[sdiy] power supply decoupling capacitors

[jhaible at debitel.net]

I don't know the ultimate truth for decoupling caps either,
but here are my 2 Euro-Cents:

> ok, what's the difference in putting a cap to V- instead to gnd?

In many popular opamps, what you need most is a cap from V-
*to* GND. Thi sis because often the negative rail is used
as the reference node for the voltage gain stage of the
opamp. For DC, th eloop is closed and this is irrelevant.
For HF, the integrating behaviour of the voltage gain stage
will make it sensitive, because the correction thru the
closed loop is slowed down by the integrator.
I often use (in addition to electrolytics which are shared
by several opamps) a single ceramic cap from V- to GND to
ensure stability:

> This should also prevent oscillations via feedback through

This is what I referred to so far. Keeping HF away from
an opamp.

But bypass caps can serve other functions, too. Like preventing
HF from being *spread* by an opamp that drives a considerable amount
of current.

Example1: opamp drives a load to GND. You want to bypass *this*
GND node where th eload is connected to the V+ and V- of the driving
opamp, to have a small current loop for HF (fast transients).
This is irrelevant if you drive 100k "loads", of course.
But with low impedance / wide band / low noise circuits it becomes
a factor.

Example 2: opamp 1 drives a load into a virtual GND summing node
og opamp 2. Obviously opamp 2 must sink the same current that
opamp 1 sources, and vice versa. So you want a decoupling cap
from V+ of opamp 1 to V- of opamp 2, and a cap from V- of opamp 1
to V+ of opamp 2. Again, you only need this if keeping the HF 
currents in a shrt loop is relevant in your circuit.

One of the big advantages of discrete circuits is that it's
much easier to see where you want to connect your capacitors.
Opamps always feign simplicity, while they are actually the more
complex circuits.

JH. 

-------------------------------------------------
debitel.net Webmail