[Ian Iveson]

Can anyone explain the relationships between OPT specs and
bandwidth for my circlotron?

Output stage sketch is he

http://www.ivesonaudio.pwp.blueyonde...lotronel84.gif

Apologies for rather random component values and cheated
bias circuit, as was when I first posted it. If anyone's
interested in developing it further, I can easily redraw,
retest, and post again.

Thanks

Ian

[Ian Iveson]

Ian Iveson wrote

Can anyone explain the relationships between OPT specs and
bandwidth for my circlotron?

Output stage sketch is he

http://www.ivesonaudio.pwp.blueyonde...lotronel84.gif

Apologies for rather random component values and cheated
bias circuit, as was when I first posted it. If anyone's
interested in developing it further, I can easily redraw,
retest, and post again.

Hmm, obviously not.

Perhaps a simpler starter question. What should the primary
impedance of the OPT be so that each valve sees a 4k load?

Then, maybe, what should the primary inductance be to put
the lower -3dB point somewhere around 10Hz?

But maybe it's just not an interesting exercise?

I'd just like to raise its spirits, if I'm going to make it,
so it doesn't end up with just me in it. Amps you build
yourself sound better for a while. Maybe amps you build with
other people sound even better, for longer?

Anyway, I can't remember how it got to the particular state
it's in. Somehow, a requirement of 100V drive for that max
25W class B doesn't seem right...

Ian

[John Byrns]

In article ,
"Ian Iveson" wrote:

Ian Iveson wrote

Can anyone explain the relationships between OPT specs and
bandwidth for my circlotron?

Output stage sketch is he

http://www.ivesonaudio.pwp.blueyonde...lotronel84.gif

Apologies for rather random component values and cheated
bias circuit, as was when I first posted it. If anyone's
interested in developing it further, I can easily redraw,
retest, and post again.

Hmm, obviously not.

Perhaps a simpler starter question. What should the primary
impedance of the OPT be so that each valve sees a 4k load?

Is that class A or class B?

Then, maybe, what should the primary inductance be to put
the lower -3dB point somewhere around 10Hz?

It depends on the inductance of all those inductors, as well as the
answer to the first question.

But maybe it's just not an interesting exercise?

It's too weird to be interesting, it would be more interesting if you
ditched the inductors.

I'd just like to raise its spirits, if I'm going to make it,
so it doesn't end up with just me in it. Amps you build
yourself sound better for a while. Maybe amps you build with
other people sound even better, for longer?

Its spirits would probably be higher if it looked more like a standard
circlotron, or even a totem pole, either of which it could
advantageously morph into.

Anyway, I can't remember how it got to the particular state
it's in. Somehow, a requirement of 100V drive for that max
25W class B doesn't seem right...

Seems reasonable to me.


Regards,

John Byrns

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

[Ian Iveson]

Thanks, John

Ian Iveson wrote

Can anyone explain the relationships between OPT specs
and
bandwidth for my circlotron?

Output stage sketch is he

http://www.ivesonaudio.pwp.blueyonde...lotronel84.gif

Apologies for rather random component values and
cheated
bias circuit, as was when I first posted it. If
anyone's
interested in developing it further, I can easily
redraw,
retest, and post again.

Hmm, obviously not.

Perhaps a simpler starter question. What should the
primary
impedance of the OPT be so that each valve sees a 4k
load?

Is that class A or class B?

17V bias for EL84 is B, at least for my valve model. 25W
from two valves also suggests B.

Then, maybe, what should the primary inductance be to put
the lower -3dB point somewhere around 10Hz?

It depends on the inductance of all those inductors, as
well as the
answer to the first question.

They are all cross-coupled, either in pairs as shown, or as
a transformer with four equal windings.

But maybe it's just not an interesting exercise?

It's too weird to be interesting, it would be more
interesting if you
ditched the inductors.

Then it wouldn't be interesting to me. Anyway, inductors
have got to be somewhere. What will be interesting to me is
comparing this design with an otherwise identical ordinary
circlotron.

The cross-coupling should ensure that for the purpose of the
exercise you can ignore the inductors, except for their
winding resistances which are shown. It can be assumed that
their effects will be sufficiently distant not to play a
part in determining the main details of the OPT.

I'd just like to raise its spirits, if I'm going to make
it,
so it doesn't end up with just me in it. Amps you build
yourself sound better for a while. Maybe amps you build
with
other people sound even better, for longer?

Its spirits would probably be higher if it looked more
like a standard
circlotron, or even a totem pole, either of which it could
advantageously morph into.

These ideas don't seem to have cheered it up much. What
would be the advantage?

Anyway, I can't remember how it got to the particular
state
it's in. Somehow, a requirement of 100V drive for that
max
25W class B doesn't seem right...

Seems reasonable to me.

OK, which maybe implies that the turns ratio of the OPT is
reasonable?

Ian

[John Byrns]

In article ,
"Ian Iveson" wrote:

Thanks, John

Ian Iveson wrote

Can anyone explain the relationships between OPT specs
and
bandwidth for my circlotron?

Output stage sketch is he

http://www.ivesonaudio.pwp.blueyonde...lotronel84.gif

Apologies for rather random component values and
cheated
bias circuit, as was when I first posted it. If
anyone's
interested in developing it further, I can easily
redraw,
retest, and post again.

Hmm, obviously not.

Perhaps a simpler starter question. What should the
primary
impedance of the OPT be so that each valve sees a 4k
load?

Is that class A or class B?

17V bias for EL84 is B, at least for my valve model. 25W
from two valves also suggests B.

Then, maybe, what should the primary inductance be to put
the lower -3dB point somewhere around 10Hz?

It depends on the inductance of all those inductors, as
well as the
answer to the first question.

They are all cross-coupled, either in pairs as shown, or as
a transformer with four equal windings.

But maybe it's just not an interesting exercise?

It's too weird to be interesting, it would be more
interesting if you
ditched the inductors.

Then it wouldn't be interesting to me. Anyway, inductors
have got to be somewhere. What will be interesting to me is
comparing this design with an otherwise identical ordinary
circlotron.

The cross-coupling should ensure that for the purpose of the
exercise you can ignore the inductors, except for their
winding resistances which are shown. It can be assumed that
their effects will be sufficiently distant not to play a
part in determining the main details of the OPT.

I'd just like to raise its spirits, if I'm going to make
it,
so it doesn't end up with just me in it. Amps you build
yourself sound better for a while. Maybe amps you build
with
other people sound even better, for longer?

Its spirits would probably be higher if it looked more
like a standard
circlotron, or even a totem pole, either of which it could
advantageously morph into.

These ideas don't seem to have cheered it up much. What
would be the advantage?

Anyway, I can't remember how it got to the particular
state
it's in. Somehow, a requirement of 100V drive for that
max
25W class B doesn't seem right...

Seems reasonable to me.

OK, which maybe implies that the turns ratio of the OPT is
reasonable?

Hi Ian,

Can you explain the advantages of your circlotron over the standard
circlotron? I assume that you are using a circlotron type circuit
because you want a class B amplifier, and the circlotron circuit
eliminates the problems caused by leakage inductance between the primary
halves, or quarters depending on which ordinary circuit we are talking
about?

What I don't understand is why you don't simply use the standard
circlotron circuit, which is simpler, probably costs less, and is likely
to perform better?


Regards,

John Byrns

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

[Ian Iveson]

John Byrns wrote

Can anyone explain the relationships between OPT
specs
and
bandwidth for my circlotron?

