yeah.... i cant share..

pancho_hideboo
can you do a favor(when u have really free time and relaxed ) ,,, could u check iip3 oip3 in cadence for device biased in class B and tell me how u make the class B obey the text book equations of oip3= iip3+gain , iip3 = 9.6 + p1db as derived in razavi ..whether these equations are valid for class B since its output current is half sinusoid....

I am really stuck

First of all, you have to understand difference between one large signal drive and two large signal drive correctly.
True P1dBin is defined for one large signal drive.
So P1dBin defined under two large signal drive has to be corrected.
True P1dBin will be 3~6dB larger than P1dBin defined under two large signal drive.
I usually use following relation.
P1dBin_One_Input=P1dBin_Two_Inputs+4.5dB

Some people set "Pin_dBm-3dB" for each tone's power in two large signal drive.
This is appropriate if you evaluate correct P1dBin from two large signal drive.
But this is never appropraite if you evaluate IM3 or IP3.
In evaluation of IM3 or IP3, you have to set "Pin_dBm" not "Pin_dBm-3dB".
For example, starter of the following thread are misunderstanding.
http://www.designers-guide.org/Forum/YaBB.pl?num=1262828862/26#26


The followings are summary of my simulation results of Class-B Amplifier.

(1) One Large Tone HB Analysis
P1dBin=24dBm
This P1dBin is true value.

If I roughly estimate P1dBin as Class-A operation from P1dB as Class-B Operation, it will be P1dBin=24-6=18dBm.


(2) Two Large Tone HB Analysis
P1dBin=19dBm
IIP3in=28.275dBm

Here this P1dBin is defined under two large signal drive, so this is not correct value.
So we have to correct this P1dBin value.
In this case, P1dBin_One_Input=P1dBin_Two_Inputs+4.5=19+4.5=23.5dBm.
This corrected value is well matched to result of (1).


(3) One Large Tone + One Small Signal HB Analysis
P1dBin=24dBm
IIP3=27.114dBm


The followings are conclusion for analysis of nonlinear amplifier such as Class-B and Class-C Amplifiers.

(1) IP3 can be defined for only linear system essentially.
You have to use IM3[dBc] for characterization for nonlinear amplifier such as Class-B and Class-C Amplifiers.

(2) IM3 output is not subjected to gradient of 3dB/dB line except for some region of input power level.

(3) IM3 characteristics of nonlinear amplifier such as Class-B and Class-C Amplifiers is affected by not only AM/AM but also AM/PM characteristics.
On the other hand, P1dBin is pure AM/AM characteristics.
Well-known relation of "IIP3=P1dBin+9.6dB" is derived under only AM/AM characteristics.

(4) Null of IM3 is true phenomena which is also observed in actual measurement.
This extremely small IM3 is a result of exquisite balance of AM/AM and AM/PM characteristics.

(5) You can not apply slave Small Signal Analysis subjected to master Large signal Analysis such as HBSS, PSS/PAC, QPSS/QPAC for nonlinear amplifier such as Class-B and Class-C Amplifiers.
Especially you must not apply them for evaluation of IM3.
Again see http://www.designers-guide.org/Forum/YaBB.pl?num=1277203226/3#3

(6) If you use estimated P1dBin as Class-A operation and IIP3 where IM3 is relative subjected to gradient of 3dB/dB line,
IIP3=P1dBin+9.6dB can be satisfied.

In this case, P1dBin=18dBm and IIP3=28.275dBm.
P1dBin+9.6dB=27.6dBm is close to IIP3=28.275dBm.

If I replace true FET with behavioral model FET of Class-B operation like following,
    Idrain=a1*Vgs+a3*Vgs³   for Vgs>=0.0
    Idrain=0.0                    for Vgs<0.0
IM3 of HB2 and HB1SS are coincident and are subjected to 3dB/dB line completely regardless of half sinusoid output current.

This model is :
 - No frequency characteristics
 - Simple AM/AM characteristics having only a1 and a3.
 - No AM/PM characteristcics

Well-known relation of "IIP3=P1dBin+9.6dB" is derived under this condition.

http://edocs.soco.agilent.com/display/ads2009/Amplifier2+%28RF+System+Amplifier%...

IM3 characteristics of nonlinear amplifier such as Class-B and Class-C Amplifiers is affected by not only AM/AM characteristics due to a1 and a3 but also AM/PM characteristics and AM/AM characteristics due to a2, a4, a5, a6, a7,.....etc.
Here "ai" is coefficient of order "i" in input/output relation.

The following is not appropriate.
pancho_hideboo wrote on Jun 23^rd, 2010, 4:15am:

Ah ha, it seems a little "EDA tool play" is instructive sometimes if, of course, one knows what one is doing!

For my case ,in Spectre, if the amplifier is biased in class A and Class AB mode the classical relationship
1)OIP3 = IIP3+Gain
 2)IIP3=P1db+9.6
are fully obeyed .

I used
a)Two Tone PSS  
b)Two Tone QPSS
c)PAC+PSS Sweep
and all of them give same results.

Another strange thing is I am getting is the same values of OIP3 and IIP3 for class A and class AB .


For Class A,VGS=1.2V ( 1 ,2 Obeyed)
P1dB(IN) = 18.6dBm, P1dB(OUT)=26.9dBm,Gain = 8.65
IIP3 =24.18dBm,OIP3=33dBm

For Class AB,VGS= 0.8V ( 1 ,2 Obeyed)
P1dB(IN) = 15.4dBm, P1dB(OUT)=23dBm,Gain = 7.5
IIP3 =24dBm,OIP3=33dBm

For Class  B ,VGS=0.55V( 2  Not  Obeyed, ??1 obeyed ??)
P1dB (IN) = 23dBm , P1dB(OUT)=27.65dBm,Gain = 6.6 dB
IIP3 =12.2dBm,OIP3=17.98dBm

Long story short:
The Spectre includes the IM null while extrapolating the Intercept lines . Since the BIG NULL in Class B makes the amplifier not obey the classical equations..Whereas for Class A , Class AB PA's ,LNA etc IM null is in-significant or near to the amplifier saturation and so the 3db slope is not effected and the classical equations obeyed.

I am attaching pics for Class A and Class B ..The IIP3 , OIP3 are independent of Input Power Level Sweep.
.

I have used higher input power levels also..you can see in Class A(Plot 2)
and Class B (Plot 2)

Also the results are same for PAC & PSS ,Two tone PSS as well..
Only the non-linearity of Class B makes the difference

What extrapolation point should be used in Spectre then?? EP should be low so as to have 3 db slope at low power levels.

I dont have any experience in measurement .

The FET I am using is BPW_N_25_RF from UMC 90nm.
here'my qpss setup in pics..

what to use in place of Direct Form??