Hi folks,
Although having used CFB in a few projects with great success, I'm a bit shy on the exact theory how a tube, for example pentode or tetrode stage operates when the cathode is attached to a separate cathode feedback winding. Now at first sight, two things can happen:
1. The feedback windings feeds a inverted portion into the cathode at low impedance, hence trough negative feedback it reduces the output impedance of the stage.
Or
2. The cathode itself swings a portion of voltage of low impedance through the cathode feedback winding, contributing to the output swing, power and low impedance of the stage.
Or
3. Both?
Bonus question. I understand the cathode winding is part of the effective primary turns, hence it is desired to respect the same rules considering low leakage inductance related to the secondary layers. But how do the cathode winding to primary winding interract each other? Is a low leakage inductance between the CFB winding and primary windings desired?
All the best,
Alexander.
Although having used CFB in a few projects with great success, I'm a bit shy on the exact theory how a tube, for example pentode or tetrode stage operates when the cathode is attached to a separate cathode feedback winding. Now at first sight, two things can happen:
1. The feedback windings feeds a inverted portion into the cathode at low impedance, hence trough negative feedback it reduces the output impedance of the stage.
Or
2. The cathode itself swings a portion of voltage of low impedance through the cathode feedback winding, contributing to the output swing, power and low impedance of the stage.
Or
3. Both?
Bonus question. I understand the cathode winding is part of the effective primary turns, hence it is desired to respect the same rules considering low leakage inductance related to the secondary layers. But how do the cathode winding to primary winding interract each other? Is a low leakage inductance between the CFB winding and primary windings desired?
All the best,
Alexander.
I'm now use only 33% UL+12% CFB in my KT88/6550 amps. You definitely will need higher voltage swing to drive output tubes (check it with simulation in LTSpice). You can take CFB winding diagram from Patrick Turner design for example, or Sansui AU111 / Quad II transformers. Built several variants, with both linear and toroidal coils, never had problem with oscillation.
Thanks LinuksGuru, however I was looking more for a theoretical explanation of the mechanics within a typical CFB stage. Someone hinted on papers of N. Crowhurst on this, but I'm not sure for which types to look for.
Last edited:
If typical is a low turn winding like the speaker secondary , than the cathode feedback voltage can be considered as induced from plate primary , but when you move up to equal number of turns like in unity coupled McIntosh is more complicated , the cathode is a generator by itself and the Raa is reduced by 4 times .
Leakage inductance should be small in all cases
Leakage inductance should be small in all cases
Thanks Depanatoru,
In a summary, can we assume that the CFB effect mechanics vary with the % amount of it related to the primary?
It makes sense that a low ratio CFB winding makes the cathode see a low-impedance winding, hence is more driven by the plate primary.
However at a high CFB turn ratio, the cathode looks at a higher impedance windings and becomes more of a generator, instead of being fed inverted signal by a low-CFB ratio, low-impedance winding?
Can we conclude that a higher impedance cathode winding transacts from less CFB effect to shared load instead, like the unity coupled McIntosh concept?
In a summary, can we assume that the CFB effect mechanics vary with the % amount of it related to the primary?
It makes sense that a low ratio CFB winding makes the cathode see a low-impedance winding, hence is more driven by the plate primary.
However at a high CFB turn ratio, the cathode looks at a higher impedance windings and becomes more of a generator, instead of being fed inverted signal by a low-CFB ratio, low-impedance winding?
Can we conclude that a higher impedance cathode winding transacts from less CFB effect to shared load instead, like the unity coupled McIntosh concept?
The unity coupled has the most feedback , voltage gain is 1 as the name would suggest .
So low turn cathode winding = no or very weak generator and low feedback also.
So low turn cathode winding = no or very weak generator and low feedback also.
A generalization:
In 99% of all output transformers, it is desireable to have low leakage inductance between all of the windings.
Except for the amount of screen current (pentodes, beam power tubes, and True Tetrodes), plate current = cathode current.
This is a Benchmark, remember it.
The only exception to this is if you draw control grid, g1 current. And I do not listen to my amplifiers at levels that have g1 grid current.
Whenever there is a winding for the plate, and a separate winding for the cathode, the power contribution of the plate versus the power contribution of the cathode is as follows:
Peak Power:
Plate current squared x plate winding impedance
Cathode current squared x cathode winding impedance
That means that the ratio of power delivered from the plate versus the power delivered from the cathode, is simply the ratio of the impedances.
McIntosh Unity Coupled, Zplate winding = Zcathode winding, the plate signal power = the cathode signal power.
10 Watts plate signal power, 10 Watts cathode signal power = 20 Watts to the load.
An output stage that has 3200 plate winding and 16 Ohm output tap that connects to the cathode . . .
