For our products, we use AD8429ARZ to amply small signals at different gain, 1 , 5, or 25 using analog MUX ADG1404YRUZ. The noise density is around 5 nV/sqrt(Hz) at gain of 25 with shorting input. I tried to use AD8421 and AD8221 to reduce power consumption and leakage current, but both of them show spikes (about 20~30nV/sqrt(Hz), around 13Hz and its harmonics) in spectrum which was caused by some periodic spikes in time series. I did many tests (including removing MUX) and finally figured out the reason is that external noise is coupled into AD8421 and AD84221's Rg loop area. But AD8429 does not have this issue. If I directly solder the Rg on pin 2, 3, there are no such spikes. If I move the board to less noise environment instead of the test bench with PC, there are no such spikes. During layout, I already paid attention to reduce the loop area of Rg, the MUX is as close as possible to AD8421 footprint. I read all the support cases related to AD8421 at analog.com, it suggested not to use analog Mux to switch Rg, instead, adding some R on both ends of Mux to reduce capacitance effect on pin 2,3, I will do that for next revision. My question is as following: 1. Each board has identical 6~8 channels with identical layout and parts, all the channels with AD8429 do not show such spikes while channels with AD8421, 8221 have problem. why there is such difference? Is there any way to improve my design/layout so that AD8421 can be used?