Diode modeling in HSPICE

[MamadJun]

Hello everyone

My aim is to model my diode. I want to use voltage controlled current source (look up table) in HSPICE parallel to a capacitor. (For the sake of easiness I decided to not use capacitor now). When I generated my HSPICE model by the code below I connected the V+ port of the VCCS to I+ port of the VCCS and V- port of the VCCS to I- port of the VCCS in ADS.

My code for desiging look up table VCCS:
.subckt wowd2 1 2 3 4
gn 3 4 VCCS PWL(1) 1 2
* VOLTAGE CURRENT
+ -99.7042012 -1.00E-05
+ -99.7041755 -9.99E-06
.ends wowd2

The problem is that when I put my model between two probe ( namely, Term in ADS ) and I bias forward the model I expect to see a short circuit behaviour So I expect to see a 50 Ohm in Smithchart but I see an open circuit behaviour in Smithchart. However there is no problem for extracting I-V curve and the model acts perfectly in that manner.

I guess because the output impedance of the VCCS is infinity and that is why the Sparamters just see open circuit even though the model is generating current.

Anyone know how should I model my diode by look up table method without having this kind of problem?

 

[dick_freebird]

What besides simplicity do you expect from your table model?
Why not just fit the diode model from data? There's not that
many params. And that table looks like you have nothing for
forward conduction so you probably have simply a 10Mohm
resistor outside the span of declared voltage-points. That
may as well just be a resistor, no?

If you care about RF-y things such as harmonic distortion
and intercept points, you want the reality of diode nonlinear
behavior to be expressed. Maybe that's for later. But why
waste effort in developing and debugging something so far
short of useful (let alone complete)?

 

[MamadJun]

What besides simplicity do you expect from your table model?
Why not just fit the diode model from data? There's not that
many params. And that table looks like you have nothing for
forward conduction so you probably have simply a 10Mohm
resistor outside the span of declared voltage-points. That
may as well just be a resistor, no?

If you care about RF-y things such as harmonic distortion
and intercept points, you want the reality of diode nonlinear
behavior to be expressed. Maybe that's for later. But why
waste effort in developing and debugging something so far
short of useful (let alone complete)?

I get your point it was simply my curiosity.
However, do u think to extract the parameters ( probably you mean curve fitting ) the how much frequency bandwidth measurement is important?
I am exciting at 2 GHz and loloking till the 5th harmonics So my are of interst is 2 GHZ ~ 10 GHz.
now that I want to model my diode should I also measure low frequency e.g., 100 MHZ ? or should I also look higher frequnecy 20 GHz?

 

[BradtheRad]

A diode adjusts its internal resistance, according to the V/I curve.
That is how it produces a voltage drop when forward biased.

The voltage drop is not a constant. It is greater as current goes up.
I once did a lot of testing with real diodes, to find out more about diode behavior.
I filled in a table of voltage-versus-ampere results.
I did calculations with different equations, to find a curve which fit the data.

For an ordinary silicon diode, the equation which seemed to fit was:

A = ( V * 1.25 ) ^ 20

This is good for a range from 0V to 0.9V.

So the resistance at .5V works out to 6044 ohms.
At .6V it is 189 ohms.
At .7V it is 10.1 ohms.
Etc.

The scalars might be higher or lower, depending on the diode.

This is only a simple formula which I use in modeling a diode. It does not derive from the Shockley equation for diodes.