Yo, Patrick

Super analysis. I've saved it. Thanks.

Make sure you get the Stockman article on NFB in triodes that John
Stewart posted at ABSE. I couldn't find it there but he kindly
e-mailed me a copy. It's one of those once in half a century
keep-forever papers.

I have no problem with either John's 14dB at my choice of 5K6 load on
the 300B anode, or your statement that at around 3K5 it would be 18dB.
That the NFB falls with increasing load may in fact account for the
fact that higher anode loads please more, as less feedback may mean
fewer add-on artifacts.

Andre Jute


Patrick Turner wrote:

Andre Jute wrote:

Reply to message

Yo, John

The key thing you say in your post is 'local NFB similar to the 14 dB
or so that is inherent in the 300B at one of the operating points you
suggested'. How did you calculate this number or where did you come by
it?

****

Indeed, I have wondered about the internal FB number myself.

Maybe I could answer the question myself.

Let us suppose the 300B was a hypothetical perfect pentode, with a screen
to prevent any FB at all.
Then the Ra would be infinite.
But you'd still have a device with a fixed gm
And gain, A, for all tubes = u x RL / ( Ra + RL )

also, u = gm x Ra for all tubes.

So we can say A = (gm x Ra) x RL / ( Ra + RL )
Trouble is, this only works for finite values of Ra, and where infinite values
of Ra
exist, its too darn hard to work out what to do to make a simple formula work,
but when we measure nearly perfect pentodes with Ra = 1M or greater,
and gm say 5 to 10 mA/V, we find A = gm x RL, near enough, where RL = say 10k

The maximum gm of the 300B occurs when the load is a short circuit, and there
is thus no anode voltage change, and hence no internal applied electrostatic
negative feedback.

At Ea = 400v, and Ia = 80 ma,
Ra = 670 ohms, and u = 5, so gm = 5 / 670 = 7.5 mA/V.

So we could say that if the RL was say 5k, then the hypothetical pentode gain
would be 5,000 x 0.0075 = 37.5.

But in practice with a real 300B triode we get triode u = 5, Ra = 670 ohms
so gain = 5 x 5,000 / (5,000 + 670 ) = 4.4.

We see a gain reduction with a 5k load of 4.4 / 37.5, or about 8.5 times,
so that's about -18dB.

So what is causing the reduction of gain from 37.5 to 4.4?

Its the NFB of course.

So what amount of series voltage FB would have to be applied to a
current source with gm = 7.5 mA/V to give a gain reduction of 18 dB
into a 5 k RL?

If you had a normal resistor network to deliver conventional NFB to the
hypothetical 300B pentode, we would have to ignore the loading effects
of the resistance network, so let's have a hypothetical divider, so we can
establish what the B is for the feedback network, B being the fraction of
the output voltage fed back in series with the input to the grid.
The gain with FB applied is A' = A / (1 + [ A x B ] ).
Thus we get 4.4 = 37.5 / ( 1 + [ 37.5 x B ] ).

Therefore, 37.5 = 4.4 + 165B.
B = 0.20.

Therfore, if we had a perfect pentode, gm = 7.5 mA/V,
and a divider network so that 0.2 x the anode voltage was applied
in series with the grid voltage at the cathode, you'd have the equivalent
circuit for a 300B triode.

So if you have +122 vrms at the 5 k load, ( 3 watts ),
and you had a perfect pentode, you'd have -24.4 volts of NFB applied to the
cathode,
and 3.26 vrms applied between the grid and cathode, so you'd need
-27.66v at the grid.

122 / 27.66 = 4.4, the gain of the triode, so I have verified the hypothetical
operation.

The Ra of a pentode is reduced by the NFB so that
Ra' = Ra / 1 + [ u x B ] ).

Infinity doesn't rest well in our equations, so let's choose
Ra = 20k...
u = gm x Ra = 0.0075 x 20k = 150.

So Ra' = 20,000 / 1 + [150 x 0.2 ] ) = 263 ohms

This result is what we'd expect with a 6550, set up as a tetrode, with a 20%
CFB winding.....