Output stage sketch is he

http://www.ivesonaudio.pwp.blueyonde...lotronel84.gif

Apologies for rather random component values and
cheated
bias circuit, as was when I first posted it. If
anyone's
interested in developing it further, I can easily
redraw,
retest, and post again.

Hmm, obviously not.

Perhaps a simpler starter question. What should the
primary
impedance of the OPT be so that each valve sees a 4k
load?

Is that class A or class B?

17V bias for EL84 is B, at least for my valve model. 25W
from two valves also suggests B.

Then, maybe, what should the primary inductance be to
put
the lower -3dB point somewhere around 10Hz?

It depends on the inductance of all those inductors, as
well as the
answer to the first question.

They are all cross-coupled, either in pairs as shown, or
as
a transformer with four equal windings.

But maybe it's just not an interesting exercise?

It's too weird to be interesting, it would be more
interesting if you
ditched the inductors.

Then it wouldn't be interesting to me. Anyway, inductors
have got to be somewhere. What will be interesting to me
is
comparing this design with an otherwise identical
ordinary
circlotron.

The cross-coupling should ensure that for the purpose of
the
exercise you can ignore the inductors, except for their
winding resistances which are shown. It can be assumed
that
their effects will be sufficiently distant not to play a
part in determining the main details of the OPT.

I'd just like to raise its spirits, if I'm going to
make
it,
so it doesn't end up with just me in it. Amps you
build
yourself sound better for a while. Maybe amps you
build
with
other people sound even better, for longer?

Its spirits would probably be higher if it looked more
like a standard
circlotron, or even a totem pole, either of which it
could
advantageously morph into.

These ideas don't seem to have cheered it up much. What
would be the advantage?

Anyway, I can't remember how it got to the particular
state
it's in. Somehow, a requirement of 100V drive for that
max
25W class B doesn't seem right...

Seems reasonable to me.

OK, which maybe implies that the turns ratio of the OPT
is
reasonable?

Hi Ian,

Can you explain the advantages of your circlotron over the
standard
circlotron? I assume that you are using a circlotron type
circuit
because you want a class B amplifier, and the circlotron
circuit
eliminates the problems caused by leakage inductance
between the primary
halves, or quarters depending on which ordinary circuit we
are talking
about?


I was originally drawn to any archetecture that might offer
parallel operation in PP, as a way of minimising the optimum
turns ratio for driving low impedance speakers, without
unduly loading the driver stage.

I'm not taken with designs using several power supplies.
Seems like driving a car with two engines. I was quite taken
by Ariels for a time when I had a Red Hunter, and still
remember the feeling of disappointment when a friend
demonstrated the weakness of the Square Four, that wrecked
itself after shedding teeth on the pinions that link the two
parallel twins.

I'm not keen on the idea of class B, either. It hadn't
really occured to me when I had the idea and I can't
remember why I tried it in simulation. It looks like I was
exploring every possible extreme.

What I don't understand is why you don't simply use the
standard
circlotron circuit, which is simpler, probably costs less,
and is likely
to perform better?


It began as an excercise to show that a circlotron doesn't
need two power supplies for the output stage. It was first
introduced here in fun, because someone had just said, as
they have again recently, that two are necessary. I believe
it was under the heading "perverse circlotron".

Then it occured to me that it was the neatest and most
profound statement of the dialectic that a valve amplifier
might ever contrive to embody. I felt I shouldn't dismiss it
without full analysis, which process I would like to share.
I'm so taken with the idea that I will build it even if it
isn't the perfect audio machine, as long as it is feasible
and offers as good a sound and bandwidth as any other valve
design when driving low impedance speakers.

So you have said it is not interesting, and that an ordinary
circlotron would be better. I can quite understand that
you're not interested, but what are its disadvantages?

It will probably weigh more, but cost about the same, it
seems to me, but that's just a vague top-down feeling rather
than a bottom-up analysis.

Ian

[John Byrns]

In article ,
"Ian Iveson" wrote:

John Byrns wrote

Hi Ian,

Can you explain the advantages of your circlotron over the
standard
circlotron? I assume that you are using a circlotron type
circuit
because you want a class B amplifier, and the circlotron
circuit
eliminates the problems caused by leakage inductance
between the primary
halves, or quarters depending on which ordinary circuit we
are talking
about?

I was originally drawn to any archetecture that might offer
parallel operation in PP, as a way of minimising the optimum
turns ratio for driving low impedance speakers,

Wouldn't a totem pole circuit also accomplish that goal?

Also I am not a transformer guy, can you explain the theory behind
"minimising the optimum turns ratio for driving low impedance speakers",
or maybe Patrick can jump in here and explain what the advantage of
minimizing the turns ratio is?

without
unduly loading the driver stage.

I don't follow you here, how does your circuit relate to the "loading"
of the driver stage?

I'm not taken with designs using several power supplies.

Why not? It doesn't take any more copper in the Power Transformer than
a single supply and your circuit already includes the extra electrolytic
capacitors a second power supply would require, so your circuit looses
on cost and complexity, and pointless complexity often opens the door to
unexpected problems so why go down that road, what is the advantage?

Seems like driving a car with two engines. I was quite taken
by Ariels for a time when I had a Red Hunter, and still
remember the feeling of disappointment when a friend
demonstrated the weakness of the Square Four, that wrecked
itself after shedding teeth on the pinions that link the two
parallel twins.

Sounds like a bad analogy to me, a better analogy would be to liken the
two power supplies to the two pistons in a single twin, with the OPT
being the equivalent of the crankshaft in the twin, no "pinions" being
required.

I'm not keen on the idea of class B, either. It hadn't
really occured to me when I had the idea and I can't
remember why I tried it in simulation. It looks like I was
exploring every possible extreme.

What I don't understand is why you don't simply use the
standard
circlotron circuit, which is simpler, probably costs less,
and is likely
to perform better?


It began as an excercise to show that a circlotron doesn't
need two power supplies for the output stage. It was first
introduced here in fun, because someone had just said, as
they have again recently, that two are necessary. I believe
it was under the heading "perverse circlotron".

I will have to look that up, it sounds like an apt title.

Then it occured to me that it was the neatest and most
profound statement of the dialectic that a valve amplifier
might ever contrive to embody.

Huh, what does that mean?

I felt I shouldn't dismiss it
without full analysis, which process I would like to share.
I'm so taken with the idea that I will build it even if it
isn't the perfect audio machine, as long as it is feasible
and offers as good a sound and bandwidth as any other valve
design when driving low impedance speakers.

So you have said it is not interesting, and that an ordinary
circlotron would be better. I can quite understand that
you're not interested, but what are its disadvantages?

All the extra iron that adds nothing but cost and weight, while reducing
low frequency performance as a result of the additional shunt inductive
reactance across the loudspeaker load.

It will probably weigh more, but cost about the same, it
seems to me, but that's just a vague top-down feeling rather
than a bottom-up analysis.

My vague feeling is that it would cost more than simply using two power
supplies, but how much depends on how much ripple filtering you feel is
needed in the power supply(s), what problems are created by combining
all four of your inductors onto a single core, or even two cores, and
there are problems, and finally the fact that your OPT will of necessity
require a higher inductance to maintain the same LF performance when
shunted by all those inductors.

If I am following what you have said, your main design goal is to
minimize the turns ratio of the OPT, and that class B operation is not a
primary design goal?


