, and the bridge recalibrated. The existing R scale is about 5% too long for my pot, so I produced a new linear Noise Bridge R scale which reconciles well right down to 0Ω with the displaced zero due to the 10Ω resistor.
© Copyright: Owen Duffy 1995, 2017. All rights reserved. Disclaimer.