I did this in my Unity-Coupled amp. I just bought four little Hammond filament transformers, one for each output tube. I direct-connected the center tap to the cathode, but now that you mention a resistor, seems like there would be a little less capacitive loading on the cathode that way. Thanks for the tip.
Bandersnatch, SpreadSpectrum,
being a Mechanical Engineer and not an EE makes me sometimes guess some reasons behing others' choices: one independent PT for each heater because sharing the same core would make the variation in reference voltage of each heater winding to influence ("modulate") other windings ac output?
Thanks in advance
Roberto
being a Mechanical Engineer and not an EE makes me sometimes guess some reasons behing others' choices: one independent PT for each heater because sharing the same core would make the variation in reference voltage of each heater winding to influence ("modulate") other windings ac output?
Thanks in advance
Roberto
I think it is because there is a maximum possible voltage between heater and cathode. Normally the cathode is close to gnd potential and stays there, but in the UNSET it is being driven so its potential towards a grounded heater can be quite high.
For the GU50 the max cathode to heater voltage is 200V. Quite good I think?
http://www.ok1rr.com/tubes/gu-50.pdf
For the GU50 the max cathode to heater voltage is 200V. Quite good I think?
http://www.ok1rr.com/tubes/gu-50.pdf
Yes the Vh-k limit is clear to me, my question is: can a single custom PT with all heaters windings be ok, or all the variable reference voltages for the different heater windings, shared under a single core, will "modulate" the AC of other windings?
I would say it won't affect others' windings, but both Bandersnatch and SpreadSpectrum underlined the fact of using single transformers, so I don't know if it's a matter of simplicity using standard commercial PT or a technical choice.
I would say it won't affect others' windings, but both Bandersnatch and SpreadSpectrum underlined the fact of using single transformers, so I don't know if it's a matter of simplicity using standard commercial PT or a technical choice.
It is easier to get a multi-secondary heater tx than trying to protect the h-k voltage...I had not thought to get one TX with multiple secondaries; I just jumped right in with separate ones...LOL Either way, a good bit of resistance between heater winding CT and the cathode should do for either choice. I do like the idea of reducing the coupling by using multiple two-bay heater tx's...but I suspect it matters little.
cheers,
Douglas
cheers,
Douglas
Thanks Douglas, I'm looking for a customized PT (both for this SE and for the PP), so it would be easier for me to have multiple secondaries with the 20k resistor you suggested from each cathode to the central tap of each secondary (plus an elevated heater for SE's driver and PP's phase splitter).
In some applications the parasitic coupling (mostly capacitance) between individual windings on the same transformer could allow signal to transfer from one winding to another. If the cathodes were operating into a high impedance (seen in some phase inverters and followers especially in guitar amps) this could degrade channel separation in a stereo amp.
In the case of the UNSET the cathodes are being driven by a very low impedance (a few ohms) source so any transfer would be inconsequential. In my case I am not using separate windings, or separate transformers, I am using one SMPS to power all the output tube heaters. All output tube heaters are wired directly in parallel which would be worse case for coupling. There is no loss of channel separation out to at least 50KHz.
The capacitance between H and K of a TV sweep tube is not often specified, but it is on the order of 30 pF drive that with a few ohms and the coupling in the audio range is minimal.
The main consideration here would be H-K breakdown in a push pull or stereo amp with a high feedback ratio where the drive voltage could approach H-K breakdown due to a different drive signal in each cathode with heaters wired in parallel. In this case the real SOA ratings of the mosfet might be the limiting factor.
In the case of the UNSET the cathodes are being driven by a very low impedance (a few ohms) source so any transfer would be inconsequential. In my case I am not using separate windings, or separate transformers, I am using one SMPS to power all the output tube heaters. All output tube heaters are wired directly in parallel which would be worse case for coupling. There is no loss of channel separation out to at least 50KHz.
The capacitance between H and K of a TV sweep tube is not often specified, but it is on the order of 30 pF drive that with a few ohms and the coupling in the audio range is minimal.
The main consideration here would be H-K breakdown in a push pull or stereo amp with a high feedback ratio where the drive voltage could approach H-K breakdown due to a different drive signal in each cathode with heaters wired in parallel. In this case the real SOA ratings of the mosfet might be the limiting factor.
Yes, they are very tough and cheap tubes, I've just receipt 14 non matched ones plus their special sockets/cages at 4 euros each tube+socket/cage.
