Possible for a Class C Cascode power amplifier

[sarge]

Why do you think a cascode would increase the gain of a class C amp?

Cascode transistor allows to decrease miller effect, input capacitance will be smaller with cascode. For gain it is approximately equal gain of common source.
Therefore gain not better with cascode, but attenuation of input signal lower than without it. Or am I wrong?

 

[natnoraa]

Hi mtiweg,

Thanks a lot for your input!

I must agree with you it provides better S12 and of course higher output impedance (if I'm not wrong) which is what I want.
Firstly, power gain has correlation with Fmax. with a cascode structure, a tunable inductor between the 2 transistors has been proven to increase Fmax by about 25% at 90nm tech node on a class A (sadly) PA. I am just bringing the idea across to try it on a class C because the biasing really made the power gain suffer by a lot. I got the gain expansion with a single transistor but it's just not enough.

I have already got the simplest Class C to be working at about 5GHz with a quick simulation. However I'm raising the bar up a few notches as I need it to be done at my frequency of interest. From what I'm seeing, instead of resting at close to zero current, it swings to negative current etc at millimeter wave frequency and the phase shift and parasitic caps are causing problems because they serve as 'leakage' paths for the signal.

Natnoraa

 

[mtwieg]

Hi mtiweg,
I must agree with you it provides better S12 and of course higher output impedance (if I'm not wrong) which is what I want.
Firstly, power gain has correlation with Fmax. with a cascode structure, a tunable inductor between the 2 transistors has been proven to increase Fmax by about 25% at 90nm tech node on a class A (sadly) PA. I am just bringing the idea across to try it on a class C because the biasing really made the power gain suffer by a lot. I got the gain expansion with a single transistor but it's just not enough.

That sounds interesting, could you link to some literature on the topic? I suspect that it would be difficult to implement such techniques in a class C amp since class C operation involves harmonic behavior, not just fundamental. And moving from class A towards class C is always going to reduce your available power gain.

I have already got the simplest Class C to be working at about 5GHz with a quick simulation. However I'm raising the bar up a few notches as I need it to be done at my frequency of interest. From what I'm seeing, instead of resting at close to zero current, it swings to negative current etc at millimeter wave frequency and the phase shift and parasitic caps are causing problems because they serve as 'leakage' paths for the signal.

When you describe transistor current, are you sure you're looking at the actual channel current, or the combined channel current and junction capacitance current? When looking at current waveforms to see if an amplifier is operating in the right class, then the former is desirable, not the latter.

In principle I don't see why a cascoded class C can't work (I've even seen some references to cascoded class E). Have you tried implementing the cascode as a dual gate FET instead of two separate channels?

 

[natnoraa]

Hi mtwieg,

"Algorithmic Design of CMOS LNAs and PAs for 60-GHz Radio", JSSC May 2007. Sorry am a tad lazy and inexperience to put them in IEEE format. It's in fact used in LNA but given the biasing condition, it's effectively in a Class A if we design for high output power.

I have also tried doing a harmonic filter in between the 2 cascode transistors but to no avail.

The cascoded class C works at 5GHz (i've mentioned that i tested and it did work) but when the frequency gets higher, it just does not have enough gain (or even -ve gain) and thus the -ve PAE. Nope I have not tried on a dgfet but the PDK does not have one with a dgfet for rf applications. The problem is that the top FET is always going into subthrehold region together with the bottom FET as the Vd of the bottom FET is high. The supply voltage typically given to the process is only 1.2V to begin with.

Natnoraa