ACWWong
|
It depends... if you are using standard source coupled pair with/without R degeneration to linearise your input then it will obey the 10dB rule more or less and the third order intermodulation distortion will grow fairly linearly at 3dB/1dB until one nears compression. So in this case I only use V1dB as its usually much quicker to simulate and yes V1dB*3.16 will give IIP3.
IF however you have a differnent linear gm structure, then the 10dB rule might not fit, ESPECIALLY under specific signal levels (which is often the case in a well defined system)... a classic example is using the multi-tanh (offset source-coupled pairs) to linearise your input.... in this gm structure the IIP3 is not simply something growing 3dB for 1dB input increase in a linear fashion (although overall trend is correct), it follows a harmonic signature which has growth gradient which varies drastically (ie not 3) depending on absolute input signal level.... infact it also has third order intermodulation minima as well... now in this case you cannot just assume that the IIP3 is 3.16*V1dB at any given V1dB. Please google Gilbert's classic multi-tanh princple to see this specific example of why 3.16*V1dB assumption might be dangerous.
So what I would do is determine in one simulation the IIP3 and V1dB swept over input power to confirm the harmonic signature, then when optimising linearity, you only need to consider V1dB, before finally returning to an IIP3 simulation to confirm you meet your spec.
|