[John Byrns]

This post is not the start of a tutorial on negative feedback; it is a
question that I have been meaning to ask for several years, that I don't
think has been addressed by any of the many experts that have posted
tutorials here on negative feedback and how to stabilize feedback
amplifiers.

The phase plot produced by my "HF Step Network Calculator" reminded me
again of this question.

My question is inspired by the fact that many solid-state amplifiers use
dominant pole stabilization, while vacuum tube amplifiers typically use
a High Frequency Gain Stepping Network of the type championed in this
group by Patrick Turner.

The question is what are the advantages of each of these two seemingly
similar techniques, and why is one used primarily with solid-state
amplifiers, while the other is favored with vacuum tube amplifiers?

The High Frequency Gain Stepping Network is very similar in form to the
dominant pole technique, with a resistor added in series with the
capacitor to add a high frequency zero to the response. As the
frequency increases, the loss of the dominant pole compensation network
continues to increase and the phase approaches -90 degrees, while with
the High Frequency Gain Stepping Network the loss stops increasing at a
frequency determined by the added zero in the response, and the phase
tends towards zero at higher frequencies.

So what is it about these two networks that cause each to be favored in
its particular venue?


Regards,

John Byrns

--
Surf my web pages at, http://fmamradios.com/

[Phil Allison]

"John Byrns"

This post is not the start of a tutorial on negative feedback; it is a
question that I have been meaning to ask for several years, that I don't
think has been addressed by any of the many experts that have posted
tutorials here on negative feedback and how to stabilize feedback
amplifiers.

The phase plot produced by my "HF Step Network Calculator" reminded me
again of this question.

My question is inspired by the fact that many solid-state amplifiers use
dominant pole stabilization, while vacuum tube amplifiers typically use
a High Frequency Gain Stepping Network of the type championed in this
group by Patrick Turner.

The question is what are the advantages of each of these two seemingly
similar techniques, and why is one used primarily with solid-state
amplifiers, while the other is favored with vacuum tube amplifiers?

The High Frequency Gain Stepping Network is very similar in form to the
dominant pole technique, with a resistor added in series with the
capacitor to add a high frequency zero to the response. As the
frequency increases, the loss of the dominant pole compensation network
continues to increase and the phase approaches -90 degrees, while with
the High Frequency Gain Stepping Network the loss stops increasing at a
frequency determined by the added zero in the response, and the phase
tends towards zero at higher frequencies.

So what is it about these two networks that cause each to be favored in
its particular venue?


** At great personal risk - I will hazard a reply, cos this looks like a
genuine question.

The short answer is that SS amps *are* different to tube amps - as they use
power transistors to directly drive the speaker and have no output
transformer.

In a hi-fi tube amp, the OT is the dominant cause of high frequency
ll-off - but in a given design all examples are the same and stay the
same for life. So the designer can tailor the frequency compensation very
accurately.

In a hi-fi SS amp, the dominant cause of high frequency roll-off is the
power transistors - which have wide manufacturing variations and change
their effective bandwidth with applied voltage, current flow and
temperature.

So it ain't generally possible to precisely tailor the frequency
compensation - except maybe on an individual amp by amp basis - and have
all examples remain reliably stable under all conditions.

Luckily, a SS amp can be configured to have huge open loop voltage gain so
allowing for huge amounts of NFB to be applied. By using single (ie
dominant) pole compensation at a sufficiently low frequency, production and
operating condition variations affecting the bandwidth of the output stage
of the amp can be rendered harmless.

( Plus, with the addition of a simple LC output network, the amp becomes
unconditionally stable as well. )

It is no big problem to produce a SS amp where the open loop THD is around
1% (across a the audio band) with an 8 ohm load - using dominant pole
compensation, commonly 60 dB of NFB is applied at 1kHz dropping to 34dB at
20 kHz.

This then reduces the 1% open loop figure to 0.001 % at 1 kHz - rising to
only 0.02 % at 20 kHz.



...... Phil

[Alex]

"John Byrns" wrote in message
...

