What characteristics might I hope for in the response of my amp to brief
excursions beyond its capability?

Where are the problem areas in a typical valve circuit, and how might I avoid
them?

If these are the wrong questions, I would be grateful for answers to the right
ones.

cheers, Ian

On Tue, 02 May 2006 01:46:05 GMT, "Ian Iveson"
wrote:

What characteristics might I hope for in the response of my amp to brief
excursions beyond its capability?

Where are the problem areas in a typical valve circuit, and how might I avoid
them?

If these are the wrong questions, I would be grateful for answers to the right
ones.

On the contrary, excellent questions.

Others will respond, doublessly better than I could, about
driving grids into forward conduction and RC time constants
and such that you already know.

So let me turn the question back to you, and ask you
to first define "capability". Methinks that definition
is along the lines of what you're after.( And, I don't
want to get ambushed again thank you very much.)

All good fortune,

Chris Hornbeck
"Qian li zou dan qi"

Chris Hornbeck wrote

What characteristics might I hope for in the response of my amp to brief
excursions beyond its capability?

Where are the problem areas in a typical valve circuit, and how might I avoid
them?

If these are the wrong questions, I would be grateful for answers to the right
ones

On the contrary, excellent questions.

Thanks.

Others will respond, doubtlessly better than I could, about
driving grids into forward conduction and RC time constants
and such that you already know.

Not sure I do, completely. To tell you the truth, the whole world seems fuzzy
and indeterminate to me, I fear I am running out of time, and often I can't
remember if I used to know more or less or anything, ever.

So let me turn the question back to you, and ask you
to first define "capability". Methinks that definition
is along the lines of what you're after.( And, I don't
want to get ambushed again thank you very much.)

Well, the common measures of amplitude capability I know of are rated maximum
power into a given load (but it is not often said what happens beyond the limit
other than some kind of unacceptable distortion), and slew rate, which I take to
be a measure of HF limitation at full power. I assume there is an equivalent
expression of a LF full-power limit...saturation?

The first rarely distinguishes between the various possible sources of
limitation...power supply, active device characteristics, etc, so how much it
relates to average power, and how much to instantaneous, or anywhere between,
and what the difference is, is not often clear.

The other two can often be reduced to the first: amplitude limitations in
closed-loop circuits operating beyond the bandwidth of the open loop.

Perhaps its a question about sequence. As I raise the input voltage, or approach
a frequency limit at full rated power, what should fail first? Can I manage the
failures to make them more euphonic?

Also about recovery...how long it takes and how euphonic it is.

Jim, if you remember, had and probably has the beautifully simple view that all
problems could be avoided by building the amp many times bigger than you think
you need, and this seems to be a fashionable approach again in mainstream hi-fi.
My argument was always that, if you solve all the problems by adding headroom,
then we have nothing left to talk about. I had this quaint notion that it should
be possible and desirable to shape an amp precisely around the requirements of
the task, without room to spare.

I've come across a few interesting approaches. Adding class C stages to
accommodate peak demands, for example, or schemes involving active or floating
bias.

As for ambush: merely an unfortunate accident of style and perception, I hope.

All good fortune

I wish I could depend on it.

cheers, Ian

flipper wrote:

The 'tube vs transistor' debate often goes:

Transistor amplifiers usually incorporate lots of global NFB to
linearize them (and achieve those vanishingly small distortion
numbers) and that causes clipping to be abrupt, as if the peaks were
sliced off by a sharp knife, resulting in a large odd order harmonic
content creating a harsh, dissonant, sound. (plus lots of discussion
about masking effects, et al.)

But tube amplifiers with little, or no, NFB 'soft clip' introducing
even order harmonics that resemble, or compliment, a 'natural'
distribution and, so, 'sound good'. (along with discussions of
'synthetic' bass, et al.)

Actually it's not about even or odd order harmonics when cliiping.

Soft clipping simply produces fewer *high order* harmonics.

Graham

On Tue, 02 May 2006 14:45:00 GMT, "Ian Iveson"
wrote:

Not sure I do, completely. To tell you the truth, the whole world seems fuzzy
and indeterminate to me, I fear I am running out of time, and often I can't
remember if I used to know more or less or anything, ever.

Very cool.


Well, the common measures of amplitude capability I know of are rated maximum
power into a given load (but it is not often said what happens beyond the limit
other than some kind of unacceptable distortion), and slew rate, which I take to
be a measure of HF limitation at full power. I assume there is an equivalent
expression of a LF full-power limit...saturation?

If we can agree to stick to conventional electronics usages
of terms, then we can divide the "clipping" and "slew-rate
limiting" very cleanly. The former is classically thought
of as in-band and a threshhold effect, and the latter is,
by definition, a limiting condition to out-of-band overdrive.

The former arises from a mechanism that has no effect below
its threshhold; the latter arises from a mechanism that
operates at all levels and all frequencies.

The definition of slew-rate limit is analogous to the
limiting condition of an amplifier driven by an in-band
signal into completely square-waved clipped output.


As for ambush: merely an unfortunate accident of style and perception, I hope.

Of course. I refuse to endanger emoticons for civilian
purposes. Just the way I was raised.

Much thanks, as always,

Chris Hornbeck
"Qian li zou dan qi"

Pooh Bear wrote

But tube amplifiers with little, or no, NFB 'soft clip' introducing
even order harmonics that resemble, or compliment, a 'natural'
distribution and, so, 'sound good'. (along with discussions of
'synthetic' bass, et al.)

Actually it's not about even or odd order harmonics when clipping.

Soft clipping simply produces fewer *high order* harmonics.

Yes, odd or even depends on how symmetrical the clipping is, rather than how
sharp. Perhaps valve amps tend to have more SE stages? I dunno much about SS amp
circuits so I can't say.

Also, how hard a clip is depends partly on the characteristics of the devices,
as well as the amount of feedback.

Anyway, what interests me more here is not how the clipping is, but rather how
it changes, and how it comes and goes according to how long the brief overdrive
lasts.

Grid current in a normal cap-coupled output stage, for example, has a particular
set of onset and recovery characteristics...a particular clipping envelope, if
you like, in response to a brief overdrive. If I use big coupling caps it lasts
longer, on average, but isn't quite as bad while its there. Which should I
choose and why?

A sagging power supply has a different set of characteristics. An issue which
still excites debate is the size of the final PS capacitor before the power
stage. Some argue the bigger the better, and some say too big sounds bad. I
assume this has something to do with the shape of the sag, how it comes and
goes, as well as its depth.

Then there is the bias "pumping" of cathode bias, which tends to exacerbate an
overdrive condition, and has its own little life cycle.

Another consideration is the way feedback can not only sharpen the onset of and
recovery from clipping, but also modify all these other effects. Also its
absence at frequency extremes multiplies the amplitude handled by some stages, a
point I often see ignored. A brief overdrive outside the bandwidth of the open
loop is likely then to provoke a different cycle of failures than one in the
middle.

I was determined to add an extra stage to my amps and try some global feedback,
just for the hell of it, even if it only means I will better appreciate its
absence. But *so much* will change as a result, I think I would be better off
starting from scratch. Dammit, they took ages to make.

cheers, Ian

On Wed, 03 May 2006 03:39:14 GMT, "Ian Iveson"
wrote:

Grid current in a normal cap-coupled output stage, for example, has a particular
set of onset and recovery characteristics...a particular clipping envelope, if
you like, in response to a brief overdrive. If I use big coupling caps it lasts
longer, on average, but isn't quite as bad while its there.

Now, *this* will require some extrapolation.

Thanks, as always,

Chris Hornbeck
"Qian li zou dan qi"

Chris Hornbeck wrote

Very cool.

