Discussion:
[EE] amplifier compensation and stability
Bob Blick
2017-08-24 16:32:54 UTC
Permalink
Hi y'all,

I'm trying to build some audio power amplifiers in order to use up some parts I have left over from a previous life.

In the past I've always built the input stages from discrete parts but this one uses a conventional 4558-style opamp. Then I follow it with voltage and current boosting parts and finally enclose the whole mess in negative feedback.

In this aforementioned past life I designed a lot of power amps so I know to design each stage to have as linear and balanced response as reasonable, keep the gain low around the slowest components etc.

Of course with the added gain and relatively low speed of the output stage, the internally compensated opamp is not able to keep things stable, so I've added an external compensation capacitor. It does work fine, and I'm not looking for super high fidelity, but it's always nice to make simple changes that reap big benefits.

So my question is, any ideas about compensation or changes I could make?

I've attached a block diagram. Note that my output stage is the right-hand amplifier block, it contains many discrete components, and most importantly, it is inverting. Therefore I am using the non-inverting input of the opamp for the negative feedback.

Thank you!

Bob
John Ferrell
2017-08-24 22:00:05 UTC
Permalink
Please keep posting on the project, I have already learned something I
should have already known...


On 8/24/2017 12:32 PM, Bob Blick wrote:
>
> Bob
>
>

--
John Ferrell W8CCW
Julian NC 27283
It is better to walk alone,
than with a crowd going the wrong direction.
--Diane Grant

--
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
Sean Breheny
2017-08-25 03:28:02 UTC
Permalink
Hi Bob,

My main suggestion is to characterize the behavior by providing various
input signals like impulses, steps, white noise, swept sine, etc. and
looking at the output on a scope. You might also vary the output load. This
gives you a good idea that you are not close to going unstable and you
aren't experiencing "peaking" around one frequency band or a null in one
band.

I've had problems in the past with circuits like this where the output
stage has a nonlinearity about zero (like crossover distortion) which can
make the gain be reduced substantially near zero. This can result in
instability by causing integrator wind-up while the output is in the
low-gain region, followed by massive overshoot when it exits the low-gain
region, followed by a compensatory wind-up in the other direction due to
the overshoot, etc. This problem can be dealt with by making sure that the
transistors in the output are always biased with some minimum current so
their input-output gain never goes below a certain value.

Note, too, that the op-amp can be treated as it's own finite gain stage,
where you close the loop locally around it and then close the loop again
around the whole system. This can make analysis of the entire loop easier
because once you guarantee that the op-amp gain stage is itself stable, you
can then treat it as a pure gain block or as a simple compensator (like an
integrator or integrator plus proportional gain or a lead compensator or
lead-lag compensator, etc.) You are sorta doing that already in that the
compensation cap is turning the op-amp into a fast integrator.

I have developed a "theorem" of simple control design (which I am sure has
been stated before but I've never seen it put this way) - given any stable
system P, one can always close the loop around P with an integrator H such
that the closed-loop system will have zero DC error and be stable, for some
value of integrator gain K in H, and the system will continue to be stable
and exhibit zero DC error for any integrator gain Kprime < K. To put it
more simply, you can always control a stable plant using a slow integrator
and obtain at least the improvement of zero DC error, but also usually some
amount of improvement in servo tracking and regulation against
disturbances.

Sean


On Thu, Aug 24, 2017 at 12:32 PM, Bob Blick <***@outlook.com> wrote:

> Hi y'all,
>
> I'm trying to build some audio power amplifiers in order to use up some
> parts I have left over from a previous life.
>
> In the past I've always built the input stages from discrete parts but
> this one uses a conventional 4558-style opamp. Then I follow it with
> voltage and current boosting parts and finally enclose the whole mess in
> negative feedback.
>
> In this aforementioned past life I designed a lot of power amps so I know
> to design each stage to have as linear and balanced response as reasonable,
> keep the gain low around the slowest components etc.
>
> Of course with the added gain and relatively low speed of the output
> stage, the internally compensated opamp is not able to keep things stable,
> so I've added an external compensation capacitor. It does work fine, and
> I'm not looking for super high fidelity, but it's always nice to make
> simple changes that reap big benefits.
>
> So my question is, any ideas about compensation or changes I could make?
>
> I've attached a block diagram. Note that my output stage is the right-hand
> amplifier block, it contains many discrete components, and most
> importantly, it is inverting. Therefore I am using the non-inverting input
> of the opamp for the negative feedback.
>
> Thank you!
>
> Bob
> --
> http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
> View/change your membership options at
> http://mailman.mit.edu/mailman/listinfo/piclist
>
>
--
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
James Burkart
2017-08-25 07:59:44 UTC
Permalink
I've got so much to learn. Sean, what's your education and background? I'm
a junior working on my EET and only have a hobbyist background.

