Opamp power supply bypass

[Daniel Haude]

Hi folks,
in some commercially available products boasting (and really
having) good low-noise performance, I've seen the supply pins of each
singe opamp bypassed with 10ohm resistors and 10uF/100nF caps. Like this:

+15V
|
10R
|
+-10uF--GND
|
+-.1uF--GND
|
OPAMp

....and the same for the negative supply. Of course I know what this is
supposed to accomplish, and in what way, but I've never seen it suggested
in any electronics text. So I suspect that it may be unnecessary, or
despite some merit may have negative side effects that I (and the
designer) am unaware of.

Thanks for any input,
--Daniel

 

[Mario Trams]

Hi folks,
in some commercially available products boasting (and really
having) good low-noise performance, I've seen the supply pins of each
singe opamp bypassed with 10ohm resistors and 10uF/100nF caps. Like this:

+15V
|
10R
|
+-10uF--GND
|
+-.1uF--GND
|
OPAMp

...and the same for the negative supply. Of course I know what this is
supposed to accomplish, and in what way, but I've never seen it suggested
in any electronics text. So I suspect that it may be unnecessary, or
despite some merit may have negative side effects that I (and the
designer) am unaware of.

I've seen such kind of low-pass filters in various data sheets
(not neccessarily related to OpAmps) as recommended power supply
handling - and I followed them.

When you know what this is supposed for, why do you think that it
may be not neccessary?

Regards,
Mario

 

[Jeroen Belleman]

Hi folks,
in some commercially available products boasting (and really
having) good low-noise performance, I've seen the supply pins of each
singe opamp bypassed with 10ohm resistors and 10uF/100nF caps.
[...] I suspect that it may be unnecessary [...]

I have settled to using much the same circuit. I just use 2R2
instead of 10R. It's not intended as a low-pass RC filter, even
though it does look like one. The idea is to spoil the Q of
parasitic resonators that consist of two bypass networks and
the inductance of the power supply trace in between.

Of course, if the power supply distribution layout is done
well, this shouldn't happen, but in practice, it sometimes
does. So go ahead, call me a paranoiac ;-) I've seen a good
deal worse.

A nice side effect is that it is easy to measure the individual
stage supply currents, simply by measuring the drop across the
resistor. It makes trouble-shooting so much easier.

Regards,
Jeroen Belleman

 

[Ban]

I've seen such kind of low-pass filters in various data sheets
(not neccessarily related to OpAmps) as recommended power supply
handling - and I followed them.

When you know what this is supposed for, why do you think that it
may be not neccessary?

Regards,
Mario

If you need a bit better low frequency(hum) performance You could try this
circuit which makes a second order filter. It might be called a gyrator or
simulated coil. It is efficient especially on single supply circuits, or to
filter the bias for electret mikes. Beware of the voltage drop!

100u Tantal
|#
+----|#---------+
| |# |
15V | | +14.2V
o--+-------)------- -----+---o
| | \ ^ 2N3904
| | ---
| ___ | ___ |
+-|___|-+-|___|--+
2k2 2k2 | +
###
---
100u Tantal|
|
===
GND
ciao Ban

 

[Roy McCammon]

Hi folks,
in some commercially available products boasting (and really
having) good low-noise performance, I've seen the supply pins of each
singe opamp bypassed with 10ohm resistors and 10uF/100nF caps. Like this:

+15V
|
10R
|
+-10uF--GND
|
+-.1uF--GND
|
OPAMp

...and the same for the negative supply. Of course I know what this is
supposed to accomplish, and in what way, but I've never seen it suggested
in any electronics text. So I suspect that it may be unnecessary, or
despite some merit may have negative side effects that I (and the
designer) am unaware of.

some opamps have lousy power supply rejection ratio.
I had to this one timeto keep 1 KHz noise out of a
high gain stage.

 

[Fred Bloggs]

in some commercially available products boasting (and really
having) good low-noise performance, I've seen the supply pins of each
singe opamp bypassed with 10ohm resistors and 10uF/100nF caps. Like this:

+15V
|
10R
|
+-10uF--GND
|
+-.1uF--GND
|
OPAMp

So I suspect that it may be unnecessary...