This post is not the start of a tutorial on negative feedback; it is a
question that I have been meaning to ask for several years, that I don't
think has been addressed by any of the many experts that have posted
tutorials here on negative feedback and how to stabilize feedback
amplifiers.

The phase plot produced by my "HF Step Network Calculator" reminded me
again of this question.

My question is inspired by the fact that many solid-state amplifiers use
dominant pole stabilization, while vacuum tube amplifiers typically use
a High Frequency Gain Stepping Network of the type championed in this
group by Patrick Turner.

The question is what are the advantages of each of these two seemingly
similar techniques, and why is one used primarily with solid-state
amplifiers, while the other is favored with vacuum tube amplifiers?

The High Frequency Gain Stepping Network is very similar in form to the
dominant pole technique, with a resistor added in series with the
capacitor to add a high frequency zero to the response. As the
frequency increases, the loss of the dominant pole compensation network
continues to increase and the phase approaches -90 degrees, while with
the High Frequency Gain Stepping Network the loss stops increasing at a
frequency determined by the added zero in the response, and the phase
tends towards zero at higher frequencies.

So what is it about these two networks that cause each to be favored in
its particular venue?


Regards,

John Byrns

Agree with Phil's answer.
Generally, a SS amplifier would consist of at least 5 stages, counting
Darlingtons, cascodes, etc. Above certain frequency, transistor stages tend
to add phase shift, which quckly amounts to quite large numbers. In this
situation winning 40...50 degrees by a stepping network would result in a
very small gain-bandwidth improvement, besides a precise matching of a zero
to an existing pole is required. The latter is also difficult due to large
transistor performance variations.

Therefore it is so much simpler just to kill the loop gain to 0dB with one
integrator before the combined phase shift gets out of control.

In a tube amplifier we deal with 2-3 poles at quite stable and predictable
frequencies. Compensating one of them with a null is worth doing.

Regards,
Alex

[Phil Allison]

"Alex"

Agree with Phil's answer.


** The fat ******* sings !!!

Maybe he's not such a ******* after all.

I'll leave the fat part for more forensic evidence .......





...... Phil

[Stewart Pinkerton]

[quote=Alex;809047]"John Byrns" wrote in message
...

This post is not the start of a tutorial on negative feedback; it is a
question that I have been meaning to ask for several years, that I don't
think has been addressed by any of the many experts that have posted
tutorials here on negative feedback and how to stabilize feedback
amplifiers.

The phase plot produced by my "HF Step Network Calculator" reminded me
again of this question.

My question is inspired by the fact that many solid-state amplifiers use
dominant pole stabilization, while vacuum tube amplifiers typically use
a High Frequency Gain Stepping Network of the type championed in this
group by Patrick Turner.

The question is what are the advantages of each of these two seemingly
similar techniques, and why is one used primarily with solid-state
amplifiers, while the other is favored with vacuum tube amplifiers?

The High Frequency Gain Stepping Network is very similar in form to the
dominant pole technique, with a resistor added in series with the
capacitor to add a high frequency zero to the response. As the
frequency increases, the loss of the dominant pole compensation network
continues to increase and the phase approaches -90 degrees, while with
the High Frequency Gain Stepping Network the loss stops increasing at a
frequency determined by the added zero in the response, and the phase
tends towards zero at higher frequencies.

So what is it about these two networks that cause each to be favored in
its particular venue?


Regards,

John Byrns

Horses for courses. As the others have stated, an essential diference between typical tube and SS amps is the massive amount of DC/LF gain present in the SS amp, necessitating a 'sledgehammer' approach to NFB if HF stability is to be maintained. Of course, this doesn't have to be global feedback - local degeneration will achieve the same result, and will leave a stage-for-stage topology which is not far different from that of the typical tube amp, where degeneration within the envelope gives low gain with relatively wide bandwidth, so that the HF gain stepping approach is practical.

[Ian Thompson-Bell]

Phil Allison wrote:
"John Byrns"
This post is not the start of a tutorial on negative feedback; it is a
question that I have been meaning to ask for several years, that I don't
think has been addressed by any of the many experts that have posted
tutorials here on negative feedback and how to stabilize feedback
amplifiers.

