"by pass" or "decouple" capacitors?

Discussion in 'Electronic Basics' started by Guest, Jul 27, 2003.

1. GuestGuest

When referring to capacitors with the term "bypass capacitor"
or "decoupling capacitor" , is that the same thing ?

2. tempus fugitGuest

I always understood bypass to refer to the cap used to bypass a resistor,
say at the emitter of a common emitter amplifier to increase the gain.
Decoupling refers to the caps connected between the hot and ground of a
voltage source to keep the voltage free of noise.

3. Roy McCammonGuest

The terms are used interchangeably, but I think
that "bypass" is more accurate.

4. John LarkinGuest

I'd say no. A coupling capacitor lets signal through, and a decoupling
or bypass cap tries to kill signal that's not supposed to be there.
The terminology is obviously not exact. "Bypass" can also mean
"shunted across", so one can "bypass" a zener diode in a level
shifting situation, to ensure that signals get through!

OK: let's say that a "decoupling" cap is intended specifically to zap
undesired signal feed-through, as from a supply rail to ground. And a
"bypass cap" means, most generally, "a capacitor shunted across
something".

But the words don't matter; the actual function does.

John

5. nwsGuest

OK: let's say that a "decoupling" cap is intended specifically to zap
How would you call the capacitor that we use at the output of an electret
microphone, decoupling or coupling ?

6. Robert MonsenGuest

That is a coupling capacitor. It couples the AC signal to the next stage.

A bypass cap is used to bypass AC around a load, which may in fact be
developing voltage for the next stage.

They are both forms of a voltage divider, consisting of a resistor and a
capacitor.

For the bypass, you take the voltage across the capacitor, and for coupling,
you take the voltage across the resistor. The formulas are

Xc = 1/(2*pi*freq*C) { reactance of the cap at freq)
Zt = sqrt(R^2 + Xc^2) { Impedance of the circuit at freq }
I = Vt/Zt { Current at freq }
Vc = IXc { Voltage across cap at freq }
Vr = IR { Voltage across resistor at freq }

using complex notation, you can see that its just a form of ac voltage
divider:

Zc = 1/(j*2*pi*freq*C) {Impedance of cap at freq}
Vc = Vt (Zc / (R + Zc)) {Voltage across cap at freq}
Vr = Vt (R / (R + Zc)) { Voltage across R at freq }

Note that Vc + Vr = Vt

As I recall, Grob's advice is to make sure Xc is 1/10 of R for both at the
'cutoff frequency'.

Regards
Bob Monsen

7. John PopelishGuest

There is nothing wrong with it. I was just saying that just because a
cap is a pretty low impedance, it doesn't necessarily isolate one
circuit form another nearby circuit, all by itself. It will have some
ability to keep ripple produced by one chip from interfering with
another chip, but adding a little intentional impedance between them
completes a decoupling filter. I see lots of old tube amplifier
circuits, where the decoupling concept was first emphasized (I
think). Two successive stages of amplification often run off the same
B+ bypass capacitor, because their interaction (if any) is in the form
of negative feedback, at least at low frequencies. But add a third
stage, and the interaction thorough the B+ becomes positive feedback,
and at that point a decoupling filter is usually necessary, consisting
of a choke or series resistor in the B+ supply line, with a second
bypass capacitor. That resistor and capacitor are often referred to
as a decoupling network (for the positive feedback).

8. Roy McCammonGuest

It is a matter of point of view. I suppose the 100uF
electrolytic sitting on the board and working against
the impedance in the power supply cabling might
"decouple" the the board from the power supply noise
a little (more so if there is an intentional inductor),
but all those little ceramics, in my point of view, are
bypassing high freq currents from the vcc plane to the
ground plane and vice versa.

--
Achilles: I wish my wish would not be granted.
< an undescribable event occurs >
Achilles: What happened? Where's my Genie?
Tortoise: Our context got restored incorrectly.
Achilles: What does that cryptic comment mean?
Tortoise: The system crashed.

To email me send to :

rb <my last name> AT ieee DOT org

9. John PopelishGuest

I see it a different way. If the cap is intended to improve the
quality (stability) of the voltage for a single device, it is a bypass
cap, but if it is intended to keep some signal generated by one
circuit from showing up in a second circuit, it is a decoupling cap.
Bypass caps serve a single customer, while decoupling caps isolate two
customers. It is just the difference between the words "across" and
"between".

10. Roy McCammonGuest

Well, I know what you mean if you say either term, but I
notice that Howard Johnson uses the term "bypass".

11. Rich GriseGuest

OK, my turn. I remember bypass caps from RF stuff, and essentially
it provided an AC ground for the "cold" end of the tank (which
could be "hot", if it's in the plate circuit) and kept the RF
out of the power supply.

Decoupling caps don't have an intervening tuned circuit - they
provide a temporary low-impedance supply for good switching times.

And a common-cathode or common-emitter or common-source amplifier
with emitter/etc bias could have the emitter resistor bypassed.
That definitely wouldn't be called decoupling, at least not by
me.

Hope this throws a little LOX on the bagel. ;-)

Cheers!
Rich