--
Sincerely,

James Burkart
*Filmmaker & Documentarian*

*Burkart Studios*
925.667.7175 | Personal
415.738.2071 | Office

*Web:* burkartstudios.com
*Facebook:* facebook.com/burkartstudios

On Thu, Aug 24, 2017 at 9:28 PM, Sean Breheny <***@cornell.edu> wrote:

> Hi Bob,
>
> My main suggestion is to characterize the behavior by providing various
> input signals like impulses, steps, white noise, swept sine, etc. and
> looking at the output on a scope. You might also vary the output load. This
> gives you a good idea that you are not close to going unstable and you
> aren't experiencing "peaking" around one frequency band or a null in one
> band.
>
> I've had problems in the past with circuits like this where the output
> stage has a nonlinearity about zero (like crossover distortion) which can
> make the gain be reduced substantially near zero. This can result in
> instability by causing integrator wind-up while the output is in the
> low-gain region, followed by massive overshoot when it exits the low-gain
> region, followed by a compensatory wind-up in the other direction due to
> the overshoot, etc. This problem can be dealt with by making sure that the
> transistors in the output are always biased with some minimum current so
> their input-output gain never goes below a certain value.
>
> Note, too, that the op-amp can be treated as it's own finite gain stage,
> where you close the loop locally around it and then close the loop again
> around the whole system. This can make analysis of the entire loop easier
> because once you guarantee that the op-amp gain stage is itself stable, you
> can then treat it as a pure gain block or as a simple compensator (like an
> integrator or integrator plus proportional gain or a lead compensator or
> lead-lag compensator, etc.) You are sorta doing that already in that the
> compensation cap is turning the op-amp into a fast integrator.
>
> I have developed a "theorem" of simple control design (which I am sure has
> been stated before but I've never seen it put this way) - given any stable
> system P, one can always close the loop around P with an integrator H such
> that the closed-loop system will have zero DC error and be stable, for some
> value of integrator gain K in H, and the system will continue to be stable
> and exhibit zero DC error for any integrator gain Kprime < K. To put it
> more simply, you can always control a stable plant using a slow integrator
> and obtain at least the improvement of zero DC error, but also usually some
> amount of improvement in servo tracking and regulation against
> disturbances.
>
> Sean
>
>
> On Thu, Aug 24, 2017 at 12:32 PM, Bob Blick <***@outlook.com> wrote:
>
> > Hi y'all,
> >
> > I'm trying to build some audio power amplifiers in order to use up some
> > parts I have left over from a previous life.
> >
> > In the past I've always built the input stages from discrete parts but
> > this one uses a conventional 4558-style opamp. Then I follow it with
> > voltage and current boosting parts and finally enclose the whole mess in
> > negative feedback.
> >
> > In this aforementioned past life I designed a lot of power amps so I know
> > to design each stage to have as linear and balanced response as
> reasonable,
> > keep the gain low around the slowest components etc.
> >
> > Of course with the added gain and relatively low speed of the output
> > stage, the internally compensated opamp is not able to keep things
> stable,
> > so I've added an external compensation capacitor. It does work fine, and
> > I'm not looking for super high fidelity, but it's always nice to make
> > simple changes that reap big benefits.
> >
> > So my question is, any ideas about compensation or changes I could make?
> >
> > I've attached a block diagram. Note that my output stage is the
> right-hand
> > amplifier block, it contains many discrete components, and most
> > importantly, it is inverting. Therefore I am using the non-inverting
> input
> > of the opamp for the negative feedback.
> >
> > Thank you!
> >
> > Bob
> > --
> > http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
> > View/change your membership options at
> > http://mailman.mit.edu/mailman/listinfo/piclist
> >
> >
> --
> http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
> View/change your membership options at
> http://mailman.mit.edu/mailman/listinfo/piclist
>
--
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
Richard Prosser
2017-08-25 08:53:46 UTC
Permalink
Hi James

I'm a professional electronics designer, 2 months from "retirement" age
after being in the game for at least 40 years am and still learning every
day!

When I stop learning I think it'll be time to stop working.

