Cathode Feedback from Secondary for PP Amp

[gingertube]

I have a little Push Pull Amp (ECL86) - the Output Transformer has 0-4-16 Ohm Secondary. By grounding the 4 Ohm output I can wire the 0 and 16 Ohm lines into the cathodes of the output tubes for a bit of cathode balanced feedback .

My concern is that I have 4 Ohm speakers (actually 6 Ohm nominal) and will therefore have the speaker connected across one cathode feedback winding ONLY. Transformer Theory says that this should'nt matter as the speaker load will be reflected equally to the other side. Unfortunately transformers are NOT ideal components and so I don't know if the theory is good enough especially since the particular output transformers in question are very "middle of the road" with regards quality.

Anyone tried this. Is it OK to wire the speaker load to just one side of the secondary when also using it for cathode feedback to the output stage? Anyone want to hazzard a guess ?

Any feedback (pardon the pun) gratefully received.

Cheers,
Ian

[sajti]

Audio Research used this configuration in many amplifiers. It works properly, only some loss in balance.

sajti

[EC8010]

Yes, you can do it. Your more likely problem is instability. I recently built a PP amplifier with an ultra-linear output stage having cathode feedback, but once I wrapped a little bit of global feedback around the amplifier, achieving stability became tricky. I had 6dB of CFB, but you are likely to have rather less, so stability might not be as tricky. Check your 10kHz square wave response...

[SY]

The trick to stabilizing is to run the feedback symmetrically from the 16 and 0 taps.

In reality, this connection is of little benefit for triode and UL- what you gain in distortion reduction from the few dB of feedback, you lose again since the driver stage has to (roughly) double its swing. For pentode stages, there is some net benefit in source impedance reduction.

[gingertube]

Thanks Guys,
SY - Point taken BUT it was actually damping factor I was after.

The amp in question is my variation on the Yves Monmagnon little ECL86 Amp (see the "Hey Planet-10" Thread) - its on the edge of serious gorgeousness with just a little bit of bass boom ONLY evident with some music selections. My variation already runs 12dB balanced shunt feedback from the output tube anodes and 20% Ultralinear to the screens. I wanted to reduce global feedback (currently about 9dB ) to get the "liveness" and imaging improved. However I'm pushing the limits of the diff amp driver already (and sensitivity is not great) so its a major redesign or just enjoy it "as is". I can cure the bass boom with a bit more global feedback BUT this compresses the image and makes the amp subjectively slower. It was better in this regard when I was running only 6dB global feedback. It is the speed, attack and "liveness" this little amp I've found so impressive and I'm convinced this is due to the low global feedback.

For 8W per channel it remains a "gaint killer" amp. I thought if I could just finness it that last little bit it would make the best "newbie" amp going, after all requests for the best amp for a beginner to build seems to be VERY common. It leaves my EL84 "Morgan Jones" Ultralinear Amp for dead and perhaps more surprisingly gives my 20W per channel 845 SET serious competition.

I intend to rebuild the "Morgan Jones" 6DJ8 /EL84 amp to this circuit by replacing the 6DJ8 (CC/split load inverter) with a 12AX7 (Current Source biased diff amp) as a final test of the schematic and then perhaps publish it here. The output trannies on the MJ amp are far superior and have 43% UL rather than 20% - That MAY be sufficient to achieve my ends.

BTW you probably already know BUT screen resistor value is critical in stabilising Ultralinear connected output tubes i.e. Its grid stopper function is more important than its g2 dissipation limiting function.

Cheers,
Ian

[EC8010]

If you really want to go all out on the next version of your amplifier, why not consider 7F7 instead of ECC83? It's slightly lower mu (70, as opposed to 100) but a lovely valve. If you really need the gain, then put a semiconductor CCS in each anode...

[SY]

I haven't tried the 7F7, but the easy-to-source 12AT7 has a similar mu and shows surprisingly low distortion on test. I think it would make a very nice diff amp for an EL84 output stage.

Your point on UL screen stoppers is absolutely correct.

[gingertube]

Did the final fine tune last night. I managed to get rid of the slight bass boom and the slight lack of "liveness" by decreasing local balanced shunt feedback - the 15K shown below (final value) was 2x 33K in parallel ( = 16K5) so the adjustment required was tiny. I found the same thing with the global feedback and the 100K value. 82K was too much 120K was not enough.

So for your edification here is the final ECL86 version. Its final because I've screwed the base on.

Ther must be 1000's of ECL82 and ECL86 Amps out there which could benefit from this circuit change.

Will let you know how the 12AT7 EL84 version goes. Will start it this weekend.

Woops who changed the image rules - back to photoshop to get it down to less than 1000 x 1000

Cheers,
Ian

[gingertube]

One more evenings fiddling:

The balanced shunt feedback is nowhere near the 12dB stated above. Confirmed by Zout measurements with it in and out. The feedback voltage is being divided by the rp of the diffamp tubes.

Came down to an exersize in juggling the 15K which sets local feedback and the 100K which sets global feedback. The more local the less global required.

I found 2 "sweet spots"
1) 47K for local and 150K for global
AND
2) 27K for local and 100K for global

Settled on the 2nd of these 27K/100K after some further listening tests. The amp has now gone to its new owner - my cleaning lady to run her Polk R5s.

As far as rebuilding the Morgan Jones EL84 Amp to this schematic goes it means that I'm stuck with 12AX7 for the diff amp - I NEED that high ra to maintain local feedback levels.

My view after all these experiments is that you need about 6dB of local feedabck to push the output tube/transformer combination one octave lower (lower rp to drive primary inductance better) and one octave higher higher (lower rp drives leakage inductance and primary capacitance better). After that the sonic improvements are due to the current sourced diff amp front end AND perhaps more importantly the reduction in required global feedback.
P.S I ditched the original "Ring of 2" current source on the diff amp and replaced it with the Cascode Current Source, LED referenced, shown in the schematic above and got noticable improvement. I also tried dual LED reference in the CCS and found NO benefit. While this allowed doubling of the emitter resistor, by the time you multiply the emitter resistor by hfe of the bottom transistor and then by the hfe of the top transistor you are already passed the point of the Zout being dominated by device capacitance and so dual series LEDs for a higher reference voltage are of no benefit.

Cheers,
Ian

[SY]

I've also been quite pleased with the cascode CCS; I've got at least one in every amp in the house and several in test jigs. It hits the wall of diminishing returns quite nicely, as you observed.