The phase plot produced by my "HF Step Network Calculator" reminded me
again of this question.

My question is inspired by the fact that many solid-state amplifiers use
dominant pole stabilization, while vacuum tube amplifiers typically use
a High Frequency Gain Stepping Network of the type championed in this
group by Patrick Turner.

The question is what are the advantages of each of these two seemingly
similar techniques, and why is one used primarily with solid-state
amplifiers, while the other is favored with vacuum tube amplifiers?

The High Frequency Gain Stepping Network is very similar in form to the
dominant pole technique, with a resistor added in series with the
capacitor to add a high frequency zero to the response. As the
frequency increases, the loss of the dominant pole compensation network
continues to increase and the phase approaches -90 degrees, while with
the High Frequency Gain Stepping Network the loss stops increasing at a
frequency determined by the added zero in the response, and the phase
tends towards zero at higher frequencies.

So what is it about these two networks that cause each to be favored in
its particular venue?


** At great personal risk - I will hazard a reply, cos this looks like a
genuine question.

The short answer is that SS amps *are* different to tube amps - as they use
power transistors to directly drive the speaker and have no output
transformer.

In a hi-fi tube amp, the OT is the dominant cause of high frequency
ll-off - but in a given design all examples are the same and stay the
same for life. So the designer can tailor the frequency compensation very
accurately.


I know very little about tube power amps. At what frequency typically is
the the pole of an output transformer?

Cheers

Ian

[Phil Allison]

"Ian Thompson-Bell"


I know very little about tube power amps.


** Ain't that da truth.

And about many more things......


At what frequency typically is the the pole of an output transformer?


** With rated load attached, the best * tube output transformers* make 50
to 100 kHz at the -3dB point.

Non descript types for tube guitar amps etc make 10 to 20 kHz.



....... Phil

[Arny Krueger]

"Phil Allison" wrote in message


It is no big problem to produce a SS amp where the open
loop THD is around 1% (across a the audio band) with an 8
ohm load - using dominant pole compensation, commonly
60 dB of NFB is applied at 1kHz dropping to 34dB at 20
kHz.

Agreed. Transistor amplifiers don't have to be as nonlinear as the disciples
of Russell Hamm made them out to be back in the 1970s.

It is easy to build a solid state amp with like you say 1% THD or less at
full power, and far less at lower powers, across the audio band. Many tubed
amps would be lucky to do this well, even with some loop feedback applied.

The fact that SS technology makes it entirely feasible to have far more
open-loop gain, and greater basic bandwidth, is what makes the modern SS
power amps with unconditional stability and less than 0.03% THD so common
and so inexpensive.

The idea of that SS amps have to be massively nonlinear, and require massive
feedback to have acceptable linearity, is an old wive's tale.

[Phil Allison]

"Arny Krueger"
"Phil Allison"

It is no big problem to produce a SS amp where the open
loop THD is around 1% (across a the audio band) with an 8
ohm load - using dominant pole compensation, commonly
60 dB of NFB is applied at 1kHz dropping to 34dB at 20
kHz.

Agreed. Transistor amplifiers don't have to be as nonlinear as the
disciples of Russell Hamm made them out to be back in the 1970s.

(snip)

The idea of that SS amps have to be massively nonlinear, and require
massive feedback to have acceptable linearity, is an old wive's tale.


** Absolutely - I can just see an ancient, timber kitchen table - with
fuel stove glowing in the background and a dozen fishermen's wives sipping
cups of tea.

While the tube heads among them are studiously perusing tea leaves in the
bottom of their cups, the others are engaged in a hot debate re the merits
of global v. nested feedback in BJT power amps ......

Riveting stuff.



....... Phil

[Ian Iveson]

John Byrns wrote:

This post is not the start of a tutorial on negative
feedback; it is a
question that I have been meaning to ask for several
years, that I don't
think has been addressed by any of the many experts that
have posted
tutorials here on negative feedback and how to stabilize
feedback
amplifiers.

