[SI-LIST] Re: Placement of decoupling capacitors
- From: Ray Anderson <Raymond.Anderson@xxxxxxx>
- To: si-list@xxxxxxxxxxxxx
- Date: Tue, 31 Jul 2001 16:53:13 -0700 (PDT)
>Thanks for your input. I am not very clear on what you mean by
>the frequency of interest. Is it the frequency range where the cap.
>is most effective (frequency of resonance, ex. 232 MHz for a 470 pf
>cap.) or the freq. that the device is running at. If it is the freq. the
>device is running at then what I understand from your email, all the
>capacitors should be at the same distance from the IC because they
>are all supplying the same IC.
>
>According to you the distance between the cap. and IC does not
>effect the loop inductance. If this is the case then why can we get
>away by placing the smallest capacitor closest to the pin and the
>once with the bigger values away?
>
When one selects decaps for a design, one method that we have
found to be very effective is to carefully select the values of
the capacitors such that the composite (or superposition) of the
individual capacitor impedance profiles produces a relatively flat
impedance profile whose value is below some target impedance.
The number of decaps required is a function of the target impedance,
the caps ESR and the mounted loop inductance. We have developed
in-house tools to aid in designing the decoupling system based on
this principle. There is now a commercial tool available that implements
this function also.
As you mentioned, various values of capacitance will resonate at
various frequencies. (the capacitance resonates with the mounted
loop inductance). It is this resonant frequency that you need to
pay attention to when determining how far from the decoupled chip
you can be. Small value caps with high resonant frequencies need to
be real close to the part that is being decoupled. Large valued
caps with relatively low resonant frequencies can be much farther
away and still be effective. In a typical design you may have 10-20
different decap values, each with their own resonant frequency. Each
value will have different "effective radius of influence".
Taking a step backwards to look at the big picture, a power distribution
system is basically composed of a source of current (a VRM for example), bulk
capacitors, ceramic capacitors, power planes, and the current consuming load.
The VRM can supply current transients up to a MHz or so, the bulk caps are
effective in somewhat the same range (a fast VRM can obviate the need for a
lot of big bulk caps). The ceramic decaps are effective from a few MHz up to
perhaps a hundred or two hundred MHz. Beyond that, the power planes are your
only real effective means of decoupling.
The power planes also serve as a conduit between your decoupling capacitors
and the devices consuming the current. There is a loop inductance associated
with the capacitors and the planes. There is also a loop inductance
associated with the current consuming load (IC) and the planes. These are
two distict loop inductances. Each should be minimized, but they are
not related to each other. The loop inductance associated with the
caps limits the current the caps can supply to the planes. The loop
inductance associated with load limits the current the planes can
supply to the load.
-Ray Anderson
Sun Microsystems Inc.
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