RP

On 25 August 2017 at 19:59, James Burkart <***@burkartstudios.com> wrote:

> I've got so much to learn. Sean, what's your education and background? I'm
> a junior working on my EET and only have a hobbyist background.
>
> --
> Sincerely,
>
> James Burkart
> *Filmmaker & Documentarian*
>
> *Burkart Studios*
> 925.667.7175 | Personal
> 415.738.2071 | Office
>
> *Web:* burkartstudios.com
> *Facebook:* facebook.com/burkartstudios
>
> On Thu, Aug 24, 2017 at 9:28 PM, Sean Breheny <***@cornell.edu> wrote:
>
> > Hi Bob,
> >
> > My main suggestion is to characterize the behavior by providing various
> > input signals like impulses, steps, white noise, swept sine, etc. and
> > looking at the output on a scope. You might also vary the output load.
> This
> > gives you a good idea that you are not close to going unstable and you
> > aren't experiencing "peaking" around one frequency band or a null in one
> > band.
> >
> > I've had problems in the past with circuits like this where the output
> > stage has a nonlinearity about zero (like crossover distortion) which can
> > make the gain be reduced substantially near zero. This can result in
> > instability by causing integrator wind-up while the output is in the
> > low-gain region, followed by massive overshoot when it exits the low-gain
> > region, followed by a compensatory wind-up in the other direction due to
> > the overshoot, etc. This problem can be dealt with by making sure that
> the
> > transistors in the output are always biased with some minimum current so
> > their input-output gain never goes below a certain value.
> >
> > Note, too, that the op-amp can be treated as it's own finite gain stage,
> > where you close the loop locally around it and then close the loop again
> > around the whole system. This can make analysis of the entire loop easier
> > because once you guarantee that the op-amp gain stage is itself stable,
> you
> > can then treat it as a pure gain block or as a simple compensator (like
> an
> > integrator or integrator plus proportional gain or a lead compensator or
> > lead-lag compensator, etc.) You are sorta doing that already in that the
> > compensation cap is turning the op-amp into a fast integrator.
> >
> > I have developed a "theorem" of simple control design (which I am sure
> has
> > been stated before but I've never seen it put this way) - given any
> stable
> > system P, one can always close the loop around P with an integrator H
> such
> > that the closed-loop system will have zero DC error and be stable, for
> some
> > value of integrator gain K in H, and the system will continue to be
> stable
> > and exhibit zero DC error for any integrator gain Kprime < K. To put it
> > more simply, you can always control a stable plant using a slow
> integrator
> > and obtain at least the improvement of zero DC error, but also usually
> some
> > amount of improvement in servo tracking and regulation against
> > disturbances.
> >
> > Sean
> >
> >
> > On Thu, Aug 24, 2017 at 12:32 PM, Bob Blick <***@outlook.com>
> wrote:
> >
> > > Hi y'all,
> > >
> > > I'm trying to build some audio power amplifiers in order to use up some
> > > parts I have left over from a previous life.
> > >
> > > In the past I've always built the input stages from discrete parts but
> > > this one uses a conventional 4558-style opamp. Then I follow it with
> > > voltage and current boosting parts and finally enclose the whole mess
> in
> > > negative feedback.
> > >
> > > In this aforementioned past life I designed a lot of power amps so I
> know
> > > to design each stage to have as linear and balanced response as
> > reasonable,
> > > keep the gain low around the slowest components etc.
> > >
> > > Of course with the added gain and relatively low speed of the output
> > > stage, the internally compensated opamp is not able to keep things
> > stable,
> > > so I've added an external compensation capacitor. It does work fine,
> and
> > > I'm not looking for super high fidelity, but it's always nice to make
> > > simple changes that reap big benefits.
> > >
> > > So my question is, any ideas about compensation or changes I could
> make?
> > >
> > > I've attached a block diagram. Note that my output stage is the
> > right-hand
> > > amplifier block, it contains many discrete components, and most
> > > importantly, it is inverting. Therefore I am using the non-inverting
> > input
> > > of the opamp for the negative feedback.
> > >
> > > Thank you!
> > >
> > > Bob
> > > --
> > > http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
> > > View/change your membership options at
> > > http://mailman.mit.edu/mailman/listinfo/piclist
> > >
> > >
> > --
> > http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
> > View/change your membership options at
> > http://mailman.mit.edu/mailman/listinfo/piclist
> >
> --
> http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
> View/change your membership options at
> http://mailman.mit.edu/mailman/listinfo/piclist
>
--
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
Sean Breheny
2017-08-25 13:03:07 UTC
Permalink
Hi James,

I am also a long term hobbyist. I have an undergraduate degree in RE and an
MS in mechatronics/feedback control systems. I've been working in
industrial robotics for 12 years.

I see from your Sig that you are a filmmaker - does that have any
connection with your interest in electronics or totally separate?