In practice we see the Ra of the 300B resting at 670 ohms,
( Svetlana data )

What if Ra = 10 M? Surely this would be close enough to a perfect pentode.
u = 10,000,000 x 0.0075 = 75,000.
Ra would remain at 0.0075 A/V.

Ra' = 10,000,000 / ( 1 + [ 75,000 x 0.2 ] ) = 666 ohms

Now this is the value of the 300B triode Ra, so
we could assume the 300B was in fact like a perfect pentode/tetrode,
but with all this NFB inside it.

The load determines the gain of any tube,
but what if the load was say 1M?

Then its very close to a CCS.

Then if you consider the 300B as a perfect pentode
with NFB, and B = 0.2 within it, and note that this B figure is 1/u, where u is
the triode u,
then the pentode gain without NFB would be a gi-normous figure.
So even with only 0.2 of the output fed back in series with the input
the gain reduction is a huge number, so you get the maximum
benefit of the very large amount of NFB applied within the 300B
when the load = CCS, or a huge number of times of the triode Ra.

Triodes are not perfect, and nor does a perfect pentode actually exist,
let alone a hypothetical one, since the electrode structure and the bts holding
it in place
are never perfect. There are stray electrostatic effects as well as the main
ones we are
theroretically dealing with.

The relationship of plate current vs plate voltage isn't a linear one.
The plate data curves show this non linear increase in plate current for a given
increase of plate
voltage.

Yet its this very non linear function on the tube which delivers the voltage
feedback
within the triode.
So one might expect that with so much NFB within a triode when a CCS load is
used
that the non-linearity of the divider element delivering the NFB would be
eliminated. IN practice we see that indeed the triode attempts to do this,
but the process isn't perfect, and infact some distortion still occurs with a
CCS load,
but its a tiny fraction of what you get with a normal power tube load where
RL = 4 Ra.

That's the other benefit of NFB.
Any nonlinearity of the active elements in a chain enclosed by a FB loop
tend to be compensated for by the NFB loop.
So the thd of a driver stage and the output stage is equally reduced if the loop

of NFB includes the two cascaded stages.

Let's see what sort of figures you guys come up with.

Patrick Turner.

PS,

I don't have pistols for a dawn shoot-out any more.
I have so many holes in my feet I can hardly stand up.
So I gave them away
to a brash young man who didn't survive long.......








The rest of this is smalltalk, of merely transient interest.

Yeah, I think you definitely had Pinkerton confused with someone else.
We were all misled by the fact that he rode in on Arnie's coattails
into believing he would at least know something about transistors.

I do actually know that transistors cannot operate in Class A1; that
was my very first disappointment with them. I was just making a mild
funny at the expense of a humourless twerp who stormed into RAT and
immediately offered a pale shadow my current project, which he
couldn't deliver, and then started hurling racist abuse at me, backed
up with long-discredited Magnequest Scum smears.

I think it would be a waste of time to wait for Pinkerman's so-called
design. Well before his latest debacle in the NFB thread which
dissipated (pun intended) your remaining faith in him, Patrick
patiently demonstrated (at greater length than Hugh Hefner's
'philosophy', and in my threads -- don't do it again, Patrick; once is
funny, twice calls for pistols at dawn) that Pinky doesn't really have
much idea of Class A amps either. This poor gatecrasher's insistence
on an emitter follower was just about the final straw for everyone,
and he knew it for he took his foolish promise away for 'tests'.

Therefore I suggest that if your interest has been piqued, you publish
the design you found on your netsite and then those interested can
discuss it without reference to Pinky's aborted efforts.

Truth is, I think you're all on a hiding to nothing though, one never
knows, you might find the journey entertaining. My own moment of
revelation in Class A discrete transistor amps may amuse mildly. The
zero-return point of marginality in transistors arrived when I
realized that for the price of the heatsink I was having cast I could
build an entire and very nice tube amp. On a very hot day in a foundry
that was it. I cancelled the order and was out of tranny DIY for good.
Well, a little twenty minute opamp amp now and again hardly counts as
an addiction.

Andre Jute

John Byrns wrote:

In article ,
(Andre Jute) wrote:

Those aren't my rules, John. Iain Churches, who invited this rude
fellow Pinkerton to submit a design in return for a little courtesy
(which we haven't seen) set the rules. I'm merely supplying
information to fit the rules Iain made.