Actually there is real tragedy in old age, and I am not looking forward to it. I
have been tempted to raise the issue of senility, which given the nature of the
group is sadly on topic.

Well, the common measures of amplitude capability I know of are rated maximum
power into a given load (but it is not often said what happens beyond the
limit
other than some kind of unacceptable distortion), and slew rate, which I take
to
be a measure of HF limitation at full power. I assume there is an equivalent
expression of a LF full-power limit...saturation?

If we can agree to stick to conventional electronics usages
of terms, then we can divide the "clipping" and "slew-rate
limiting" very cleanly. The former is classically thought
of as in-band and a threshold effect, and the latter is,
by definition, a limiting condition to out-of-band overdrive.

Happy to stick to convention, yes please. What's the difference(s) between a
threshold effect and a limiting condition? For the rest of my questions, see
below.

The former arises from a mechanism that has no effect below
its threshold; the latter arises from a mechanism that
operates at all levels and all frequencies.

I am struggling to follow this. I get the first part OK, but the second part is
hazy. The same language problem, probably.

The definition of slew-rate limit is analogous to the
limiting condition of an amplifier driven by an in-band
signal into completely square-waved clipped output.

Er...once I'm lost I'm lost it seems. Got up too early maybe. You must admit
your sentences can be quite complicated, and there are not enough of them.

As for ambush: merely an unfortunate accident of style and perception, I hope.

Of course. I refuse to endanger emoticons for civilian
purposes. Just the way I was raised.

I was raised entirely free of emoticons. I experimented briefly with sentiment
but was embarrassed by the pathos. Now I use them occasionally, strictly for
tactical purposes.

cheers, Ian

below:


Good, thanks. I have been fishing for a better definition for some time. The few
books I have that mention it (J.L Hood, Morgan Jones) are uncharacteristically
evasive.

Other than from there, I got most of my idea from here. A mechanism cited once
or twice has been a cathode follower driving a capacitor.

The picture I have so far arose from the following considerations:

An assumption that slew rate limiting is an absolute limit to the slope, ie rate
of change, of output voltage.

For a sine wave, the slope is proportional to the product of frequency and
amplitude.

A frequency limit is imposed by the transient response of a linear system, or
for a system handling small signals. But this is a fixed frequency regardless of
amplitude, so is not proportional to slew rate.

Since slew rate limiting cannot happen with a small signal, or with a linear
system, I took a leap at this point and decided it must be an amplitude
phenomenon in some sense. Reductionism at its worst, I know, but it is tempting
to envisage a fixed frequency increasing in amplitude until a slew rate limit is
reached. Equally, it could be seen as a signal of fixed amplitude, rising in
frequency until that limit is reached.

The result in either case appears as the opposite of clipping on the voltage
waveform...which becomes triangulated...more pointy rather than less, as the
slope from peak to peak is minimised.

The next leap was to assume that a triangular voltage must be the consequence of
a clipped current, in a situation where the current waveform is shifted by 90
degrees from the voltage.

Perhaps where I go wrong is to think of current clipping, in a context in which
the output is defined as a voltage?

Anyway, there are several ways to think of the CF example. I prefer not to think
in terms of current limiting in this context, on reflection. Wasn't happy at the
time I first came across it (in an historic debate here, years ago), not happy
now, uneasy in between. "Can't supply current" just means "has a high effective
output impedance" as far as I now see. Just at the moment I am thinking of it
like this:

The transient response of the circuit is largely determined by the output
resistance of the CF, and the value of the capacitor. However, the former varies
with amplitude, and hence the transient response does too. Whereas, for a linear
system, transient response cannot be the culprit because it lacks the necessary
dependence on amplitude, now it has an amplitude dimension, and so could be the
perpetrator of SRL, or rather could itself *be* SRL. The concept of a dynamic TR
is tricky for me (there doesn't seem to be a general theory of nonlinear
systems), so I could be barking up the wrong tree. (or simply barking)

Anyway, the CF arrangement approaches slew rate limiting, if that's what it is,
as the CF approaches an amplitude limit, where its output impedance is at a
maximum. The effect is multiplied by the presence of feedback, or perhaps would
not be present without feedback, because it would resolve to the ordinary
considerations of frequency response and headroom.

So, I wonder if a dynamic transient response can be resolved into a fixed
frequency response and a slew-rate limit, somehow, by superposition, such that
the two together tell the whole story. Somehow I doubt it, but maybe there is
some merit in the idea.

I would be grateful for an alternative explanation, in terms of that or another
valve circuit, to illustrate the classic definition. Also, if the above example
is not of SRL, what is it?

Try to be less enigmatic.

Ian

Ian Iveson wrote:

What characteristics might I hope for in the response of my amp to brief
excursions beyond its capability?

Where are the problem areas in a typical valve circuit, and how might I avoid
them?

If these are the wrong questions, I would be grateful for answers to the right
ones.

cheers, Ian

If your amp is not DC coupled it may not be concrned much at all, since clipping
due to RC time constants in the OP tube grids will simply increase the -ve grid
bias.

But tweeters, if you are using them may be damaged due to the high frequencies in
the amp output due to the fast edges resulting from clipping. I've heard of
tweeter failure in conditions like this!

My thoughts, anyway. John Stewart

Ian Iveson wrote:

What characteristics might I hope for in the response of my amp to brief
excursions beyond its capability?

Where are the problem areas in a typical valve circuit, and how might I avoid
them?

If these are the wrong questions, I would be grateful for answers to the right
ones.

cheers, Ian

If your amp is not DC coupled it may not be concerned much at all, since clipping
due to RC time constants in the OP tube grids will simply increase the -ve grid
bias.

But tweeters, if you are using them may be damaged due to the high frequencies in
the amp output due to the fast edges resulting from clipping. I've heard of
tweeter failure in conditions like this!

My thoughts, anyway. John Stewart

On Wed, 03 May 2006 15:01:40 GMT, "Ian Iveson"
wrote:

Actually there is real tragedy in old age, and I am not looking forward to it. I
have been tempted to raise the issue of senility, which given the nature of the
group is sadly on topic.

Arf! / Ouch!


What's the difference(s) between a
threshold effect and a limiting condition?

How about this?: A threshold effect doesn't matter (however
we define that term) "below" (it's in our Judeo-Christian
upbringing) some threshold. To use your doubtlessly better
spelling. Until an amplifier clips, there's no clipping
"effect". Like that.

A limiting condition is much farther/further/father down
the food chain. The limiting condition of an amplifier
clipping is to be overdriven in-band into square waves,
where they don't get any squarer or bigger or nuttin'.

Slewing is the limiting condition where the amplifier
is driven out-of-band with signal whose dV/dT is so
large that output is no longer related to input; it
becomes only a function of charge rate of capacitors.


Try to be less enigmatic.

The cathode follower example is too complicated and obscure
to be useful to us. Slew rate is a concept from op-amps
and for a simple and elegant explanation you could do
worse than to read the Walt Jung books about op-amp theory
and practice. IMO, all of us would benefit from the
perspective, because it rationalizes several otherwise
difficult models, especially stability issues and the
difference between small-signal and large-signal behavior
with fast signals.

To (poorly) summarize slewing using op-amp terms, meaning
amplifiers with long-loop feedback and a single dominant
pole open-loop response: the amplifier's response (same
number for open- or closed- loop!) to a step input can be
expressed equally well as a time rate of change of output
voltage or as a ratio of charging current to the value of
the compensation capacitor. Same-same.

As you might be wondering, this translates only in raggedy-
assed ways to valve amplifiers. Low open loop gains and
low long-loop feedback make a horse of a different color.

