Gentlemen,

plate curves for generic triode / tetrode / pentode mode
can easily be measured and drawn using a statical approach.

For a few power output tubes like for example EL34, also
some UL mode operation plate curves for certain plate/screen
tap ratios are given in their spec sheets. Almost no plate
curves are available for CFB mode.

But how were the few available UL op plate curves measured
or derived?

Tnx2u,

Tom

--

Tom Schlangen wrote:

Gentlemen,

plate curves for generic triode / tetrode / pentode mode
can easily be measured and drawn using a statical approach.

For a few power output tubes like for example EL34, also
some UL mode operation plate curves for certain plate/screen
tap ratios are given in their spec sheets. Almost no plate
curves are available for CFB mode.

But how were the few available UL op plate curves measured
or derived?

Not hard i would say.

Ra can be plotted for any tube by just setting up a tube with a plate
supply that can be varied
and a specific curve of Ra gained for the fixed values of grid bias
voltage chosen.

So for tetrodes and pentodes the screen is also kept at a chosen fixed
value
and a set of Ra curves obtained for each Eg2 value.

For triode the screen is connected to the anode and both screen and
anode
voltage is adjusted between say 0V and +500v and the Ra lines you see
in the data sheets are plotted.
To get UL curves, the screen voltage is varied at say 43% of the anode
voltage
by means of using a transformer, and voila, the UL curves for Ra appear,

and btw with Ra lines about equal to the slopes of the usual RL values.
The transformer used for such displays of Ra can include a cathode
winding,
and the Ra lines for CFB with a fixed screen voltage or UL tap can be
plotted.

The main thing is to keep the screen DC potential from drifting during
tests to
simulate the real world conditions of the tube.
But using a mains tranny with a 2:1 ratio would give a low impedance
voltage source of 480vrms here in Oz with our 240v mains, and this is
maybe enough
to give the required anode voltage change.
Taps for screen and cathode windings can be included.
The idea is that you force the tube anode current to change by changing
the anode voltage but with fixed G1 voltages, and a graph can be drawn
for Va vs Ia, or the curve shown on a CRO using the XY mode
used for transfer curve displays.

The time taken for testing a tube like this shouldn't be too long
since the Pa diss or G2 diss might be exceeded, but using some sort of
digital recording and storage of the Va/Ia current changes should
be easy and quick enough before a tube might cook.

I am sure a capable man as yourself will be able to
work all this out.

Patrick Turner.






Tnx2u,

Tom

--

Hi Patrick,

thank you for giving the suggestions how to do that.

Reason I am asking is, that I would feel much better
with my 6550PSE tetrode mode w/ CFB project, when I
would have a sheet with plate curves for CFB op to
draw a load line on...

Try & error is not the right way to do such an ambitious
(for me, that is) project.

I am sure a capable man as yourself will be able to
work all this out.

Thank you for the flowers, but I shamefully must admit
that sometimes I still stumble over most basic issues
of tube tech.

Well, it is a hobby to me, after all.

Tom

--
To err is human - to purr feline.
- R. Byrne

"Tom Schlangen" wrote

Reason I am asking is, that I would feel much better
with my 6550PSE tetrode mode w/ CFB project, when I
would have a sheet with plate curves for CFB op to
draw a load line on...

There's an UL set for 6550A here

http://www.triodeel.com/6550ap5.gif

AFAIK you won't find sheets for distributed load (CFB)...unlike UL
there is no optimum ratio, so no reason to show one rather than any
other.

I made a spice model and then simulated curves for various
connections for my 6CH6 distributed load amp. However, the 6550
doesn't model so well (less evenly-spaced curves and a sharp knee)
so accuracy won't be so good.

For a general idea of how the curves are pulled about, see

http://www.ivesonaudio.pwp.blueyonder.co.uk/modal.gif

cheers, Ian

in message ...

What you really want to know is how to choose the best loading and
operating point, right?

AFAIK, distributing the load makes no difference to the choice.
Having chosen, you can then calculate the difference in drive
voltage requirement and effective input and output impedance of the
stage.

So the optimum distributed load OPT is the same as that for pentode
mode, exept the primary is split into two parts.

Someone is going to argue, I can feel it...

cheers, Ian

Try & error is not the right way to do such an ambitious
(for me, that is) project.

I am sure a capable man as yourself will be able to
work all this out.

Thank you for the flowers, but I shamefully must admit
that sometimes I still stumble over most basic issues
of tube tech.

Well, it is a hobby to me, after all.

Tom

--
To err is human - to purr feline.
- R. Byrne

Tom Schlangen wrote:

Hi Patrick,

thank you for giving the suggestions how to do that.

Reason I am asking is, that I would feel much better
with my 6550PSE tetrode mode w/ CFB project, when I
would have a sheet with plate curves for CFB op to
draw a load line on...

Try & error is not the right way to do such an ambitious
(for me, that is) project.

I used trial and error for the SEUL amps using a 13E1
with UL taps.
I tested the tube with an anode feed choke in a mock set up,
and used various R across it to get the best loads for
triode and tetrode while monitoring thd and
composing graphs of power at clipping
due to G1 current or cut off with various loads.
When I had all that info after a morning's work,
I reckoned on about what load I'd need for UL
and I just wound a tranny with taps, and for an impedance
ratio allowing for triode or UL.
It woked out beautifully; I couldn't have got it any better even if I'd
had all the curves.