Regards,

John Byrns

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

[Ian Iveson]

John Byrns wrote:


Hi Ian,

Can you explain the advantages of your circlotron over
the
standard
circlotron? I assume that you are using a circlotron
type
circuit
because you want a class B amplifier, and the
circlotron
circuit
eliminates the problems caused by leakage inductance
between the primary
halves, or quarters depending on which ordinary circuit
we
are talking
about?

I was originally drawn to any archetecture that might
offer
parallel operation in PP, as a way of minimising the
optimum
turns ratio for driving low impedance speakers,

Wouldn't a totem pole circuit also accomplish that goal?

A number of possibilities would fit the bill, including
nearly all SS amps. Swings and roundabouts; I like my idea.

Also I am not a transformer guy, can you explain the
theory behind
"minimising the optimum turns ratio for driving low
impedance speakers",
or maybe Patrick can jump in here and explain what the
advantage of
minimizing the turns ratio is?

Fat chance, from my viewpoint. I'd rather trust a snake.

The optimum turns ratio is the one that presents the optimum
load to the valves for a given speaker impedance. This
optimum is reduced with falling Zaa, and minimised by using
valves in parallel. Lower turns ratio allows wider
bandwidth, all other things being equal.

Menno van der Veen is a useful source, if you have his book
or papers.

without
unduly loading the driver stage.

I don't follow you here, how does your circuit relate to
the "loading"
of the driver stage?

The most immediately obvious way of using parallel valves in
PP is to use several valves in parallel on each side of an
ordinary PP circuit. This decreases the input resistance of
the stage because each valve has a minimum value of grid
leak resistor. The lower input resistance is a greater load
for the driver to drive. Perhaps I have misused the term?


I'm not taken with designs using several power supplies.

Why not? It doesn't take any more copper in the Power
Transformer than
a single supply and your circuit already includes the
extra electrolytic
capacitors a second power supply would require, so your
circuit looses
on cost and complexity, and pointless complexity often
opens the door to
unexpected problems so why go down that road, what is the
advantage?

The 'lytics may be its downfall, I would be the first to
admit. There are also the details of the bias circuit to
sort out. I'm not keen on so many complexities attached to
that single loop (or figure of 8 if you like).

Advantages from a narrow engineering perspective...hmm...how
about perfect matching of HT voltage for each valve? Also,
it has a neat simplicity to it. I don't know why you see it
as complicated. If the power suppies were included, I would
say the ordinary dual supply looked more complicated, just
because it has two connections to the outside that require
matching. If you mean complex, as in too clever by half,
then that seems appropriate for a circlotron, which is
something of a contrivance in any case.

It would be interesting to compare the effects of imbalanced
drive signals or poorly matched valves. Sound wise, I
suspect that the constraints of all that cross-bracing, as
it were, may be essentially un-musical.

Seems like driving a car with two engines. I was quite
taken
by Ariels for a time when I had a Red Hunter, and still
remember the feeling of disappointment when a friend
demonstrated the weakness of the Square Four, that
wrecked
itself after shedding teeth on the pinions that link the
two
parallel twins.

Sounds like a bad analogy to me, a better analogy would be
to liken the
two power supplies to the two pistons in a single twin,
with the OPT
being the equivalent of the crankshaft in the twin, no
"pinions" being
required.

OK, of course. Although I have seen the analogy between a
transformer and a gearbox expressed a fair few times, I
don't think it's a good one, and anyway the pinions would
act between the crank and the output shaft, or between the
primary and the secondary... It's a bit interesting,
pursuing the analogy hopelessly, that gear couplings don't
like high tooth ratios.

I'm not keen on the idea of class B, either. It hadn't
really occured to me when I had the idea and I can't
remember why I tried it in simulation. It looks like I
was
exploring every possible extreme.

What I don't understand is why you don't simply use the
standard
circlotron circuit, which is simpler, probably costs
less,
and is likely
to perform better?


It began as an excercise to show that a circlotron
doesn't
need two power supplies for the output stage. It was
first
introduced here in fun, because someone had just said, as
they have again recently, that two are necessary. I
believe
it was under the heading "perverse circlotron".

I will have to look that up, it sounds like an apt title.

Thanks. I always try to make my titles apt.

Then it occured to me that it was the neatest and most
profound statement of the dialectic that a valve
amplifier
might ever contrive to embody.

Huh, what does that mean?

Well, er...

An impetuous dialectical fundamentalist would have had my
guts for garters before reaching the word "contrive".

The dialectic is the single dynamic that drives progress;
that forces the constant change necessary for being. It
concerns the interaction of opposites, and has always been
the central problem of philosophy. Hence such ideas as a
single god being the synthesis of the opposites of body and
spirit as in the Holy Trinity, or "for every action there is
an equal and opposite reaction", or e=mc^2; whatever
religion or science you can think of has the interdependence
of opposites at its core.

However, the opposites significant to the dialectic are
generally not simple arithmetic inversions. A significant
pair of dialectical opposites interact to arrive at a
synthesis, rather than a null.

The opposites in my circlotron may be trivial, but there are
lots of them, and each pair acts across or around that
single figure of 8. So I quite like it all the same.

I felt I shouldn't dismiss it
without full analysis, which process I would like to
share.
I'm so taken with the idea that I will build it even if
it
isn't the perfect audio machine, as long as it is
feasible
and offers as good a sound and bandwidth as any other
valve
design when driving low impedance speakers.

So you have said it is not interesting, and that an
ordinary
circlotron would be better. I can quite understand that
you're not interested, but what are its disadvantages?

All the extra iron that adds nothing but cost and weight,
while reducing
low frequency performance as a result of the additional
shunt inductive
reactance across the loudspeaker load.

Ah, well, that's the kind of wondering that led me to raise
the matter. I hoped that, having sorted the OPT whilst
ignoring the effects of the weirdness, attention could then
pass to the inductors and caps, so we could duly evaluate
the cost of deviant contrivance.

It will probably weigh more, but cost about the same, it
seems to me, but that's just a vague top-down feeling
rather
than a bottom-up analysis.

My vague feeling is that it would cost more than simply
using two power
supplies, but how much depends on how much ripple
filtering you feel is
needed in the power supply(s), what problems are created
by combining
all four of your inductors onto a single core, or even two
cores, and
there are problems, and finally the fact that your OPT
will of necessity
require a higher inductance to maintain the same LF
performance when
shunted by all those inductors.

I was hoping the inductor could be easily made using a pair
of bifilar windings. What problems do you perceive?

I had rather assumed they are in series with the
transformer, and I guess others may have done so too. I
don't think a higher inductance for the OPT should be
necessary. If the zeros can't be kept far enough apart, the
design loses its appeal for me.

The details, such as CMRR and PSRR and minor poles and
zeros, and cap voltage and ESL and ESR, etc etc, aren't
immediately obvious to me. I hope to get the basics
optimised first, and then consider how sensitive it is to
imperfections and parasitics.

If I am following what you have said, your main design
goal is to
minimize the turns ratio of the OPT, and that class B
operation is not a
primary design goal?

I think the minimising of optimum turns ratio is why PPP
architectures found favour, even if their afficionados don't
know it, and is linked to the tendency towards falling
speaker impedance. I might go as far as to say that falling
speaker impedance has been the death of traditional valve
hi-fi.

Ian

[John Byrns]

In article ,
"Ian Iveson" wrote:

John Byrns wrote:

Also I am not a transformer guy, can you explain the
theory behind
"minimising the optimum turns ratio for driving low
impedance speakers",
or maybe Patrick can jump in here and explain what the
advantage of
minimizing the turns ratio is?

Fat chance, from my viewpoint. I'd rather trust a snake.

