# How to calculate the size for differential amp

In the baker's book (CMOS. Circuit Design, Layout, and Simulation) ,it just teach us how to calculate the gain,but it does't teach us how to calculate the size for a amp,take the differential amp for example,if i want to let all mos be in the saturation region,how do i calculate it,because if the size is wrong,then the amp will not work.

I really want to know,because i just don't want to guess the size,and waste lots of time.Thanks all. • Those books teach you to analyze circuits, designing them is something completely different. When designing / dimensioning circuits the first thing you need to have is a specification. As a designer I need to know: what is: VDD, required gm (Iout/Vin), Bandwidth, maximum current consumption, commonmode range, distortion etc etc. Also very important I want to know the IC design manufacturing process and I need models of these transistors. Only then can I start. I often start with 10u/1u transistors and see what I get. It is lots of trial-and-error and only few calculations. – Bimpelrekkie Dec 1 '17 at 12:40
• So you cannot just calculate all device sizes, there are far too many interactions going on. Even if you get the size wrong the amp can still work, only performance will be different. I suggest that instead of this amplifier, start with a simple current mirror and after that an inverter. Start with a mirror of NMOS both 10u/1u and check bandwith, output voltage range etc. Then make them 10u/0.5u and check what changes. Those 2 transistor circuits will be challenging enough for a beginner. Also, what you learn from those you can apply to other circuits. – Bimpelrekkie Dec 1 '17 at 12:47
• Well,i set NMOS=10u/1u,PMOS=20u/1u,but the gain is from -140dB to -23dB,the VDD is 3.3V,i give bias3 and bias4 for 650mV – Shine Sun Dec 1 '17 at 13:31
• If the gain is 1000,VDD is 3.3V,and the other is whatever – Shine Sun Dec 1 '17 at 13:32
• Using basic models we often use hand calculations for three things: 1) first estimate of sizes (if you know how to go from w/l to gm you can do the same backwards). 2) understanding effects of changes (the math can give understanding into what the effects will be in changing a parameter or what will be best to try) and finally estimates on what is possible (we can plug in the numbers for a spec and see what we find for transistors sizes/currents. For example: in order for us to be able to use a certain architecture we know we need a gain of 100db but find that we need 2mm long transistors...) – Joren Vaes Dec 1 '17 at 16:38