From: Michael Nudelman (email@example.com)
Date: Mon Nov 13 2000 - 07:50:42 PST
I am currently dealing with some BGA chips, where PLL VCC is DEEP inside the BGA
grid, and the chip being small (0.5"x0.5") the grid is without the gridless area
inside the chip for the de-coup caps to put in. Which results in about 200 mils of a
wire from PLL VCC to the cap.
The manufacturer however said that this is all right, and the wire will act as
inductance to additionally clean the VCC.
ANyway, I discovered, that I had low-freq. component affecting it the most (DCDC
switching freq. remnants, really low), and what actually cleaned it was a ferrite
plus 47uF cap before the hi-freq. filtering element.
The output jitter (the part was a CDR) became really small.
> My approach is to put bypass capacitors as close as possible to the power pins
> on a chip. Whether we are using a multilayer card with ground and power planes,
> or a double-sided card with ground gridding, we will almost always have more
> connections to and more copper for ground than for any one supply voltage. Thus
> the path from a bypass capacitor to ground is usually shorter and wider than the
> path to a supply voltage. This results in:
> * Faster response by the capacitor, due to shorter transit time in the
> microstrip/stripline between the chip and the
> capacitor(s)--about 1/6 ns per inch for FR-4 boards.
> * More return-paths to the chip for the transient current, which reduces the
> inductance, impedance, L * dI/dt drop, and
> maybe radiated emissions by the magnetic fields of the various paths
> partially cancelling one another.
> * Smaller loop area for the transient currents, ditto.
> Since transient currents are my major concern, I try to put the highest
> frequency-response capacitors (typically 220pF NPO ceramics for clock and
> phase-locked loops (PLL's)) right next to their corresponding power pins. Then
> I put lower frequency-response capacitors (typically a 100nF X7R ceramic for
> each power pin or cluster of power pins) as close as possible to their
> corresponding power pins. Next I consider how to route traces/vias to bring
> power and ground to the power-pin/ground-pin/bypass capacitor cluster. If I can
> I will bring in power and ground right next to each other, but I don't worry too
> much if the power and ground connections are on opposite sides of a cluster.
> Depending on the complexity and package size of the chip, I will also put 1 to 4
> bulk ceramic capacitors (typically 2.2uF Y5V ceramics) on each supply voltage
> within 1 inch of its power pins, trying to "surround" the chip. If I have a
> bunch of small chips in a small area, I may use just one bulk ceramic capacitor
> for the entire clump.
> Finally I put bulk aluminum electrolytic or tantalum capacitors (10uF and up):
> * Near the power connector(s) to a card.
> * Near power connectors to other cards/devices.
> * Near the corners of the supply-voltage domain.
> * Near "power hog" components, trying to have one within 2 inches of every
> power pin on that supply voltage ("coffee
> cup" rule).
> John Barnes Advisory
> Lexmark International
> **** To unsubscribe from si-list or si-list-digest: send e-mail to
> firstname.lastname@example.org. In the BODY of message put: UNSUBSCRIBE
> si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
> si-list archives are accessible at http://www.qsl.net/wb6tpu
**** To unsubscribe from si-list or si-list-digest: send e-mail to
email@example.com. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
This archive was generated by hypermail 2b29 : Tue May 08 2001 - 14:30:05 PDT