The phase plot produced by my "HF Step Network Calculator"
reminded me
again of this question.

Why the phase plot, in particular?

My question is inspired by the fact that many solid-state
amplifiers use
dominant pole stabilization, while vacuum tube amplifiers
typically use
a High Frequency Gain Stepping Network of the type
championed in this
group by Patrick Turner.

I wonder how universal this characterisation is? Maybe I am
not alone here in having no experience of SS amp design. A
link to a typical design illustrating your point would be
useful.

The question is what are the advantages of each of these
two seemingly
similar techniques, and why is one used primarily with
solid-state
amplifiers, while the other is favored with vacuum tube
amplifiers?

The High Frequency Gain Stepping Network is very similar
in form to the
dominant pole technique, with a resistor added in series
with the
capacitor to add a high frequency zero to the response.
As the
frequency increases, the loss of the dominant pole
compensation network
continues to increase and the phase approaches -90
degrees, while with
the High Frequency Gain Stepping Network the loss stops
increasing at a
frequency determined by the added zero in the response,
and the phase
tends towards zero at higher frequencies.

So what is it about these two networks that cause each to
be favored in
its particular venue?

It's often hard to say why other ppl do things the way they
do. History isn't always rational, it seems to me.

My own view, inspired by Morgan Jones, is that the measures
you use to ensure stability fall in this order of
preference:

1. Minimise the number of poles
2. Move all remaining poles to as high a frequency as
possible
3. If the dominant pole is not far enough from the next,
slug it.
4. If slugging would leave you with insufficient bandwidth,
fudge with a stepping network.

Where "far enough" is according to the rule that the loop
gain must not be greater than the ratio of the two most
dominant time constants. "Slugging" is moving a pole to a
lower frequency.

Presumably that last resort is least favoured because it
results in a wriggly phase response?

My impression is that the output transformer of a valve amp
combines two poles which tend to be, as a consequence of its
manufacture, close enough to the audio bandwidth so it may
be that they can't be separated far enough by slugging to
allow the required amount of feedback.

Perhaps with a typical SS amp, all the poles but one can be
moved far enough from the audio band so that slugging does
the job, in spite of all the gain in the loop?

A key determinant of available unmolested bandwidth appears
to be the "quality factor" of the output transformer, being
Lprimary/Lleakage. To address an issue I feel I left hanging
in an earlier conversation, it appears that for a given type
of transformer of a given quality of manufacture, the
quality factor is strongly related to the turns ratio.

Ian

[Ian Thompson-Bell]

Phil Allison wrote:
"Ian Thompson-Bell"


I know very little about tube power amps.


** Ain't that da truth.

And about many more things......


At what frequency typically is the the pole of an output transformer?


** With rated load attached, the best * tube output transformers* make 50
to 100 kHz at the -3dB point.

Non descript types for tube guitar amps etc make 10 to 20 kHz.



Thanks, so it is entirely possible this pole would be close to Miller
pole elsewhere in the circuit.

Cheers

Ian

[John Byrns]

In article ,
"Ian Iveson" wrote:

John Byrns wrote:

This post is not the start of a tutorial on negative
feedback; it is a
question that I have been meaning to ask for several
years, that I don't
think has been addressed by any of the many experts that
have posted
tutorials here on negative feedback and how to stabilize
feedback
amplifiers.

The phase plot produced by my "HF Step Network Calculator"
reminded me
again of this question.

Why the phase plot, in particular?

Because when I first thought about asking this question several years
ago, it was the realization that the phase shift returns to zero at high
frequencies, unlike with dominant pole compensation, that inspired the
question, the phase plot show this in a visual way remining me of my old
question.

My question is inspired by the fact that many solid-state
amplifiers use
dominant pole stabilization, while vacuum tube amplifiers
typically use
a High Frequency Gain Stepping Network of the type
championed in this
group by Patrick Turner.

I wonder how universal this characterisation is? Maybe I am
not alone here in having no experience of SS amp design. A
link to a typical design illustrating your point would be
useful.

