[SI-LIST] : Power Noise

Jon Keeble (j.keeble@fairlightesp.com.au)
Thu, 18 Mar 1999 10:08:06 +1000

For mixed 5v / 3v3 systems I put another pair of planes close to the =
other side of the board, and bypass that paifrom that side. If the GND =
planes are on the inside, return current for all inner routing layers =
run in GND, making routing more flexible.

Don't forget that the bypass caps do more than provide a low impedance =
for power: they also, and equally importantly, provide a path for return =
current to get from the plane they return in to the appropriate power =
pin on the driver chip. Driving a high going edge means that return =
current has to get into the power pin of the driver: the only path is =
via a close by capacitor.

The other reason for enclosing tracks within a pair of GND planes means =
that vias don't require bypass
caps nearby to provide a path for return currents.

So, I place caps in an array around the chip, not being concerned about =
how close to the power pins they are, no more than about an inch apart.

Placement of caps has more effect on the length of path the return =
signal has to traverse. How different this path is governs the size of =
the loop and how much raditation results. (Another very good reason to =
provide as much C between planes as possible).

BTW, it doesn't really matter whether the bypass cap looks like a C or =
L, as long as the magnitude of Z is low.

The effectiveness of small SM caps relates to the ratio of length to =
width: the 0805 package is the best of the normal packages.

There is a number of low inductance cap designs. One is 0612 (the =
opposite of 1206) that provide a better ration of L/W.

Another is actually multiple caps with the current flowing in opposite =
directions to reduced inductance.

Jon Keeble

------=_NextPart_000_00F0_01BE7127.37C56860
Content-Type: text/html;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE HTML PUBLIC "-//W3C//DTD W3 HTML//EN">

>Now, if the highest inductance part of the path is the pad/via, = would=20 that
>favor running (wide) traces from the cap directly to the = device,=20 since that
>eliminates two vias between the cap and the device? A = wide,=20 short path only
>has a few nH inductance (I think).

Hi=20 Doug 
 
There's no argument that minimising = the length,=20 and maximising the width, of the connection from C to planes minimises = the=20 inductance. Same goes for connecting component pads to planes. Both are=20 important.
 
Between connections on a plane there = is really a=20 very low impedance.
 
The C between planes can be a very useful component = as well,=20 if they placed close together.
 
From the above, I layup my boards so that the power = planes are=20 adjacent, and close to one surface of the board, and place the bypass = caps on=20 that side. For a .060" thick board, if the planes are in the = middle, that=20 adds a total of .060" trace length (often through a narrow via) = which is a=20 much larger inductance.
 
For mixed 5v / 3v3 systems I put another pair of = planes close=20 to the other side of the board, and bypass that paifrom that side. If = the GND=20 planes are on the inside, return current for all inner routing layers = run in=20 GND, making routing more flexible.
 
Don't forget = that the bypass=20 caps do more than provide a low impedance for power: they also, and = equally=20 importantly, provide a path for return current to get from the plane = they return=20 in to the appropriate power pin on the driver chip. Driving a high going = edge=20 means that return current has to get into the power pin of the driver: = the only=20 path is via a close by capacitor.
 
The other reason for enclosing = tracks within a=20 pair of GND planes means that vias don't require bypass
 caps nearby to provide a path = for =20 return currents.
 
So, I place caps in an array around the chip, not = being=20 concerned about how close to the power pins they are, no more than about = an inch=20 apart.
 
Placement of caps has more effect on = the length=20 of path the return signal has to traverse. How different this path is = governs=20 the size of the loop and how much raditation results. (Another very good = reason=20 to provide as much C between planes as possible).
 
 
BTW, it doesn't really matter = whether the bypass=20 cap looks like a C or L, as long as the magnitude of Z is = low.
 
The effectiveness of small SM caps relates to the = ratio of=20 length to width: the 0805 package is the best of the normal=20 packages.
 
There is a number of low inductance cap designs. One = is 0612=20 (the opposite of 1206) that provide a better ration of L/W.
 
Another is actually multiple caps with the current = flowing in=20 opposite directions to reduced inductance.
 
Jon Keeble
 
------=_NextPart_000_00F0_01BE7127.37C56860-- **** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****