For our purposes, behavior long, long before slewing is all
that matters, although the ability to charge and discharge
stray capacitances linearly is still "in-band" for us.

And, "clipping" is trivial; just say no.

Much thanks, as always,

Chris Hornbeck
"Qian li zou dan qi"

Chris Hornbeck wrote:

To (poorly) summarize slewing using op-amp terms, meaning
amplifiers with long-loop feedback and a single dominant
pole open-loop response: the amplifier's response (same
number for open- or closed- loop!) to a step input can be
expressed equally well as a time rate of change of output
voltage or as a ratio of charging current to the value of
the compensation capacitor. Same-same.

As you might be wondering, this translates only in raggedy-
assed ways to valve amplifiers. Low open loop gains and
low long-loop feedback make a horse of a different color.

Actually feedback is largely irrelevant to slew rate.

As you say it's determined ( ususally by one dominant stage ) by the ability to
charge a capacitive node. You can't slew faster by applying feedback ( although you
can extend bandwidth - which is different ).

Graham

Pooh Bear wrote

Actually feedback is largely irrelevant to slew rate.

As you say it's determined ( ususally by one dominant stage ) by the ability
to
charge a capacitive node. You can't slew faster by applying feedback (
although you
can extend bandwidth - which is different ).

No, but you do need infinite gain. To be useful in any circuit, feedback would
be a necessary addition.

Or so it seems to me. For the transient response to be truly a straight line,
there must be a discontinuity...an instantaneous change in dv/dt, requiring
infinite d2v/dt2.

cheers, Ian

Ian Iveson wrote:

Pooh Bear wrote

Actually feedback is largely irrelevant to slew rate.

As you say it's determined ( ususally by one dominant stage ) by the ability
to
charge a capacitive node. You can't slew faster by applying feedback (
although you
can extend bandwidth - which is different ).

No, but you do need infinite gain.

Pardon ? Since when does any circuit have 'infinite gain' ?

To be useful in any circuit, feedback would
be a necessary addition.

Sorry, addition to what ?

Or so it seems to me. For the transient response to be truly a straight line,
there must be a discontinuity...an instantaneous change in dv/dt, requiring
infinite d2v/dt2.

I see what you mean, but how then do you generate an input signal with infinite
d2V/dt2 ? Not possible. All siganls are band limited.

Graham

Ian Iveson wrote:

Pooh Bear wrote

But tube amplifiers with little, or no, NFB 'soft clip' introducing
even order harmonics that resemble, or compliment, a 'natural'
distribution and, so, 'sound good'. (along with discussions of
'synthetic' bass, et al.)

Actually it's not about even or odd order harmonics when clipping.

Soft clipping simply produces fewer *high order* harmonics.

Yes, odd or even depends on how symmetrical the clipping is, rather than how
sharp. Perhaps valve amps tend to have more SE stages? I dunno much about SS amp
circuits so I can't say.

Also, how hard a clip is depends partly on the characteristics of the devices,
as well as the amount of feedback.

Anyway, what interests me more here is not how the clipping is, but rather how
it changes, and how it comes and goes according to how long the brief overdrive
lasts.

Grid current in a normal cap-coupled output stage, for example, has a particular
set of onset and recovery characteristics...a particular clipping envelope, if
you like, in response to a brief overdrive. If I use big coupling caps it lasts
longer, on average, but isn't quite as bad while its there. Which should I
choose and why?

A sagging power supply has a different set of characteristics. An issue which
still excites debate is the size of the final PS capacitor before the power
stage. Some argue the bigger the better, and some say too big sounds bad. I
assume this has something to do with the shape of the sag, how it comes and
goes, as well as its depth.

Then there is the bias "pumping" of cathode bias, which tends to exacerbate an
overdrive condition, and has its own little life cycle.

Norman Crowhurst has pointed out that the cathode biased, CR coupled output stage
was a better solution in regards to OD problems than a similar fixed bias, CR
coupled output stage. His cure for the fixed bias OP stage is CF drivers to each of
the OP control grids. That avoids blocking caused by the extra charge on the grid
side of the coupling cap after the OD state has passed.

I've used that remedy several times with good success. It really works very well.

Cheers, John Stewart

Hi John.

Then there is the bias "pumping" of cathode bias, which tends to exacerbate
an
overdrive condition, and has its own little life cycle.

Norman Crowhurst has pointed out that the cathode biased, CR coupled output
stage
was a better solution in regards to OD problems than a similar fixed bias, CR
coupled output stage.

The two pumping effects cancel I guess.

His cure for the fixed bias OP stage is CF drivers to each of
the OP control grids. That avoids blocking caused by the extra charge on the
grid
side of the coupling cap after the OD state has passed.

I've used that remedy several times with good success. It really works very
well.

How do you work out the size of cap to use in such a case?

cheers, Ian

Pooh Bear wrote

Actually feedback is largely irrelevant to slew rate.

As you say it's determined ( ususally by one dominant stage ) by the
ability
to
charge a capacitive node. You can't slew faster by applying feedback (
although you
can extend bandwidth - which is different ).

No, but you do need infinite gain.

Pardon ? Since when does any circuit have 'infinite gain' ?

I was responding to Chris's contention that SRL is independent of amplitude. For
that truly to be the case, then it must be possible even with an input signal of
infinitessimally small amplitude...for small signal analysis. It's a theoretical
point. If it does happen following such a signal, and the signal is a perfect
step, then it follows from the ensuing straight line response that there is no
limit to the steady state output. A straight line goes up to infinity as long as
its gradient is not zero. In a closed loop circuit, the gain would depend only
on the feedback proportion.

This is pretty close to what happens with opamps. I think the ideal opamp has
infinite gain. Obviously real opamps don't. Perhaps they are close enough,
however, so that SLR becomes a dominant effect in some circumstances. The
postulation of the perfect allows engineers to analyse. That's what the
principle of superposition is for.

To be useful in any circuit, feedback would
be a necessary addition.

Sorry, addition to what ?

To infinite gain.

Or so it seems to me. For the transient response to be truly a straight line,
there must be a discontinuity...an instantaneous change in dv/dt, requiring
infinite d2v/dt2.

I see what you mean, but how then do you generate an input signal with
infinite
d2V/dt2 ? Not possible. All siganls are band limited.

Of course. To clear up a matter of possible confusion, just because the
transient response displays an infinite d2V/dt2, doesn't mean that the input
signal does too. Not as a matter of definition, anyway.

However, in this case I was responding to Chris's point about response to a step
input, as square as can be no squarer. As it happens, the idealised
slew-rate-limited opamp would respond with an infinite d2V/dt2 no matter what.
To any real input, the total response would be triangular, at the frequency of
the input fundamental, and an amplitude depending entirely on that frequency, as
far as supply voltage would allow. A rail-to-rail trapezium.

For myself, I think that if the term "slew rate limiting" is applicable only to
the perfect case, then it has no application. If it applies to an imperfect
case, then why should it not be applied to a circuit which is not an opamp at
all? Also, to even approximate to this perfection, in the sense that an opamp
does, must surely involve clipping of some kind, somewhere. To put it in Chris's
terms, *all* SRL is "raggedy-assed".

What does "clipping" mean anyway? Can it happen to current, or only voltage?

cheers, Ian

On Mon, 08 May 2006 13:51:10 GMT, "Ian Iveson"
wrote:

Pooh Bear wrote

Actually feedback is largely irrelevant to slew rate.

As you say it's determined ( ususally by one dominant stage ) by the
ability
to
charge a capacitive node. You can't slew faster by applying feedback (
although you
can extend bandwidth - which is different ).

No, but you do need infinite gain.

Pardon ? Since when does any circuit have 'infinite gain' ?