Good question, the solid state designs I am familiar with are older
designs, but I assume the technology hasn't really advanced much in the
last 30 years.

The question is what are the advantages of each of these
two seemingly
similar techniques, and why is one used primarily with
solid-state
amplifiers, while the other is favored with vacuum tube
amplifiers?

The High Frequency Gain Stepping Network is very similar
in form to the
dominant pole technique, with a resistor added in series
with the
capacitor to add a high frequency zero to the response.
As the
frequency increases, the loss of the dominant pole
compensation network
continues to increase and the phase approaches -90
degrees, while with
the High Frequency Gain Stepping Network the loss stops
increasing at a
frequency determined by the added zero in the response,
and the phase
tends towards zero at higher frequencies.

So what is it about these two networks that cause each to
be favored in
its particular venue?

It's often hard to say why other ppl do things the way they
do. History isn't always rational, it seems to me.

I don't know, Phil and Alex seem to have presented a lot of good
insight, although I suspect there is a bit more that skipped.

My own view, inspired by Morgan Jones, is that the measures
you use to ensure stability fall in this order of
preference:

1. Minimise the number of poles
2. Move all remaining poles to as high a frequency as
possible
3. If the dominant pole is not far enough from the next,
slug it.
4. If slugging would leave you with insufficient bandwidth,
fudge with a stepping network.

Where "far enough" is according to the rule that the loop
gain must not be greater than the ratio of the two most
dominant time constants. "Slugging" is moving a pole to a
lower frequency.

Presumably that last resort is least favoured because it
results in a wriggly phase response?

What's wrong with that assuming it is true? While the phase response
with dominant pole compensation may very well be less "wriggly", it
isn't clear to me that the maximum phase error isn't greater with
dominant pole compensation.

My impression is that the output transformer of a valve amp
combines two poles which tend to be, as a consequence of its
manufacture, close enough to the audio bandwidth so it may
be that they can't be separated far enough by slugging to
allow the required amount of feedback.

Well then the zero of the HF gain stepping network may neatly cancel one
of the transformer's poles as Alex suggested, pretty clever I'd say.

Perhaps with a typical SS amp, all the poles but one can be
moved far enough from the audio band so that slugging does
the job, in spite of all the gain in the loop?

Presumably that is the case, and you simply "slug" it enough to get the
gain down to less than one by the time the phase hits 180 degrees.

A key determinant of available unmolested bandwidth appears
to be the "quality factor" of the output transformer, being
Lprimary/Lleakage. To address an issue I feel I left hanging
in an earlier conversation, it appears that for a given type
of transformer of a given quality of manufacture, the
quality factor is strongly related to the turns ratio.

Is there a reference available on the web that discusses this issue, why
the bandwidth is inversely proportional to the turns ratio?


Regards,

John Byrns

--
Surf my web pages at, http://fmamradios.com/

[Ian Iveson]

John Byrns wrote:

This post is not the start of a tutorial on negative
feedback; it is a
question that I have been meaning to ask for several
years, that I don't
think has been addressed by any of the many experts
that
have posted
tutorials here on negative feedback and how to
stabilize
feedback
amplifiers.

The phase plot produced by my "HF Step Network
Calculator"
reminded me
again of this question.

Why the phase plot, in particular?

Because when I first thought about asking this question
several years
ago, it was the realization that the phase shift returns
to zero at high
frequencies, unlike with dominant pole compensation, that
inspired the
question, the phase plot show this in a visual way
remining me of my old
question.

OK.

My question is inspired by the fact that many
solid-state
amplifiers use
dominant pole stabilization, while vacuum tube
amplifiers
typically use
a High Frequency Gain Stepping Network of the type
championed in this
group by Patrick Turner.

I wonder how universal this characterisation is? Maybe I
am
not alone here in having no experience of SS amp design.
A
link to a typical design illustrating your point would be
useful.

Good question, the solid state designs I am familiar with
are older
designs, but I assume the technology hasn't really
advanced much in the
last 30 years.

