EMBEDDEDNET Archives

May 2005

EmbeddedNet@IPC.ORG

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Richard Ulrich <[log in to unmask]>
Reply To:
D-50 Embedded Devices Committee Forum <[log in to unmask]>, Richard Ulrich <[log in to unmask]>
Date:
Wed, 4 May 2005 11:18:38 -0500
Content-Type:
text/plain
Parts/Attachments:
text/plain (104 lines)
I've gotten reports of attachment problems, so here is the text alone  
without the two figures.

Rick

for “Embedded Passive Update” column, CircuiTree, July 2005


My Capacitor Can Beat Up Your Capacitor



Rick Ulrich



One of the reasons for switching from surface mount passives to  
embedded is the claim that the latter exhibits less parasitic  
electrical behavior.  Ideally, a given passive would be a pure  
component; a resistor would show only resistance without any hint of  
capacitance or inductance.  But, in reality, each passive always has  
some amount of the other two, and these tend to become apparent at  
certain frequencies. Comparing the electrical performance of surface  
mount and embedded resistors reveals that their parasitcs are not  
very different.  Same for inductors.  What improvement there is for  
embedded R's and L's comes not from the components themselves being  
purer, but because some or all horizontal and, especially, vertical  
interconnections may be eliminated.

However, embedded capacitors can exhibit far less parasitic  
inductance than SM caps, and this is a very important advantage for  
some applications, especially decoupling.   Parasitic inductance  
causes any capacitor, SM or embedded, to cease being a cap and become  
an inductor above some frequency.  The significantly lower parasitic  
inductance of the embedded capacitor expands the useful frequency  
range relative to a SM part.

There are two reasons for this, and the first is related to the  
connection issue mentioned above.  Figure 1 shows how inductive vias  
and traces can be eliminated by placing the dielectric directly in  
between the power and ground plane instead of inside a SM package on  
top of the board.  This distributed capacitance has been known and  
used for decades, but a few mils of FR4 provides very little  
capacitance density, less than 0.1 nF/cm2.  The challenge today is to  
place higher k and thinner materials in this space.





Figure 1.  Decreased parasitic inductance due to eliminating  
interconnects to the capacitor.



The other effect has to do with the reduction of the self-inductance  
of the structure by mutual inductance.  In the conventional SM  
capacitor on the left side of Figure 2, current travels from left to  
right in the plates of both polarities.  In the embedded capacitor on  
the right, the connections to the plates are arranged so that current  
flows in opposite directions in the plates, thereby canceling some of  
the structure’s self-inductance by mutual inductance.  The fields for  
the SM add, and the fields for the embedded capacitor cancel.




Figure 2.  Field cancellation in embedded capacitors



The result is a return circuit, which can be shown to have an  
inductance that is directly proportional to the thickness of the  
dielectric. This is best expressed in Henrys/square just as  
resistance can be expressed in W/square.  For a parallel plate  
embedded capacitor with the current entering and leaving the same  
side and the contacts distributed over that entire side so there is  
no spreading inductance,


parasitic inductance in pH/square = 1.26(dielectric thickness in  
microns)



If this were the only significant source of inductance, then using a  
50% thinner dielectric would give a 41% higher self-resonance  
frequency, while simultaneously doubling the capacitance. The makers  
of SM low-ESL capacitors are utilizing this principle by shepherding  
currents in their parts to achieve at least some degree of field  
cancellation, at higher cost associated with the more complex  
interior structure that is required.

The total parasitic inductance of an installed capacitor is the sum  
of the conductor to reach it (interconnects and vias), the contacts  
to the device, and the inductance of the part itself.  Embedded  
structures eliminate the first one and, if multiple or line contacts  
are used, can result in a part that has almost immeasurably small  
inductance, less than around 10 pH, far below that of even expensive  
low-inductance SM parts.  Decoupling is as much about inductance as  
it is capacitance, and the inherent performance advantages of  
embedded capacitors gives them a clear edge in local power management  
for high-current IC’s.

ATOM RSS1 RSS2