I was responding to Chris's contention that SRL is independent of amplitude. For
that truly to be the case, then it must be possible even with an input signal of
infinitessimally small amplitude...for small signal analysis. It's a theoretical
point. If it does happen following such a signal, and the signal is a perfect
step, then it follows from the ensuing straight line response that there is no
limit to the steady state output. A straight line goes up to infinity as long as
its gradient is not zero. In a closed loop circuit, the gain would depend only
on the feedback proportion.

This is pretty close to what happens with opamps. I think the ideal opamp has
infinite gain. Obviously real opamps don't. Perhaps they are close enough,
however, so that SLR becomes a dominant effect in some circumstances. The
postulation of the perfect allows engineers to analyse. That's what the
principle of superposition is for.

To be useful in any circuit, feedback would
be a necessary addition.

Sorry, addition to what ?

To infinite gain.

Or so it seems to me. For the transient response to be truly a straight line,
there must be a discontinuity...an instantaneous change in dv/dt, requiring
infinite d2v/dt2.

I see what you mean, but how then do you generate an input signal with
infinite
d2V/dt2 ? Not possible. All siganls are band limited.

Of course. To clear up a matter of possible confusion, just because the
transient response displays an infinite d2V/dt2, doesn't mean that the input
signal does too. Not as a matter of definition, anyway.

However, in this case I was responding to Chris's point about response to a step
input, as square as can be no squarer. As it happens, the idealised
slew-rate-limited opamp would respond with an infinite d2V/dt2 no matter what.
To any real input, the total response would be triangular, at the frequency of
the input fundamental, and an amplitude depending entirely on that frequency, as
far as supply voltage would allow. A rail-to-rail trapezium.

For myself, I think that if the term "slew rate limiting" is applicable only to
the perfect case, then it has no application. If it applies to an imperfect
case, then why should it not be applied to a circuit which is not an opamp at
all? Also, to even approximate to this perfection, in the sense that an opamp
does, must surely involve clipping of some kind, somewhere. To put it in Chris's
terms, *all* SRL is "raggedy-assed".

What does "clipping" mean anyway? Can it happen to current, or only voltage?

cheers, Ian


I think you are allowing perfection to become an enemy of the useful
here. SRL is a clipping phenomenon. It is clipping of the input stage
current feeding the dominant pole capacitor in an op amp.

So what if D2V/dv2 isn't infinite. It just means that the transition
into slew rate limiting isn't a perfect corner - it has a little
roundness to it. Is this a reason to dismiss SRL out of hand? I don't
think so. It is still a vital term in the analysis of any circuit
involving capacitors and current sources.

In no way can SRL be described as "raggedy-assed" - it is far to
useful in the real world.

d

--
Pearce Consulting
http://www.pearce.uk.com

"Ian Iveson" a écrit dans le message de
news: ...
Hi John.

Then there is the bias "pumping" of cathode bias, which tends to
exacerbate an
overdrive condition, and has its own little life cycle.

Norman Crowhurst has pointed out that the cathode biased, CR coupled
output stage
was a better solution in regards to OD problems than a similar fixed
bias, CR
coupled output stage.

The two pumping effects cancel I guess.

His cure for the fixed bias OP stage is CF drivers to each of
the OP control grids. That avoids blocking caused by the extra charge on
the grid
side of the coupling cap after the OD state has passed.

I've used that remedy several times with good success. It really works
very well.

How do you work out the size of cap to use in such a case?

cheers, Ian


Hi,

I'm not a brilliant theorician, but i've read Crowhurst and others 'bout
overdrive and blocking being exacerbated by feedback.
Look at he

http://www.dissident-audio.com/PP_6L6/Page.html

Keys a
-Just enough gain.
-Only one link cap at a point where grid current is unlikely to occur.
-Regulated PSU for screens and driver.
-Just enough feedback.

Seems coherent to what already pointed in this post ;)

Yves.

On Mon, 08 May 2006 13:51:10 GMT, "Ian Iveson"
wrote:

I was responding to Chris's contention that SRL is independent of amplitude.

The slew rate limit is defined as a charge rate of a capacitor.
This can be translated into a variety of external, but related
numbers. It is, however, an *internal* function, and in that
context is unrelated to, rather than independent of, amplitude.
Does that make sense?

For that truly to be the case, then it must be possible even with an input signal of
infinitessimally small amplitude..

Exactimundo! The connection between small-signal and large-signal
was explored in-depth in the mid-1970's by folks like Matti Otala,
Jan Lohstroh, Robert Cordell, and Walt Jung's troupe. I've doubtlessly
forgotten others who contributed significantly.

However, in this case I was responding to Chris's point about response to a step
input, as square as can be no squarer.

It appears on first reading that you've conflated comments
about slewing with other comments about clipping. My poor
explanation skills are to blame, no doubt.

Thanks, as always,

Chris Hornbeck
"Qian li zou dan qi"

On Mon, 08 May 2006 14:04:06 GMT, (Don Pearce)
wrote:

In no way can SRL be described as "raggedy-assed" - it is far to
useful in the real world.

I'm being misquoted in the second derivative.

The mechanism that generates "SLR" is the same mechanism
that causes ordinary distortion into reactive loads in
ordinary circuits. Managing drive into real-world reactive
loads is the crux of the biscuit, indeed.

My contention is simply that the concept of slew rate limit
is *not* a useful one for studying valve audio amplifiers, per se.
Individual stages... yeah, you betcha, a (somewhat nebulous)
benchmark.

But valve audio amplifiers are so different from op-amps
as to make the global concept ragged-assed.

That's my story, an' I'm stickin' to it.

Thanks, as always,

Chris Hornbeck
"Qian li zou dan qi"

On Tue, 09 May 2006 03:42:50 GMT, Chris Hornbeck
wrote:

On Mon, 08 May 2006 14:04:06 GMT, (Don Pearce)
wrote:

In no way can SRL be described as "raggedy-assed" - it is far to
useful in the real world.

I'm being misquoted in the second derivative.

The mechanism that generates "SLR" is the same mechanism
that causes ordinary distortion into reactive loads in
ordinary circuits. Managing drive into real-world reactive
loads is the crux of the biscuit, indeed.

My contention is simply that the concept of slew rate limit
is *not* a useful one for studying valve audio amplifiers, per se.
Individual stages... yeah, you betcha, a (somewhat nebulous)
benchmark.

But valve audio amplifiers are so different from op-amps
as to make the global concept ragged-assed.

That's my story, an' I'm stickin' to it.

Thanks, as always,

Chris Hornbeck
"Qian li zou dan qi"

I haven't done the sums, but do typical valve amps slew rate limit
into typical speaker reactances with audio signals? I've always found
voltage clipping to be the dominant factor.

d

--
Pearce Consulting
http://www.pearce.uk.com

Chris Hornbeck wrote

In no way can SRL be described as "raggedy-assed" - it is far to
useful in the real world.

I'm being misquoted in the second derivative.

The mechanism that generates "SLR" is the same mechanism
that causes ordinary distortion into reactive loads in
ordinary circuits. Managing drive into real-world reactive
loads is the crux of the biscuit, indeed.

My contention is simply that the concept of slew rate limit
is *not* a useful one for studying valve audio amplifiers, per se.
Individual stages... yeah, you betcha, a (somewhat nebulous)
benchmark.

But valve audio amplifiers are so different from op-amps
as to make the global concept ragged-assed.

That's my story, an' I'm stickin' to it.


Oi!

The specific circuit I cited was a CF with a capacitor load. I took it that your
"raggedy-assed" comment applied to that.

I would be happy with the conclusion that all SRL is RA, in the sense that it is
an idealised concept applied to a real phenomenon. "Triangle" would be another
example. There are no real triangles, and yet the word is useful for describing
things that approach the ideal.