Sure the curves are nice to contemplate over morning tea,
but just setting up a tube and trying things does work for me ok
because I have a reasonable idea of what to expect from experience.

If you measure the Ro without GNFB at the speaker output,
and then multiply that by the ZR of the OPT, you will know what the Ra
is for the tube.
So with one 6550 you will find Ra with beam tet is say 18k, but in
triode its say 1.1k,
and in UL, its about 4k, depending onm the tapping point.
Then you can say the slopes of the Ra curves will be at 4k too.

But all that's needed for the UL only tests is a plate choke with taps
for UL screen connection
say at convenint points at say 30%, 40%, 50%, 60% of the winding.


I am sure a capable man as yourself will be able to
work all this out.

Thank you for the flowers, but I shamefully must admit
that sometimes I still stumble over most basic issues
of tube tech.

Testing tubes to get their curves isn't easy.
I don't have a plotter to print out the curves or any software or old
fashioned
tube gear to do what they did in the labs in bygone times.
Tube curves plotted from real tubes are unobtanium, getting them is
rather like tryin
to get Nicole Kidman to come around and strip for me.

So I just set up a tube, and play with it, and finally I get a working
point that
can't be any better, and from the distortion figures and the phase of
the 2H in particular,
I can estimate how the slopes of the Ra lines I know exist change with
values of grid voltage.

Well, it is a hobby to me, after all.

But don't let humility get in the way of fascination with the glassware
in front of you.

They are just things that turn current on and off to allow music, and
finding
out a good way of using them isn't difficult.

You will find for example that if Ea becomes too high, its indicated by
the tube
displaying a tendency for not remaining biased properly; it tends to run
away,
like a 6CM5 in triode with Ea = more than 400V.
But at 375V, they stay biased well for 20 years.
Used at Ea = 200V, the Ra line where Eg1 = 0V restricts the Emin of the
anode swing due to grid current.
So somewhere is a good spot for Ea between 200V and 400V.

Some larger oddball transmitter tetrode tubes I was given are useless;
they need about
3,000V to begin to work at all, and I gave up on them.

And what if a tube has more or less distortion than a set of published
curves may indicate?
I still think one needs to play around with electrode voltages and loads
and examine it carefully
to find out a good working point.

At least I have the thd measurring gear to confirm if I am getting
anywhere,
and I'd be lost without a CRO.
The use of a dual trace CRO with XY function is great for looking at a
tube set
up and finding the best bias point for a load value just by
watching the CRO and swinging the grid bias.
The straighter the xy line the better.

PP amps can be observed the same way but then its a case of
getting the least S in the XY line, and the least wriggles in the middle

for biasing AB amps.
This can be confirmed with thd measurements.

Patrick Turner.




Tom

--
To err is human - to purr feline.
- R. Byrne

Ian Iveson wrote:

"Tom Schlangen" wrote

Reason I am asking is, that I would feel much better
with my 6550PSE tetrode mode w/ CFB project, when I
would have a sheet with plate curves for CFB op to
draw a load line on...

There's an UL set for 6550A here

http://www.triodeel.com/6550ap5.gif

AFAIK you won't find sheets for distributed load (CFB)...unlike UL
there is no optimum ratio, so no reason to show one rather than any
other.

I made a spice model and then simulated curves for various
connections for my 6CH6 distributed load amp. However, the 6550
doesn't model so well (less evenly-spaced curves and a sharp knee)
so accuracy won't be so good.

For a general idea of how the curves are pulled about, see

http://www.ivesonaudio.pwp.blueyonder.co.uk/modal.gif

cheers, Ian

in message ...

What you really want to know is how to choose the best loading and
operating point, right?

AFAIK, distributing the load makes no difference to the choice.
Having chosen, you can then calculate the difference in drive
voltage requirement and effective input and output impedance of the
stage.

So the optimum distributed load OPT is the same as that for pentode
mode, exept the primary is split into two parts.

Someone is going to argue, I can feel it...

You'd be right, they'd argue that for distributed load there is an
optimum
load and it ain't the straight pentode or beam tetrode load, but
somewhere
between pentode and the triode load.

But in fact loads presented by speakers vary between say 3 and 30 ohms
with an average of say 6 ohms
So you have to know if the amp has sufficient current for
3 ohms and sufficient voltage for 30 ohms and that the thd doesn't
become
absurdly huge for the range of loads that the amp will have to face.

For SE, setting up for lowest thd and maximum power for
6 ohms may well be ideal, knowing that thd will rise
for loads either side of the ideal.

So one may find that the anode load of 4k is a good load for an SE 6550,

for say Ea = 380V, and the Pd is about 2/3 of max rating,
and thus the Ia can be worked out and tried.
Then adjusting the RL to 2k or 8k should still give a useful amount
of power at low thd and therefore the 4k would be about right.

PP amps are quite different to load because their max po load is one
where the speaker would have its lowest Z.
So 6 ohms is a good load to design to get say 40 watts,
but you'd know you'd get 60 watts into 3ohms and 35 watts
into 8 ohms and that what matters is the amount of class A, so
one may design for 1/2 the 6ohm power to be in class A, the rest in
class AB,
then the 3 ohm power is mainly all AB, and the 8 ohm po is mainly class
A.

Some PP amps have their max po to be produced at the rated average load
of say 6 ohms, but they are unneccessarily distorted when a load of 3
ohms is used, and many
modern speakers rated for 6 ohms will have part of the lead at 3 ohms.