A snake? Snakes know less than I do about transformers, Patrick on the
other hand is quite knowledgeable when it comes to transformers.

The optimum turns ratio is the one that presents the optimum
load to the valves for a given speaker impedance. This
optimum is reduced with falling Zaa, and minimised by using
valves in parallel. Lower turns ratio allows wider
bandwidth, all other things being equal.

Menno van der Veen is a useful source, if you have his book
or papers.

I have neither his books nor his papers, so he is useless to me.

without
unduly loading the driver stage.

I don't follow you here, how does your circuit relate to
the "loading"
of the driver stage?

The most immediately obvious way of using parallel valves in
PP is to use several valves in parallel on each side of an
ordinary PP circuit. This decreases the input resistance of
the stage because each valve has a minimum value of grid
leak resistor. The lower input resistance is a greater load
for the driver to drive. Perhaps I have misused the term?

I think I see the point you are trying to make, but there are a couple
of flaws in your logic. The circlotron may have higher value grid
resistors loading the driver stage, when using an OPT with a given
primary/secondary ratio, than a PPP circuit does, but it also requires a
considerably higher drive voltage which doesn't come for free. Given
that you have the excess drive voltage available, as you must for a
circlotron, you can bring the value of the grid resistance loading the
driver up in the PPP design by simply adding a series resistor between
each driver plate and following grid circuit to form a 2:1 voltage
divider, that will give the PPP the same grid resistance loading the
driver as the circlotron has. Or given that the circlotron driver stage
will require significantly greater voltage gain than the PPP circuit
does, the PPP driver can compensate by using valves with lower gain and
plate resistance to facilitate driving the lower grid resistance, or
alternatively if the circlotron driver gets its extra gain by using more
valves, those can be repurposed as cathode followers to drive the lower
value grid resistors in the PPP design. The second flaw is that the PPP
design will have twice the power output capability of the circlotron, so
the two designs aren't directly comparable in the first place.

I'm not taken with designs using several power supplies.

Why not? It doesn't take any more copper in the Power
Transformer than
a single supply and your circuit already includes the
extra electrolytic
capacitors a second power supply would require, so your
circuit looses
on cost and complexity, and pointless complexity often
opens the door to
unexpected problems so why go down that road, what is the
advantage?

The 'lytics may be its downfall, I would be the first to
admit. There are also the details of the bias circuit to
sort out. I'm not keen on so many complexities attached to
that single loop (or figure of 8 if you like).

Advantages from a narrow engineering perspective...hmm...how
about perfect matching of HT voltage for each valve?

How close do the HT voltages have to be matched? The valves themselves
won't be perfectly matched, the two HT voltages are likely to be quite
close, especially relative to the matching of the two valves, I think
you are worrying about a nonexistent problem here.

Also,
it has a neat simplicity to it. I don't know why you see it
as complicated.

Nor do I see the simplicity that you do.

If the power suppies were included, I would
say the ordinary dual supply looked more complicated, just
because it has two connections to the outside that require
matching.

Draw the circuit with two power supplies, it looks simpler to me than it
does with your inductors.

If you mean complex, as in too clever by half,
then that seems appropriate for a circlotron, which is
something of a contrivance in any case.

It's not obvious to me that the straight circlotron is a "contrivance".

It would be interesting to compare the effects of imbalanced
drive signals or poorly matched valves. Sound wise, I
suspect that the constraints of all that cross-bracing, as
it were, may be essentially un-musical.

What cross-bracing are you talking about, are you talking about your
modified circlotron?

It will probably weigh more, but cost about the same, it
seems to me, but that's just a vague top-down feeling
rather
than a bottom-up analysis.

My vague feeling is that it would cost more than simply
using two power
supplies, but how much depends on how much ripple
filtering you feel is
needed in the power supply(s), what problems are created
by combining
all four of your inductors onto a single core, or even two
cores, and
there are problems, and finally the fact that your OPT
will of necessity
require a higher inductance to maintain the same LF
performance when
shunted by all those inductors.

I was hoping the inductor could be easily made using a pair
of bifilar windings. What problems do you perceive?

OK, sounds like an idea, while you are at it why not ditch the OPT, and
add an additional winding on the "inductor" to drive the speaker without
a separate OPT?

I had rather assumed they are in series with the
transformer, and I guess others may have done so too. I
don't think a higher inductance for the OPT should be
necessary.

The combination of all the inductors is directly in parallel with the
OPT.

If the zeros can't be kept far enough apart, the
design loses its appeal for me.

The details, such as CMRR and PSRR and minor poles and
zeros, and cap voltage and ESL and ESR, etc etc, aren't
immediately obvious to me. I hope to get the basics
optimised first, and then consider how sensitive it is to
imperfections and parasitics.

I'm not sure what you are trying to say about the zeros, and minor poles
and zeros here?

If I am following what you have said, your main design
goal is to
minimize the turns ratio of the OPT, and that class B
operation is not a
primary design goal?

I think the minimising of optimum turns ratio is why PPP
architectures found favour, even if their afficionados don't
know it, and is linked to the tendency towards falling
speaker impedance.

If an optimum turns ratio is the important thing, why not just connect
all the valves in parallel in a simple PSE circuit? One of the left
over valves from the push-pull driver circuit can then be repurposed as
a cathode follower to solve the low grid resistance "problem".

I might go as far as to say that falling
speaker impedance has been the death of traditional valve
hi-fi.

That view depends on your thesis that a slightly higher than optimum
turns ratio is really a serious compromise.


Regards,

John Byrns

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

[Patrick Turner]

John Byrns wrote:

In article ,
"Ian Iveson" wrote:

John Byrns wrote:

Also I am not a transformer guy, can you explain the
theory behind
"minimising the optimum turns ratio for driving low
impedance speakers",
or maybe Patrick can jump in here and explain what the
advantage of
minimizing the turns ratio is?

Fat chance, from my viewpoint. I'd rather trust a snake.

A snake? Snakes know less than I do about transformers, Patrick on the
other hand is quite knowledgeable when it comes to transformers.

Hi John, Iveson doesn't like me very much because
I insist he make sure all his posts would be clear even to a moderately
educated school boy.

Iveson gave a URL to a schematic of a Circlotron but its very unclear
how it all works.

I recall the original Circlotron built by Electovoice used a pair of 6V6
to make 20 watts
and had two B+ supplies per channel, and as each supply produces 1/2 the
VA of a single supply
for the channel considered, the inconvenience of two B+ floating
supplies
isn't much greater than having one B+, mainly because we don't have to
use tube rectifiers any more.

However, the best circlotrons have shielded tranny windings, or B+
windngs
wound on C cores with mains one side and B+ on the other, and kept apart
due to stray C between the
mains and B+ windings and btween each B+ winding.

The effect of the two supplies and the Circlotron design means that
the two 6V6 output tubes work in paralel upon the load at all times,
and the turn ratio when in class B is the same as 1/2 a normal primary
in a series
and conventional arrangement.

Allow me to clarify further.

The Wiiliamson OPT as everyone here should know by now had 4,400 turns
in its
primary winding.
The OP tubes are in series effectively with a the load a-a = 10k.

For 8 ohms at the sec, the ZR = 1,250:1, TR = 37.35:1.

But while working in class B, the W has an effective TR of half the OPT
because one tube is cut off.
So the TR becomes 17.67:1, and the load each tube sees for each 1/2 wave
cycle = 2,500 ohms.