But that's just to say we live in a RA world, which we could have taken for
granted at the outset. RASRL is tautological.

I am sure also that I was told that SRL is nothing to do with clipping, by
several ppl.

If there is one cloud in the sky and I say "look at that triangular cloud", it
would be pointless to argue about triangularity. How RA must SRL be before it
ceases to answer to the description? Surely that comes down to common sense? All
that's lacking here is mutual respect.

Thanks everyone for helping me to explore this stuff. Here are some tentative
conclusions and a few more begging questions.

**Slew rate limiting is a condition in which the rate of change of output
voltage is absolutely limited by a constant value other than zero.

**This condition may arise for only a portion of the output voltage. For
example, it may occur only beyond a certain amplitude, in which case the
response to a sine wave may have a flat imposed at some point other than the
zero-crossing point, where input dv/dt is greatest.

**A simple way of producing SRL is by placing a constant current source in
series with the signal path, and following it with a capacitor as a load. Is
this part of the definition? Or is it simply the *only* way of imposing a
constant dv/dt? Or neither? I suggest that the *cause* of SRL is only an
accidental part of
its definition, perhaps because it was the way it first occurred. If there is
another way of producing an absolute limit to output dv/dt, is it still SRL?

**A CCS can result if the output impedance and voltage output of a circuit are
proportional, so that V/R is constant. Saturation (forcing *both* to a constant
value) is one way of achieving this, but can it occur without saturation? I
think it should be possible. I think a CF into a capacitive load achieves SRL on
both counts, depending on voltage amplitude. The output, in response to a sine
input, becomes triangular. The triangle is never perfect...about as triangular
as a triangular cloud.

**SRL is necessarily a condition arising from non-linearity. The response is
neither a transient response (which always reduces to zero eventually) nor a
steady state response (which is always, er, steady), so has nothing to do with
linear systems.

**I had in mind the idea that, if one linear system had a transient response
which was complimentary to another, then the wiggles of one would be cancelled
by those of the other when the two were placed in series, resulting in a
straight line. I now think this is nonsense, but I'm not sure why. I do know
that transient responses don't have sharp corners

**With respect to the relationship between feedback and SRL I would like to
explore the following idea a bit further:

The steady state output of a linear system with a sine wave input is a sine wave
with some delay which arises from the transient response of the system. This
constant delay translates into a phase shift that increases with frequency.

The output may then be expressed as a sum of cosine and sine. The proportion of
the two then varies, from all sine at the centre of bandwidth, to all cosine
when the output is shifted by 90 degrees, beyond which -sine begins to appear,
etc.

It is also true that d(sin(t))dt = cos(t).

It can therefore be said that feedback increasingly includes a proportion of
dv/dt, until when the output of a closed loop system is shifted by 90 degrees,
the feedback is entirely dv/dt. In that case the effective input is V-K.dV/dt,
where V is the input voltage and K is some constant.

It seems on the face of it that, given enough feedback, dV/dt can become the
dominant feature of the input, such that, when it is large, the effective input
is also large. Hence if the system is amplitude limited, then the limit can be
reached as a result of a signal with a dV/dt beyond a certain limit.

Perhaps it could equally be said that SRL is clipping at an angle.

Just a raggedy-assed way of looking at the same thing, maybe.

cheers, Ian

On Tue, 09 May 2006 16:27:31 GMT, "Ian Iveson"
wrote:


The specific circuit I cited was a CF with a capacitor load. I took it that your
"raggedy-assed" comment applied to that.

I would be happy with the conclusion that all SRL is RA, in the sense that it is
an idealised concept applied to a real phenomenon. "Triangle" would be another
example. There are no real triangles, and yet the word is useful for describing
things that approach the ideal.

But that's just to say we live in a RA world, which we could have taken for
granted at the outset. RASRL is tautological.

Voltage clipping has non-sharp corners too. Would you call that RA? It
is really just a silly term for describing stuff that already has a
perfectly adequate cogent description. Just drop it.

I am sure also that I was told that SRL is nothing to do with clipping, by
several ppl.

It is clipping. It is just current clipping rather than voltage
clipping. It looks different because it is being integrated by a
capacitor. You are witnessing the integral of a clipped waveform.

If there is one cloud in the sky and I say "look at that triangular cloud", it
would be pointless to argue about triangularity. How RA must SRL be before it
ceases to answer to the description? Surely that comes down to common sense? All
that's lacking here is mutual respect.

Please drop the RA in favour of something that means something. If you
mean soft-cornered, say so. SRL is defined by the straight line, not
how clean the corners are. They simply define the onset of current
clipping.

Thanks everyone for helping me to explore this stuff. Here are some tentative
conclusions and a few more begging questions.

**Slew rate limiting is a condition in which the rate of change of output
voltage is absolutely limited by a constant value other than zero.

No, it is a condition where the rate of change of output voltage is
inversely proportional to the capacitance being charged.

**This condition may arise for only a portion of the output voltage. For
example, it may occur only beyond a certain amplitude, in which case the
response to a sine wave may have a flat imposed at some point other than the
zero-crossing point, where input dv/dt is greatest.

No it is not amplitude related. It is rise-time related. The amplitude
is immaterial.

**A simple way of producing SRL is by placing a constant current source in
series with the signal path, and following it with a capacitor as a load. Is
this part of the definition? Or is it simply the *only* way of imposing a
constant dv/dt? Or neither? I suggest that the *cause* of SRL is only an
accidental part of
its definition, perhaps because it was the way it first occurred. If there is
another way of producing an absolute limit to output dv/dt, is it still SRL?

SRL can have a variety of causes, but in and around audio almost the
only one you ever encounter is the constant current source of the
input stage charging the dominant pole capacitor. You will only
encounter it these days in an incompetently designed amplifier. All
amplifiers should have a frequency response that limits signal rise
time to a value less than that which will trigger SRL.

**A CCS can result if the output impedance and voltage output of a circuit are
proportional, so that V/R is constant. Saturation (forcing *both* to a constant
value) is one way of achieving this, but can it occur without saturation? I
think it should be possible. I think a CF into a capacitive load achieves SRL on
both counts, depending on voltage amplitude. The output, in response to a sine
input, becomes triangular. The triangle is never perfect...about as triangular
as a triangular cloud.

Very muddled. a SRL results from an impedance going very high, not
proportional to voltage - this is a definition of a current source.
The triangular output in response to a sine wave won't have sharp
corners. Why should it? There will be no SRL near the peaks of the
sine wave where the rate of change of voltage is low.

**SRL is necessarily a condition arising from non-linearity. The response is
neither a transient response (which always reduces to zero eventually) nor a
steady state response (which is always, er, steady), so has nothing to do with
linear systems.

SRL is a result of current clipping - as non linear as you can get. As
for the transient response, I think you mean the error reduces to
zero, not the response. A perfect transient response would be 1 - the
output is exactly equal to the input at all times.

**I had in mind the idea that, if one linear system had a transient response
which was complimentary to another, then the wiggles of one would be cancelled
by those of the other when the two were placed in series, resulting in a
straight line. I now think this is nonsense, but I'm not sure why. I do know
that transient responses don't have sharp corners

Not nonsense at all. This is done all the time in radio - google for
all-pass networks. As for transient responses and sharp corners - they
need an infinite frequency response, so no, you won't see them.


**With respect to the relationship between feedback and SRL I would like to
explore the following idea a bit further:

The is no relationship whatever.

The steady state output of a linear system with a sine wave input is a sine wave
with some delay which arises from the transient response of the system. This
constant delay translates into a phase shift that increases with frequency.

