The Designer's Guide Community Forum
Modeling >> Passive Devices >> Inductor Selection

Message started by aaron_do on Jul 17th, 2009, 12:14am

Title: Inductor Selection
Post by aaron_do on Jul 17th, 2009, 12:14am

Hi all,

suppose I want to design an LC tank and I have two options:

1. Use a large inductor with a low self-resonant frequency. At the resonant frequency of the tank, the L of the inductor is higher than its DC inductance. The parallel capacitor is either an MIM or the gate of a transistor.

2. Use a small inductor with a high self-resonant frequency. The inductance at the resonant frequency of the tank is the DC inductance of the inductor. Again, the parallel capacitor is either an MIM or the gate of a transistor.

Assume both designs have the same Q. So my question is, which option will result in less process variation? From my understanding, if we use option 2, the DC inductance will be quite accurate. However, i'm not sure if that matters since i'm connecting it to a parallel cap which will vary with the process. So it basically comes down to this: is the parasitic capacitance variation of an integrated inductor more or less than the capacitance variation of an MIM cap, or the cap at the gate of a transistor?


Title: Re: Inductor Selection
Post by RFICDUDE on Jul 19th, 2009, 4:33am

For case #1 it all depends on which process parameters control the self resonant frequency of the inductor. The inductance is higher than at DC because there is a resonance between the interwinding capacitance (and maybe the substrate) and the low frequency inductance, so the value of effective inductance depends on the distance of the inductor traces and dielectric constant of the SiO dielectric.

Small changes in the self resonant frequency will make significant changes in the effective inductance depending on how close you are operating near resonance, so you need to take this sensitivity into account when calculating the required range of tuning capacitance to compensate (if you have the luxury of compensating).

For circuits requiring wide tuning range the increasing inductance with frequency characteristic can cause a lot of difficulty in designing a tuning network with small enough parasitics to permit reliable operation at high frequencies (high inductance low capacitance) and low frequencies (low inductance high capacitance). For VCOs this really makes it difficult to achieve a resonably constant KVCO.

At least with case #2, all you generally need to worry about is making it work with the known variation of the tuning caps.

Hope this was helpful.

Title: Re: Inductor Selection
Post by aaron_do on Jul 20th, 2009, 6:59pm

thanks for the reply. I hadn't yet considered the cases with tuning, but i suppose they are important.


Title: Re: Inductor Selection
Post by RFICDUDE on Jul 21st, 2009, 3:35am

More directly to your question regarding variation ...

I really don't know if the back end of the line variation is worse for self resonant inductor or not. The fab models probably are not super accurate at resonance, and they may not even model process variation very accurately. If you design your own inductor, then you have to figure how to properly model the inductor over process or else you will not know what the variation will be until you seen enough hardware measurements to determine the variation.

If you operate away from resonance. The models are less sensitive to parasitic variation around the inductor and the operation will be more correctly predicted by well characterized variation in the capacitors.

In other words, the variation you do know is prefered over the variation you don't know.

If the inductors are well characterized over process then you can make the comparisons yourself.

Hmm, now you have me curious (it is like a song someone mentions and you can't get it out of your head). I'll post if I find anything out worth mentioning.

Title: Re: Inductor Selection
Post by aaron_do on Jul 27th, 2009, 5:25am


since my original post I've done a little more reading. It seems that most inductors are modeled as some variation of the PI model. The PI model is only accurate over a limited frequency range. So I guess part of the answer to my question would have to be that the inductors are not so well modeled around self resonance...

thanks for you input,


The Designer's Guide Community Forum » Powered by YaBB 2.2.2!
YaBB © 2000-2008. All Rights Reserved.