Patrick Turner.





cheers, Ian

Try & error is not the right way to do such an ambitious
(for me, that is) project.

I am sure a capable man as yourself will be able to
work all this out.

Thank you for the flowers, but I shamefully must admit
that sometimes I still stumble over most basic issues
of tube tech.

Well, it is a hobby to me, after all.

Tom

--
To err is human - to purr feline.
- R. Byrne

oops...

http://www.ivesonaudio.pwp.blueyonder.co.uk/modal.htm

Ian

On Sun, 14 Aug 2005 14:43:17 +0200, Tom Schlangen
wrote:

plate curves for generic triode / tetrode / pentode mode
can easily be measured and drawn using a statical approach.

For a few power output tubes like for example EL34, also
some UL mode operation plate curves for certain plate/screen
tap ratios are given in their spec sheets. Almost no plate
curves are available for CFB mode.

Triodes operate into exactly the same plate loading
independent of the load's distribution between plate-
to-ground and cathode-to-ground. Feedback is just
feedback.

Tetrodes/ pentodes have the added complication that
G2 is referenced to the cathode. Referencing G2 to
ground is effectively tapping G2 "up" the output
transformer exactly like "Ultralinear" does. This
can be useful.


But how were the few available UL op plate curves measured
or derived?

Probably tediously hand-plotted combined with some
combination of wishful thinking and educated curve smoothing,
just like all engineering.

Good fortune,

Chris Hornbeck

Chris Hornbeck wrote:

On Sun, 14 Aug 2005 14:43:17 +0200, Tom Schlangen
wrote:

plate curves for generic triode / tetrode / pentode mode
can easily be measured and drawn using a statical approach.

For a few power output tubes like for example EL34, also
some UL mode operation plate curves for certain plate/screen
tap ratios are given in their spec sheets. Almost no plate
curves are available for CFB mode.

Triodes operate into exactly the same plate loading
independent of the load's distribution between plate-
to-ground and cathode-to-ground. Feedback is just
feedback.

Tetrodes/ pentodes have the added complication that
G2 is referenced to the cathode. Referencing G2 to
ground is effectively tapping G2 "up" the output
transformer exactly like "Ultralinear" does. This
can be useful.

Quad II is an example. Its a very good way to set up any output tube.

But to get the pentode to work as a pentode but with CFB, simply
bypass the screen to the cathode and supply the screen current
via the usual supply resistor.




But how were the few available UL op plate curves measured
or derived?

Probably tediously hand-plotted combined with some
combination of wishful thinking and educated curve smoothing,
just like all engineering.

I would suppose that someone measured the tube and plotted the curves
just as I
said in another earlier post.

There are a few UL curves around; I have seen one for KT88.
And its not just guesswork.

It would be difficult to guess the curves.

Patrick Turner.





Good fortune,

Chris Hornbeck

"Patrick Turner" wrote

... they'd argue that for distributed load there is an
optimum
load and it ain't the straight pentode or beam tetrode load, but
somewhere
between pentode and the triode load.

Rubbish.

I should however clarify my original point, as Chris has pointed out
an ambiguity.

The optimum loading for a pentode with a distributed load is the
same as for one with all its load at the anode. This is true as long
as it remains in pentode mode, ie with constant Vsk. For PP, this
can be achieved by connecting screens to taps as for UL, but
reversing the connections. Haven't tried SE...a cap from screen to
cathode looks awkward to me because it requires either a high
impedance supply to the screen, or a very big cap.

But this proviso actually makes sod all difference. In fact the
optimum load for a pentode isn't much different whatever mode it is
in. The difference between UL and triode optima isn't worth
bothering about.

Conceivably, a sweet spot resulting from fortuitous wriggles in the
curves for a particular valve will shift so that a small change in
load must be used to find the same spot with different feedback to
the screen. But such sweet spots aren't worth much in an output
stage because the load varies and the valves age quickly.

cheers, Ian

Ian Iveson wrote:

"Patrick Turner" wrote

... they'd argue that for distributed load there is an
optimum
load and it ain't the straight pentode or beam tetrode load, but
somewhere
between pentode and the triode load.

Rubbish.

I should however clarify my original point, as Chris has pointed out
an ambiguity.

The optimum loading for a pentode with a distributed load is the
same as for one with all its load at the anode.

That's what I have been maintaining all along.
The load the tube "sees" is the optimum, load experienced by the tube
between
cathode and anode.
But distributed loading with a fixed screen voltage is in effect
like a tube with a UL tap, so that like Quad II if you have 10%
of the primary turns in cathode windings and a fixed G2 voltage supply,
then the tube acts as if it is loaded like pentode/beam but has a 10%
tapping
for UL.

The loading issue is quite a separate issue to the action of the
cathode-G1 feedback.
With KT66/6L6/5881 in SE class A beam tet the load should be about 2.5k.

But triode mode would be perhaps 5k, and UL or distributed RL =
somewhere between.
Ea would be greater for triode, Ia less.
UL op points are somewhere between.

But if you have beam tet with CFB, the G2 is bypassed to the cathode
and the tube is in beam tet mode with CFB and 2.5k is the load
and Ea is the same as for plain beam mode without the CFB.


This is true as long
as it remains in pentode mode, ie with constant Vsk.

OK.

For PP, this
can be achieved by connecting screens to taps as for UL, but
reversing the connections.