The output may then be expressed as a sum of cosine and sine. The proportion of
the two then varies, from all sine at the centre of bandwidth, to all cosine
when the output is shifted by 90 degrees, beyond which -sine begins to appear,
etc.

It is also true that d(sin(t))dt = cos(t).

It can therefore be said that feedback increasingly includes a proportion of
dv/dt, until when the output of a closed loop system is shifted by 90 degrees,
the feedback is entirely dv/dt. In that case the effective input is V-K.dV/dt,
where V is the input voltage and K is some constant.

It seems on the face of it that, given enough feedback, dV/dt can become the
dominant feature of the input, such that, when it is large, the effective input
is also large. Hence if the system is amplitude limited, then the limit can be
reached as a result of a signal with a dV/dt beyond a certain limit.

Didn't follow any of that - but the conclusion was wrong, so I presume
so was the reasoning.

Perhaps it could equally be said that SRL is clipping at an angle.


No. It is just clipping. The angle you see is the result of
integrating that clipping with a capacitor.

Just a raggedy-assed way of looking at the same thing, maybe.

cheers, Ian


There's that adjective again. Do you use it to describe absolutely
everything?

d

--
Pearce Consulting
http://www.pearce.uk.com

Don Pearce wrote

The specific circuit I cited was a CF with a capacitor load. I took it that
your
"raggedy-assed" comment applied to that.

I would be happy with the conclusion that all SRL is RA, in the sense that it
is
an idealised concept applied to a real phenomenon. "Triangle" would be another
example. There are no real triangles, and yet the word is useful for
describing
things that approach the ideal.

But that's just to say we live in a RA world, which we could have taken for
granted at the outset. RASRL is tautological.

Voltage clipping has non-sharp corners too. Would you call that RA? It
is really just a silly term for describing stuff that already has a
perfectly adequate cogent description. Just drop it.

Dropping it is my purpose. You have failed to see the context.

I am sure also that I was told that SRL is nothing to do with clipping, by
several ppl.

It is clipping. It is just current clipping rather than voltage
clipping. It looks different because it is being integrated by a
capacitor. You are witnessing the integral of a clipped waveform.

Indeed. If the current-source-into-a-capacitor is the only mechanism.

If there is one cloud in the sky and I say "look at that triangular cloud", it
would be pointless to argue about triangularity. How RA must SRL be before it
ceases to answer to the description? Surely that comes down to common sense?
All
that's lacking here is mutual respect.

Please drop the RA in favour of something that means something. If you
mean soft-cornered, say so. SRL is defined by the straight line, not
how clean the corners are. They simply define the onset of current
clipping.

I am trying to arrive at a notion of *partial* SRL. Raggedy-assed is not my
expression. I am looking for a way of categorising it: non-sharp corners;
not-quite-straight line; part of the output waveform only...whatever
imperfections are possible, whilst still remaining answerable to the description
of SRL

Thanks everyone for helping me to explore this stuff. Here are some tentative
conclusions and a few more begging questions.

**Slew rate limiting is a condition in which the rate of change of output
voltage is absolutely limited by a constant value other than zero.

No, it is a condition where the rate of change of output voltage is
inversely proportional to the capacitance being charged.

That is a possible cause, and maybe the only possible cause, but it is not a
description of the effect. Slew, rate, and limit all have meanings in everyday
language, after all. All it means is that the rate of voltage slew is limited.

**This condition may arise for only a portion of the output voltage. For
example, it may occur only beyond a certain amplitude, in which case the
response to a sine wave may have a flat imposed at some point other than the
zero-crossing point, where input dv/dt is greatest.

No it is not amplitude related. It is rise-time related. The amplitude
is immaterial.

Not if it only occurs beyond a certain amplitude. I know what you mean, but you
are getting carried away.

A rounded corner on the leading edge of the straight line indicates that SRL
only comes into effect beyond a certain amplitude. It is in that sense
amplitude-related. Take the classic opamp derivation if you like. You said SRL
can be said to occur in the non-ideal case, and I agree. In the real, non-ideal
case, a very small signal may be insufficient to cause saturation of the first
stage. Hence SRL only happens beyond a certain amplitude.

**A simple way of producing SRL is by placing a constant current source in
series with the signal path, and following it with a capacitor as a load. Is
this part of the definition? Or is it simply the *only* way of imposing a
constant dv/dt? Or neither? I suggest that the *cause* of SRL is only an
accidental part of
its definition, perhaps because it was the way it first occurred. If there is
another way of producing an absolute limit to output dv/dt, is it still SRL?

SRL can have a variety of causes, but in and around audio almost the
only one you ever encounter is the constant current source of the
input stage charging the dominant pole capacitor. You will only
encounter it these days in an incompetently designed amplifier. All
amplifiers should have a frequency response that limits signal rise
time to a value less than that which will trigger SRL.

....at full amplitude, presumably. Yes, quite. Unless some other way can be found
to actively manage the situation.

**A CCS can result if the output impedance and voltage output of a circuit are
proportional, so that V/R is constant. Saturation (forcing *both* to a
constant
value) is one way of achieving this, but can it occur without saturation? I
think it should be possible. I think a CF into a capacitive load achieves SRL
on
both counts, depending on voltage amplitude. The output, in response to a sine
input, becomes triangular. The triangle is never perfect...about as triangular
as a triangular cloud.

Very muddled. a SRL results from an impedance going very high, not
proportional to voltage - this is a definition of a current source.

A sufficient but not a necessary condition. A current source results whenever
V/R is constant. Making both constant, or making R very large, are just two ways
of achieving it. Actually we don't generally rely on very large R alone, because
then we need very large V to generate appreciable current. We use active
circuits, which work by holding V/R constant. That is, they have a very high
*dynamic* resistance.

The triangular output in response to a sine wave won't have sharp
corners. Why should it? There will be no SRL near the peaks of the
sine wave where the rate of change of voltage is low.

**SRL is necessarily a condition arising from non-linearity. The response is
neither a transient response (which always reduces to zero eventually) nor a
steady state response (which is always, er, steady), so has nothing to do with
linear systems.

SRL is a result of current clipping - as non linear as you can get. As
for the transient response, I think you mean the error reduces to
zero, not the response. A perfect transient response would be 1 - the
output is exactly equal to the input at all times.

No. Total system response is the *sum*, not the product, of transient response
and steady state response. The transient response is the path the system takes
on the way to its steady state. If it achieves its steady state instantly, then
there is no transient response. If it achieves its steady state instantly, then
there is no error. In order for there to be no error in response to a change in
input, then the transient response must be zero. I am absolutely correct on this
and will defend my position with as many quotes from as many authoritative works
as you wish. Perhaps you can find a single authoritative source that upholds
your position? I honestly believe you should change your own conception here.

**I had in mind the idea that, if one linear system had a transient response
which was complimentary to another, then the wiggles of one would be cancelled
by those of the other when the two were placed in series, resulting in a
straight line. I now think this is nonsense, but I'm not sure why. I do know
that transient responses don't have sharp corners

Not nonsense at all. This is done all the time in radio - google for
all-pass networks. As for transient responses and sharp corners - they
need an infinite frequency response, so no, you won't see them.


**With respect to the relationship between feedback and SRL I would like to
explore the following idea a bit further:

The is no relationship whatever.

Not necessarily, always, but there may be, sometimes. That is why I am exploring
it.

The steady state output of a linear system with a sine wave input is a sine
wave
with some delay which arises from the transient response of the system. This
constant delay translates into a phase shift that increases with frequency.

The output may then be expressed as a sum of cosine and sine. The proportion
of
the two then varies, from all sine at the centre of bandwidth, to all cosine
when the output is shifted by 90 degrees, beyond which -sine begins to appear,
etc.

