Changing W of "only" M4 will not keep the topology differential any more. But if you change the width of the active load pair of M3-M4 you are driving them with more gm, or pushing them to low vdsat, which will reduce the ouput impedance of the active load and your gain will decrease.
Changing W of "only" M4 will not keep the topology differential any more. But if you change the width of the active load pair of M3-M4 you are driving them with more gm, or pushing them to low vdsat, which will reduce the ouput impedance of the active load and your gain will decrease.
I think it is wrong. True that assymmetry between M3 and M4 creates offset, but it will be still a differential amplifier. And if you increase the width of M3 and M4 the output impedance of the single ended output won't decrease, because gain is: Au=Gm1,2*(ro1,2 || ro3,4), and ro3,4 depends only from the length of the devices (by channel length modulation), Gm3,4 is not affected.
If you increase the width of M3 and M4 the Gm of devices will increase, but they rather affect the ICMR and output swing, not the gain.
ro ~= 1/(\lambda * I_D), which is inversely proportional to width, no?
I think if you increase width of M3,-M4, gm1,2 stays the same, ro3,4 would decrease, gain would decrease but overdrive voltage and VSD,sat3,4 would decrease allowing a wider output swing.