Very tricky, and the UL tap % of winding should be
equal to the turns in the cathode winding.
Then caps can be placed from anode winding to cathode winding
to make the coupling more direct....
If the screens cop more voltage than the CFB voltage ther is a state of
positive FB....


Haven't tried SE...a cap from screen to
cathode looks awkward to me because it requires either a high
impedance supply to the screen, or a very big cap.

For ream pentode/beam tet mode with CFB, the g2 can be bypassed with
say 100uF to the cathode and the g2 supplied with a choke from the B+,
so that Eg2 = Ea.
But it is ok to supply a lower Eg2 than Ea in most tubes, so that for an
EL34, 6L6 et all
the Ea might be 350V and Eg2 is 300V, thus allowing Eg1 to be less to
get the wanted idle current.
So therfore a resistor feed to g2 is OK, and although Eg2 sags under
sine wave conditions,
its OK with music signal, and if you don't like that the bypass cap cab
be shunted with
zeners, the R can be reduced a bit so the Eg2 is shunt regged.
The R might be 1k, so that 5mA can flow to g2 from the B+ and you get
the 50V drop.
The cathode windings have a low output impedance and the loading effect
of the 1k at the cathode
is negligible.



But this proviso actually makes sod all difference. In fact the
optimum load for a pentode isn't much different whatever mode it is
in. The difference between UL and triode optima isn't worth
bothering about.

I think it definitely IS worth bothering about.
Who in their right mind would load an EL34 in triode with 2.5k?.

In fact with 20% CFB and a fixed g2 supply a bit below the B+ as stated
above
the EL34 becomes about as linear as you can get it without
enduring excessive g1 drive voltages, and the the FB which exerts itself

between k and g2 acte well to reduce the odd order distortion products
and the combination of the two paths for FB, the k to g1 and k to g2
act together in a unique way to much improve the performance
past ordinary UL or straight pentode.

I built an SECFB amp last year configured just like this using a quad of
parallel EH6CA7.
The owner is as happy as a mudlark.
The Ea is about 360V, Eg2 about 300V, Ia about 55mA, Ig2 is about 5mA.
The max po = 35 watts.
it also has 8dB of global NFB in addition to the approx 8db of local
CFB.
There is less than 0.1% thd at any load between 3 ohms and 12 ohms up to
10 watts.

The load for the 35 watts from the 4 tubes is ideally 1.2k, so RL =
approx 4.8k per tube, way higher than the 2.5k used with Ea = 250V for
pure pentode mode.

I am happy with the efficiency of the circuit, getting about
9 watts per tube max at the anode and Pda = 20 watts, so
max efficiency = 45%.
The tubes are very happy, and should give long life.



Conceivably, a sweet spot resulting from fortuitous wriggles in the
curves for a particular valve will shift so that a small change in
load must be used to find the same spot with different feedback to
the screen. But such sweet spots aren't worth much in an output
stage because the load varies and the valves age quickly.

The wriggles in the transfer you speak of don't much occur in
SE designs at low levels; SE amps shine at low levels because the
thd tends to be mainly plain 2H perhaps with some 3H,
and little else.

Its the "little else" that i don't like to see; the traces of 4H, 5H,
6H, et all
that can be seen in PP amps not set up with enough class A.

The essentially straight low level transfer curve is true certainly of
pentodes or beam tubes
regardelss of the NFB arrangements.
In fact 2H varies in phase sometimes.
At low levels its like triode 2H, with greater -ve going V swings
than +ve swings. 3H is low.
The as RL rises the 2H reduces and sometimes to zero, with increasing
3H.
The as RL rises above where the 2H = 0, the 2H re-appears, but with the
opposite phase.
Meanwhile the 3H may increase.
Its all due to the slopes of the Ra curves not being equal spaced or
parallel
for given Eg1 values.
But as the level increases, many other harmonics both odd and even begin
to appear
as the transfer curve begins to make its asymetrical S shape.
With pure pentode + CFB, the HORRIBLE spectral mix
of SE pentodes or beam tubes is merely reduced by the NFB and because
not much
NFB is involved, a few more higher numbered harmonics are created by
the intermodulation process, so that by 10 watts the signal
is not worth having.
I once had a client bring me an SE amp made in Thailand with Ea = 450V,
RL = 2.5k for 8 ohms, and no CFB, just 12dB of GNFB, and it was only OK
for
1/2 a watt.
Above that it was worse than most transistor amps, which is why the
unhappy dude
had me do what I could with the horror.
thd measurements were 3 times what you'd get with a triode amp without
FB.
Wrong Ea and Ia and RL.
I give the Thai cowboy 1/10.
The subtle effect of the CFB with a fixed G2 voltage tends to remove
odd order harmonics.
The 2H does tend to remain high with plain UL, but it is much reduced
by the CFB, say from 6% like you may get with UL at 10 watts down
to 2% with the CFB.
Thus at the end of the day you have a better outcome with the fixed g2
supply and CFB.
This then maximizes the 2H cancelation which will naturally occur
between the
driver triode and the output stage because the driver will have a
similar
rise in 2H as the output stage right up to clipping.
The majority of thd cancelation can be tailored to occur
most just where one wants it, for loads of about 4 to 6 ohms,
or the difficult sections of "8 ohm" speakers.

