On Wednesday, 7 November 2012 16:32:59 UTC+11, Alex Pogossov wrote:
Alex: I was aware of your cathode feedback design. It requires a specially made transformer. What I am wondering, do people often use output tubes as "inverting op-amp" with resistive feedback from plate to grid to make a rather deep and always stable local NFB around the output stage?
Using shunt FB with R from anode to cathode then R to some driver source has rarely ever been done in commercially made amplifiers even where output tube gain is high as is the case for normal pure pentode or tetrode output stages.
The reason is that the technique usually requires the driver tubes ahead of the OP stage to have low Ra and to make say 3 times Vg1 applied to OP grid and to cope with R1 which might be say 47k, with R2 to anode being a much higher R and cap coupled to anode. Damn, too many extra R and C and using more expensive drive tubes, so bean counters vomited on the idea. There is an alternative, see pagess 333 and 334 of shunt FB techniques where the high anode Ra of a driver pentode works like R1 of the shunt FB network. ß can then become usefully high as about 0.5, and pentode gain is still high enough. But this idea never took hold either, and the simple cheaper solution of GNFB could never be proven to sound inferior to all manner of other ideas tried by egotists wanting to be original. I tried the idea, see by circuits page for the stuff about Balanced Shunt FB,
http://www.turneraudio.com.au/miscel...chematics2.htm
All good tube audio power amps require a "specially made OPT". The non specially made OPTs are all inferior. Quad's OPT for Quad-II was specially made, and no more difficult to make than any other specially made OPT - made by ppl able to do horrid stuff while everyone else is bone lazy.
But, even Quad-II OPTs have no advertised ow loss way to match the amp to 4 ohms, which is needed in 2012 because modern speakers are not often 16 or 9 ohms, but often 5 ohms. So with 5 ohms connected across the 9 ohm setting, RLa-a becomes 2k2, way too low for good fidelity, and then KT66 gain is low, making the 10% CFB fairly ineffective. But its possible to remove Quad-II OPTs from their case, and to adjust sec wires and install an extra terminal to allow waste free sec windings for 4k0 : 4 ohms, so that RLa-a with 5 ohms becomes 5k0, and then the losses reduce and although still high, fain and FB is more effective and overall function much better. Using 8 ohms connected to a 4 ohm OPT setting gives RLa-a = 8k0, and then the Quads function far better, although PO is limited, but its mostly pure class A.
My 300 Watt amps have large OPTs with 6 sec sections 5 primary sections. The core is 110mm stack of 50T E&I GOSS. Np = 1,060 turns in 10 layers so each P section = 2 layers. One complete 2 layer section of P is devoted to CFB = 20% with CT which is far more effective than the 10% of Quad. Now there is NO extra special effort to provide the 20% CFB winding. The OPT can be used to give a wide range of primary use, plain UL, UL with CFB, CFB only, whatever, its doable, and the secs have each section of one layer being 48t + 24t which allow a considerable number of waste free low loss sec connections which maintain the possible bandwidth at 250Watts = 20Hz to 270kHz. The 300W OPT isn't hard to wind using such large dia wire, but there are many connections to make, all of which make CEOs and accountants have heart attacks.
I was once offered awful OPTs by someone for a price way too high, and I said, "Have you ever thought of re-naming your company to 'Mediocre Transformers P/L'? They were not amused, and I just wound all my own after that. They went broke 4 years later, an all too common occurence because nothing is specially done for anyone. For them, special = ruiniation.
Patrick Turner.