The rest of your remarks come down to a suggestion that I
retroactively alter the design of the KISS 300B I am embarked on to
suit the needs of an interloper. That's a bit tortuous! I think I'll
just stick to the straight and narrow of my business and ignore the
PinkyTron Borg'o'Blaster until it comes to its inevitable fall.


Hi Andre,

I went to the trouble of rummaging through many of the posts in the sub
threads that I hadn't read, it appears to me that Iain Churches only made
the suggestion that Pinkerton should design a transistor KISS amplifier to
contrast with your vacuum tube design, it appears that the detailed
specifications for the transistor KISS amplifier were provided by you when
you stated, "the main parameters he should match are single-ended output,
3W, all of it in Class A1, zero negative feedback". Needless to say my
faith in Pinkerton's ability to design a suitable transistor KISS
amplifier was badly shaken by his recent demonstration of a lack of
understanding of the workings of "NFB". I think I may have had Pinkerton
confused with someone else.

I was not at all suggesting that you redesign your KISS amplifier to
eliminate the feedback that is inherent in triode tubes, what I was
suggesting was that the transistor KISS amplifier also be allowed to use
local NFB similar to the 14 dB or so that is inherent in the 300B at one
of the operating points you suggested. Also class "A1" operation applies
to vacuum tubes and does not make sense in connection with transistors,
better to simply say the transistor KISS amplifier should operate in class
A. I would sum a specification for the transistor KISS amplifier up this
way, "the main parameters a transistor KISS amplifier should match are
single-ended output, 3W, all of it in Class A, with only local negative
feedback at audio frequencies."

As an example of a transistor KISS amplifier I will have to scan the old
article from Audio magazine that I mentioned and put it on my web site for
all to comment on. The design has a number of problems, including the use
of late 1950's transistors, and a very high damping factor, but it is a
starting point for a design that would eliminate these problems and
others.


Regards,

John Byrns


Surf my web pages at, http://users.rcn.com/jbyrns/

Andre Jute wrote:

Yo, Patrick

Super analysis. I've saved it. Thanks.

Make sure you get the Stockman article on NFB in triodes that John
Stewart posted at ABSE. I couldn't find it there but he kindly
e-mailed me a copy. It's one of those once in half a century
keep-forever papers.

I have the one lone Stockman article from 1953 in pdf, probably
JS emailed it to me years ago.
I am having a sl of probelms with my ISP, who now says he has a dud server,
and the saga of poor news group access continues, so I cannot
see the most recent article posted by JS, but I did see where a guy
complained about the quality of the scan of the article.

I found the Stockman article I di have to be almost incomprehensible, and what I said
about the operation
of a 300B as a current source, like a perfect pentode but with a loop of FB was
based on my simpleton's grasp of the concepts, and some very basic gain/feedback
formulae,
which aren't all that accurate since I don't comply with exact application of the
signage
for u and gain, eg, u should always be a negative number, since all amplifying devices

invert an input signal, so -1v in gives + 4v out....



I have no problem with either John's 14dB at my choice of 5K6 load on
the 300B anode, or your statement that at around 3K5 it would be 18dB.

Its actually more likely to be the other way around, 14 dB for 3.5k, and
18 dB with 5.6k.
The higher the RL, the greater the amount of applied NFB.
Hence we see less thd as we raise RL.
This is usally true of all amps.
With a pentode amp, the output tubes have gain = u when the amp is unloaded,
so the dB of applied NFB is enormously greater than when the rated load is used.
This often takes the amp to a point of instability, so gain should be controlled,
at least at HF, so hence R&C networks are placed at V1 load, across each 1/2 of the
OPT,
and so on, and then it don't matter if we don't have a load connected.
Pentode thd is often very bad when no load is connected, so a NFB network
won't lower the thd as much as one would expect when the applied NFB is high,
say 35 dB with no load.
Dynaco bootstrap their input pentode's RL to force its gain up, and with it
the thd, but luckily for them the open loop thd still isn't that bad,
and the input tube's thd is all reduced by the global loop
of NFB, ( so they reckon, but I am yet to build a voltage amp this way,
and I prefer to squander an extra tube and socket to a an all triode driver amp ).



