thank you erikl
you always helping me with my layout issue
from your array you mean to say I have to repeat it three times to get my desired ratios.
one more thing, I am thinking that the layout of my OTA becoming a little complicated, do you think so or it is common to have like this?
thank you again
I think it's enough to have a good matching between the input transistors, a very good matching (large W/L, common centroid -- as good as possible -- dummies around to keep the physical environment), if you need small offset voltage. For the mirror transistors, common centroid is not so important, I think, as their matching accuracy is not important for the OTA function. Just keep them close together, and if possible interdigitize the fingers of 2 equal transistors to be paired (load and cascode transistors).
Perhaps you should try and find layouts from other OTAs to study how they are designed. Searching for "OTA layout" in G00GLE images will show you 2 of those in the first row, one of them even including the schematic topology.
thank you erikl
you are right regarding the importance of the input transistors and the mirror transistors. but for me the accuracy of the mirror is never the less is also important as my OTA is current mirror topology.
for the array arrangement, I am restricted in my design to use A,B,C with m=3 and this is why I asked you that I must repeat your suggested array three times. is that possible ?
Thank you in advance
for the array arrangement, I am restricted in my design to use A,B,C with m=3 and this is why I asked you that I must repeat your suggested array three times. is that possible ?
If we make one transistor =3 fingers then it is difficult to share the diffusion or you can say connection with other devices. Otherwise you need to break up the diffussion which increases the parasitic and not a good idea for matching in lower tech nodes.
I would say always make one transistor = Even fingers.
You are totally right, bgangur, but he (Junus, or Senan) told to be restricted to m=3 (for a single multiple transistor, he didn't tell why). So I suggested to use f=3 instead, which will certainly decrease the parasitic capacitance in comparison to an m=3 construction.