by repositioning the dut to the non-inverting end.
This approach has some severe limitations, as I found out.
As the attenuation goes up with a low ESR part, the gain on the amplifier has to go up significantly to overcome the non-linear region of the diodes. For a reasonably fast part like ne5532, 30x is about its practical top end. That means the input signal to the opamp needs to be 3mv or so if schottky diodes are used in the feedbck loop.
With 3vpp output of the oscillator, and 110ohm on the input of the attenuator, the minimum this approach can reasonably measure is 0.1ohm - I found out that anything shy of 0.5ohm will be over-estimated by the meter.
It can be solved via 1) the use of a fast opamp in the precision rectifier; or 2) additional gain stage(s).
I think a better approach is to go with a passive rectifier (as shown earlier), if you want voltage read-out.
I also experimented with using an instrumentation amplifier to read out the differential voltage from the rectifiers but that has the same issue with attenuation. More complexity but minimum gain.