3200 / 16 = 200:1
Plate signal power 2 Watts, cathode power 0.01 Watts (10 milliwatts).
Just about everything else pretty much is in-between those limits.
In 99% of all output transformers, it is desireable to have low leakage inductance between all of the windings.
Except for the amount of screen current (pentodes, beam power tubes, and True Tetrodes), plate current = cathode current.
This is a Benchmark, remember it.
The only exception to this is if you draw control grid, g1 current. And I do not listen to my amplifiers at levels that have g1 grid current.
Whenever there is a winding for the plate, and a separate winding for the cathode, the power contribution of the plate versus the power contribution of the cathode is as follows:
Peak Power:
Plate current squared x plate winding impedance
Cathode current squared x cathode winding impedance
That means that the ratio of power delivered from the plate versus the power delivered from the cathode, is simply the ratio of the impedances.
McIntosh Unity Coupled, Zplate winding = Zcathode winding, the plate signal power = the cathode signal power.
10 Watts plate signal power, 10 Watts cathode signal power = 20 Watts to the load.
An output stage that has 3200 plate winding and 16 Ohm output tap that connects to the cathode . . .
3200 / 16 = 200:1
Plate signal power 2 Watts, cathode power 0.01 Watts (10 milliwatts).
Just about everything else pretty much is in-between those limits.
If g2 is ac coupled to the cathode (cap from g2 to k) the tetrode/pentode works as tetrode/pentod, CFB provides fb in Uin/Ug1relation.
If g2 is ac coupled to ground you get, as above,the Uin/Ug1 related fb, but additionally the tetrode/pentode gets Ug2 fb, the tube is essentially also
connected for an intermediate between terode/pentode, the so called "ultralinear" operation.
CFB also reduces the effective capacitances of the transformer, one of the big advantages of a McIntosh style OPT often overlooked.
Also, in case you wound a transformer with too much interleaving, CFB could be all you need to prevent it from going to the garbage bin 😉
If g2 is ac coupled to ground you get, as above,the Uin/Ug1 related fb, but additionally the tetrode/pentode gets Ug2 fb, the tube is essentially also
connected for an intermediate between terode/pentode, the so called "ultralinear" operation.
CFB also reduces the effective capacitances of the transformer, one of the big advantages of a McIntosh style OPT often overlooked.
Also, in case you wound a transformer with too much interleaving, CFB could be all you need to prevent it from going to the garbage bin 😉
Last edited:
This is the way to look at it. UL is negative voltage feedback from anode to g2, while CFB is negative voltage feedback from anode to g1 (applied through cathode).
By applying cathode feedback, you are also varying the UL feeeback, because the tube sees a different voltage on the cathode (that is its voltage reference for g2 as well).
You could also connect g2 as pentode and apply only CFB.
Just to sort out the confusio there often is, the unity coupled stage has ä voltage gain of ideally 1 ( but usually quite a bit less), true.
But the same tube is also suppying the anode curcuiit, so from this poin the voltage gain has to double up . After all, it is not a cathode follower, its split loaded and should be looked upon as being a voltage "generator" at the cathode side, and a "current generator" at the anode side. This very fortune in pp, because one sides voltage source paralelled with the current source from the other side works very well togetherl.
3. Yes
Bonus q: Because the cathode winding is part of the primary winding, you've answered your own question. Leakage inductance, as we define it, is a measure of coupling to the secondary. However, your real question seems to be: how should I, designing a transformer, weight the relative importance of the cathode windings' coupling to the rest of the primary vs. (because everything is a tradeoff) its coupling to the secondary?
It's a very subtle question, and might require more knowledge of intended use. In a non-loop-feedback amplifier all you really need is a fairly well controlled response at and a little above audible, but an amplifier with feedback around the OPT needs a simple response, one that's easy for the loop to deal with over a good margin of possible loads.
Don't modern OPT designs tend towards higher shunt capacitance and lower leakage L anyway? This would lean the coupling importance towards the primary side. But high feedback amplifier designs might benefit from leaning towards prioritizing the coupling to the secondary.
Very interesting subject, thanks, and all good fortune,
Chris
A thought about the importance of paying attention to leakage inductance.
(Yes, often it is defined as from primary to the secondary; but that is not always the only important leakage inductance).
Just for discussion and analysis, make a special single ended Ultra Linear output transformer.
For flexibility, the Ultra Linear windings are separate from the plate windings. That way, we can use two different B+ voltages, one for the screen, and one for the plate.
I have seen Ultra Linear circuits that use these separate plate and screen windings.
1. Single ended Ultra Linear:
So we wind the transformer this way:
Laminations
Screen: with B+ #2 at the laminations, and screen at the end of the winding
Secondaries: Common next to screen winding, 4, 8, then 16 Ohm at the end of the winding.