That the NFB falls with increasing load may in fact account for the
fact that higher anode loads please more, as less feedback may mean
fewer add-on artifacts.

The ppl who came before me to explain the phenomena of NFB's ability
to cause the number of products appearing at the output to rise
had some pertinent conclusions to make.
They said that 10 dB of NFB around an amp with 10% thd open loop before clipping
would likely make the sound worse.
So they said either use no FB, or the equivalent of what is now considered to be a
shirtload of NFB, say over 20 dB.
The alternative is to make sure the thd is low to begin with, around 1%,
and then there was no need to worry for the "second order products"
that are generated by the IMD process of the applied low quantity of NFB.
Williamson started that ball rolling with his use of KT66 in triode, hi-fi for the
masses,
(well, well healed masses), and by applying 20 dB around an amp with 1% open loop max,
he
then was able to get 0.01% at normal levels.
Not to many folks complained about the sound or the use of NFB.
Probably, 20 dB is over cooking it.

Thus in a preamp we can happily use an unbypassed cathode resistor in a triode gain
stage
without worrying about the miniscule extra 3H etc that will definately occur.
In a 12AU7 set up in CF, there is about 20 dB NFB in action, and the 3%
of 2H normally present in the plate loaded condition at say 50 vo does get reduced to
about
0.5% 2H, but a little dash of 3H etc appears.
I am not sure if the level of 3H is more or less after the CF.
But the only time you'd have 30v from such a triode is in a power amp driver, maybe
for class AB2, and then the 3H from other avenues would swamp the IMD
caused 3H in the CF.

If one examines the spectra of the 300B, I am sure there is some 3H there,
and maybe because of the FB internally.
The plate current - voltage character depends on a cube root of something squared,
and the slight S curve that generates 3H isn't found in the linear usable portion of
a triodes working range.
If its there its way below the 2H product, my CRO tells me.

When a pair of these tubes is used in *pure* class A, with a 10k to 20k RLa-a,
its surprising how low the thd becomes, but its all mostly 3H.


I recall seeing a british magazine cover from around 1993 where the header cried out
"28 watts, 0.2% thd, no NFB" with a photo of a PP 300B amp in the background.
But I think the 0.2% was at lower levels than 28 watts, which must have been
class AB watts.

I think we can forgive the God Of Triodes for providing us with NFB
locked up inside the tube and tamperproof before we woke up to how to do it with
external loopery tricks.
Not only should we forgive Him, we oughta thank Him.
But He hasn't allowed the secrets of a bjt with internal FB be known to mankind.
Gods are prone to saying suffering is good for the soul.....
I don't argue, they have high voltages in their Control.
Some Gods are sentimentalists, and like the glow of their allowances,
and are loathe to let us have anything much better.


Patrick Turner.



Andre Jute

Patrick Turner wrote:

Andre Jute wrote:

Reply to message

Yo, John

The key thing you say in your post is 'local NFB similar to the 14 dB
or so that is inherent in the 300B at one of the operating points you
suggested'. How did you calculate this number or where did you come by
it?

****

Indeed, I have wondered about the internal FB number myself.

Maybe I could answer the question myself.

Let us suppose the 300B was a hypothetical perfect pentode, with a screen
to prevent any FB at all.
Then the Ra would be infinite.
But you'd still have a device with a fixed gm
And gain, A, for all tubes = u x RL / ( Ra + RL )

also, u = gm x Ra for all tubes.

So we can say A = (gm x Ra) x RL / ( Ra + RL )
Trouble is, this only works for finite values of Ra, and where infinite values
of Ra
exist, its too darn hard to work out what to do to make a simple formula work,
but when we measure nearly perfect pentodes with Ra = 1M or greater,
and gm say 5 to 10 mA/V, we find A = gm x RL, near enough, where RL = say 10k

The maximum gm of the 300B occurs when the load is a short circuit, and there
is thus no anode voltage change, and hence no internal applied electrostatic
negative feedback.

At Ea = 400v, and Ia = 80 ma,
Ra = 670 ohms, and u = 5, so gm = 5 / 670 = 7.5 mA/V.