Sean

On Aug 25, 2017 4:00 AM, "James Burkart" <***@burkartstudios.com> wrote:

> I've got so much to learn. Sean, what's your education and background? I'm
> a junior working on my EET and only have a hobbyist background.
>
> --
> Sincerely,
>
> James Burkart
> *Filmmaker & Documentarian*
>
> *Burkart Studios*
> 925.667.7175 | Personal
> 415.738.2071 | Office
>
> *Web:* burkartstudios.com
> *Facebook:* facebook.com/burkartstudios
>
> On Thu, Aug 24, 2017 at 9:28 PM, Sean Breheny <***@cornell.edu> wrote:
>
> > Hi Bob,
> >
> > My main suggestion is to characterize the behavior by providing various
> > input signals like impulses, steps, white noise, swept sine, etc. and
> > looking at the output on a scope. You might also vary the output load.
> This
> > gives you a good idea that you are not close to going unstable and you
> > aren't experiencing "peaking" around one frequency band or a null in one
> > band.
> >
> > I've had problems in the past with circuits like this where the output
> > stage has a nonlinearity about zero (like crossover distortion) which can
> > make the gain be reduced substantially near zero. This can result in
> > instability by causing integrator wind-up while the output is in the
> > low-gain region, followed by massive overshoot when it exits the low-gain
> > region, followed by a compensatory wind-up in the other direction due to
> > the overshoot, etc. This problem can be dealt with by making sure that
> the
> > transistors in the output are always biased with some minimum current so
> > their input-output gain never goes below a certain value.
> >
> > Note, too, that the op-amp can be treated as it's own finite gain stage,
> > where you close the loop locally around it and then close the loop again
> > around the whole system. This can make analysis of the entire loop easier
> > because once you guarantee that the op-amp gain stage is itself stable,
> you
> > can then treat it as a pure gain block or as a simple compensator (like
> an
> > integrator or integrator plus proportional gain or a lead compensator or
> > lead-lag compensator, etc.) You are sorta doing that already in that the
> > compensation cap is turning the op-amp into a fast integrator.
> >
> > I have developed a "theorem" of simple control design (which I am sure
> has
> > been stated before but I've never seen it put this way) - given any
> stable
> > system P, one can always close the loop around P with an integrator H
> such
> > that the closed-loop system will have zero DC error and be stable, for
> some
> > value of integrator gain K in H, and the system will continue to be
> stable
> > and exhibit zero DC error for any integrator gain Kprime < K. To put it
> > more simply, you can always control a stable plant using a slow
> integrator
> > and obtain at least the improvement of zero DC error, but also usually
> some
> > amount of improvement in servo tracking and regulation against
> > disturbances.
> >
> > Sean
> >
> >
> > On Thu, Aug 24, 2017 at 12:32 PM, Bob Blick <***@outlook.com>
> wrote:
> >
> > > Hi y'all,
> > >
> > > I'm trying to build some audio power amplifiers in order to use up some
> > > parts I have left over from a previous life.
> > >
> > > In the past I've always built the input stages from discrete parts but
> > > this one uses a conventional 4558-style opamp. Then I follow it with
> > > voltage and current boosting parts and finally enclose the whole mess
> in
> > > negative feedback.
> > >
> > > In this aforementioned past life I designed a lot of power amps so I
> know
> > > to design each stage to have as linear and balanced response as
> > reasonable,
> > > keep the gain low around the slowest components etc.
> > >
> > > Of course with the added gain and relatively low speed of the output
> > > stage, the internally compensated opamp is not able to keep things
> > stable,
> > > so I've added an external compensation capacitor. It does work fine,
> and
> > > I'm not looking for super high fidelity, but it's always nice to make
> > > simple changes that reap big benefits.
> > >
> > > So my question is, any ideas about compensation or changes I could
> make?
> > >
> > > I've attached a block diagram. Note that my output stage is the
> > right-hand
> > > amplifier block, it contains many discrete components, and most
> > > importantly, it is inverting. Therefore I am using the non-inverting
> > input
> > > of the opamp for the negative feedback.
> > >
> > > Thank you!
> > >
> > > Bob
> > > --
> > > http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
> > > View/change your membership options at
> > > http://mailman.mit.edu/mailman/listinfo/piclist
> > >
> > >
> > --
> > http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
> > View/change your membership options at
> > http://mailman.mit.edu/mailman/listinfo/piclist
> >
> --
> http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
> View/change your membership options at
> http://mailman.mit.edu/mailman/listinfo/piclist
>
--
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
Sean Breheny
2017-08-25 13:04:25 UTC
Permalink
Degree in EE not RE

On Aug 25, 2017 9:03 AM, "Sean Breheny" <***@cornell.edu> wrote:

Hi James,

I am also a long term hobbyist. I have an undergraduate degree in RE and an
MS in mechatronics/feedback control systems. I've been working in
industrial robotics for 12 years.
--
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
Bob Blick
2017-09-03 17:24:23 UTC
Permalink
Hi Sean,
Sorry not to have replied sooner. Thanks very much for your input. I'd pretty much already done a lot of the things you suggested, so it was great to get the reinforcement. I'm trying to squeak a little too much from too few parts and mostly getting away with it. Unfortunately I don't have all the same test equipment that I used to, and what I have now is more RF than audio, and I don't have a low distortion oscillator anymore. But as I said, I'm not looking for super low distortion.

