Opamp stability question

[MRW]

Hi all:

I found this PDF from Intersil:

http://www.intersil.com/data/an/an9415.pdf

I also found this page talking about op amp feedback:

http://www.ecircuitcenter.com/Circuits/opfeedback1/opfeedback1.htm

Intersil states that it "is observed that if the magnitude of the loop
gain, Aβ, can achieve one while
the phase equals -180 degrees, the closed loop gain becomes infinity
because of division by zero," so it is unstable. However, the other
site mentions "If there is a frequency where the AC magnitude around
the loop is 1 and the total phase is -360 or 0 deg, the circuit is
unstable."

Which one is it?

Thanks!

 

[Noway2]

Hi all:

I found this PDF from Intersil:

http://www.intersil.com/data/an/an9415.pdf

I also found this page talking about op amp feedback:

http://www.ecircuitcenter.com/Circuits/opfeedback1/opfeedback1.htm

Intersil states that it "is observed that if the magnitude of the loop
gain, Aβ, can achieve one while
the phase equals -180 degrees, the closed loop gain becomes infinity
because of division by zero," so it is unstable. However, the other
site mentions "If there is a frequency where the AC magnitude around
the loop is 1 and the total phase is -360 or 0 deg, the circuit is
unstable."

Which one is it?

Thanks!

I believe that they are saying the same thing, just differently. An op
amp in the inverting configuration inherently has a -180 deg phase
shift. Hence, per stability criterion, if the loop gain has adds an
(additional) phase shift of -180 degrees (or a total phase shift of
-360, 0, or 360 whatever you want to call it) at the 0db gain (db as 20
* log V/V) or a gain of 1, then you have an unstable system.

 

[John Popelish]

Hi all:

I found this PDF from Intersil:

http://www.intersil.com/data/an/an9415.pdf

I also found this page talking about op amp feedback:

http://www.ecircuitcenter.com/Circuits/opfeedback1/opfeedback1.htm

Intersil states that it "is observed that if the magnitude of the loop
gain, Aβ, can achieve one while
the phase equals -180 degrees, the closed loop gain becomes infinity
because of division by zero," so it is unstable. However, the other
site mentions "If there is a frequency where the AC magnitude around
the loop is 1 and the total phase is -360 or 0 deg, the circuit is
unstable."

Which one is it?

They are both right. The first assumes that at low
frequencies, there is a 180 degree phase shift (signal
inversion) to start with, and an additional 180 degrees, as
frequency rises, brings the total phase shift around the
loop to 360. The total, including the normal low frequency
inversion, is described in the second one.

 

[MRW]

They are both right. The first assumes that at low
frequencies, there is a 180 degree phase shift (signal
inversion) to start with, and an additional 180 degrees, as
frequency rises, brings the total phase shift around the
loop to 360. The total, including the normal low frequency
inversion, is described in the second one.

Thank you folks!

 

[MRW]

....Just an additional question.

In reference to this IC:
http://datasheets.maxim-ic.com/en/ds/MAX4465-MAX4469.pdf
MAX4469

Regarding the phase and gain charts given on page 3, at what opamp
input is the test signal normally applied?

I'm curious because if I input a signal at the inverting input, then
I'm technically adding a 180 degree phase shift instantly (right?). If
that's the case, then at 100kHz I would already get a -180 - 120 = -360
degree phase?

Thanks!

 

[John Popelish]

...Just an additional question.

In reference to this IC:
http://datasheets.maxim-ic.com/en/ds/MAX4465-MAX4469.pdf
MAX4469

Regarding the phase and gain charts given on page 3, at what opamp
input is the test signal normally applied?

I'm curious because if I input a signal at the inverting input, then
I'm technically adding a 180 degree phase shift instantly (right?). If
that's the case, then at 100kHz I would already get a -180 - 120 = -360
degree phase?

Stability has little to do with where you inject an input
signal. It has everything to do with a signal loop that can
regenerate an echo.

The graphs on page 3 show the open loop gain and phase
shift, if you apply a signal between the + and - input (with
the phase referenced to the + input), magically balanced to
produce an average output voltage of Vcc/2 (with a 100k load
resistor terminated to Vcc/2, no load capacitive load and
with 100 pF capacitive load).