All I know is from JL Hood's "Valve and Transistor Audio
Amplifiers", which spans a long period but not up to the
present. He's dead now, sadly. But if they didn't need step
networks 30 yrs ago, I wouldn't expect they would use them
now, unless there is actually something good about step
networks.

The question is what are the advantages of each of
these
two seemingly
similar techniques, and why is one used primarily with
solid-state
amplifiers, while the other is favored with vacuum tube
amplifiers?

The High Frequency Gain Stepping Network is very
similar
in form to the
dominant pole technique, with a resistor added in
series
with the
capacitor to add a high frequency zero to the response.
As the
frequency increases, the loss of the dominant pole
compensation network
continues to increase and the phase approaches -90
degrees, while with
the High Frequency Gain Stepping Network the loss stops
increasing at a
frequency determined by the added zero in the response,
and the phase
tends towards zero at higher frequencies.

So what is it about these two networks that cause each
to
be favored in
its particular venue?

It's often hard to say why other ppl do things the way
they
do. History isn't always rational, it seems to me.

I don't know, Phil and Alex seem to have presented a lot
of good
insight, although I suspect there is a bit more that
skipped.

Other side of the coin, yes. Interesting to know why SS amps
can't easily use step networks, and why valve amps can.
Doesn't explain why they do, though, unless you assume that
step networks are preferable. I doubt it, but I may be
wrong. The most important question, I believe, is why many
valve amps can't cope without them.

My own view, inspired by Morgan Jones, is that the
measures
you use to ensure stability fall in this order of
preference:

1. Minimise the number of poles
2. Move all remaining poles to as high a frequency as
possible
3. If the dominant pole is not far enough from the next,
slug it.
4. If slugging would leave you with insufficient
bandwidth,
fudge with a stepping network.

Where "far enough" is according to the rule that the loop
gain must not be greater than the ratio of the two most
dominant time constants. "Slugging" is moving a pole to a
lower frequency.

Presumably that last resort is least favoured because it
results in a wriggly phase response?

What's wrong with that assuming it is true? While the
phase response
with dominant pole compensation may very well be less
"wriggly", it
isn't clear to me that the maximum phase error isn't
greater with
dominant pole compensation.


Well, Morgan warns that it is relatively easy to stabilise
an amp, but whether it sounds any good as a result is
another matter. Unfortunately he doesn't say why but,
considering he immediately follows that remark with the
priority list I reported (more or less), and appears to
present step networks as a last-ditch refuge for no-hopers,
I assume his scorn is related to sound quality. Menno
stresses the importance of what AFAIR he calls differential
phase error, on the grounds that a constant time delay,
leading to constant rate of change of phase with respect to
frequency, cannot be heard. Intuitively, I would expect
sudden departures from the ideal slope to have more audible
effects than gradual ones.

My impression is that the output transformer of a valve
amp
combines two poles which tend to be, as a consequence of
its
manufacture, close enough to the audio bandwidth so it
may
be that they can't be separated far enough by slugging to
allow the required amount of feedback.

Well then the zero of the HF gain stepping network may
neatly cancel one
of the transformer's poles as Alex suggested, pretty
clever I'd say.

Maybe, but can it be neat enough? Lp varies but that doesn't
impinge on HF. Does the leakage vary too? Effective primary
resistance changes a bit depending on frequency and
amplitude but maybe that's not very significant. Valve
output resistance varies with amplitude, does that matter?
Varies with age, too, and from one set of valves to another.
How neat does it need to be? Even if it is perfectly neat,
it may not sound good, perhaps. I don't get to listen to
many different amps.

Perhaps with a typical SS amp, all the poles but one can
be
moved far enough from the audio band so that slugging
does
the job, in spite of all the gain in the loop?

Presumably that is the case, and you simply "slug" it
enough to get the
gain down to less than one by the time the phase hits 180
degrees.

That's my assumption, and a bit more to give an adequate
margin.

A key determinant of available unmolested bandwidth
appears
to be the "quality factor" of the output transformer,
being
Lprimary/Lleakage. To address an issue I feel I left
hanging
in an earlier conversation, it appears that for a given
type
of transformer of a given quality of manufacture, the
quality factor is strongly related to the turns ratio.