Plate: B+ #1 next to the 16 Ohm secondary, and the plate at the end of the winding.
You will notice that there may be a lot of leakage inductance from the screen winding to the plate winding.
What might this do? The leakage inductance causes a phase difference between the plate signal, and the negative feedback signal to the screen.
We can change the order of the windings to get lower leakage reactance:
Screen
Plate
Secondaries
That should work OK, because the screen only provides a very small amount of power to the secondary; almost all of the power comes from the plate.
(Yes, often it is defined as from primary to the secondary; but that is not always the only important leakage inductance).
Just for discussion and analysis, make a special single ended Ultra Linear output transformer.
For flexibility, the Ultra Linear windings are separate from the plate windings. That way, we can use two different B+ voltages, one for the screen, and one for the plate.
I have seen Ultra Linear circuits that use these separate plate and screen windings.
1. Single ended Ultra Linear:
So we wind the transformer this way:
Laminations
Screen: with B+ #2 at the laminations, and screen at the end of the winding
Secondaries: Common next to screen winding, 4, 8, then 16 Ohm at the end of the winding.
Plate: B+ #1 next to the 16 Ohm secondary, and the plate at the end of the winding.
You will notice that there may be a lot of leakage inductance from the screen winding to the plate winding.
What might this do? The leakage inductance causes a phase difference between the plate signal, and the negative feedback signal to the screen.
We can change the order of the windings to get lower leakage reactance:
Screen
Plate
Secondaries
That should work OK, because the screen only provides a very small amount of power to the secondary; almost all of the power comes from the plate.
Last edited:
This was my take on the subject:
https://www.diyaudio.com/community/threads/virtual-otl-se-amplifier-idea.386963/post-7041399
It didn't generate a great deal of enthusiasm....
https://www.diyaudio.com/community/threads/virtual-otl-se-amplifier-idea.386963/post-7041399
It didn't generate a great deal of enthusiasm....
Thanks for answering, Chris,
In my work, I usually keep the CFB layers close to the secondary layers. The best winding low ratio CFB configuration for best HF performance IMHO is connecting all CFB layers in parallel, and using finer wire if necessary to achieve a higher CFB to primary Z winding ratio. Connecting them in parallel permits them to act as capacitance shunts, that way you gain free lunch screening effect between the primary and secondary, which allows to connect the secondaries in series without detriments to frequency response resonances.
For example, in a configuration such as P3-S-P6-S-P3, that would translate to:
P3-CFB-S-CFB-P6-CFB-S-CFB-P3
In a situation where secondary layers are always connected in parallel, such as S-P6-SS-P6-S, considering the internal SS layers are the ones, and the outer others are to be series connected, then one could make it S-CFB-P6-SS-P6-CFB-S, at the expense of more leakage related to the CFB layers.
All the best,
Alexander
Preface: I know nothing about winding transformers. Saw it done once, at McIntosh in Birmingham about half a century ago, but the ladies (*) doing it were focused and I was too dumb to ask useful questions anyway.
Wondering if this couldn't be modeled pretty well by extrapolating from the classic model of an idealized "magical perfect inner transformer" with parasitic reactances, keeping the conventional main primary to secondary parasitics, but adding CFB windings to secondary parasitics and CFB to main primary parasitics. That's how we would measure the finished product in the real world, but is it possible in whatever software you young whippersnappers use?
And, you kids get off my lawn!
Chris
(*) At the time, the managers (guys, of course) told me that they often tried to hire men to work hand assembly, but they couldn't take it - only lasted a few days. The main assembly room that day was all women, no exceptions. I guess I shouldn't have been surprised, but was and still remember it.
Wondering if this couldn't be modeled pretty well by extrapolating from the classic model of an idealized "magical perfect inner transformer" with parasitic reactances, keeping the conventional main primary to secondary parasitics, but adding CFB windings to secondary parasitics and CFB to main primary parasitics. That's how we would measure the finished product in the real world, but is it possible in whatever software you young whippersnappers use?
And, you kids get off my lawn!
Chris
(*) At the time, the managers (guys, of course) told me that they often tried to hire men to work hand assembly, but they couldn't take it - only lasted a few days. The main assembly room that day was all women, no exceptions. I guess I shouldn't have been surprised, but was and still remember it.
I use a simplified winding arrangement in my CFB PP amplifier. The cathodes of the output tubes are directly connected to the secondary of the output transformer, but cross-connected. This realizes some <100% negative feedback, meaning the g1 voltage should be nearly double. My primary goal was to reduce distortion of the output tubes+output transfomer complex. Like any other NFB, it reduces the output impedance too.