Another good point about them is that 10% a-g1 feedback seems to give triode-like curves, whilst KT88 (Vh-k max 200V) and EL34 (Vh-k max 100V, Rh-k max 20k) seem to need at least 20%, and below that is UL-like curves like 6A3sUMMER noticed in his post.
Again I'm guessing the impedance variation across the frequency spectrum. Zobel filters will flatten the impedance at the higher frequencies, and some specifically designed crossovers can flatten the frequencies below (as planet10 was explanining few days ago in another thread). So a monophonic sound won't show any elliptical loadline, even with the worst impedance plot ever. For polyphonic sounds, a stable impedance will show a narrow ellipse.
So a triode, with a wide area of parallel tube curves around (on top&bottom of) the ideal linear loadline, will suffer less from a wider elliptic loadline.
UL with its wider knee will help a bit but will show convex linear and concave curves.
Pentode is the worst of the three having a very narrow knee not very friendly with elliptic loadlines.
So we need to set the a-g1 feedback and loadline in order to have enough "area" of parallel curves around (again on top&bottom of) the ideal linear loadline.
Am I right?
Last edited:
In push pull the curves of the output tubes are rotated from each other by 180 degrees.
Plot that, and then set a load line on that pair of curves.
Then center an ellipse around that load line.
Operate only in the Class A region (because next, below, you have to operate in the Class A region for the single ended amp).
In single ended there is only one tube curve.
Plot that, and then set the same load line on it.
Then center the same ellipse around the load line.
Now, trace the path of the ellipse on the push pull, and trace the path of the ellipse on the single ended amplifiers.
You will find a more linear elliptical path as it crosses the plate curves on the push pull amp, versus on the single ended amp.
Or, if you like to do simulations, and graphics, just do both a push pull, and a single ended output stage.
Use a simple RC load to get the load line to be elliptical.
Example, use a 1kHz sine wave, and a load of 8 Ohms in series with 20uF, on the 8 Ohm tap of both the push pull amp, and the single ended amp.
Hint: with both the push pull and single ended amps in Class A, use a plate to plate primary for the push pull that is 2 times the plate primary for the single ended amp.
For example, 6k p-p for push pull, and 3k for single ended. That way, all 3 output tubes will drive 3k (it takes a little understanding of the circuits to believe that, but it is true).
I believe you will be able to see it visually.
And, use the simulation to calculate the first 5 harmonics.
Plot that, and then set a load line on that pair of curves.
Then center an ellipse around that load line.
Operate only in the Class A region (because next, below, you have to operate in the Class A region for the single ended amp).
In single ended there is only one tube curve.
Plot that, and then set the same load line on it.
Then center the same ellipse around the load line.
Now, trace the path of the ellipse on the push pull, and trace the path of the ellipse on the single ended amplifiers.
You will find a more linear elliptical path as it crosses the plate curves on the push pull amp, versus on the single ended amp.
Or, if you like to do simulations, and graphics, just do both a push pull, and a single ended output stage.
Use a simple RC load to get the load line to be elliptical.
Example, use a 1kHz sine wave, and a load of 8 Ohms in series with 20uF, on the 8 Ohm tap of both the push pull amp, and the single ended amp.
Hint: with both the push pull and single ended amps in Class A, use a plate to plate primary for the push pull that is 2 times the plate primary for the single ended amp.
For example, 6k p-p for push pull, and 3k for single ended. That way, all 3 output tubes will drive 3k (it takes a little understanding of the circuits to believe that, but it is true).
I believe you will be able to see it visually.
And, use the simulation to calculate the first 5 harmonics.
Thanks 6A3sUMMER,
So will in SE a lower voltage and steeper loadline (lower Ra) perform better than an higher volytage and Ra for this specific aspect?
So is there a general rule on the preferred zone of the plate max dissipation hyperbole to place the working point in SE? It will for sure change between triode connected (plate swing is limited so working point is shifted towards high voltage area), UL and pentode connected (plate can easily swing down to 50V and you need to "hit the knee" with the loadline), and this CED/UNSET topology (the working point can be above the max plate dissipation curve, because the pmosfet on bottom will contribute to the total dissipation of the tube+pmosfet system, and plate will swing down to 50V as pentodes do).
So will in SE a lower voltage and steeper loadline (lower Ra) perform better than an higher volytage and Ra for this specific aspect?