So we could say that if the RL was say 5k, then the hypothetical pentode gain
would be 5,000 x 0.0075 = 37.5.

But in practice with a real 300B triode we get triode u = 5, Ra = 670 ohms
so gain = 5 x 5,000 / (5,000 + 670 ) = 4.4.

We see a gain reduction with a 5k load of 4.4 / 37.5, or about 8.5 times,
so that's about -18dB.

So what is causing the reduction of gain from 37.5 to 4.4?

Its the NFB of course.

So what amount of series voltage FB would have to be applied to a
current source with gm = 7.5 mA/V to give a gain reduction of 18 dB
into a 5 k RL?

If you had a normal resistor network to deliver conventional NFB to the
hypothetical 300B pentode, we would have to ignore the loading effects
of the resistance network, so let's have a hypothetical divider, so we can
establish what the B is for the feedback network, B being the fraction of
the output voltage fed back in series with the input to the grid.
The gain with FB applied is A' = A / (1 + [ A x B ] ).
Thus we get 4.4 = 37.5 / ( 1 + [ 37.5 x B ] ).

Therefore, 37.5 = 4.4 + 165B.
B = 0.20.

Therfore, if we had a perfect pentode, gm = 7.5 mA/V,
and a divider network so that 0.2 x the anode voltage was applied
in series with the grid voltage at the cathode, you'd have the equivalent
circuit for a 300B triode.

So if you have +122 vrms at the 5 k load, ( 3 watts ),
and you had a perfect pentode, you'd have -24.4 volts of NFB applied to the
cathode,
and 3.26 vrms applied between the grid and cathode, so you'd need
-27.66v at the grid.

122 / 27.66 = 4.4, the gain of the triode, so I have verified the hypothetical
operation.

The Ra of a pentode is reduced by the NFB so that
Ra' = Ra / 1 + [ u x B ] ).

Infinity doesn't rest well in our equations, so let's choose
Ra = 20k...
u = gm x Ra = 0.0075 x 20k = 150.

So Ra' = 20,000 / 1 + [150 x 0.2 ] ) = 263 ohms

This result is what we'd expect with a 6550, set up as a tetrode, with a 20%
CFB winding.....

In practice we see the Ra of the 300B resting at 670 ohms,
( Svetlana data )

What if Ra = 10 M? Surely this would be close enough to a perfect pentode.
u = 10,000,000 x 0.0075 = 75,000.
Ra would remain at 0.0075 A/V.

Ra' = 10,000,000 / ( 1 + [ 75,000 x 0.2 ] ) = 666 ohms

Now this is the value of the 300B triode Ra, so
we could assume the 300B was in fact like a perfect pentode/tetrode,
but with all this NFB inside it.

The load determines the gain of any tube,
but what if the load was say 1M?

Then its very close to a CCS.

Then if you consider the 300B as a perfect pentode
with NFB, and B = 0.2 within it, and note that this B figure is 1/u, where u is
the triode u,
then the pentode gain without NFB would be a gi-normous figure.
So even with only 0.2 of the output fed back in series with the input
the gain reduction is a huge number, so you get the maximum
benefit of the very large amount of NFB applied within the 300B
when the load = CCS, or a huge number of times of the triode Ra.

Triodes are not perfect, and nor does a perfect pentode actually exist,
let alone a hypothetical one, since the electrode structure and the bts holding
it in place
are never perfect. There are stray electrostatic effects as well as the main
ones we are
theroretically dealing with.

The relationship of plate current vs plate voltage isn't a linear one.
The plate data curves show this non linear increase in plate current for a given
increase of plate
voltage.

Yet its this very non linear function on the tube which delivers the voltage
feedback
within the triode.
So one might expect that with so much NFB within a triode when a CCS load is
used
that the non-linearity of the divider element delivering the NFB would be
eliminated. IN practice we see that indeed the triode attempts to do this,
but the process isn't perfect, and infact some distortion still occurs with a
CCS load,
but its a tiny fraction of what you get with a normal power tube load where
RL = 4 Ra.

