Bypass cap's

[Arthur Harrison]

Steve Varner:


>Well, it seems you guys all read the same bypass cap book  ;-)
>I'm mainly using analog synth modules and have a regulated power supply
>with a bipolar +15/-15 and a +5V power output. I'm using a design that is
>the same as the Electronotes PCC and Thomas Henry's power supply. I run 4
>wires (18 awg) up through the synth and tap in using those blue tap-in
>connectors. The 4 lines (+15 / -15 / +5 / Gnd) that were tapped out then
>run to a hard-drive-type power connector and clip into to a connector
>mounted on the board.
>Sooo... Mark got me a bit worried when he said:

Mark Amundson:

>> 3. Use 10uf to 100uf tantalum or electrolytic caps for local supply sag
>> issues. Do not use these caps if you have regulated supplies and good
>> beefy wiring as they may cause more power supply hum than their curative
>> effects.

Steve Varner:

>Exactly how do these "larger" caps on the power supply inputs to the module
>cause power supply hum?

Art Harrison:

Bill Layer had a though-provoking post last month in this regard, in which
he responded
to a comment from J. Larry Hendry. I've been meaning to do some tests to
investigate
the relationship of PS capacitance to line-frequency noise, and specifically
see if a
real-world example of  "more is worse" could be created.  Keep in mind that
the
capacitor at issue in the following post is the pre-regulator "bulk" filter,
not the capacitor
at the regulator output.

Anyway, I am taking the liberty of reposting Larry's question and Bill's
response
(I hope y'all don't mind):

Bill Layer:

Hi Larry, All,

J. Larry Hendry:

>I was having trouble following the logic that a linear power supply using
>solid state diodes would cause distortion of this type.  Then the obvious
>hit me like a brick --  the diode does not start conducting until the
>voltage magnitude of the sine wave overcomes the voltage required to cause
>the diode to connect.  So, at each zero crossing, you have a shut off
>period on both sides where neither diode is connecting.

Bill Layer:

This is true, but the zero-crossing is not always the ugliest hour for the
SS rectfier. Imagine a simple single-wave rectified supply, with a single
diode feeding a single filter cap. Assume that this supply is unloaded, the
voltage across the capacitor now rises to the peak of the AC waveform,
about 1.414 x RMS. Assuming that no current is drawn (not the case, for a
real cap) the voltage remains constant AND the diode never conducts again.

But when current is drawn, something else ocurrs. The voltage across the
cap falls, and now the diode wants to conduct, BUT ONLY at the peak of the
waveform. Now, the filter cap tends oppose that change in voltage (the
bigger the cap - the bigger the opposition... read on), and when that diode
does switch on - look out. The AC waveform is at the near peak of it's
power cycle, and when the diode switches on, the instantanious current is
large. If the supply voltage now falls to 1.2 X peak, this means that for
over 10% of the waveform, the transformer and rectifier are pushing against
a brick wall.

To an extent, this also relates to the fact that given two otherwise
identical power supplies, more current may be drawn from an inductive 1st
filter than a capacative 1st filter design. Peak currents are lower in the
inductive filter.

What this means to me, is that the guidelines for filter cap size in power
design may need to be re-thougt, on the basis of 'bigger is not better'.
How would you like to find out that that ultra-quiet 10X overfiltered
supply you built is doing more harm to the overall sound of your system,
than good for the device it is powering! Obviously the extreme case, but
the logical end nonetheless.

Any thoughts?

Bill Layer
Sales Technician
<b.layer at vikingelectronics.com>


--------end archived post---------

Steve Varner:

>It seems that a lot of other people on this list are using this size cap on
>the power lines and smaller ones into each IC. I was under the impression
>that the caps on the ICs were usually only necessary for digital circuits
>and for op-amps that handle low-level signals.

Art Harrison:

Ironically, some digital circuits built on power/ground plane boards need no
bypass
caps at all.  I was informed that the stored energy exists due to the
topology of
the planes.  Something about the "H" field, but I'm no expert in these
matters, so
perhaps a lurking physicist may care to comment.

In some instances, the use of bypass capacitors for linear integrated
circuits is
desirable for stability, regardless of power supply noise.

Last thought:  In many instances, large output capacitors can actually cause
instabilities
in a regulator.  !00uF of tantalum loading a 7812, for example, has cause
oscillations,
in my experience.