The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> Linearity/variation tradeoff in LNA design
https://designers-guide.org/forum/YaBB.pl?num=1145603520

Message started by aaron_do on Apr 21^st, 2006, 12:11am

Title: Linearity/variation tradeoff in LNA design
Post by aaron_do on Apr 21^st, 2006, 12:11am

Hi all.

To minimize process and temperature variation in differential LNAs I have been using a tail resistor between the sources of my two input transistors. However there is a tradeoff between IIP3 and temperature/process invariance since the tail current sets the DC current level and hence the maximum current level through either input transistor. I have two possible methods for getting around this...either

1) adjust the tail resistor to get the desired IIP3 and settle for whatever temperature/process variation you get or

2) have two gain modes where the low gain mode is implemented by a separate LNA.

does anybody have any hints on a better method to get around this tradeoff?

Title: Re: Linearity/variation tradeoff in LNA design
Post by ACWWong on Apr 21^st, 2006, 5:56am

if the your differential LNA is a source coupled pair and the resistor is used between the sources and ground, then you should bias your LNA devices as a mirror, with the bias referenced device scaled in size and degenerated with the same (although scaled) resistor. If then the input bias current (to the now mirror biased LNA deivces) is then inversely proportional to that type of resistor, you will maintain the correct tail current over process (r variation) and therfore the linearity improvements.
choice of input bias current temperature coeffiecent can also be optimised to minimise varaition.

Title: Re: Linearity/variation tradeoff in LNA design
Post by aaron_do on Apr 21^st, 2006, 6:20am

thanks for the reply,

I understand how the process variation would be improved using your biasing method. You are saying I should use a scaled current mirror with resistive degeneration so that since the two resistors are matched, any change in my resistance value will oppositely change the gate bias voltage and hence the same current will be maintained right?

However this will not improve the differential mode linearity since the DC current in the LNA will always be fixed. For example if the plus side is maximum and the minus side is minimum, the current through the minus side cannot be less than zero and the current through the plus side cannot be more than the DC current fixed by the tail resistor. Hence the current will be clipped off. The larger the tail resistor, the worse the IIP3.

thanks again for the reply,
Aaron

Title: Re: Linearity/variation tradeoff in LNA design
Post by ACWWong on Apr 21^st, 2006, 11:25am

what one normally would do is to use a pi network of R's in lieu of one R. that is there are three resistors, two of which connect from each source to ground, and one across.

S-Rd-S
| |
Rs Rs
| |
gnd gnd

So Rs is used in the biasing mirror to keep the current constant, and now Rd provides the bulk of the lineraity imporvement (at the cost of noise & gain). Typically Rd would be replaced with a spiral L to give the linearity with less noise (now at the cost of area).
It often helps to transform the pi to t (star - delta impedance transform) to get a clearer idea as to how the values of Rd and Rs contribute to the linearity, this is especially true if Rs and Rd become comparable. At one extreme, if Rs >>Rd (usually where one aims) then the linearity/IIP3 is now dominated by Itail*Rd (in addition to Vod= vgs-vth).

Title: Re: Linearity/variation tradeoff in LNA design
Post by aaron_do on Apr 21^st, 2006, 10:37pm

Yes...I understand your method with the tradeoff between gain and linearity. In fact the linearity trades directly with gm/IDS when a tail current/resistor is used. For example if I lower my device width and increase VOD (gate overdrive) then I can keep my current constant while lowering gm (gm/IDS=1/VOD). As a result the output current would saturate at a higher input level => higher IIP3 (input referred).

Ideally I would like to keep my high gain and improve the linearity while getting good process variation. This might be possible with some kind of amplitude feedback to a tail current, but i'm not familiar with any easy way to do this...

I will keep what you said in mind since it might be the best way to fix the probem.

thanks for the feedback,

Aaron