I was able to get better thd figures with the SE35 with 5 ohms compared
to a PP amp.
And the driver tube, which is an EL84 in triode is not at all tortured
into producing enough
thd to cancel that in the output stage; it is optimally loaded.
I don't believe creating distortion to get rid of it elsewhere.
The SE35 has 20dB less thd at a few watts into 5 ohms than
when using SEUL without CFB, even though the SEUL
has triode like thd character, and has 3 dB more total NFB.

My client likes the clarity and detail and musicality he says.
To me, tubes sing when thd is minimal without having too much
NFB applied.
Many SET amps with transmitter tubes measure excellently because
of 2H cancellations betwen the driver and output, and thus they sing
well.
Its slightly more difficult to arrange good cancellation in multi grid
tubes.

Patrick Turner.




cheers, Ian

"Patrick Turner" wrote

This is true as long
as it remains in pentode mode, ie with constant Vsk.

OK.

For PP, this
can be achieved by connecting screens to taps as for UL, but
reversing the connections.

Very tricky,

How so?

and the UL tap % of winding should be
equal to the turns in the cathode winding.

Of course.

Then caps can be placed from anode winding to cathode winding
to make the coupling more direct....

Likely an unnecessary complication. An ideal application for bifilar
winding, giving more capacitance and closer coupling.

If the screens cop more voltage than the CFB voltage ther is a
state of
positive FB....

No. Different gain in cathode loop from that of screen. Product of
gain and signal voltgage would have to be greater at the screen for
net positive feedback.

eg,
http://www.ivesonaudio.pwp.blueyonder.co.uk/modal.htm
shows examples of normally coupled and cross-coupled UL screens used
with CFB. Compare the curves for UL30, CF-10; and UL-30, CF-10

Both work OK in practice, although the latter has a greater positive
signal to the screen than the negative one to the cathode. It still
works OK if the screens are connected to opposite anodes. This is
because the gain at the cathode is much greater than at the screen.

Two condiderations militate against adopting cross coupling in my
case. First, the comparitively knee-less 6CH6 is a poor choice for
such coupling because 0V gridline is miles from 0Vak axis so screens
fry at low Vak. Second, the UL30, CF-10 option gives the best
damping for my particular speakers.

Arguably the screen connection should not be called feedback
anyway...only so far as the pentode can be seen as a dual input
device.

cheers, Ian

On Tue, 16 Aug 2005 17:44:44 GMT, "Ian Iveson"
wrote:

Then caps can be placed from anode winding to cathode winding
to make the coupling more direct....

Likely an unnecessary complication. An ideal application for bifilar
winding, giving more capacitance and closer coupling.

Which leads to the Frank McIntosh design of 1949, with screens
tied to the opposite plate. Then the Circlotron, with
floating power supplies replacing bifilar winding.
Tempting for single-ended use...


Arguably the screen connection should not be called feedback
anyway...only so far as the pentode can be seen as a dual input
device.

Thank you! That really, really needed to be said.

Good fortune,

Chris Hornbeck

Ian Iveson wrote:

"Patrick Turner" wrote

This is true as long
as it remains in pentode mode, ie with constant Vsk.

OK.

For PP, this
can be achieved by connecting screens to taps as for UL, but
reversing the connections.

Very tricky,

How so?

I have tried it, and it can lead to HF instability
since the parasitic leakages and stray C do things.......



and the UL tap % of winding should be
equal to the turns in the cathode winding.

Of course.

Then caps can be placed from anode winding to cathode winding
to make the coupling more direct....

Likely an unnecessary complication. An ideal application for bifilar
winding, giving more capacitance and closer coupling.

Nope, the caps shunting windings with equal signals but different DC
potentials is a good technique and used in GE designs
and even in EAR509 etc.
The C entirely shunts the leakage inductance.



If the screens cop more voltage than the CFB voltage ther is a
state of
positive FB....

No. Different gain in cathode loop from that of screen. Product of
gain and signal voltgage would have to be greater at the screen for
net positive feedback.



eg,
http://www.ivesonaudio.pwp.blueyonder.co.uk/modal.htm
shows examples of normally coupled and cross-coupled UL screens used
with CFB. Compare the curves for UL30, CF-10; and UL-30, CF-10

I didn't see a schematic there with instananeous signal voltages shown
at each point in the output circuit.



Both work OK in practice, although the latter has a greater positive
signal to the screen than the negative one to the cathode. It still
works OK if the screens are connected to opposite anodes. This is
because the gain at the cathode is much greater than at the screen.

Neville Thiele must be about 85 now, but a few years ago I phoned him
in Sydney about speaker matters and we had a long discussion on tubes as
well and he asked me
if I'd tried positive FB to the screens from OPT taps and with CFB
windings.
Its all been done before because he was quite familiar with the
technique that the "in crowd"
was using back in the 1950s.

So. You boost the tube gain with positive FB, and increase the Ra and
THD,
but then because you have more gain, there is more gain reduction due to
the aplied CFB
at the same time, so the boosted Ra is reduced and thd reduced, and
exactly what
thd/Ra outcome one gets is down to what CFB% and screen tappings one
uses.

I wouldn't bother with the +ve screen FB idea because I suspect the
spectral outcome is
dirtier than with just a fixed screen supply and CFB only.



Two condiderations militate against adopting cross coupling in my
case. First, the comparitively knee-less 6CH6 is a poor choice for
such coupling because 0V gridline is miles from 0Vak axis so screens
fry at low Vak. Second, the UL30, CF-10 option gives the best
damping for my particular speakers.

Arguably the screen connection should not be called feedback
anyway...only so far as the pentode can be seen as a dual input
device.