This obviously bandlimits incoming single-ended power supply disturbance
to ~1.5KHz where the OA PSRR remains high, and -apparently- less obvious
attenuates the OA current loading on the rails with the same bandwidth.
The values shown allow the power supply wiring as much as 4000uH of
inductance for a damped response to any wideband current loading profile
the OA may produce.

 

[Fred Bloggs]

This obviously bandlimits incoming single-ended power supply disturbance
to ~1.5KHz where the OA PSRR remains high, and -apparently- less obvious
attenuates the OA current loading on the rails with the same bandwidth.
The values shown allow the power supply wiring as much as 4000uH of
inductance for a damped response to any wideband current loading profile
the OA may produce.

Then there may be an issue of crosstalk between OA's which is greatly
attenuated also- a simple diagram is like so- shows that crosstalk can
be reduced to next nothing:

Please view in a fixed-width font such as
Courier.




I R R
----->-----/\/\---/\/\----> VCT
| |
|C |C
=== ===
| |
+----------------+



VCT 0.5
--- = -----------------------------
I f f
( 1 + j ---- ) ( 1 + j ----)
1.5K 750

 

[Fred Bloggs]

Then there may be an issue of crosstalk between OA's which is greatly
attenuated also- a simple diagram is like so- shows that crosstalk can
be reduced to next nothing:

Please view in a fixed-width font such as
Courier.




I R R
----->-----/\/\---/\/\----> VCT
| |
|C |C
=== ===
| |
+----------------+



VCT 0.5
--- = -----------------------------
I f f
( 1 + j ---- ) ( 1 + j ----)
1.5K 750

That should be:

Please view in a fixed-width font such as
Courier.




I R R
----->-----/\/\---/\/\----> VCT
| |
|C |C
=== ===
| |
+----------------+



VCT 0.5
--- = -----------------------------
I f f
( 1 + j ---- ) ( j --- )
1.5K 15K

 

[John Larkin]

Hi folks,
in some commercially available products boasting (and really
having) good low-noise performance, I've seen the supply pins of each
singe opamp bypassed with 10ohm resistors and 10uF/100nF caps. Like this:

+15V
|
10R
|
+-10uF--GND
|
+-.1uF--GND
|
OPAMp

...and the same for the negative supply. Of course I know what this is
supposed to accomplish, and in what way, but I've never seen it suggested
in any electronics text. So I suspect that it may be unnecessary, or
despite some merit may have negative side effects that I (and the
designer) am unaware of.

Thanks for any input,
--Daniel

That may be overkill. Some people like to pepper everything with
bypasses, and IC manufacturers seem to think their 25-cent gadget is
the center of the universe, so recommend insane amounts of bypassing
in their appnotes. Spec sheets give PSRR v. frequency ratings, so the
thing is predictable... it may be more economic to buy a better opamp
than to add all this stuff.

Low-level stages may need some PSRR help, but it's unlikely that every
stage would. Sounds like somebody's "practise". I knew a guy who made
*every* analog stage differential, no matter what.

I have used RC or ferrite-bead-C rail filters in very low-level,
bipolar opamp stages, to keep digital spikes from bleeding through to
the opamp front end.

John

 

[Ken Smith]

+15V
If you are working with a muliplexed signal, this may add some odd sort of
crosstalk. The OP-amps current varies with the signal input. The
10R's drop changes with current. The 10uF spreads the changed voltage into
a different channel's time slice.

 

[Daniel Haude]

On Thu, 07 Aug 2003 10:09:00 -0700,

Low-level stages may need some PSRR help, but it's unlikely that every
stage would. Sounds like somebody's "practise".

That's what it looks like, because everey single opamp in the entire thing
is bypassed that way. That's why I asked.

Also, I should have read Analog Device's App note AN-202 more carefully
because the concept is explained in detail there.

--Daniel