The circlotron has a centre tapped primary that has 1/2 the P turns of
the W, say 2,200 turns for the same
tubes and power outcome.
Each tube conducts in class B through ALL the P turns.
Each tube in its turn to to work on either +ve or -ve waves
causes a current flow in alternate directions in ALL P turns for each
+ve and -ve wave 1/2 cycle.

The Circlotron OPT is thus easier to wind than a Williamson.
The McIntosh works in a similar manner, but has the same total turns as
the Williamson example.


If the Circlotron is made to work in class A, each tube sees a 5k load,
and the load of each is in parallel, so the total primary load is 2.5k.

The Williamson when working in class A has each tube still seeing 5k,
but the loads are in series rather than parallel,
so 10k is the RLa-a, not 5k or 2.5k.



The optimum turns ratio is the one that presents the optimum
load to the valves for a given speaker impedance. This
optimum is reduced with falling Zaa, and minimised by using
valves in parallel. Lower turns ratio allows wider
bandwidth, all other things being equal.

Menno van der Veen is a useful source, if you have his book
or papers.

I have neither his books nor his papers, so he is useless to me.

without
unduly loading the driver stage.

I don't follow you here, how does your circuit relate to
the "loading"
of the driver stage?

The most immediately obvious way of using parallel valves in
PP is to use several valves in parallel on each side of an
ordinary PP circuit. This decreases the input resistance of
the stage because each valve has a minimum value of grid
leak resistor. The lower input resistance is a greater load
for the driver to drive. Perhaps I have misused the term?

I think I see the point you are trying to make, but there are a couple
of flaws in your logic. The circlotron may have higher value grid
resistors loading the driver stage, when using an OPT with a given
primary/secondary ratio, than a PPP circuit does, but it also requires a
considerably higher drive voltage which doesn't come for free. Given
that you have the excess drive voltage available, as you must for a
circlotron, you can bring the value of the grid resistance loading the
driver up in the PPP design by simply adding a series resistor between
each driver plate and following grid circuit to form a 2:1 voltage
divider, that will give the PPP the same grid resistance loading the
driver as the circlotron has. Or given that the circlotron driver stage
will require significantly greater voltage gain than the PPP circuit
does, the PPP driver can compensate by using valves with lower gain and
plate resistance to facilitate driving the lower grid resistance, or
alternatively if the circlotron driver gets its extra gain by using more
valves, those can be repurposed as cathode followers to drive the lower
value grid resistors in the PPP design. The second flaw is that the PPP
design will have twice the power output capability of the circlotron, so
the two designs aren't directly comparable in the first place.

The circlotron like the McIntosh or any other amp with lots of cathode
FB in the output stage
can bootstrap its grid bias R to the cathode ac voltage.
This effectively will raise the loading value of the bias resistor
to the driving stage, and especially so in a class A circ' or McI type
amp.
thus a 47k appears as say near 470k to a driver tube if op tube gain is
10.
It means the bootstrapping allows lower values of grig bias and
more reliable control of bias dc voltages, something may makers
have failed to realise because in older tubes there is reverse grid
current even at idle,
and a dangerously high vdc can deveop across the bias R of an output
tube.
Quad-II amps suffer in this respect badly, with 680k biasing R, which is
way too high.




I'm not taken with designs using several power supplies.

Why not? It doesn't take any more copper in the Power
Transformer than
a single supply and your circuit already includes the
extra electrolytic
capacitors a second power supply would require, so your
circuit looses
on cost and complexity, and pointless complexity often
opens the door to
unexpected problems so why go down that road, what is the
advantage?

The 'lytics may be its downfall, I would be the first to
admit. There are also the details of the bias circuit to
sort out. I'm not keen on so many complexities attached to
that single loop (or figure of 8 if you like).

Advantages from a narrow engineering perspective...hmm...how
about perfect matching of HT voltage for each valve?

How close do the HT voltages have to be matched?

Not very close.

The valves themselves
won't be perfectly matched, the two HT voltages are likely to be quite
close, especially relative to the matching of the two valves, I think
you are worrying about a nonexistent problem here.

Could be.

Also,
it has a neat simplicity to it. I don't know why you see it
as complicated.

Nor do I see the simplicity that you do.

If the power suppies were included, I would
say the ordinary dual supply looked more complicated, just
because it has two connections to the outside that require
matching.

Draw the circuit with two power supplies, it looks simpler to me than it
does with your inductors.

If you mean complex, as in too clever by half,
then that seems appropriate for a circlotron, which is
something of a contrivance in any case.

It's not obvious to me that the straight circlotron is a "contrivance".

Its a fine way to make an amp.

It would be interesting to compare the effects of imbalanced
drive signals or poorly matched valves. Sound wise, I
suspect that the constraints of all that cross-bracing, as
it were, may be essentially un-musical.

What cross-bracing are you talking about, are you talking about your
modified circlotron?

It will probably weigh more, but cost about the same, it
seems to me, but that's just a vague top-down feeling
rather
than a bottom-up analysis.

My vague feeling is that it would cost more than simply
using two power
supplies, but how much depends on how much ripple
filtering you feel is
needed in the power supply(s), what problems are created
by combining
all four of your inductors onto a single core, or even two
cores, and
there are problems, and finally the fact that your OPT
will of necessity
require a higher inductance to maintain the same LF
performance when
shunted by all those inductors.

I was hoping the inductor could be easily made using a pair
of bifilar windings. What problems do you perceive?

OK, sounds like an idea, while you are at it why not ditch the OPT, and
add an additional winding on the "inductor" to drive the speaker without
a separate OPT?

I had rather assumed they are in series with the
transformer, and I guess others may have done so too. I
don't think a higher inductance for the OPT should be
necessary.

The combination of all the inductors is directly in parallel with the
OPT.

If the zeros can't be kept far enough apart, the
design loses its appeal for me.

The details, such as CMRR and PSRR and minor poles and
zeros, and cap voltage and ESL and ESR, etc etc, aren't
immediately obvious to me. I hope to get the basics
optimised first, and then consider how sensitive it is to
imperfections and parasitics.

I'm not sure what you are trying to say about the zeros, and minor poles
and zeros here?

If I am following what you have said, your main design
goal is to
minimize the turns ratio of the OPT, and that class B
operation is not a
primary design goal?

I think the minimising of optimum turns ratio is why PPP
architectures found favour, even if their afficionados don't
know it, and is linked to the tendency towards falling
speaker impedance.

If an optimum turns ratio is the important thing, why not just connect
all the valves in parallel in a simple PSE circuit? One of the left
over valves from the push-pull driver circuit can then be repurposed as
a cathode follower to solve the low grid resistance "problem".

I might go as far as to say that falling
speaker impedance has been the death of traditional valve
hi-fi.

That view depends on your thesis that a slightly higher than optimum
turns ratio is really a serious compromise.

Falling speaker Z was accompanied by falling sensitivity.

This meant that most owners of old tube gear were left high and dry
with their amps that had only 16 ohm outlets and which made only 10
watts.

To accomodate modern speakers, make all PP amps with a class AB loading
to match no higher than 5 ohms.
Thus modern speakers of "8" ohms with dips to 3 ohms will be
accomodated.
And use a pair of KT88 instead of EL84.
I'd suggest the cost of tubes in terms of a fraction of a week's wage
is far lower now than it was in 1960.

A quad of EL34 or 6L6 also work wonders.

Patrick Turner.



Regards,

John Byrns

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

[Ian Iveson]

John Byrns wrote:


Also I am not a transformer guy, can you explain the
theory behind
"minimising the optimum turns ratio for driving low
impedance speakers",
or maybe Patrick can jump in here and explain what the
advantage of
minimizing the turns ratio is?

