on-chip spiral inductor geometric parameters

[m.n.h]

Hi
am working on the optimization of the spiral inductor and i need a geometric parameter constraints for a real application ..i made a search but all the data sheets i'd found give me only the quality factor ,L,F my be n ..but not the bounds of n,s,w,d..can anyone help me?

 

[volker_muehlhaus]

One geometry constraint is the minimum allowed line width and line spacing. These values depend on the metal layer where you put your spiral, and can be found in the technology documentation (layout rules). Typically, we want to place the spiral as far away from the substrate as possible, so we use the topmost metal layer, which usually has the thickest metal and smallest resistance. For such thick top metal, a typical value for min line width and min spacing would be 2µm. Max line width can be something like 20µm, depending on layer and technology.

The number of turns is a compromise between Q factor and inductor size. Large inductors with few turns tend to be better, but need more expensive chip area. Often, the circuit design will specify a minimum possible Q factor and then the smallest inductor is used which meets that requirement.

To get some real world values, including results, have a look at this book: "Integrated Inductors and Transformers" by Angelo Scuderi et al, CRC Press, 2011, ISBN 9781420088441

 

[m.n.h]

thank u for ur reply

the book u gave me i can't get it how can i get it ?

also i want to say: the process of optimization not difficult matter since i use pso tech ..the first step of put the constraints is the thing that made me stop..i used a specific constraints from paper (without mention to application) :

n 2~ 8
w(μm) 5~ 25
s (μm) 1 ~3
dout(μm) 40~ 400 (for 10nH)

but my supervisor didnt accept it that because the fo was 2GHz and he see it doesnt a practical freq.

 

[volker_muehlhaus]

the book u gave me i can't get it how can i get it

You can order it from a book store, or try to get it through your university library.

the process of optimization not difficult matter since i use pso tech ..

I don't understand. What is "pso tech" ?

The technology and the related stackup (oxide thickness, metal and substrate conductivity) are very important, and will change results.

i used a specific constraints from paper (without mention to application) :

n 2~ 8
w(μm) 5~ 25
s (μm) 1 ~3
dout(μm) 40~ 400 (for 10nH)

The limits for n,w,s make sense to me. The dout range is quite large, an inductor with 400µm size will be expensive due to the large area needed.

but my supervisor didnt accept it that because the fo was 2GHz and he see it doesnt a practical freq.

Note sure what he means. Does he want application specific frequencies like 2.4GHz?

 

[m.n.h]

PSO: particle swarm optimization it is a recently used algorithm i used matlab to write the code, i enter the constraints to it and it give me the values of the four geometric parameters that give the optimal Q..so the difficult point is the constraints..

exactly 2.4GHz or any other freqs. known used in applications like LNA. VCO.filters.etc.

 

[volker_muehlhaus]

PSO: particle swarm optimization it is a recently used algorithm i used matlab to write the code, i enter the constraints to it and it give me the values of the four geometric parameters that give the optimal Q..so the difficult point is the constraints..

This sounds really interesting! I am working on EM extraction of RFIC inductors, and know that many users are interested in synthesis/optimization based on L and Q target values.

https://muehlhaus.com/products/spiral-inductor-assistant
https://muehlhaus.com/archives/1003

What I do not understand yet is what method you use for simulating the inductors. Maybe you can give some details about that?

Regarding your target values and limits, I could send you one or two testcases (requirements) by email <klick here>
However, I can't send the related process technology data, which is confidential.

 

[m.n.h]

i used only the mathlab for optimization(ok i insert the equations of the spiral within the algorithm of the PSO)..then i will need for an em simulator(like sonnet which i should directly enter the values of pso output) to verify my PSO output.

- - - Updated - - -

i used only the mathlab for optimization(ok i insert the equations of the spiral within the algorithm of the PSO)..then i will need for an em simulator(like sonnet which i should directly enter the values of pso output) to verify my PSO output.

 

[volker_muehlhaus]

the equations of the spiral

That's the part that confuses me: we usually need an EM solver to calculate the inductor behavior. The inductance is quite easy to calculate, but series resistance is not so easy (skin effect, proximity effect) and the substrate shunt effects seem quite complicated to do based on equations. So I think there are two challenges: not only the the optimization strategy, but also the equation based calculation of inductor performance.

 

[m.n.h]

right but i have used the commonly used pi-model as simple model just to explain my optimization. i send u email ..did you receive it? ..i wait to send to me the information (test acses)

 

[volker_muehlhaus]

i send u email ..did you receive it?

No, I have not received your email. Can you try again?