... so I have a bunch of these nice "super" 6BG6-GA (7027A in disguise) gathering dust, just burning for me to build a nice amplifier around them.
For this design, I want to try zero global feedback, which means extracting the maximum performance from the output configuration & local feedback. Distortion trumps power, but I'd still like to get reasonable efficiency. Triode class AB2 is one option, but I'm thinking I would rather run ultra-linear, with a separate screen winding to allow a higher B+ on the plates, and another separate winding for cathode feedback. Having not tried the latter before, and wanting to 'get it right' (or close to right) with a set of custom-wound OPTs, all sorts of questions about distributed load winding ratios come to mind
For a start, the optimal Pentode loading appears to be 6.6k, which SY also suggests is a good loading for Triode strapped operation. One might then assume then that this would also be a good load for UL operation - would that be a correct statement?
Second, I'm having a hard time figuring out the optimal UL% to use. I am presently reading R. Moers' article on calculating the optimal UL%, but it seems that 43% is commonly recommended by the original manufacturers for this broad family of valves, apparently on the vague premise that about 25% gives most of the distortion reduction, but there is more to gain with yet more screen feedback. I am also uncertain how the screen FB would be affected, if at all, by running a separate winding at a lower DC potential than the plates.
Third, I haven't yet decided how much winding to put under the cathodes. What Crowhurst calls "modified ultra-linear" subtracts the full UL percentage from the primary and places it in the cathode winding, such that the screens simply tie straight to a regulated supply. This achieves the maximum cathode feedback without reversing the polarity of the screen drive (and neatly eliminates one winding), but at a typical UL ratio of 43%, it demands an awful lot of voltage from the driver stage. Drive swing and heater breakdown considerations may set a practical upper bound, but I wonder if there is an optimization / distortion null (maybe an IMD minima?) to be obtained by choosing the right ratio of cathodes to screen winding, or if the benefit of the larger 43% vs. a lesser UL% is largely wasted, unnecessarily burdening the drive stage? I note that the Quad II apparently uses only a 10% winding at the cathode.
Finally, in some distributed load implementations (MacIntosh, etc.), I have seen the appearance of an extra small-value choke at the plates of the output tubes, and I'm curious as to the purpose of this. Stability? Reduced IMD?
Between plate load impedance, bias point, plate & screen supply voltages, screen winding % and cathode winding %, I don't think I could possibly explore the universe of possible solutions on my own. I'll keep reading as much reference material as I can lay my hands on, but I'm hoping the resident experts will come out of the woodwork and help guide me towards a good configuration for some custom OPTs.
For this design, I want to try zero global feedback, which means extracting the maximum performance from the output configuration & local feedback. Distortion trumps power, but I'd still like to get reasonable efficiency. Triode class AB2 is one option, but I'm thinking I would rather run ultra-linear, with a separate screen winding to allow a higher B+ on the plates, and another separate winding for cathode feedback. Having not tried the latter before, and wanting to 'get it right' (or close to right) with a set of custom-wound OPTs, all sorts of questions about distributed load winding ratios come to mind
For a start, the optimal Pentode loading appears to be 6.6k, which SY also suggests is a good loading for Triode strapped operation. One might then assume then that this would also be a good load for UL operation - would that be a correct statement?
Second, I'm having a hard time figuring out the optimal UL% to use. I am presently reading R. Moers' article on calculating the optimal UL%, but it seems that 43% is commonly recommended by the original manufacturers for this broad family of valves, apparently on the vague premise that about 25% gives most of the distortion reduction, but there is more to gain with yet more screen feedback. I am also uncertain how the screen FB would be affected, if at all, by running a separate winding at a lower DC potential than the plates.
Third, I haven't yet decided how much winding to put under the cathodes. What Crowhurst calls "modified ultra-linear" subtracts the full UL percentage from the primary and places it in the cathode winding, such that the screens simply tie straight to a regulated supply. This achieves the maximum cathode feedback without reversing the polarity of the screen drive (and neatly eliminates one winding), but at a typical UL ratio of 43%, it demands an awful lot of voltage from the driver stage. Drive swing and heater breakdown considerations may set a practical upper bound, but I wonder if there is an optimization / distortion null (maybe an IMD minima?) to be obtained by choosing the right ratio of cathodes to screen winding, or if the benefit of the larger 43% vs. a lesser UL% is largely wasted, unnecessarily burdening the drive stage? I note that the Quad II apparently uses only a 10% winding at the cathode.
Finally, in some distributed load implementations (MacIntosh, etc.), I have seen the appearance of an extra small-value choke at the plates of the output tubes, and I'm curious as to the purpose of this. Stability? Reduced IMD?
Between plate load impedance, bias point, plate & screen supply voltages, screen winding % and cathode winding %, I don't think I could possibly explore the universe of possible solutions on my own. I'll keep reading as much reference material as I can lay my hands on, but I'm hoping the resident experts will come out of the woodwork and help guide me towards a good configuration for some custom OPTs.