Fat chance, from my viewpoint. I'd rather trust a snake.

A snake? Snakes know less than I do about transformers,
Patrick on the
other hand is quite knowledgeable when it comes to
transformers.

He knows how to make them.

The optimum turns ratio is the one that presents the
optimum
load to the valves for a given speaker impedance. This
optimum is reduced with falling Zaa, and minimised by
using
valves in parallel. Lower turns ratio allows wider
bandwidth, all other things being equal.

Menno van der Veen is a useful source, if you have his
book
or papers.

I have neither his books nor his papers, so he is useless
to me.

You'd enjoy his book. He derives a complete transfer
function for a typical PP amp, and a load of parameters
relating output transformers to audio performance. Also
considers various proportions of distributed load, several
simple amp designs, and a good design procedure for ensuring
stability.

From his transfer function he derives a formula for the
ratio of LF and HF bandwidth limits, influenced by both the
impedance ratio and the turns ratio.

without
unduly loading the driver stage.

I don't follow you here, how does your circuit relate
to
the "loading"
of the driver stage?

The most immediately obvious way of using parallel valves
in
PP is to use several valves in parallel on each side of
an
ordinary PP circuit. This decreases the input resistance
of
the stage because each valve has a minimum value of grid
leak resistor. The lower input resistance is a greater
load
for the driver to drive. Perhaps I have misused the term?

I think I see the point you are trying to make, but there
are a couple
of flaws in your logic. The circlotron may have higher
value grid
resistors loading the driver stage, when using an OPT with
a given
primary/secondary ratio, than a PPP circuit does, but it
also requires a
considerably higher drive voltage which doesn't come for
free. Given
that you have the excess drive voltage available, as you
must for a
circlotron, you can bring the value of the grid resistance
loading the
driver up in the PPP design by simply adding a series
resistor between
each driver plate and following grid circuit to form a 2:1
voltage
divider, that will give the PPP the same grid resistance
loading the
driver as the circlotron has. Or given that the
circlotron driver stage
will require significantly greater voltage gain than the
PPP circuit
does, the PPP driver can compensate by using valves with
lower gain and
plate resistance to facilitate driving the lower grid
resistance, or
alternatively if the circlotron driver gets its extra gain
by using more
valves, those can be repurposed as cathode followers to
drive the lower
value grid resistors in the PPP design. The second flaw
is that the PPP
design will have twice the power output capability of the
circlotron, so
the two designs aren't directly comparable in the first
place.

I didn't intend to give the impression that the circlotron
is unique in that respect. Yes, there are other ways of
achieving pretty much anything...swings and roundabouts, no
free lunch etc.

I haven't put it forward as the best design of amplifier
from an analytical point of view. Far from it: as I made
clear in the section on dialectic, which you cut. My only
hope was that someone might be interested in optimising the
OPT, and then maybe considering the effect of the other
inductors.

I'm not taken with designs using several power
supplies.

Why not? It doesn't take any more copper in the Power
Transformer than
a single supply and your circuit already includes the
extra electrolytic
capacitors a second power supply would require, so your
circuit looses
on cost and complexity, and pointless complexity often
opens the door to
unexpected problems so why go down that road, what is
the
advantage?

The 'lytics may be its downfall, I would be the first to
admit. There are also the details of the bias circuit to
sort out. I'm not keen on so many complexities attached
to
that single loop (or figure of 8 if you like).

Advantages from a narrow engineering
perspective...hmm...how
about perfect matching of HT voltage for each valve?

How close do the HT voltages have to be matched? The
valves themselves
won't be perfectly matched, the two HT voltages are likely
to be quite
close, especially relative to the matching of the two
valves, I think
you are worrying about a nonexistent problem here.


How equal the two power supply voltages of an ordinary
circlotron need to be is something none of us knows, and
words like "not very close" don't mean much to me. I will
compare with an ordinary circlotron, with various
imperfections such as mismatched valves, drive voltages,
etc., eventually.

Also,
it has a neat simplicity to it. I don't know why you see
it
as complicated.

Nor do I see the simplicity that you do.

If the power supplies were included, I would
say the ordinary dual supply looked more complicated,
just
because it has two connections to the outside that
require
matching.

Draw the circuit with two power supplies, it looks simpler
to me than it
does with your inductors.

It's no different from any other two-valve circlotron AFAIK
so I can picture someone else's. Hmm....looks complicated to
me.

If you mean complex, as in too clever by half,
then that seems appropriate for a circlotron, which is
something of a contrivance in any case.

It's not obvious to me that the straight circlotron is a
"contrivance".

It looks like a smarty-pants indulgence and, considering the
world isn't full of circlotrons, I guess it can be equalled
by more obvious designs. I'm interesting in pursuing it
because it fascinates me. What appears to be a contrivance
depends on your cognitive framework, I suppose.

It would be interesting to compare the effects of
imbalanced
drive signals or poorly matched valves. Sound wise, I
suspect that the constraints of all that cross-bracing,
as
it were, may be essentially un-musical.

What cross-bracing are you talking about, are you talking
about your
modified circlotron?


Yes. By crossbracing, I mean linking two points across a
structure such that they are constrained to move in unison.
An analogy from mechanics referring to the way the inductors
link each side of the circuit with the other, particularly
if all four windings are on one core.

It will probably weigh more, but cost about the same,
it
seems to me, but that's just a vague top-down feeling
rather
than a bottom-up analysis.

My vague feeling is that it would cost more than simply
using two power
supplies, but how much depends on how much ripple
filtering you feel is
needed in the power supply(s), what problems are
created
by combining
all four of your inductors onto a single core, or even
two
cores, and
there are problems, and finally the fact that your OPT
will of necessity
require a higher inductance to maintain the same LF
performance when
shunted by all those inductors.

I was hoping the inductor could be easily made using a
pair
of bifilar windings. What problems do you perceive?

OK, sounds like an idea, while you are at it why not ditch
the OPT, and
add an additional winding on the "inductor" to drive the
speaker without
a separate OPT?

Cripes.


The idea of combining all five inductors is interesting, but
for the moment I'm looking at the original three-inductor
idea.

The way I probably saw them, they are a pair of differential
mode chokes used to isolate the signal from the power
supply. It didn't occur to me that they could be combined
into a single inductor and I haven't tried the circuit in
that form; it was pointed out in discussion and seems so
plausible that I have accepted the idea without much
thought.

I hadn't particularly noticed that the combined inductor
could be seen as in shunt with the output transformer and,
the way it's drawn, that may not have struck anyone else
who looked.

I had rather assumed they are in series with the
transformer, and I guess others may have done so too. I
don't think a higher inductance for the OPT should be
necessary.

The combination of all the inductors is directly in
parallel with the
OPT.

Yes, well spotted. As the design appears in my diagram,
there are two
inductors. Ignoring the caps, they are connected in series
and in parallel with each other, and together they are
connected across the OPT.

If the zeros can't be kept far enough apart, the
design loses its appeal for me.

The details, such as CMRR and PSRR and minor poles and
zeros, and cap voltage and ESL and ESR, etc etc, aren't
immediately obvious to me. I hope to get the basics
optimised first, and then consider how sensitive it is to
imperfections and parasitics.

I'm not sure what you are trying to say about the zeros,
and minor poles
and zeros here?

