limit of on chip resistors matching without trimming

[wxsimon]

Without trimming, could <0.25% mismatch of on chip resistors be achieved?

 

[erikl]

Without trimming, could <0.25% mismatch of on chip resistors be achieved?

Here is an example for a double-poly 0.18µ CMOS process:

High-resistive poly matching parameter AR for the Pelgrom matchinq equation (s. below) is AR=1.5% (related to W and L in [µm]). So with
(ΔR/R) = σ * AR / √(W*L)
... with W=0.72µm (4*min.size) and L=20*25=500µm you can achieve a ΔR/R of about 0.22% for a 3σ-deviation, perfect centroid layout and external dummy protection assumed, of course.