Note the broad frequency band that produces roughly -90
degrees phase shift as the gain falls from a dominant roll
off pole. Note also that the phase shift is still somewhere
between -90 and -180 degrees phase shift as the frequency
that produces 0 db gain is reached.

The negative feedback would normally go to the - input,
which adds another -180 degrees to the phase shift shown.

 

[MRW]

Stability has little to do with where you inject an input
signal. It has everything to do with a signal loop that can
regenerate an echo.

Ok, I'll remember this in the future. Thanks, John!

The graphs on page 3 show the open loop gain and phase
shift, if you apply a signal between the + and - input (with
the phase referenced to the + input), magically balanced to
produce an average output voltage of Vcc/2 (with a 100k load
resistor terminated to Vcc/2, no load capacitive load and
with 100 pF capacitive load).

Note the broad frequency band that produces roughly -90
degrees phase shift as the gain falls from a dominant roll
off pole. Note also that the phase shift is still somewhere
between -90 and -180 degrees phase shift as the frequency
that produces 0 db gain is reached.

The negative feedback would normally go to the - input,
which adds another -180 degrees to the phase shift shown.

Ahh, I see. So at audio frequencies, I should still have a stable
condition.

By the way, is phase affected by a change in gain? I'm curious because
if I change the gain of my opamp to 20V/V, its bandwidth decreases to
about 10kHz. I'm not sure if that affects the phase, too.

 

[John Popelish]

Ok, I'll remember this in the future. Thanks, John!




Ahh, I see. So at audio frequencies, I should still have a stable
condition.

At closed loop gains all the way to 0 db (unity gain
follower) they should be stable. Closed loop gains of 10 or
more (20db or greater) where the phase shift at the
frequency where the closed loop gain goes through 0db
(including the -20 db or more of attenuation in the feedback
that gets added to the open loop gain) is 90 degrees or less
will not only be stable, but not produce a frequency peak
(tendency to ring) at any frequency. But gains higher than
20 db will produce a flat response band less than 20 kHz, so
for the full 20 kHz audio range, 20 db is as high as you can
go with these.

By the way, is phase affected by a change in gain? I'm curious because
if I change the gain of my opamp to 20V/V, its bandwidth decreases to
about 10kHz. I'm not sure if that affects the phase, too.

To find the phase shift at the point where the closed loop
gain hits 0 db, add the open loop gain of the amp with the
attenuation of the feedback network. If you divide the
output signal by a factor of 10 (-20db) to close the loop,
then then you subtract that 20 db from the open loop gain
(move the gain curve down 20 db) and see where this shifted
curve (amplifier gain factor times feedback attenuation
factor to complete the loop or amplifier gain in db plus
attenuation in negative db) passes through 0 db. Then you
look directly below, at that frequency, to see what the
amplifier phase shift is at this reduced o db closed loop
operating point.

 

[MRW]

At closed loop gains all the way to 0 db (unity gain
follower) they should be stable. Closed loop gains of 10 or
more (20db or greater) where the phase shift at the
frequency where the closed loop gain goes through 0db
(including the -20 db or more of attenuation in the feedback
that gets added to the open loop gain) is 90 degrees or less
will not only be stable, but not produce a frequency peak
(tendency to ring) at any frequency. But gains higher than
20 db will produce a flat response band less than 20 kHz, so
for the full 20 kHz audio range, 20 db is as high as you can
go with these.

Sorry, John, you lost me. I guess I'll need to sleep on this..... Where
are you getting the -20dB attenuation from the feedback? Where are you
getting this information: "But gains higher than 20 db will produce a
flat response band less than 20 kHz, so for the full 20 kHz audio
range, 20 db is as high as you can go with these." I'd really like to
know how you figure this out.

To find the phase shift at the point where the closed loop
gain hits 0 db, add the open loop gain of the amp with the
attenuation of the feedback network. If you divide the
output signal by a factor of 10 (-20db) to close the loop,
then then you subtract that 20 db from the open loop gain
(move the gain curve down 20 db) and see where this shifted
curve (amplifier gain factor times feedback attenuation
factor to complete the loop or amplifier gain in db plus
attenuation in negative db) passes through 0 db. Then you
look directly below, at that frequency, to see what the
amplifier phase shift is at this reduced o db closed loop
operating point.