By minor poles and zeros, I mean those other than the
dominant pole and zero. If the inductance shunting the OPT
is great enough, then its effect shouldn't be significant.
However, there is also the matter of the load reflected by
each winding, also combined in shunt. I remember this is the
point where I originally put the design to one side coz I
couldn't be bothered to think through what that might be...

If I am following what you have said, your main design
goal is to
minimize the turns ratio of the OPT, and that class B
operation is not a
primary design goal?

I think the minimising of optimum turns ratio is why PPP
architectures found favour, even if their aficionados
don't
know it, and is linked to the tendency towards falling
speaker impedance.

If an optimum turns ratio is the important thing, why not
just connect
all the valves in parallel in a simple PSE circuit? One
of the left
over valves from the push-pull driver circuit can then be
repurposed as
a cathode follower to solve the low grid resistance
"problem".

Then it would be an entirely different thing! Anyway,
there's the matter of the large idle current and big gapped
transformer that will have difficulty in providing accurate
bass, and other absent advantages of PP. PPP designs allow
PP and parallel with just two valves.

I might go as far as to say that falling
speaker impedance has been the death of traditional valve
hi-fi.

That view depends on your thesis that a slightly higher
than optimum
turns ratio is really a serious compromise.

How serious is serious? My thesis isn't as you have put
it...you put in "slightly" and "serious" to put it in the
worst light possible, and it's not my thesis anyway.

Given a trend towards lower speaker impedance, what is the
best response? You might reduce the turns on the secondary,
or increase the turns on the primary, both of which increase
the turns ratio, or you might reduce the source resistance,
keeping the turns ratio the same. Which do you think would
be best when considering bandwidth?

The simplification, from the point of view of the signal, of
considering the inductors in shunt is useful, thanks for
taking the trouble to notice it. My thoughts are now that
the OPT primary should present about 2.5k, and the four
other windings in the primary circuit should present a
considerably higher inductance. At present I notice the OPT
ended up with 50H, and the others with 10H each, so I may
well have been exploring the point where they begin to
dominate the LF limit.

Perhaps you could submit your own circlotron for comparison?
Considering the present EL84 is not an ideal choice, I could
change to KT88, or whatever's more suitable that we can find
a model for.

Ian

[Patrick Turner]

Ian Iveson wrote:

John Byrns wrote:


Also I am not a transformer guy, can you explain the
theory behind
"minimising the optimum turns ratio for driving low
impedance speakers",
or maybe Patrick can jump in here and explain what the
advantage of
minimizing the turns ratio is?

Fat chance, from my viewpoint. I'd rather trust a snake.

A snake? Snakes know less than I do about transformers,
Patrick on the
other hand is quite knowledgeable when it comes to
transformers.

He knows how to make them.

The optimum turns ratio is the one that presents the
optimum
load to the valves for a given speaker impedance. This
optimum is reduced with falling Zaa, and minimised by
using
valves in parallel. Lower turns ratio allows wider
bandwidth, all other things being equal.

Menno van der Veen is a useful source, if you have his
book
or papers.

I have neither his books nor his papers, so he is useless
to me.

You'd enjoy his book. He derives a complete transfer
function for a typical PP amp, and a load of parameters
relating output transformers to audio performance. Also
considers various proportions of distributed load, several
simple amp designs, and a good design procedure for ensuring
stability.

From his transfer function he derives a formula for the
ratio of LF and HF bandwidth limits, influenced by both the
impedance ratio and the turns ratio.

without
unduly loading the driver stage.

I don't follow you here, how does your circuit relate
to
the "loading"
of the driver stage?

The most immediately obvious way of using parallel valves
in
PP is to use several valves in parallel on each side of
an
ordinary PP circuit. This decreases the input resistance
of
the stage because each valve has a minimum value of grid
leak resistor. The lower input resistance is a greater
load
for the driver to drive. Perhaps I have misused the term?

I think I see the point you are trying to make, but there
are a couple
of flaws in your logic. The circlotron may have higher
value grid
resistors loading the driver stage, when using an OPT with
a given
primary/secondary ratio, than a PPP circuit does, but it
also requires a
considerably higher drive voltage which doesn't come for
free. Given
that you have the excess drive voltage available, as you
must for a
circlotron, you can bring the value of the grid resistance
loading the
driver up in the PPP design by simply adding a series
resistor between
each driver plate and following grid circuit to form a 2:1
voltage
divider, that will give the PPP the same grid resistance
loading the
driver as the circlotron has. Or given that the
circlotron driver stage
will require significantly greater voltage gain than the
PPP circuit
does, the PPP driver can compensate by using valves with
lower gain and
plate resistance to facilitate driving the lower grid
resistance, or
alternatively if the circlotron driver gets its extra gain
by using more
valves, those can be repurposed as cathode followers to
drive the lower
value grid resistors in the PPP design. The second flaw
is that the PPP
design will have twice the power output capability of the
circlotron, so
the two designs aren't directly comparable in the first
place.

I didn't intend to give the impression that the circlotron
is unique in that respect. Yes, there are other ways of
achieving pretty much anything...swings and roundabouts, no
free lunch etc.

I haven't put it forward as the best design of amplifier
from an analytical point of view. Far from it: as I made
clear in the section on dialectic, which you cut. My only
hope was that someone might be interested in optimising the
OPT, and then maybe considering the effect of the other
inductors.

I'm not taken with designs using several power
supplies.

Why not? It doesn't take any more copper in the Power
Transformer than
a single supply and your circuit already includes the
extra electrolytic
capacitors a second power supply would require, so your
circuit looses
on cost and complexity, and pointless complexity often
opens the door to
unexpected problems so why go down that road, what is
the
advantage?

The 'lytics may be its downfall, I would be the first to
admit. There are also the details of the bias circuit to
sort out. I'm not keen on so many complexities attached
to
that single loop (or figure of 8 if you like).

Advantages from a narrow engineering
perspective...hmm...how
about perfect matching of HT voltage for each valve?

How close do the HT voltages have to be matched? The
valves themselves
won't be perfectly matched, the two HT voltages are likely
to be quite
close, especially relative to the matching of the two
valves, I think
you are worrying about a nonexistent problem here.

How equal the two power supply voltages of an ordinary
circlotron need to be is something none of us knows, and
words like "not very close" don't mean much to me. I will
compare with an ordinary circlotron, with various
imperfections such as mismatched valves, drive voltages,
etc., eventually.

Also,
it has a neat simplicity to it. I don't know why you see
it
as complicated.

Nor do I see the simplicity that you do.

If the power supplies were included, I would
say the ordinary dual supply looked more complicated,
just
because it has two connections to the outside that
require
matching.

Draw the circuit with two power supplies, it looks simpler
to me than it
does with your inductors.

It's no different from any other two-valve circlotron AFAIK
so I can picture someone else's. Hmm....looks complicated to
me.

If you mean complex, as in too clever by half,
then that seems appropriate for a circlotron, which is
something of a contrivance in any case.

It's not obvious to me that the straight circlotron is a
"contrivance".

It looks like a smarty-pants indulgence and, considering the
world isn't full of circlotrons, I guess it can be equalled
by more obvious designs. I'm interesting in pursuing it
because it fascinates me. What appears to be a contrivance
depends on your cognitive framework, I suppose.

It would be interesting to compare the effects of
imbalanced
drive signals or poorly matched valves. Sound wise, I
suspect that the constraints of all that cross-bracing,
as
it were, may be essentially un-musical.

What cross-bracing are you talking about, are you talking
about your
modified circlotron?


Yes. By crossbracing, I mean linking two points across a
structure such that they are constrained to move in unison.
An analogy from mechanics referring to the way the inductors
link each side of the circuit with the other, particularly
if all four windings are on one core.