It is also true that d(sin(t))dt = cos(t).

It can therefore be said that feedback increasingly includes a proportion of
dv/dt, until when the output of a closed loop system is shifted by 90 degrees,
the feedback is entirely dv/dt. In that case the effective input is V-K.dV/dt,
where V is the input voltage and K is some constant.

It seems on the face of it that, given enough feedback, dV/dt can become the
dominant feature of the input, such that, when it is large, the effective
input
is also large. Hence if the system is amplitude limited, then the limit can be
reached as a result of a signal with a dV/dt beyond a certain limit.

Didn't follow any of that - but the conclusion was wrong, so I presume
so was the reasoning.

In what way was the conclusion wrong? The reasoning is worth exploring. I can't
know what you don't understand unless you tell me, or unless I see you talking
nonsense.


Perhaps it could equally be said that SRL is clipping at an angle.


No. It is just clipping. The angle you see is the result of
integrating that clipping with a capacitor.

In the case of a CCS into a capacitor, yes, but just because that is the cause,
doesn't make the effect false. The effect is voltage clipping at an angle.

Just a raggedy-assed way of looking at the same thing, maybe.

There's that adjective again. Do you use it to describe absolutely
everything?

I am pointing out that, if it applies equally to everything, then it is a
meaningless adjective. You have missed the context, that's all.

cheers, Ian

On Tue, 09 May 2006 18:44:16 GMT, "Ian Iveson"
wrote:

am awful lot of nonsense

Jeez, I'd forgotten about you and your "expertise". Silly me.

Out!

d

--
Pearce Consulting
http://www.pearce.uk.com

Hi Yves.

I'm not a brilliant theorician, but i've read Crowhurst and others 'bout
overdrive and blocking being exacerbated by feedback.
Look at he

http://www.dissident-audio.com/PP_6L6/Page.html

Keys a
-Just enough gain.
-Only one link cap at a point where grid current is unlikely to occur.
-Regulated PSU for screens and driver.
-Just enough feedback.

Seems coherent to what already pointed in this post ;)

Indeed. Thanks. Silly of me not to realise that John meant CF without coupling
caps.

Nice pages.

An issue that raises itself with direct coupling is bias drift...another bag of
bones. And of course whatever problems an extra stage might bring on its own
account. "Just enough fb" and "just enough gain" are related, naturally. I
wonder what "just enough" means? I remember that someone found that too little
fb can be counter-productive...Leach, Leak, H&K, someone like that.

cheers, Ian

Don Pearce wrote

am awful lot of nonsense

Freudian slip?

Jeez, I'd forgotten about you and your "expertise". Silly me.

Out!

How can I tell you have strayed beyond your depth?

You are the one who calls himself a consultant. I have not claimed expertise,
but rather I am working hard, checking out your statements by reading books and
asking others.

You appear to be the one who thinks he has nothing to learn. A terminal case of
intellectual atrophy, I fear.

cheers, Ian

Ian Iveson wrote:

Don Pearce wrote

am awful lot of nonsense

Freudian slip?

Jeez, I'd forgotten about you and your "expertise". Silly me.

Out!

How can I tell you have strayed beyond your depth?

You are the one who calls himself a consultant. I have not claimed expertise,
but rather I am working hard, checking out your statements by reading books and
asking others.

I suggest you work harder.

You appear to be the one who thinks he has nothing to learn. A terminal case of
intellectual atrophy, I fear.

And you, Ian write some of most unmitigated crap about audio electronics that it
has been my misfortune to read.

Graham

Ian Iveson wrote:

Perhaps it could equally be said that SRL is clipping at an angle.

Snort ! At an angle my arse.

Give a clown a scope and see what happens ! Reminds me slightly of another discussin
a week or so ago where another clot talked of 'slowing down' a waveform.

Graham

Pooh Bear wrote

How can I tell you have strayed beyond your depth?

You are the one who calls himself a consultant. I have not claimed expertise,
but rather I am working hard, checking out your statements by reading books
and
asking others.

I suggest you work harder.

No need. I am satisfied with my progress, thanks.

You appear to be the one who thinks he has nothing to learn. A terminal case
of
intellectual atrophy, I fear.

And you, Ian write some of most unmitigated crap about audio electronics that
it
has been my misfortune to read.

Not the same as the magazines you study, that's all.

cheers, Ian

"Pooh Bear" wrote in message
...


Ian Iveson wrote:

Perhaps it could equally be said that SRL is clipping at an angle.

Snort ! At an angle my arse.

Perhaps you are Jamaican? The English would be "My arse is at an angle". Is that
relevant?

Give a clown a scope and see what happens ! Reminds me slightly of another
discussin
a week or so ago where another clot...

Or a monkey with a simulator. But why ridicule by proxy, when I am here and the
other clot is not? Perhaps you can fault my maths? Or maybe if you have some
worthwhile objection relevant to this thread, you could explain what it is?

....talked of 'slowing down' a waveform.

Hmm, interesting! Missed that...where and when, please?

cheers, Ian

An issue that raises itself with direct coupling is bias drift...another bag of
bones. And of course whatever problems an extra stage might bring on its own
account. "Just enough fb" and "just enough gain" are related, naturally. I
wonder what "just enough" means? I remember that someone found that too little
fb can be counter-productive...Leach, Leak, H&K, someone like that.

cheers, Ian

Here ya are, Ian. A couple of Heathkit examples using CF drivers to the OP stage.

http://www.freeinfosociety.com/elect...iew.php?id=496

http://www.freeinfosociety.com/elect...ew.php?id=2380

You can read about the circuits where I used this technique in back issues of Glass
Audio & AudioXpress magazines.

Cheers, John

John Stewart wrote

An issue that raises itself with direct coupling is bias drift...another bag
of
bones. And of course whatever problems an extra stage might bring on its own
account. "Just enough fb" and "just enough gain" are related, naturally. I
wonder what "just enough" means? I remember that someone found that too
little
fb can be counter-productive...Leach, Leak, H&K, someone like that.

cheers, Ian

Here ya are, Ian. A couple of Heathkit examples using CF drivers to the OP
stage.

http://www.freeinfosociety.com/elect...iew.php?id=496

http://www.freeinfosociety.com/elect...ew.php?id=2380

You can read about the circuits where I used this technique in back issues of
Glass
Audio & AudioXpress magazines.


Thanks John. Plenty stages. I shall study them carefully.

cheers, Ian

Ian Iveson wrote:

John Stewart wrote

An issue that raises itself with direct coupling is bias drift...another bag
of
bones. And of course whatever problems an extra stage might bring on its own
account. "Just enough fb" and "just enough gain" are related, naturally. I
wonder what "just enough" means? I remember that someone found that too
little
fb can be counter-productive...Leach, Leak, H&K, someone like that.

cheers, Ian

Here ya are, Ian. A couple of Heathkit examples using CF drivers to the OP
stage.

http://www.freeinfosociety.com/elect...iew.php?id=496

http://www.freeinfosociety.com/elect...ew.php?id=2380

You can read about the circuits where I used this technique in back issues of
Glass
Audio & AudioXpress magazines.

Thanks John. Plenty stages. I shall study them carefully.

cheers, Ian

Just think of those circuits as nothing more than a slightly modified Williamson.
The phase & gain characteristics are only little changed since the CF stages have
response from DC to some frequency much higher than the OPT is capable of. The
primary benefit is the freedom from output stage clipping in the OP tube control
grid circuit. On program peaks the circuit allows problem free operation well into
Class AB2.

But the OPT Ra-a must be lowered in order to get the complete advantage of the
changes when using the CF drivers.