....And this, too. I think this is the time where I would need an
alphabet book with pictures. I still have a lot more texts to go to
before I can catch up.

 

[John Popelish]

Sorry, John, you lost me. I guess I'll need to sleep on this..... Where
are you getting the -20dB attenuation from the feedback?

To program the amplifier for a gain of 10 (20 db) you need
to divide the output voltage by a factor of 10 and connect
the divided output signal to the - input. This was a
hypothetical case.

Where are you
getting this information: "But gains higher than 20 db will produce a
flat response band less than 20 kHz, so for the full 20 kHz audio
range, 20 db is as high as you can go with these." I'd really like to
know how you figure this out.

Look at the gain curves on page 3. If you apply one of
these amplifiers as a gain 10 block (20 db) at the frequency
where the gain curve falls below 20 db, the feedback will
completely lose control of the amplifier, because beyond
that, the amplifier is not capable of a gain if 20 db. If
you program it for a closed loop gain of 100 (40 db), it
will go open loop and the gain will start to follow the open
loop curve down, beyond the frequency where the curve falls
below 40 db, which is about a decade lower than the one
where it falls below 20 db.

...And this, too. I think this is the time where I would need an
alphabet book with pictures. I still have a lot more texts to go to
before I can catch up.

You are a lot closer than you think.

 

[MRW]

To program the amplifier for a gain of 10 (20 db) you need
to divide the output voltage by a factor of 10 and connect
the divided output signal to the - input. This was a
hypothetical case.

Look at the gain curves on page 3. If you apply one of
these amplifiers as a gain 10 block (20 db) at the frequency
where the gain curve falls below 20 db, the feedback will
completely lose control of the amplifier, because beyond
that, the amplifier is not capable of a gain if 20 db. If
you program it for a closed loop gain of 100 (40 db), it
will go open loop and the gain will start to follow the open
loop curve down, beyond the frequency where the curve falls
below 40 db, which is about a decade lower than the one
where it falls below 20 db.

You are a lot closer than you think.

Thanks again, John!

 

[MRW]

Hello again:

Just another follow up question regarding op amp stability. Referring
to the MAX4469 again, on the plot on page 3 of the datasheet, it shows
that the 0db gain is at 200kHz. Now, when I go down to the the phase
plot, it shows that the phase at this frequency is -140 degrees. So my
phase would only need to change by 40 degrees to get to -180 degrees.
Why is the phase margin in the Electrical Characteristics table on the
same page showing 70 degrees?

Thanks!

 

[John Popelish]

Hello again:

Just another follow up question regarding op amp stability. Referring
to the MAX4469 again, on the plot on page 3 of the datasheet, it shows
that the 0db gain is at 200kHz. Now, when I go down to the the phase
plot, it shows that the phase at this frequency is -140 degrees. So my
phase would only need to change by 40 degrees to get to -180 degrees.
Why is the phase margin in the Electrical Characteristics table on the
same page showing 70 degrees?

Sorry, I don't have an answer for you. Normally, the phase
margin and gain margin specs refer to the follower
configuration, but they don't have to. Sometimes specs are
made to look better than they are by including some note
that a particular test circuit is used. But I don't see
that, here. The non inverting waveforms also do not
correspond to a simple follower based on the gain and phase
curves. There should be some ringing or at least overshoot
in the edges. I suspect they used a tweaked feedback
network, and 'forgot' to include that information on the
data sheet.

 

[MRW]

Sorry, I don't have an answer for you. Normally, the phase
margin and gain margin specs refer to the follower
configuration, but they don't have to. Sometimes specs are
made to look better than they are by including some note
that a particular test circuit is used. But I don't see
that, here. The non inverting waveforms also do not
correspond to a simple follower based on the gain and phase
curves. There should be some ringing or at least overshoot
in the edges. I suspect they used a tweaked feedback
network, and 'forgot' to include that information on the
data sheet.

Thanks again, John! Your input is very much appreciated.