So is there a general rule on the preferred zone of the plate max dissipation hyperbole to place the working point in SE? It will for sure change between triode connected (plate swing is limited so working point is shifted towards high voltage area), UL and pentode connected (plate can easily swing down to 50V and you need to "hit the knee" with the loadline), and this CED/UNSET topology (the working point can be above the max plate dissipation curve, because the pmosfet on bottom will contribute to the total dissipation of the tube+pmosfet system, and plate will swing down to 50V as pentodes do).
Last edited:
I do not think you can apply a single general rule.
Dealing with a varying impedance, and reactive loads, requires a consideration of all aspects of design.
The goals of the designer should determine what operating conditions to apply for a given output tube.
And the selection of the tube type, Triode, Pentode, Beam Power (Oh, and do not forget True Tetrodes).
The mode of the Pentode and Beam Power tubes, Pentode/Beam Power; Ultra Linear; Triode wired mode.
Push Pull.
Single Ended.
Local negative feedback
Global negative feedback
No negative feedback
Multiple feedback loops
Plate impedance, rp, verus plate load, RL.
Ratio of Plate voltage to Screen Voltage
Maximum Screen voltage and maximum Screen dissipation versus Pentode/Beam Power, Ultra Linear, and Triode wired modes.
The design has to make decisions, based on a number of tradeoffs, just a few the many are listed here:
Power output
Harmonic distortion
Order of distortion (dominant 2nd, dominant 3rd)
Intermodulation distortion
Damping Factor
Symmetry of damping factor
Required driver peak to peak volts out
Required driver output impedance
Fixed adjustable bias
Self Bias
Tube Life
Cost of tubes
Cost of other parts
Size
Weight
It is possible to meet the goals of the designer, without being tied down to a single combination selection of the above factors.
Multiple combinations may accomplish the same goals.
Just be sure to pick a combination that works well.
We do not want another "Tacoma Narrows Bridge" event.
I will not comment on the CED/UNSET, I leave that up to the experienced experts.
I hope that gives some food for thought, even though it is not a single or easy answer.
Dealing with a varying impedance, and reactive loads, requires a consideration of all aspects of design.
The goals of the designer should determine what operating conditions to apply for a given output tube.
And the selection of the tube type, Triode, Pentode, Beam Power (Oh, and do not forget True Tetrodes).
The mode of the Pentode and Beam Power tubes, Pentode/Beam Power; Ultra Linear; Triode wired mode.
Push Pull.
Single Ended.
Local negative feedback
Global negative feedback
No negative feedback
Multiple feedback loops
Plate impedance, rp, verus plate load, RL.
Ratio of Plate voltage to Screen Voltage
Maximum Screen voltage and maximum Screen dissipation versus Pentode/Beam Power, Ultra Linear, and Triode wired modes.
The design has to make decisions, based on a number of tradeoffs, just a few the many are listed here:
Power output
Harmonic distortion
Order of distortion (dominant 2nd, dominant 3rd)
Intermodulation distortion
Damping Factor
Symmetry of damping factor
Required driver peak to peak volts out
Required driver output impedance
Fixed adjustable bias
Self Bias
Tube Life
Cost of tubes
Cost of other parts
Size
Weight
It is possible to meet the goals of the designer, without being tied down to a single combination selection of the above factors.
Multiple combinations may accomplish the same goals.
Just be sure to pick a combination that works well.
We do not want another "Tacoma Narrows Bridge" event.
I will not comment on the CED/UNSET, I leave that up to the experienced experts.
I hope that gives some food for thought, even though it is not a single or easy answer.
Last edited:
That MOSFET SOA has me wondering just why they don't all blow up when pushed even a wee bit...
I intend to be as conservative as possible. I have a slight issue with wanting to cascode them...which makes hanging the small PCB off the heatsink a small trouble...should not be too bad in practice. In my case the power stage will be a zero bias set of power triodes with no FB( the grid is grounded ). Grid current obviates FB in the manner presented, and I am not going to use a global loop either. I have not much voltage to swing.
Unfortunately, entertainment budgets are not like I wish they were...LOL should give me time to create the solutions...heh-heh-heh
cheers,
Douglas
I intend to be as conservative as possible. I have a slight issue with wanting to cascode them...which makes hanging the small PCB off the heatsink a small trouble...should not be too bad in practice. In my case the power stage will be a zero bias set of power triodes with no FB( the grid is grounded ). Grid current obviates FB in the manner presented, and I am not going to use a global loop either. I have not much voltage to swing.
Unfortunately, entertainment budgets are not like I wish they were...LOL should give me time to create the solutions...heh-heh-heh
cheers,
Douglas
Last edited: