Re: Power/ground connections/bypassing on ICs

Mark Randol ([email protected])
Thu, 01 May 1997 14:41:14 -0700

Andy Ingraham wrote:
> So then, is there no one here who advocates running traces from IC
> power/ground pins first to a bypass capacitor, and then on to the
> power/ground planes?

Well, I'll throw this out then.

If the idea of bypassing is to allow currents to travel through
a lower impedance path back to their source, thereby lessening the
effect on other parts of the circuit by reducing the resulting
voltage (or vice versa), AND if the source is on chip, isn't it
better to put the bypassing as close to the IC as possible? For
RF circuits, almost any signal getting on the supply or return
lines can cause lots of problems. So using power and ground planes
for supply decoupling isn't really an option. Components are placed
as close as practical to the devices to try to contain currents locally
much as possible. For RF, chokes are used to further isolate devices
from the supply, while maintaining a relatively low DC impedance.
Sounds similar to caps close to the device, then vias to the power
and ground planes.

> I thought my question might have touched off some argument, or at
> least someone saying that it works in some cases. (Such as the
> person who last week told me to do this for mixed-signal devices
> that don't specifically need filtered power.)

See above for the chokes and bypass caps "explanation". It's a
sorta crude filter, tho' I suspose designing it that way could
have advantages, it generally works ok if you follow the rule
of thumb of making blocking impedances at least 10X the system
impedance, and "bypassing" impedances 10X less (100:1).

Bias tees are an example of using a large inductance in series
with the supply prevent the signal from getting on the supply.
The capacitor in this case is a DC block instead of bypassing
cap though.

> One reply mentioned removing the thermal reliefs from the vias.
> Is this a common practice? Does it really make a difference?

For RF it is. I've defeated bad layouts using them in a couple
of kludgy ways. Solder blobs and solder wick cover up the
thermal reliefs quite well until the next pass on the board.

Mark Randol, RF Measurements Engineer   | Motorola SPS, Inc.
(602)413-8052 Voice                     | M/S EL379
(602)413-4150 FAX                       | 2100 E. Elliot Road
[email protected]                | Tempe, AZ 85284