Definately the screen connection is a secondary control grid.
Its not a very efficient second input, but its easy to feed it a high
voltage from an anode winding
for feedback, and it is feedback that is being applied.
The feedback can be equal to the anode signal voltage, and then you have
triode,
which is thought to be maximal screen FB, but you could
"extend the windings" beyond the ends of the anode connections to
feedback even more signal voltage than available at the anode,
and this would further reduce Ra nd thd, but probably at
some expense to the power, since triode power is restricted by grid
current
unless you have a low impedance driver.
I've never tried this last idea.

Ordinary UL connection does a lot to improve a pentode
or beam tube's performance. ( lower Ra and lower odd order Dn )
The tendency for screens to fry may possible be avoided with having the
screens at a lower potential.
But with CFB, such schemes get overly complex, and you find yourself
chasing tiny benefits not worth the chase, and better addressed with
some global NFB,
and that's what Peter Walker must have concluded all those years ago.

I used to use UL taps with CFB but settled on 12.5% or P turns in the
cathode
as being near ideal. 25% is also OK but the drive voltage creeps up and
it has Dn which tends to negate the benefits of the
higher CFB in the output stage.
This certainly is the case with KT66,5881,6L6, KT88, 6550, KT90
which need higher drive voltages than you 6CH6, or if you used 6CL6,
6BQ5, or EL86.
The little tubes are very effective when paralleled.

Patrick Turner.





cheers, Ian

Chris Hornbeck wrote:

On Tue, 16 Aug 2005 17:44:44 GMT, "Ian Iveson"
wrote:

Then caps can be placed from anode winding to cathode winding
to make the coupling more direct....

Likely an unnecessary complication. An ideal application for bifilar
winding, giving more capacitance and closer coupling.

Which leads to the Frank McIntosh design of 1949, with screens
tied to the opposite plate. Then the Circlotron, with
floating power supplies replacing bifilar winding.
Tempting for single-ended use...

Arguably the screen connection should not be called feedback
anyway...only so far as the pentode can be seen as a dual input
device.

Thank you! That really, really needed to be said.

The UL screen connection is an application of NFB imho.
It has feathers, it quacks, it likes ponds and waterways, it swins
gracefully
and flies long distances, and accepts bread junk food from humans, it
lays eggs, and I suspect it **is**a duck.

And one could arrange the pentode or beam tube to be screen grid
controlled, but have the
feedback from the anode output circuit fed back to the normal
control grid.

Patrick Turner.




Good fortune,

Chris Hornbeck

On Wed, 17 Aug 2005 02:37:38 GMT, Patrick Turner
wrote:


Arguably the screen connection should not be called feedback
anyway...only so far as the pentode can be seen as a dual input
device.

Thank you! That really, really needed to be said.

The UL screen connection is an application of NFB imho.
It has feathers, it quacks, it likes ponds and waterways, it swins
gracefully
and flies long distances, and accepts bread junk food from humans, it
lays eggs, and I suspect it **is**a duck.

And if I don't like the music and smash the contraption
with a hammer, that's feedback too.

If I smash the duck's head with the hammer, cook and eat it,
is my **** still a duck?

If I die and the duck eats part of me, is...

Both yes and no; 's all i'm sayin',

Except as far as calling G2 tapping "feedback", which is
just Wrong. Probably immoral, and possibly illegal in the
American South.

We still maintain some standards here. A word like
"feedback" still means the differential comparison of
an input signal with a (possibly modified) replica
of the output signal, with a well understood and
time honored tradition curtailing anti-social behaivior
in the output signal.


The deliberate modification of current (real time)
behaviour by reward/punishment positive/negative
reinforcement is what we call an ass-whuppin'.

There's a difference. All just matters of phraseology,

Chris Hornbeck

Hi Patrick,

I built an SECFB amp last year configured just like
this using a quad of parallel EH6CA7.
The owner is as happy as a mudlark.
The Ea is about 360V, Eg2 about 300V, Ia
about 55mA, Ig2 is about 5mA. The max po = 35 watts.
it also has 8dB of global NFB in addition to the
approx 8db of local CFB.

With two fixed bias parallel 6550 at B+ about 405V,
Ea (to cathode) about 400V, Eg2 (to cathode) about
300V, Ia about 80mA, Ra = 3k (thus 6k for each tube)
and 10% CFB I get about 15-18 watts, which is
about 1/2 of your SECFB w/ 4x6CA7 at most similar
voltage condidions and load (4k8 per tube compared
to 6k).

This is poor compared to your amp, which, scaled down
to 2x 6CA7 (and adjusted Ra), would produce about the
same Po with much better efficiency, since I get only
about 30% while you claim about 45% for your design.

Maybe instead of further following the 6550 path, I should
reconsider changing to 2x EL34/6CA7.

What is the driver circuit like for your 6CA7 SECFB?
Those 6CA7/EL34 surely are easier to drive than 6550
I guess.

Tom

--
this is my favourite sig, since there is no reference to
Kibo, Discordianism or The Church of the Subgenius in it.

Tom Schlangen wrote:

What is the driver circuit like for your 6CA7 SECFB?

Please forget this question as you wrote about trioded
EL84 down in your article.

Tom

--
When in doubt, use brute force.
- Ken Thompson

Chris Hornbeck wrote:

On Wed, 17 Aug 2005 02:37:38 GMT, Patrick Turner
wrote:

Arguably the screen connection should not be called feedback
anyway...only so far as the pentode can be seen as a dual input
device.