That's the other benefit of NFB.
Any nonlinearity of the active elements in a chain enclosed by a FB loop
tend to be compensated for by the NFB loop.
So the thd of a driver stage and the output stage is equally reduced if the loop

of NFB includes the two cascaded stages.

Let's see what sort of figures you guys come up with.

Patrick Turner.

PS,

I don't have pistols for a dawn shoot-out any more.
I have so many holes in my feet I can hardly stand up.
So I gave them away
to a brash young man who didn't survive long.......








The rest of this is smalltalk, of merely transient interest.

Yeah, I think you definitely had Pinkerton confused with someone else.
We were all misled by the fact that he rode in on Arnie's coattails
into believing he would at least know something about transistors.

I do actually know that transistors cannot operate in Class A1; that
was my very first disappointment with them. I was just making a mild
funny at the expense of a humourless twerp who stormed into RAT and
immediately offered a pale shadow my current project, which he
couldn't deliver, and then started hurling racist abuse at me, backed
up with long-discredited Magnequest Scum smears.

I think it would be a waste of time to wait for Pinkerman's so-called
design. Well before his latest debacle in the NFB thread which
dissipated (pun intended) your remaining faith in him, Patrick
patiently demonstrated (at greater length than Hugh Hefner's
'philosophy', and in my threads -- don't do it again, Patrick; once is
funny, twice calls for pistols at dawn) that Pinky doesn't really have
much idea of Class A amps either. This poor gatecrasher's insistence
on an emitter follower was just about the final straw for everyone,
and he knew it for he took his foolish promise away for 'tests'.

Therefore I suggest that if your interest has been piqued, you publish
the design you found on your netsite and then those interested can
discuss it without reference to Pinky's aborted efforts.

Truth is, I think you're all on a hiding to nothing though, one never
knows, you might find the journey entertaining. My own moment of
revelation in Class A discrete transistor amps may amuse mildly. The
zero-return point of marginality in transistors arrived when I
realized that for the price of the heatsink I was having cast I could
build an entire and very nice tube amp. On a very hot day in a foundry
that was it. I cancelled the order and was out of tranny DIY for good.
Well, a little twenty minute opamp amp now and again hardly counts as
an addiction.

Andre Jute

John Byrns wrote:

In article ,
(Andre Jute) wrote:

Those aren't my rules, John. Iain Churches, who invited this rude
fellow Pinkerton to submit a design in return for a little courtesy
(which we haven't seen) set the rules. I'm merely supplying
information to fit the rules Iain made.

The rest of your remarks come down to a suggestion that I
retroactively alter the design of the KISS 300B I am embarked on to
suit the needs of an interloper. That's a bit tortuous! I think I'll
just stick to the straight and narrow of my business and ignore the
PinkyTron Borg'o'Blaster until it comes to its inevitable fall.


Hi Andre,

I went to the trouble of rummaging through many of the posts in the sub
threads that I hadn't read, it appears to me that Iain Churches only made
the suggestion that Pinkerton should design a transistor KISS amplifier to
contrast with your vacuum tube design, it appears that the detailed
specifications for the transistor KISS amplifier were provided by you when
you stated, "the main parameters he should match are single-ended output,
3W, all of it in Class A1, zero negative feedback". Needless to say my
faith in Pinkerton's ability to design a suitable transistor KISS
amplifier was badly shaken by his recent demonstration of a lack of
understanding of the workings of "NFB". I think I may have had Pinkerton
confused with someone else.

I was not at all suggesting that you redesign your KISS amplifier to
eliminate the feedback that is inherent in triode tubes, what I was
suggesting was that the transistor KISS amplifier also be allowed to use
local NFB similar to the 14 dB or so that is inherent in the 300B at one
of the operating points you suggested. Also class "A1" operation applies
to vacuum tubes and does not make sense in connection with transistors,
better to simply say the transistor KISS amplifier should operate in class
A. I would sum a specification for the transistor KISS amplifier up this
way, "the main parameters a transistor KISS amplifier should match are
single-ended output, 3W, all of it in Class A, with only local negative
feedback at audio frequencies."

