Improvement of Linearity in Differential Amplifier at 35Ghz

Hello

I am designing for the first time a differential lna at 35Ghz and I have a very bad linearity..The other results seem pretty good . Which are some techniques or tips to get a better result for 1DB-ICP,IIP3? Are some biasing techniques that helps at improvement, some Ic current increase could help?What do you think?

Thank you for your help!

If this amplifier is a classical differential amplifier ( I don't know the details ) the linearity can be improved by adding series resistances to the emitters of the each tail.These resistances will create a negative feedback and the linearity will be better.
There are other more complicated techniques but the first one is to apply a negative feedback.

BigBoss said:

If this amplifier is a classical differential amplifier ( I don't know the details ) the linearity can be improved by adding series resistances to the emitters of the each tail.These resistances will create a negative feedback and the linearity will be better.
There are other more complicated techniques but the first one is to apply a negative feedback.

Click to expand...

I guess you meant add inductors not resistors ?

But he has to be careful since there is a problem with source degeneration method, it reduces your gain.

Yup now it works fine

Ata_sa16 said:

I guess you meant add inductors not resistors ?

But he has to be careful since there is a problem with source degeneration method, it reduces your gain.

Click to expand...

I meant "Resistor".

BigBoss said:

I meant "Resistor".

Click to expand...

If you add resistor you will have dc voltage on emitter. Emitter is preferred to be grounded so I believe we add inductors for source degeneration method.

Moreover, when you add inductor in emitter or source you are also adding another degree of freedom which helps you in input matching

The reactive feedback from an inductor produces a broadband programmable real input impedance that can simplify matching (or even eliminate it altogether).

Also resistance in source or emitter means extra noise (it will be added to channel noise of transistor and will increase input referred noise) so it will degrade noise figure as well.

Ata_sa16 said:

If you add resistor you will have dc voltage on emitter. Emitter is preferred to be grounded so I believe we add inductors for source degeneration method.

Click to expand...

Emitter has no necessity to connect to GND.A differential pair amplifier uses generally a current source and Emitters are to be assumed as connected to "Pseudo GND". But they are not actually Grounded because of this current source.
Source degeneration method is generally used to improve the input matching and linearity is not improved too much by this method because nonlinear Drain/Collector Current vs. Vgs/Vbe so Gm(Vgs,Vds) is still nonlinear even the inductor regeneration is used.
The sole method to improve the linearity is to linearize the Gm and any feedback ( parallel or series or both ) will improve the linearity.
Or, you will eliminate the harmonics in somewhere..

BigBoss said:

Emitter has no necessity to connect to GND.A differential pair amplifier uses generally a current source and Emitters are to be assumed as connected to "Pseudo GND". But they are not actually Grounded because of this current source.
.

Click to expand...

Adding resistor in Emitter means voltage drop and biasing issues. I have never seen LNA structure with resistive source degeneration. I am hearing this from you for the first time.

You are making low noise amplifier and adding extra noise by using resistors what is the point of it ?

Ata_sa16 said:

Adding resistor in Emitter means voltage drop and biasing issues. I have never seen LNA structure with resistive source degeneration. I am hearing this from you for the first time.

You are making low noise amplifier and adding extra noise by using resistors what is the point of it ?

Click to expand...

The emitter ballast resistor is around/more or less 1-20 Ohm, they will not add too much noise but improves the linearity and bias is not impacted for few mA/uA region.