exp
Full Member level 1
I have a simple MOS diff pair with MOS loads and self biased output common mode in a 28nm process.
For example: input pair NMOS, loads: PMOS (gates connected to drain via high resistance for setting the common mode).
Linearity is important for my application.
However, all linearity analysis and discussion I find is related to the drain current nonlinearity (ids/VGS). If I ac ground the outputs (i.e., there is no swing and hence no gain) these expressions match pretty closely with my simulations. However, gain is also very important and without AC ground the linearity is orders of magnitude lower than expected (e.g. -40 dB!).
I suspect that the "traditional" ID/VGS nonlinearity is completely swamped by the ID/VDS nonlinearity (i.e., the dependence of ID on VDS).
I am wondering why I seem to be the one person in the world carying about this ... or why there is no single word in literature discussing this.
The problem is that I do not have an analytical expression and the linearity seems to be completely "arbitrary". There is no obvious dependency on GM/ID, ID, VDS etc. It is probably a complicated function of all the parameters.
But how would you design such a simple diff pair if linearity is a concern and the VDS nonlinearity is so dominant?
As a reference, with ac ground, the drain current nonlinearity follows a nice relationship with GM/ID which becomes better with low GM/ID and worse with high GM/ID (IIP3=+12 dBm for GM/ID=5 and IIP3=-5 dBm for GM/ID=24).
With the loads, I cannot get below IIP3=-13 dBm and IIP3 shows no clear relationship between GM/ID, VDS, ID etc.
Thanks!
For example: input pair NMOS, loads: PMOS (gates connected to drain via high resistance for setting the common mode).
Linearity is important for my application.
However, all linearity analysis and discussion I find is related to the drain current nonlinearity (ids/VGS). If I ac ground the outputs (i.e., there is no swing and hence no gain) these expressions match pretty closely with my simulations. However, gain is also very important and without AC ground the linearity is orders of magnitude lower than expected (e.g. -40 dB!).
I suspect that the "traditional" ID/VGS nonlinearity is completely swamped by the ID/VDS nonlinearity (i.e., the dependence of ID on VDS).
I am wondering why I seem to be the one person in the world carying about this ... or why there is no single word in literature discussing this.
The problem is that I do not have an analytical expression and the linearity seems to be completely "arbitrary". There is no obvious dependency on GM/ID, ID, VDS etc. It is probably a complicated function of all the parameters.
But how would you design such a simple diff pair if linearity is a concern and the VDS nonlinearity is so dominant?
As a reference, with ac ground, the drain current nonlinearity follows a nice relationship with GM/ID which becomes better with low GM/ID and worse with high GM/ID (IIP3=+12 dBm for GM/ID=5 and IIP3=-5 dBm for GM/ID=24).
With the loads, I cannot get below IIP3=-13 dBm and IIP3 shows no clear relationship between GM/ID, VDS, ID etc.
Thanks!