As an example of a transistor KISS amplifier I will have to scan the old
article from Audio magazine that I mentioned and put it on my web site for
all to comment on. The design has a number of problems, including the use
of late 1950's transistors, and a very high damping factor, but it is a
starting point for a design that would eliminate these problems and
others.


Regards,

John Byrns


Surf my web pages at, http://users.rcn.com/jbyrns/

Patrick Turner wrote:

Andre Jute wrote:

Yo, Patrick

Super analysis. I've saved it. Thanks.

Make sure you get the Stockman article on NFB in triodes that John
Stewart posted at ABSE. I couldn't find it there but he kindly
e-mailed me a copy. It's one of those once in half a century
keep-forever papers.

I have the one lone Stockman article from 1953 in pdf, probably
JS emailed it to me years ago.
I am having a sl of probelms with my ISP, who now says he has a dud server,
and the saga of poor news group access continues, so I cannot
see the most recent article posted by JS, but I did see where a guy
complained about the quality of the scan of the article.

I found the Stockman article I di have to be almost incomprehensible, and what I said
about the operation
of a 300B as a current source, like a perfect pentode but with a loop of FB was
based on my simpleton's grasp of the concepts, and some very basic gain/feedback
formulae,
which aren't all that accurate since I don't comply with exact application of the
signage
for u and gain, eg, u should always be a negative number, since all amplifying devices

invert an input signal, so -1v in gives + 4v out....



I have no problem with either John's 14dB at my choice of 5K6 load on
the 300B anode, or your statement that at around 3K5 it would be 18dB.

Its actually more likely to be the other way around, 14 dB for 3.5k, and
18 dB with 5.6k.
The higher the RL, the greater the amount of applied NFB.
Hence we see less thd as we raise RL.
This is usally true of all amps.
With a pentode amp, the output tubes have gain = u when the amp is unloaded,
so the dB of applied NFB is enormously greater than when the rated load is used.
This often takes the amp to a point of instability, so gain should be controlled,
at least at HF, so hence R&C networks are placed at V1 load, across each 1/2 of the
OPT,
and so on, and then it don't matter if we don't have a load connected.
Pentode thd is often very bad when no load is connected, so a NFB network
won't lower the thd as much as one would expect when the applied NFB is high,
say 35 dB with no load.
Dynaco bootstrap their input pentode's RL to force its gain up, and with it
the thd, but luckily for them the open loop thd still isn't that bad,
and the input tube's thd is all reduced by the global loop
of NFB, ( so they reckon, but I am yet to build a voltage amp this way,
and I prefer to squander an extra tube and socket to a an all triode driver amp ).


That the NFB falls with increasing load may in fact account for the
fact that higher anode loads please more, as less feedback may mean
fewer add-on artifacts.

The ppl who came before me to explain the phenomena of NFB's ability
to cause the number of products appearing at the output to rise
had some pertinent conclusions to make.
They said that 10 dB of NFB around an amp with 10% thd open loop before clipping
would likely make the sound worse.
So they said either use no FB, or the equivalent of what is now considered to be a
shirtload of NFB, say over 20 dB.
The alternative is to make sure the thd is low to begin with, around 1%,
and then there was no need to worry for the "second order products"
that are generated by the IMD process of the applied low quantity of NFB.
Williamson started that ball rolling with his use of KT66 in triode, hi-fi for the
masses,
(well, well healed masses), and by applying 20 dB around an amp with 1% open loop max,
he
then was able to get 0.01% at normal levels.
Not to many folks complained about the sound or the use of NFB.
Probably, 20 dB is over cooking it.

Thus in a preamp we can happily use an unbypassed cathode resistor in a triode gain
stage
without worrying about the miniscule extra 3H etc that will definately occur.
In a 12AU7 set up in CF, there is about 20 dB NFB in action, and the 3%
of 2H normally present in the plate loaded condition at say 50 vo does get reduced to
about
0.5% 2H, but a little dash of 3H etc appears.
I am not sure if the level of 3H is more or less after the CF.
But the only time you'd have 30v from such a triode is in a power amp driver, maybe
for class AB2, and then the 3H from other avenues would swamp the IMD
caused 3H in the CF.