It will probably weigh more, but cost about the same,
it
seems to me, but that's just a vague top-down feeling
rather
than a bottom-up analysis.

My vague feeling is that it would cost more than simply
using two power
supplies, but how much depends on how much ripple
filtering you feel is
needed in the power supply(s), what problems are
created
by combining
all four of your inductors onto a single core, or even
two
cores, and
there are problems, and finally the fact that your OPT
will of necessity
require a higher inductance to maintain the same LF
performance when
shunted by all those inductors.

I was hoping the inductor could be easily made using a
pair
of bifilar windings. What problems do you perceive?

OK, sounds like an idea, while you are at it why not ditch
the OPT, and
add an additional winding on the "inductor" to drive the
speaker without
a separate OPT?

Cripes.

The idea of combining all five inductors is interesting, but
for the moment I'm looking at the original three-inductor
idea.

The way I probably saw them, they are a pair of differential
mode chokes used to isolate the signal from the power
supply. It didn't occur to me that they could be combined
into a single inductor and I haven't tried the circuit in
that form; it was pointed out in discussion and seems so
plausible that I have accepted the idea without much
thought.

I hadn't particularly noticed that the combined inductor
could be seen as in shunt with the output transformer and,
the way it's drawn, that may not have struck anyone else
who looked.

I had rather assumed they are in series with the
transformer, and I guess others may have done so too. I
don't think a higher inductance for the OPT should be
necessary.

The combination of all the inductors is directly in
parallel with the
OPT.

Yes, well spotted. As the design appears in my diagram,
there are two
inductors. Ignoring the caps, they are connected in series
and in parallel with each other, and together they are
connected across the OPT.

If the zeros can't be kept far enough apart, the
design loses its appeal for me.

The details, such as CMRR and PSRR and minor poles and
zeros, and cap voltage and ESL and ESR, etc etc, aren't
immediately obvious to me. I hope to get the basics
optimised first, and then consider how sensitive it is to
imperfections and parasitics.

I'm not sure what you are trying to say about the zeros,
and minor poles
and zeros here?

By minor poles and zeros, I mean those other than the
dominant pole and zero. If the inductance shunting the OPT
is great enough, then its effect shouldn't be significant.
However, there is also the matter of the load reflected by
each winding, also combined in shunt. I remember this is the
point where I originally put the design to one side coz I
couldn't be bothered to think through what that might be...

If I am following what you have said, your main design
goal is to
minimize the turns ratio of the OPT, and that class B
operation is not a
primary design goal?

I think the minimising of optimum turns ratio is why PPP
architectures found favour, even if their aficionados
don't
know it, and is linked to the tendency towards falling
speaker impedance.

If an optimum turns ratio is the important thing, why not
just connect
all the valves in parallel in a simple PSE circuit? One
of the left
over valves from the push-pull driver circuit can then be
repurposed as
a cathode follower to solve the low grid resistance
"problem".

Then it would be an entirely different thing! Anyway,
there's the matter of the large idle current and big gapped
transformer that will have difficulty in providing accurate
bass, and other absent advantages of PP. PPP designs allow
PP and parallel with just two valves.

I might go as far as to say that falling
speaker impedance has been the death of traditional valve
hi-fi.

That view depends on your thesis that a slightly higher
than optimum
turns ratio is really a serious compromise.

How serious is serious? My thesis isn't as you have put
it...you put in "slightly" and "serious" to put it in the
worst light possible, and it's not my thesis anyway.

Given a trend towards lower speaker impedance, what is the
best response? You might reduce the turns on the secondary,
or increase the turns on the primary, both of which increase
the turns ratio, or you might reduce the source resistance,
keeping the turns ratio the same. Which do you think would
be best when considering bandwidth?

The simplification, from the point of view of the signal, of
considering the inductors in shunt is useful, thanks for
taking the trouble to notice it. My thoughts are now that
the OPT primary should present about 2.5k, and the four
other windings in the primary circuit should present a
considerably higher inductance. At present I notice the OPT
ended up with 50H, and the others with 10H each, so I may
well have been exploring the point where they begin to
dominate the LF limit.

Perhaps you could submit your own circlotron for comparison?
Considering the present EL84 is not an ideal choice, I could
change to KT88, or whatever's more suitable that we can find
a model for.

Ian

So Ian, if you seem so certain your amp will work, why have you not
built a sample and published all your test results on a page we can
view?

Patrick Turner.

wh

[John Byrns]

In article ,
"Ian Iveson" wrote:

John Byrns wrote:


Menno van der Veen is a useful source, if you have his
book
or papers.

I have neither his books nor his papers, so he is useless
to me.

You'd enjoy his book. He derives a complete transfer
function for a typical PP amp, and a load of parameters
relating output transformers to audio performance. Also
considers various proportions of distributed load, several
simple amp designs, and a good design procedure for ensuring
stability.

It's not obvious to me that I would enjoy his book, I think I read one
of his papers on Output Transformers once and did not find it enjoyable,
maybe I would enjoy it more with a reread, if I could figure out where I
found it the first time.

From his transfer function he derives a formula for the
ratio of LF and HF bandwidth limits, influenced by both the
impedance ratio and the turns ratio.

How about giving us a summary, mainly the bottom line on the LF & HF
bandwidth limits, no need to bother us with the derivation.

It would be interesting to compare the effects of
imbalanced
drive signals or poorly matched valves. Sound wise, I
suspect that the constraints of all that cross-bracing,
as
it were, may be essentially un-musical.

What cross-bracing are you talking about, are you talking
about your
modified circlotron?

Yes. By crossbracing, I mean linking two points across a
structure such that they are constrained to move in unison.
An analogy from mechanics referring to the way the inductors
link each side of the circuit with the other, particularly
if all four windings are on one core.

I like the idea of "cross-bracing", and especially putting all four
inductors on one core, and including the OPT too.

I was hoping the inductor could be easily made using a
pair
of bifilar windings. What problems do you perceive?

OK, sounds like an idea, while you are at it why not ditch
the OPT, and
add an additional winding on the "inductor" to drive the
speaker without
a separate OPT?

Cripes.


The idea of combining all five inductors is interesting, but
for the moment I'm looking at the original three-inductor
idea.

The way I probably saw them, they are a pair of differential
mode chokes used to isolate the signal from the power
supply. It didn't occur to me that they could be combined
into a single inductor and I haven't tried the circuit in
that form; it was pointed out in discussion and seems so
plausible that I have accepted the idea without much
thought.

I hadn't particularly noticed that the combined inductor
could be seen as in shunt with the output transformer and,
the way it's drawn, that may not have struck anyone else
who looked.

I had rather assumed they are in series with the
transformer, and I guess others may have done so too. I
don't think a higher inductance for the OPT should be
necessary.

The combination of all the inductors is directly in
parallel with the
OPT.

Yes, well spotted. As the design appears in my diagram,
there are two
inductors. Ignoring the caps, they are connected in series
and in parallel with each other, and together they are
connected across the OPT.

I am presently looking at a design with a five winding inductor,
combining the functions of all your inductors and the OPT into a single
piece of iron.

Perhaps you could submit your own circlotron for comparison?
Considering the present EL84 is not an ideal choice, I could
change to KT88, or whatever's more suitable that we can find
a model for.

Yes, I think I may very well submit my own "circlotron" design based on
your single power supply idea, but with all your inductors and the OPT
combined on a single core.


Regards,

John Byrns

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