Thank you! That really, really needed to be said.

The UL screen connection is an application of NFB imho.
It has feathers, it quacks, it likes ponds and waterways, it swins
gracefully
and flies long distances, and accepts bread junk food from humans, it
lays eggs, and I suspect it **is**a duck.

And if I don't like the music and smash the contraption
with a hammer, that's feedback too.

If I smash the duck's head with the hammer, cook and eat it,
is my **** still a duck?

Could be duck **** :-)

Hellll, I'm always ducking **** hurled in my direction.... ;-]

If I die and the duck eats part of me, is...

Well then the duck would then **** some Hornbeck.
Christ, what an end, eaten by a duck.




Both yes and no; 's all i'm sayin',

Except as far as calling G2 tapping "feedback", which is
just Wrong. Probably immoral, and possibly illegal in the
American South.

Ah, so you have a Police Academy where they teach the young dudes
about Audiologically Correct Procedures.
Rule No 69, anyone calling screen tappings feedback should not
duck the long arm of the LAW.


We still maintain some standards here. A word like
"feedback" still means the differential comparison of
an input signal with a (possibly modified) replica
of the output signal, with a well understood and
time honored tradition curtailing anti-social behaivior
in the output signal.

Well now yo gettin all big wordy there bro and sayin stuff about
dis differ rentation **** then one could sart to think that there is a
bitta
diffarantation stuff goin on in a UL circuit becaiuse when the screen
volts go
down, the grid has to come up even more than ever to get the same
plate volt going down.
Its a sort of imperfect differpants thinge.
Not maybe a duck, but perhaps a platypus......



The deliberate modification of current (real time)
behaviour by reward/punishment positive/negative
reinforcement is what we call an ass-whuppin'.

Ah, carrot and stick approach to learnin.



There's a difference. All just matters of phraseology,

Tha ya go again with a diffa-erentishus thingeme.
Can't get away from it.

Patrick Turner.



Chris Hornbeck

Tom Schlangen wrote:

Hi Patrick,

I built an SECFB amp last year configured just like
this using a quad of parallel EH6CA7.
The owner is as happy as a mudlark.
The Ea is about 360V, Eg2 about 300V, Ia
about 55mA, Ig2 is about 5mA. The max po = 35 watts.
it also has 8dB of global NFB in addition to the
approx 8db of local CFB.

With two fixed bias parallel 6550 at B+ about 405V,
Ea (to cathode) about 400V, Eg2 (to cathode) about
300V, Ia about 80mA, Ra = 3k (thus 6k for each tube)
and 10% CFB I get about 15-18 watts, which is
about 1/2 of your SECFB w/ 4x6CA7 at most similar
voltage condidions and load (4k8 per tube compared
to 6k).

This is poor compared to your amp, which, scaled down
to 2x 6CA7 (and adjusted Ra), would produce about the
same Po with much better efficiency, since I get only
about 30% while you claim about 45% for your design.

Maybe instead of further following the 6550 path, I should
reconsider changing to 2x EL34/6CA7.

I found in preliminary tests i could get 35W from 3 x 6550
with similar Ea/Ia conditions and % of CFB and RL,
but I had to dissipate more power in the 6550,
and they are capable of that easily.



What is the driver circuit like for your 6CA7 SECFB?
Those 6CA7/EL34 surely are easier to drive than 6550
I guess.

6550/KT88/KT90 are only marginally more difficult to drive, ie, they
don't need much more drive voltage.
I recal I have 17% CFB, and the max G1 input voltage is 46vrms to get 35
watts to
the rated load of 5 ohms where SE power is maximum and efficiency is
highest.

You can plot your power output vs load value at 3% thd ( discernable on
the CRO ) as you vary
loads, 24,16,12,8,6,5,4,3,2 ohms and it should look like an arch.
At the top of the arch, that is max power, and you should be able to
read off the RL.
Only at this peak will efficiency be at its maximum.
at all other points, efficiency is less.
I always like to have the max power ability of an SE amp to be into a
load value slightly
lower than the rated load, so for 8 ohm speakers,
max power is still at 5 ohms.
It means there's just a little better current ability for the Z dips
below
the rated speaker value.
And the amp will still turn out a reasonable effort into 4 ohm speakers,

as long as head banger levels are not used.

Patrick Turner.



Tom

--
this is my favourite sig, since there is no reference to
Kibo, Discordianism or The Church of the Subgenius in it.

Tom Schlangen wrote:

Tom Schlangen wrote:

What is the driver circuit like for your 6CA7 SECFB?

Please forget this question as you wrote about trioded
EL84 down in your article.

I did start off using a paralleled 12BH7.

It worked fine but then I went to using EL84 in triode,
with much more Ia, and the sound got better.
The thd stayed about the same.

Some things happen for which I have no explanation,
and I don't care why, its just nice when nice things happen.

Patrick Turner.



Tom

--
When in doubt, use brute force.
- Ken Thompson

"Patrick Turner" wrote

The UL screen connection is an application of NFB imho.
It has feathers, it quacks, it likes ponds and waterways, it swins
gracefully
and flies long distances, and accepts bread junk food from humans,
it
lays eggs, and I suspect it **is**a duck....

At which point hopefully all but one of your seasick sailors are
going home.

It doesn't quack like a duck at all.

The clear practical reason for adopting the conventional engineering
definition of feedback is illustrated throughout the thread.

