S parameter simulation of PGS inductor from momentum and spectre don't match?

[howto]

Deal all,
I am doing S parameter simulation of a smic 40nm PGS inductor in ADS momentum, But the result of the momentum don't match the S parameter obtained from cadence spectre which uses the model provided by the smic foundary,in fact , they differ much,I did many experiments, but still can't get a matched result, Can anyone help me, what may be the cause? thanks . Below are the results from momentu and spectre respectively, for obvious, I put them on the same matlab figure. Attached is the slm file i use in momentum, for they are about 40 layers in smic 40nm technology, but momentum can't tackle so many, So I averaged the dielectric layers between every two metal layer with a mean permitivity(Er).
]CONFIG]86071[/ATTACH]

 

[kspalla]

Average leads inaccuracy.
No pain no Gain

 

[howto]

But if I use the slm file with complete layer stack to simulate it, even if run it on a server, convergence problem or memory error will occur. Any way, weighted average should not bring so much difference, I once matched the TSMC 180nm inductor with spectre with the complete layer stack and averaged layer stack, they hardly have any difference, So, it is not likely caused by the averaging. The valid model simc provided for the inductor reaches 10GHz in our case, but the difference can reach 50% in this range. Thanks for your reply anyway!

 

[kspalla]

take few layers that really contribute on performance.
Let say top layer is inductor, then take till it see perfect gnd let say on 4.
though you have 12 layers but effective inductance is definded by the top 4 layers only.
Agreed?

now judge yourself and do not average till you feel the layer is contributing towards H.

Hope this helps you.

 

[howto]

Yes, I agree with what you said, but maybe you misunderstand my averaging process, for example, if there are four dielectrical layers(die1 die2 die3 die4) between two metal layers(M2 and M1), the thickness and permitivity of each dielectrical layer are T1=1um Er1=1, T2=1um Er2=2, T3=1um Er3=3, T4=1um Er4=4 respectively. So I use one layer with thickness equal to Teq=4um(i.e. T1+T2+T3+T4) and permitivity equal to Er,eq=2.5(i.e. weighted average value:[(Er1*T1+Er2*T2+Er3*T3+Er4*T4)/(T1+T2+T3+T4)]) to substitute the four dielectrical layers between M1 and M2. With this process, the layer stack is almost not changed except the dielectrical layer types. Are you still sure this will affect the result so much?

By the way, in the substrate layer setup(i.e. the slm file), the permitivity blank can be filled with type( Re, Conductivity), I am not clear how to understand the "conductivity" of a dielectrical layer, does this mean that the dielectrical layer is conductive and can conduct current? it seems unreasonable.

another question: for there has no infinit metal layer in our inductor, then how should I to declare the ground? I tried to put ground ports on GT or AA layer where they make the PGS, but still the result can't match the s parameter results obtained from Cadence Spectre.

I really appreciate your help, thanks very much!

 

[kspalla]

if possible share the design in momentum ( no approximation) and related docs indicating material properties 3d views etc..
i do not have spectra

 

[howto]

For there are something that is confidential, i can't upload it here. My email is xiaohu0602@gmail.com, you can send me an email and i will offer the necessary information, thanks.