If one examines the spectra of the 300B, I am sure there is some 3H there,

In spite of what is circulated as general information on triodes, that being that
they are plagued only with 2nd harmonic distortion, there is in fact some 3rd
present in all. I will post something tomorrow on this. Cheers, John Stewart


and maybe because of the FB internally.
The plate current - voltage character depends on a cube root of something squared,
and the slight S curve that generates 3H isn't found in the linear usable portion of
a triodes working range.
If its there its way below the 2H product, my CRO tells me.

When a pair of these tubes is used in *pure* class A, with a 10k to 20k RLa-a,
its surprising how low the thd becomes, but its all mostly 3H.

I recall seeing a british magazine cover from around 1993 where the header cried out
"28 watts, 0.2% thd, no NFB" with a photo of a PP 300B amp in the background.
But I think the 0.2% was at lower levels than 28 watts, which must have been
class AB watts.

I think we can forgive the God Of Triodes for providing us with NFB
locked up inside the tube and tamperproof before we woke up to how to do it with
external loopery tricks.
Not only should we forgive Him, we oughta thank Him.
But He hasn't allowed the secrets of a bjt with internal FB be known to mankind.
Gods are prone to saying suffering is good for the soul.....
I don't argue, they have high voltages in their Control.
Some Gods are sentimentalists, and like the glow of their allowances,
and are loathe to let us have anything much better.

Patrick Turner.

In spite of what is circulated as general information on triodes, that being that
they are plagued only with 2nd harmonic distortion, there is in fact some 3rd
present in all. I will post something tomorrow on this. Cheers, John Stewart

For all the believers & others go to ABSE for a look at 300B performance.
There are two pages, about 500K. Cheers, John Stewart

John Stewart wrote:


In spite of what is circulated as general information on triodes, that being that
they are plagued only with 2nd harmonic distortion, there is in fact some 3rd
present in all. I will post something tomorrow on this. Cheers, John Stewart

For all the believers & others go to ABSE for a look at 300B performance.
There are two pages, about 500K. Cheers, John Stewart

The info gives the full power 2H and 3H, with 3H being about 15 dB down below 2H.

What we really wanna know is the relationship at 3 watts, not 8 watts, where clipping
has begun to occur.

Patrick Turner.

Patrick Turner wrote:

John Stewart wrote:


In spite of what is circulated as general information on triodes, that being that
they are plagued only with 2nd harmonic distortion, there is in fact some 3rd
present in all. I will post something tomorrow on this. Cheers, John Stewart

For all the believers & others go to ABSE for a look at 300B performance.
There are two pages, about 500K. Cheers, John Stewart

The info gives the full power 2H and 3H, with 3H being about 15 dB down below 2H.

What we really wanna know is the relationship at 3 watts, not 8 watts, where clipping
has begun to occur.

Patrick Turner.

Go to ABSE for that relationship. JLS

John Stewart wrote:

Patrick Turner wrote:

John Stewart wrote:


In spite of what is circulated as general information on triodes, that being that
they are plagued only with 2nd harmonic distortion, there is in fact some 3rd
present in all. I will post something tomorrow on this. Cheers, John Stewart

For all the believers & others go to ABSE for a look at 300B performance.
There are two pages, about 500K. Cheers, John Stewart

The info gives the full power 2H and 3H, with 3H being about 15 dB down below 2H.

What we really wanna know is the relationship at 3 watts, not 8 watts, where clipping
has begun to occur.

Patrick Turner.

Go to ABSE for that relationship.

I don't have up to date ABSE at all.

I did get the two other pages you sent directly and they seem to be simply
a tabulated set of operating conditions for 300B at many Ea and Ia and RL combinations
and for full power output only.
There were no graphs showing the incremental rise in 2H and 3H for any of the conditions.

afaik, the levels of 3H are lower at 3 watts relative to 2H at full power.

There is no reason not to think that the 3H is produced in the triode as a result
of IMD action due to the NFB loop present.
The lower the RL, the less NFB you get, but more 2H, and more 3H......

The 3H would fall relative to the 2H as the load rises, and the amount of FB increases.
With a CCS, there isn't much 3H at all until you get to extreme voltage swings
where the plate resistance lines crowd together.

Patrick Turner.



JLS