Hello,
I am simulating quality factor Q of an inductor from TSMC 0.13um.
The inductor chose is spiral_std with the following parameters:

Inductor_Width (M): 3u
I set two variables: Inner_Radius and Number_Of_Turns.

I simulate sp with two variables sweeping as follows:
Inner_Radius : 15u - 90u
Number_Of_Turns: 0.5 - 5.5

Below is my results. Sorry for the inconvinence. I will attach each post with an image.

The problem is that I can't understand why the line with n = 0.5 have a completely different shape compared to other lines. Please explain. Thank you.

Here is the real part of inductance with different inner radius and number of turn:

Here is the Imaginary part of inductance with different inner radius and number of turn:

The reason is likely that you are using a low simulation frequency where the inductive reactance of the 0.5 turn inductor is small compared to the resistive loss.

If you repeat this sweep at a higher frequency (say 2x) then the 0.5 curve should start looking more like the other curves in the trend.


As a practical matter, this result just says there the inductance of the 0.5 inductor is too low to be of any practical use at the simulation frequency.

Hi,


The result seems a little odd to me. For the 0.5 turn case, I thought there would be much more change in the series resistance and inductance as the radius increases. What equation did you use to calculate inductance and resistance? Maybe your width is scaling with the inner radius?

Also, from memory (its been a while since I looked at inductors), the smaller inner radius values seem a little too small (15 um?). If you roughly assume about 0.75 nH per mm (from memory, this is roughly correct for straight lines), a half turn with 15 um radius would be pi*0.015*0.75n = 0.03nH. But according to your plot, the inductance is around 2.5/(2*pi*1.5e9) = 0.265nH. So my rough estimation is one order of magnitude off.

Another thing, 3um seems a little narrow to get good Q unless your metal is extremely thick...

Sometimes you cannot do parametric sweeps on PDK models. You might need to plot your data one step at a time.


regards,
Aaron