how to model the large signal model with measurement?

[purian]

large signal model

if there is no large signal model, how can i simulate the power characteristic, for example, the IP3, P1dB, etc. how to model the large signal model through the I-V ,C-V curves ,and also the fitting of the S para of the small signal model, the SDD control in ADS can be used to model the large signal model in some papers, some small signal model para are used too, can anybody give me some advices which with the given device but no large signal model, can i get it with the measurement?how can i do it ?
thanks all!

purian

 

[afz23]

large signal modeling

It is very common problem faced by designers who work in HPA or SSPA design
where a nonlinear or large signal model is a must.

What you get from a manufacturer is linear model of the device,that is nothing
but s-par measured at a fixed point? you cannot generate a large signal model
with this.

Some manufacturers are kind enough to give a spice model which is nothing but
a large signal model of the device.ADS comes with a very rich library of such models
from almost all manufacturers,If you are lucky you may get your device model there itself.
If not then it is a tough life ahead of you.

What you need is s-par measured at multiple bias points,covering almost full I-V
curve ?

For such a measurement you need to have a TRL kit for your device.

TRL measurement can be done in some standard VNA.

Once you are done with this , you need to write a program in ADS
to curve fit the data measured.This program will than estimate the values
of constants of a nonlinear equation,which best resembles the measured data.

 

[Element_115]

how to create small signal model in ads

FYI I just went to an Agilent seminar and they said that they have a new tester
that generates "X-Parameter" for High Power/ Non-Linear devices (Modeling).

It sounds like they are not planing to make these measurement for us, but
will let us buy ( or Rent ) the equipment (~$200K USD, which is worthless thanks to
our Politicians, Both parties) so we can make our own models and ADS can use these
models to do co-simulations with other model types.

Cheers

 

[ChaoFan]

mdlgred

Actually, For Agilent's X-parameter testing equipment, I had a chance to use them. In my opinion, it's not very useful. No big improvement based on our current equipments.

Added after 1 minutes:

Actually, it is easy to know the impedance instead of the large signal model....... If you know the right impedance, it's same as the model.

 

[purian]

small signal sdd model

It is very common problem faced by designers who work in HPA or SSPA design
where a nonlinear or large signal model is a must.

What you get from a manufacturer is linear model of the device,that is nothing
but s-par measured at a fixed point? you cannot generate a large signal model
with this.

Some manufacturers are kind enough to give a spice model which is nothing but
a large signal model of the device.ADS comes with a very rich library of such models
from almost all manufacturers,If you are lucky you may get your device model there itself.
If not then it is a tough life ahead of you.

What you need is s-par measured at multiple bias points,covering almost full I-V
curve ?

For such a measurement you need to have a TRL kit for your device.

TRL measurement can be done in some standard VNA.

Once you are done with this , you need to write a program in ADS
to curve fit the data measured.This program will than estimate the values
of constants of a nonlinear equation,which best resembles the measured data.

there is only small signal model paras, which are only can simulate the s para,i want to make some measurement get the I-V as you have said to fit the s para in multip-bias conditions, but it's not enough, the nonliner effect is mostly characterized by C-V curve in multi-bias, i don't know how to get the C-V curves exactly, and how to fit them, this is the issue, if getting all the small signal paras and I-V and C-V fitting equiations , i think all the simulations about the power performance can be done.the difficult thing is i don't know the details about the whole process, and all the equations about the C-V and I-V are unknown.

purian

Added after 26 minutes:

FYI I just went to an Agilent seminar and they said that they have a new tester
that generates "X-Parameter" for High Power/ Non-Linear devices (Modeling).

It sounds like they are not planing to make these measurement for us, but
will let us buy ( or Rent ) the equipment (~$200K USD, which is worthless thanks to
our Politicians, Both parties) so we can make our own models and ADS can use these
models to do co-simulations with other model types.

Cheers

but the X-para have some problems too, it can't simulate IP3,and harmonic etc, it has not popular used in industry, i don't know much about it.

 

[nccu1990]

how to formulate a signal model examples

If the user want to model the large signal model , I suggest the NMDG solution.
http://www.nmdg.be/start.html

They are the specialist in this nonlinear device area.

NMDG develops and commercializes products and services that ”characterize and analyze beyond S-parameters" the behavior of active high-frequency (HF) electrical components under realistic test conditions. The main focus is on the non-linear behavior of these components.

Added after 4 minutes:

Attachment file is the NMDG solution.

 

[afz23]

creating small signal models

i want to make some measurement get the I-V as you have said to fit the s para in multip-bias conditions, but it's not enough, the nonliner effect is mostly characterized by C-V curve in multi-bias, i don't know how to get the C-V curves exactly, and how to fit them, this is the issue, if getting all the small signal paras and I-V and C-V fitting equiations ,

We have a program modeling software called MDLGRED which is a part of LINMIC
software earlier sold by AC microwaves.
Now this software comes as a part of CST.

If you have access to this you can easily do this modeling, what you need is
the measured s-par at multi bias points,more the bias points covered more accurate the model will be. Just feed this data in MDLGRED ,it willcurve fit the all necessary curves as mentioned by you, and actually show you the whole curve fitting process,
and give you different parameter values of a standard non-linear model for your
device.

 

[jayce3390]

large-signal model

X parameters seem to be very promising for designers, you can directly get a non linear model of your device useable directly in ADS from a pna-x (NVNA option).

It is not popular in industries because it's new and expensive (56000$ only for the NVNA option). I don't understand why purian said it can't harmonics? According to agilent forum you could get harmonics as 27GHz/n.

X parameters combined S parameters(for low level) with non linear characteristics, it seems very powerfull

 

[afz23]

large+signal+model

X-parameter measuring equipment is very latest and expensive thing.

It may not beavailable to all at the moment.

Best is to ask manufacturer to give X-par of the device,as
I have heard that manufactures have started measuring
X-pars of there devices.

Another option is ADS ICCAP, which is a very power tool for
modeling active devices.

 

[purian]

how to get model measurement

i have seen some researchers characterize the large signal model by SDD in ADS, only need the intrinsic components with nonlinear equations and external components as constant extracted by small signal model, which is very useful, but i just don't understand the SDD even there are examples in ADS help, i am confused by the port expressions, does anyone have the detail informations about it ,also the way models the large sigal by SDD, are there good suggestions about the method,thanks!!

What you need is s-par measured at multiple bias points,covering almost full I-V
curve ?

quote]
could you explain the I-V curves at multiple bias points and how to get it, can you give me an example, i think what i thought before is wrong,thanks.