Even at DC an oscilloscope has an accuracy in the 1.5% to 3% range which make an oscilloscope useless for accurate amplitude measurements (unless you have ways to calibrate it and control drift).
Hi res Picos are 1% at DC (my ancient 212/100 is 1%) and PicoScope 4262 is 0,25% to 1% depending of range.
But DC accuracy is not so important here, but linearity. Also it is a relative measurement when using Pico FRA software.
But I think Andreas is referring to a fact that because of TrueRMS converter used in HP34401A, there is a dead band around +5mV/0/-5mV where measurements are not even defined and are highly non-linear. With 20mV signal you might get 10-20% error...
http://www.gellerlabs.com/34401A%20AC%20zero.htmNew 34465/70 has direct sampling (TrueVolt AC) and are superior (within frequency response range of, course)
But I realized that, for instance, my Pico 212/100 gives better results in true RMS than old school thermal RMS meters, and has 50MHz bandwidth...
I use it for noise measurements.
Regards,
Sinisa