The distinction between this conventional definition and the loose
general use of the word in common language is actually all the
original poster needs to know to answer his question.

He asked if there were data sheets depicting the transfer function
of a 6550 with cathode feedback.

The answer is that he should forget about the feedback and use the
appropriate set of curves for whatever mode the valves are operated
in. Then use the usual arithmetic to calculate how the feedback
modifies the input and output, and the usual analysis to check for
stability margin etc.

The complication arises because "cathode feedback" is ambiguous as
it says nothing about what is happening to the screen. Since I am
trying to introduce some discipline here, I will try to use the
principle of superposition properly.

The return of the output signal to the cathode has three components:

a) a return to Vgk.
b) a return to Vsk
c) a return to Vak

Of these, c) can be ignored for most of the operating range of a
pentode or tetrode; a) is conventional feedback and can be ignored
for the purpose of choosing which curves to use and the best
loadline; b) determines the mode the valve is operating in, and
hence which set of curves to use.

This is how engineering makes things plain and simple. Stick to the
conventions and you can use the conventional data. All the books,
all the data, all the methods used in design, fit within the
convention of control engineering. Stray and you will stay lost
forever.

Confusion between a) and b) is what prompted the original question,
and Chris' timely intervention.

Having said all that, it is quite *possible* to frame the circuit
analysis in such a way that the return to screen can be contrived as
conventional feedback. But then you would need datasheets contrived
for that purpose, in this case sets for screen and grid control, and
do heaps of graphical analysis to combine the two.

cheers, Ian

Ian Iveson wrote:

"Patrick Turner" wrote

The UL screen connection is an application of NFB imho.
It has feathers, it quacks, it likes ponds and waterways, it swins
gracefully
and flies long distances, and accepts bread junk food from humans,
it
lays eggs, and I suspect it **is**a duck....

At which point hopefully all but one of your seasick sailors are
going home.

It doesn't quack like a duck at all.

The clear practical reason for adopting the conventional engineering
definition of feedback is illustrated throughout the thread.

Oh is it really?

If it was, it'd lead to the idea that there is NFB inside triodes,
something
you are trying to avoid coming to.



The distinction between this conventional definition and the loose
general use of the word in common language is actually all the
original poster needs to know to answer his question.

He asked if there were data sheets depicting the transfer function
of a 6550 with cathode feedback.

The answer is that he should forget about the feedback and use the
appropriate set of curves for whatever mode the valves are operated
in. Then use the usual arithmetic to calculate how the feedback
modifies the input and output, and the usual analysis to check for
stability margin etc.

But the curves fo CFB become like triode curves which are pentode
curves when FB is applied......



The complication arises because "cathode feedback" is ambiguous as
it says nothing about what is happening to the screen. Since I am
trying to introduce some discipline here, I will try to use the
principle of superposition properly.

The return of the output signal to the cathode has three components:

a) a return to Vgk.
b) a return to Vsk
c) a return to Vak

Of these, c) can be ignored for most of the operating range of a
pentode or tetrode; a) is conventional feedback and can be ignored
for the purpose of choosing which curves to use and the best
loadline; b) determines the mode the valve is operating in, and
hence which set of curves to use.

This is how engineering makes things plain and simple. Stick to the
conventions and you can use the conventional data. All the books,
all the data, all the methods used in design, fit within the
convention of control engineering. Stray and you will stay lost
forever.

Well there are formulas which describe all the combined effects of the
various overlapping intertwined loops of the FB behave.
They were all worked out by brains better tna ours in around 1955.
Now they are forgotten, or nobody here is able to
express the effective Ra for a given pentode with a given amount of
feedback occuring in the 3 directions at once.

So if nobody can express it mathematically, then bloody measure the
*******.
If it can be measured, and if the Ra with all this FB stuff can be
displayed
on a CRO, it can be plotted, and since so many combinations are involved

it will keep a plotter going for months' worth of sundays, and keep him
out of church
and in contact with something real that the God of triodes has provided
us.




Confusion between a) and b) is what prompted the original question,
and Chris' timely intervention.

Having said all that, it is quite *possible* to frame the circuit
analysis in such a way that the return to screen can be contrived as
conventional feedback. But then you would need datasheets contrived
for that purpose, in this case sets for screen and grid control, and
do heaps of graphical analysis to combine the two.

Well yes, so just measure, plot, and then graphically the load can be
applied.
When that's done, test the circuit for thd and power etc,
and confirm what's best.

Listening with Motzartian test signals will also offer confirmation
of the correctness of the process.

But nobody achieves much in an armchair.

Patrick Turner.




cheers, Ian

On Wed, 17 Aug 2005 08:30:52 GMT, Patrick Turner
wrote:

The deliberate modification of current (real time)
behaviour by reward/punishment positive/negative
reinforcement is what we call an ass-whuppin'.

Ah, carrot and stick approach to learnin.

And, more specifically, the term "ass-whuppin" is
locally applied in the specific case of an adult
male to another adult male in the behaviour
modification at issue.

Maybe we need a whole new thread on the colloquial
terms for giving and getting one's ass kicked?


There's a difference. All just matters of phraseology,

Tha ya go again with a diffa-erentishus thingeme.
Can't get away from it.

Sorry; it's an Americanism, from _The Music Man_,
"Watch your phraseology!", from my childhood.

Only funny in context; sorry.
Thanks!

Chris Hornbeck