Cheers, John Stewart

"John Stewart" a écrit dans le message de news:
...
Ian Iveson wrote:

John Stewart wrote

An issue that raises itself with direct coupling is bias
drift...another bag
of
bones. And of course whatever problems an extra stage might bring on
its own
account. "Just enough fb" and "just enough gain" are related,
naturally. I
wonder what "just enough" means? I remember that someone found that
too
little
fb can be counter-productive...Leach, Leak, H&K, someone like that.

cheers, Ian

Here ya are, Ian. A couple of Heathkit examples using CF drivers to the
OP
stage.

http://www.freeinfosociety.com/elect...iew.php?id=496

http://www.freeinfosociety.com/elect...ew.php?id=2380

You can read about the circuits where I used this technique in back
issues of
Glass
Audio & AudioXpress magazines.

Thanks John. Plenty stages. I shall study them carefully.

cheers, Ian

Just think of those circuits as nothing more than a slightly modified
Williamson.
The phase & gain characteristics are only little changed since the CF
stages have
response from DC to some frequency much higher than the OPT is capable of.
The
primary benefit is the freedom from output stage clipping in the OP tube
control
grid circuit. On program peaks the circuit allows problem free operation
well into
Class AB2.

Hi John,
Another benefit is that the previous stage becomes virtually unloaded.
What do you think about the penthoded LTP PI ?
The idea was to drop out a gain stage (and the associated link caps !)
(rem the link): http://www.dissident-audio.com/PP_6L6/Page.html

But the OPT Ra-a must be lowered in order to get the complete advantage of
the
changes when using the CF drivers.

Yeap ! Indeed.
Anyway, this helps when driving difficult loads such as speakers with
serious impedance dips !

Yves.

Cheers, John Stewart

Hi John,
Another benefit is that the previous stage becomes virtually unloaded.
What do you think about the penthoded LTP PI ?
The idea was to drop out a gain stage (and the associated link caps !)
(rem the link): http://www.dissident-audio.com/PP_6L6/Page.html

But the OPT Ra-a must be lowered in order to get the complete advantage of
the
changes when using the CF drivers.

Yeap ! Indeed.
Anyway, this helps when driving difficult loads such as speakers with
serious impedance dips !

Yves.

Cheers, John Stewart


Hi Yves- I like your circuit very much, since as you have pointed out it avoids
too many phase shifts, there is just enough gain & limited NFB is used.

Oddly, it reminded me of an earlier circuit, but I was not sure why. I guess
what had fooled me was the way you had arranged the schematic, which after a
closer look made a lot of sense. A couple of days later I realized I had tried
to accomplish the same thing in 1965 as you had here. The primary difference
being that I had not used CF drivers to the OP tubes.

At the front end I used a pair of 6EJ7's as the diff amp. The cathode tails are
tied to another 6EJ7 connected as a triode. I guess I figured with the high G of
the 6EJ7's there would be plenty of gain. Certainly there was enough to easily
run the 20% UL 6L6GC PP output stage. The transformers are specials made for me
by Hammond. These days I try to do everything with off-the-shelf parts. But when
getting into the more exotic designs that is not always possible.

One aspect of your circuit which in my opinion is an improvement is to feed the
diff amp pentode screens from a current source. In your schematic that would be
the 56 K resistor R37. That improves the Diff Amp balance, just as an unbypassed
cathode resistor does. I bypassed the screen supply with an electrolytic,
something I would not do now.

I've noticed some people are putting a small resistor like your 100R, R11 at the
input connector. Does anyone have an explanation for putting it there? I think
it may be a potential for ground loop interference. When a program source is
connected it is possible for a small leakage current to flow in the outer
conductor of the signal lead. The resulting voltage drop in the 100R appears in
series with the input signal. Not much of a problem in a power amp such as this
one but if we did the same thing with a low level magnetic phono input, hum
problems would be a sure thing. Anybody out there know why this is sometimes
done?

Do you have a link to the S14K60, Etc? They look interesting & I assume they are
Zeners.

I built that amp & a multichannel preamp for a customer to be used for public
address, not hifi, just before going into sales at HP. Then I didn't do much
with tubes for about 30 years. With all the semiconductor devices available
today, design is a lot more satisfying, to me, anyway. I've sent a copy of the
old circuit to you for the sake of curiosity! There are 3 pages. One page has
some measurement results. I don't remember how I made the measurements but they
seem to be OK for this kind of circuit. I hope it gets to you OK!

Cheers, John

"John Stewart" a écrit dans le message de news:
...
Hi John,
Another benefit is that the previous stage becomes virtually unloaded.
What do you think about the penthoded LTP PI ?
The idea was to drop out a gain stage (and the associated link caps !)
(rem the link): http://www.dissident-audio.com/PP_6L6/Page.html

But the OPT Ra-a must be lowered in order to get the complete advantage
of
the
changes when using the CF drivers.

Yeap ! Indeed.
Anyway, this helps when driving difficult loads such as speakers with
serious impedance dips !

Yves.

Cheers, John Stewart


Hi Yves- I like your circuit very much, since as you have pointed out it
avoids
too many phase shifts, there is just enough gain & limited NFB is used.

Oddly, it reminded me of an earlier circuit, but I was not sure why. I
guess
what had fooled me was the way you had arranged the schematic, which after
a
closer look made a lot of sense. A couple of days later I realized I had
tried
to accomplish the same thing in 1965 as you had here. The primary
difference
being that I had not used CF drivers to the OP tubes.

At the front end I used a pair of 6EJ7's as the diff amp. The cathode
tails are
tied to another 6EJ7 connected as a triode. I guess I figured with the
high G of
the 6EJ7's there would be plenty of gain. Certainly there was enough to
easily
run the 20% UL 6L6GC PP output stage. The transformers are specials made
for me
by Hammond. These days I try to do everything with off-the-shelf parts.
But when
getting into the more exotic designs that is not always possible.

One aspect of your circuit which in my opinion is an improvement is to
feed the
diff amp pentode screens from a current source. In your schematic that
would be
the 56 K resistor R37. That improves the Diff Amp balance, just as an
unbypassed
cathode resistor does. I bypassed the screen supply with an electrolytic,
something I would not do now.

This trick is from a vertical amplifier for an O'scope. Just copied !


I've noticed some people are putting a small resistor like your 100R, R11
at the
input connector. Does anyone have an explanation for putting it there? I
think
it may be a potential for ground loop interference. When a program source
is
connected it is possible for a small leakage current to flow in the outer
conductor of the signal lead. The resulting voltage drop in the 100R
appears in
series with the input signal. Not much of a problem in a power amp such as
this
one but if we did the same thing with a low level magnetic phono input,
hum
problems would be a sure thing. Anybody out there know why this is
sometimes
done?

It is supposed to open ground loop due to the (already connected) earth in
the preamp and the second channel.
I hope that "ground current" will choose the "easier path".
Not yet built any way !


Do you have a link to the S14K60, Etc? They look interesting & I assume
they are
Zeners.

Nope, varistors, but they act as, just bidirectional.
Less precise (who cares !) but lower temperature drift than HV Zeners.
Here are data:
http://www.epcos.com/inf/70/db/var_01/01590173.pdf


I built that amp & a multichannel preamp for a customer to be used for
public
address, not hifi, just before going into sales at HP. Then I didn't do
much
with tubes for about 30 years. With all the semiconductor devices
available
today, design is a lot more satisfying, to me, anyway. I've sent a copy of
the
old circuit to you for the sake of curiosity! There are 3 pages. One page
has
some measurement results. I don't remember how I made the measurements but
they
seem to be OK for this kind of circuit. I hope it gets to you OK!

Pages received, going to look at.
TNX2U

Yves.

Cheers, John