Advice on modelling spiral inductor

[pigkiller]

hi all
anyone who have experience modeling spiral inductor?cmos technologies
multi-metal layers.
i have read a lot of papers,have used 2-pi&1-pi model,but the result is so poor...
who can give me some valuable advice on it,thanks a lot

 

[BigBoss]

asitic pi models

I got some documents on this subject, please wait few days.

I work also on Silicon Substarte Passive Integration.

 

[pigkiller]

Re: inductor modeling?

nice to c u ,can you tell me which structure is more efficient for you?
can you tell me some about your work?

 

[BigBoss]

Re: inductor modeling?

I post 3 thesis here( I don't know they have already been post ..if it is so, sorry )

They include inductor design , transformer design and their modellings.
Very useful and nice to get into introduction of design on silicon substrated passive circuits.

Rgrds

 

[Element_115]

Re: inductor modeling?

Some of my co-workers used Änsoft ĤSFF to model the inductors and imported the models into Cadance. I understand that this is a very good method.

Good Luck

 

[BigBoss]

Re: inductor modeling?

Some of my co-workers used Änsoft ĤSFF to model the inductors and imported the models into Cadance. I understand that this is a very good method.

Good Luck

We use a special library of Cadence called as PICS ( Passive Integrated Components on Silicon ) which is subject to our company. Then we transfom GDS2 file into Momentum and simulate.

In fact, I have personally tried HFSS9 to compare the results ( measurement and simulation ) but unfortunately HFSS results are not comparable with measurements. Indeed, only frequency is shifted but magnitude of the S-Parameters interestingly are same :roll:

Maybe I did a mistake while defining parameters or I have missed some tricky ways..

Essentially, we use this special library to build and create capacitors , resistors etc. All technology depending.
Anyway. For the time being Momentum is very close to measurements with a little bit care of setup.

Rgrds

 

[extractman]

Re: inductor modeling?

hi all
anyone who have experience modeling spiral inductor?cmos technologies
multi-metal layers.
i have read a lot of papers,have used 2-pi&1-pi model,but the result is so poor...
who can give me some valuable advice on it,thanks a lot

Did you optimise the curve-fitting?
Remember the standard flow,

lower band for main inductor value estimated
1st order self resonator freq. determine pararell parasitic capacitance.
The ESR vs. freq. is freq. dependet (Q vs freq also) , dc resistance@lower freq, but ac resistacnce vs sqrt(freq) dependent @higher band

 

[extractman]

Re: inductor modeling?

Some papers helful

 

[extractman]

Re: inductor modeling?

Some papers helful

do a good measurement before modeling extraction!

 

[khouly]

Re: inductor modeling?

try ASITIC , it is free program that can design and analysis the spiral inductor ,
also there a prgram called SISP but it is not free

asitic give pi model and can generate the layout of the inductor and calculate sparameters of it

 

[flyhigh]

Re: inductor modeling?

I did quite a lot of inductor modeling last year, not really on Si but in multilayer environment as well. My experience is that a database built by measurements or EM simulators (Sonnet) can be used to create a Neural Network model. This can be again introduced in ADS trough NeuroADS Patch. It worked OK for my inductors. Generally, you will hardly find any physical-based model, i.e. the one wit lumped L , R and C that can cover relatively complicated behaviour of multi-line inductor. Neural Networks don't care about this, they simply give S pars and it can be very effective.

Regards

FlyHigh

 

[pigkiller]

Re: inductor modeling?

many thanks for your good advice & useful papers.
Neural Network model based on large measured data is a good choice,however,i have not so much measurement result.

i used compare my simulation results based on M/omentum with measurement data. Resonence frequency is shifted,and the phases of S parameters are badly unfit.(about 20 degrees shifting)

There are some problem in my Asitic installtion(Cygwin installed on my PC & worked well,but i can't enter Asitic interface,why?)

which botherd me most are the values of Capacitance&Resistance in substrate included Cox,how can i gets the accurate values?

 

[harkonnen]

Re: inductor modeling?

If have simulated many inductors with S*NNET (2.5D) on GaAs. So far I got good results but it is tricky.
The important part is to account for the metal thickness and get a good estimate of your bulk and skin loss values.
On CMOS you probably have to model the dielectric losses very well.

 

[xlgic]

how to simulate quality factor of inductor?

 

[flyhigh]

Re: inductor modeling?

Determine equivalent series resistance of your inductor and divde this by the reactance. This is Q factor by definition - ratio of energy discipated in inductor to energy stored in it.

You can extract R and L from Spar files by fitting equivalent model to it.

There are two meanings of R depending on how you extracted Spars. When you put your inductor between two ports, and you measure S21 and extract R and L, you are determining the loaded Q, it should be re-calculated to unloaded (inductor itself is loaded with 50 ohm port impedance, this is not true Q).

There is another method to determine unloaded Q directly - from short circuted inductor (one inductor terminal to ground, the other to 50 OHm port termination). Unloaded Q is determined from S11.

flyhigh

 

[yingyang]

Re: inductor modeling?

you may like to review the lastest issue of motl on inductor modeling...

 

[MaxSU]

Re: inductor modeling?

Are there any free (+source) programs to generate equivalent circuits in SPICE format netlist (except asitic)? I'm wonder to get not compact, but extremely accurate models.

Can you help me?

ps. May be some urls?
pps. Do you have such problems?

 

[flyhigh]

Re: inductor modeling?

Hi,

of course you know for the best what you need and have to do, but just to give you some opinion if you might have overlooked one detail. In my previous work it was the goal to have accurate inductor models, but to have them scalable as well. It means that the values of the elements of the equivalent circuit can be easily fitted into set of equation to describe the influence of change in physical dimensions ( track length and width) and number of turns to the value of elements of equivalent circuit. It was our conclusion that extremely accurate models (with more than 3-4 elements) can work for a single set of parameters but it can not be made scalable.

flyhigh

 

[MaxSU]

Re: inductor modeling?

flyhigh, thank you for help
what model do u used then?
are there published material (your material) about scalable model (depends on geometric parameters)?

i'm only begginer in spiral inductor problem

 

[flyhigh]

Re: inductor modeling?

Hi MaxSU,

I did modeling of 3D helical multilayer inductors, not really spiral inductors. My equivalent model consisted of 5 elements. It was well known PI equivalent circuit. L in series with R, this two bridged by C and two Cs to ground on each port.

Although my task was different from yours, I did some reading of modeling spiral inductors, and my feeling is that it is even trickier than my case. As I wrote on early stages of this topic, if you can try neural network modeling it might be a solution for your problem.

flyhigh