Is there a reference available on the web that discusses
this issue, why
the bandwidth is inversely proportional to the turns
ratio?

I hoped you might ask again, but not because I've got a good
answer. Last time, I went ferretting to find where I'd got
the idea from, but in vain. It won't be a simple linear
proportionality, because I would assume highest quality
factor would be with a 1:1 ratio, considering transformers
are reversible (er...does that follow?), so it would be some
inverted U-shaped curve.

I tried for ages to isolate the turns ratio, T, from Menno's
convoluted but coherent formulae (see link below), but
didn't quite manage it, although I think I got as far as
convincing myself that it cannot be eliminated, and so must
have some effect. In any case, the relationship I'm looking
for may largely arise from the physical constraints of
practical transformer geometry, rather than first
principles.

So then I looked at all the data I have on particular
transformers, including a list of Plitrons, and there
appears to be a strong relationship between turns ratio and
quality factor. However, there also appears to be a strong
relationship between size (power handling ability) and
quality factor. Considering there is also quite strong
correlation between size and turns ratio (more powerful amps
tend to have parallel valves and hence lower turns ratios),
it's not easy to disentangle one cause from another. You can
see that from the very short list presented on the last page
he

http://www.plitron.com/PDF/Atcl_5_2.pdf

For the rest, Page 5 equation 4-5 is a good place to start,
then work back through the definitions of terms. Note that
the turns ratio is tangled up in Menno's parameter "a2", as
well as in beta, the impedance ratio.

So maybe you have more patience or ability than me, and can
express bandwidth as a function of T? The equations get very
long when fully unpacked.

cheers, Ian

[Iain Churches[_2_]]

"Ian Iveson" wrote in message
...

My own view, inspired by Morgan Jones, is that the measures you use to
ensure stability fall in this order of preference:

1. Minimise the number of poles
2. Move all remaining poles to as high a frequency as possible
3. If the dominant pole is not far enough from the next, slug it.
4. If slugging would leave you with insufficient bandwidth, fudge with a
stepping network.

Where "far enough" is according to the rule that the loop gain must not be
greater than the ratio of the two most dominant time constants. "Slugging"
is moving a pole to a lower frequency.

Presumably that last resort is least favoured because it results in a
wriggly phase response?

My impression is that the output transformer of a valve amp combines two
poles which tend to be, as a consequence of its manufacture, close enough
to the audio bandwidth so it may be that they can't be separated far
enough by slugging to allow the required amount of feedback.


Hello Ian. I agree, but I wonder why Morgan Jones regards
the step network as a fudge, when a series RC combination
in parallel with Rp is seen so often?

For those of us that do not have a specific electronic
engineering orienentated background, this subject of
stability is a tough one, so it is good to see it being
discussed here.

I can remember in workshop practice (more years
ago than I care to remember:-) being told that
problems with instability were greatly reduced if
the poles were kept apart by a frequency
factor of 20. But I am still not clear how they can
be "moved to as high a freequency as possible"

I have been fortunate recently in that the use of good
OPTs and tried and tested schematics have resulted in
good sounding stable amplifiers.

But I am still trying to get the steps (no pun intended) in
order for stabilising an amp.

As I see it, the first thing to do is plot the response open
loop to find the poles I recall having done this with no
step network across Rp. I also make careful note of the
frequencies at which the phase swings through 180 degrees.

Then I work on the Rp step network.

I test the stability by having the amp idle with an open
cct load, and then a cap of 0.22µF.

Then I close the loop, using a 5k multi-turn pot for Rfb,
and set the stability margin to 10dB or so. I then check
the input sensitivity. I aim for 0dB (0.775V) for a 25W
pp amp, and 0dBV for an amp of higher power.

I usually follow the Radford example of a series
RC combination in parallel with Rfb. I choose a
resistor 1/10th of the value of Rfb, and then with a
decade C box, find a value that gives an optimum
square wave at 5kHz.

Am I doing it correctly?

Best regards
Iain