Running the amplifier open loop shows it to be pretty well behaved. But I am using mosfet output devices and the driver stage is running class A, so it's definitely slew rate limited due to the low available gate current. But it sounds good and I think I can get away with it up to about 100V between power rails. Unfortunately the power transformers I have will net me 110V so I'll probably add a class AB driver stage.

Mosfets are strange creatures in audio amps. I was hoping that by using them in their linear range I would not be up against the full input capacitance and gate charge, but unfortunately it has not worked out that way. I also need to run a lot of quiescent current or else they are a little squirrely(actually more opossum-y) around zero. I suppose a fancy delta-Vbe circuit might be able to moderate the need for high quiescent current. But this is a low parts-count design.

I'll post the schematic for this and also whatever it is I end up finishing with as soon as I get a chance.

Thanks again,
Bob
________________________________________
From: piclist-***@mit.edu <piclist-***@mit.edu> on behalf of Sean Breheny
Sent: Thursday, August 24, 2017 8:28 PM
To: Microcontroller discussion list - Public.
Subject: Re: [EE] amplifier compensation and stability

Hi Bob,

My main suggestion is to characterize the behavior by providing various
input signals like impulses, steps, white noise, swept sine, etc. and
looking at the output on a scope. You might also vary the output load. This
gives you a good idea that you are not close to going unstable and you
aren't experiencing "peaking" around one frequency band or a null in one
band.

I've had problems in the past with circuits like this where the output
stage has a nonlinearity about zero (like crossover distortion) which can
make the gain be reduced substantially near zero. This can result in
instability by causing integrator wind-up while the output is in the
low-gain region, followed by massive overshoot when it exits the low-gain
region, followed by a compensatory wind-up in the other direction due to
the overshoot, etc. This problem can be dealt with by making sure that the
transistors in the output are always biased with some minimum current so
their input-output gain never goes below a certain value.

Note, too, that the op-amp can be treated as it's own finite gain stage,
where you close the loop locally around it and then close the loop again
around the whole system. This can make analysis of the entire loop easier
because once you guarantee that the op-amp gain stage is itself stable, you
can then treat it as a pure gain block or as a simple compensator (like an
integrator or integrator plus proportional gain or a lead compensator or
lead-lag compensator, etc.) You are sorta doing that already in that the
compensation cap is turning the op-amp into a fast integrator.

I have developed a "theorem" of simple control design (which I am sure has
been stated before but I've never seen it put this way) - given any stable
system P, one can always close the loop around P with an integrator H such
that the closed-loop system will have zero DC error and be stable, for some
value of integrator gain K in H, and the system will continue to be stable
and exhibit zero DC error for any integrator gain Kprime < K. To put it
more simply, you can always control a stable plant using a slow integrator
and obtain at least the improvement of zero DC error, but also usually some
amount of improvement in servo tracking and regulation against
disturbances.

Sean


On Thu, Aug 24, 2017 at 12:32 PM, Bob Blick wrote:

> Hi y'all,
>
> I'm trying to build some audio power amplifiers in order to use up some
> parts I have left over from a previous life.
>
> In the past I've always built the input stages from discrete parts but
> this one uses a conventional 4558-style opamp. Then I follow it with
> voltage and current boosting parts and finally enclose the whole mess in
> negative feedback.
>
> In this aforementioned past life I designed a lot of power amps so I know
> to design each stage to have as linear and balanced response as reasonable,
> keep the gain low around the slowest components etc.
>
> Of course with the added gain and relatively low speed of the output
> stage, the internally compensated opamp is not able to keep things stable,
> so I've added an external compensation capacitor. It does work fine, and
> I'm not looking for super high fidelity, but it's always nice to make
> simple changes that reap big benefits.
>
> So my question is, any ideas about compensation or changes I could make?
>
> I've attached a block diagram. Note that my output stage is the right-hand
> amplifier block, it contains many discrete components, and most
> importantly, it is inverting. Therefore I am using the non-inverting input
> of the opamp for the negative feedback.
>
> Thank you!
>
> Bob
Loading...