KT88 amp

[hylaphone]

It.s very strange. That mean you nfb circuit create a frequency pole in low frequency range which is very pretty unusual. Can you make a hand draw sketch how you nfb circuit is wired ? Technically the only reasonably reason could be an incomplete full bypassed cathode resistor....hmm

I just tried clipping in capacitors parallel with the existing 50uF cathode caps. 220uF yielded 40Hz cutoff. 470uF yielded 30Hz cutoff. Above that, frequency is +/-1dB well beyond 20kHz.

Here is a rough sketch of the layout, sorry for my chicken scratch. The speaker jack is isolated, always has been. OT common is grounded with the PI. As you can see I brought the shunt resistor back to the cathode side. Seeing good symmetry now.
I haven't made the grid leak and stopper changes yet, though it is a high load for this PI I am still able to drive the KT88 to full power.
Unless the 100K is somehow attenuating the low end due to NFB? I can't visualize how.

[tedmich]

TLDR but has anyone mentioned these voltages (525V plate/360V screens) looks more suitable for 6550 tubes? Real KT88s handle much higher screen voltages and are often run with no drop from the plates.

[catalin gramada]

When correctly applied, NFB lowers gain but extends frequeny response at both extremes, i.e. treble and bass.

That is an extremely simple circuit. It is almost impossible to not make it proper work. I suspect all issues comes from a not proper layout plan.

[Helmholtz]

When correctly applied, NFB lowers gain but extends frequeny response at both extremes, i.e. treble and bass.

[catalin gramada]

It.s very strange. That mean you nfb circuit create a frequency pole in low frequency range which is very pretty unusual. Can you make a hand draw sketch how you nfb circuit is wired ? Technically the only reasonably reason could be an incomplete full bypassed cathode resistor....hmm

[catalin gramada]

OK, well I rebuilt with 1.5K cathode resistors and the PI outputs are nicely balanced. Something was wrong before, I'm not sure what but it's working well now.

One other question I had --- with no feedback applied, the amp is flat down to 20Hz. With 1K/10r feedback ratio I'm seeing -3dB point of around 130Hz into a resistive load.
How should this be optimized?

I.ll think about. But at first glance try to bypass the PI cathoses with a bigger cap value, please report back.
You.re still with 100k grid leaks ? Make coupling caps twice as big- 100nf. Or better get rid off those 100k and put some decent grid leaks as 220k I suggested. Please report back.

[hylaphone]

OK, well I rebuilt with 1.5K cathode resistors and the PI outputs are nicely balanced. Something was wrong before, I'm not sure what but it's working well now.

One other question I had --- with no feedback applied, the amp is flat down to 20Hz. With 1K/10r feedback ratio I'm seeing -3dB point of around 130Hz into a resistive load.
How should this be optimized?

[catalin gramada]

Mostly like that

[catalin gramada]

I will reduce also cathode resistors a bit to 1.5k each, biasing at -2v around for better linearity...
100k plate with 1.5k in cathode with 220k grid leak, 5.6k grid stoppers and 47nf coupling caps is my suggestion to try.

[catalin gramada]

Currently have 56nf into 100K grid leaks. No problems driving the tubes to full power. Only "issue" is the inverter side imbalance which happens with and without the KT88s installed.

Again im just trying to understand this, the circuit is "close enough for rock n roll" as is. The paraphase is a cool approach, with lots of room for tweaking.

I think in terms of impedance 100k is a very hard load. 220k will work fine for kt88 as time you will keep biased into moderate power dissipation. Yo don't try - you'll never know what you loosed...but is safe as time you not overpass power dissipation over 70 percent. You may even try to reduce the 6sl7 plate resistors a bit. Over that think you output Impedances of you 6sl7 driver are diferent: one output is pretty high and plate follower output is quite low. May be all of this a reason for you imbalance...? You have to try. Take it progressive not all together...

[hylaphone]

Not sure. Put 220k grid leaks on power tubes. 5.6k grid stoppers should be enough to keep it stable and be confident nothing will blow up till you not hard overdrive the power stage...which is quite impossible with you 6sl7 driver tube.
100nF coupling caps to get effective nfb into low end region. Or may try 22-47 nF for a well overdamped speaker- that.s different tweakings by taste...

Currently have 56nf into 100K grid leaks. No problems driving the tubes to full power. Only "issue" is the inverter side imbalance which happens with and without the KT88s installed.

Again im just trying to understand this, the circuit is "close enough for rock n roll" as is. The paraphase is a cool approach, with lots of room for tweaking.

[catalin gramada]

Not sure. Put 220k grid leaks on power tubes. 5.6k grid stoppers should be enough to keep it stable and be confident nothing will blow up till you not hard overdrive the power stage...which is quite impossible with you 6sl7 driver tube. But be careful to keep it biased into moderate power dissipation, no more than 70 percent.
100nF coupling caps to get effective nfb into low end region. Or may try 22-47 nF for a well overdamped speaker- that.s different tweakings by taste...

[hylaphone]

You circuit should look quite like that and is your option to choose where you install the 10ohm resistor: in top to keep secondary "floated" or in bottom to keep secondary grounded. Think common wire as signal nfb return then as ground reference...or...think optimise the circuit path first then hit where you choose to ground it.
Sorry, due my limitation English skills cannot be more clear like that...

I understand you well, and thanks for spending so much time with my questions.

Do you have any further thoughts on the imbalance I'm seeing at higher frequencies ?

[catalin gramada]

You circuit should look quite like that and is your option to choose where you install the 10ohm resistor: in top to keep secondary "floated" or in bottom to keep secondary grounded. Think common wire as signal nfb return then as ground reference...or...think optimise the circuit path first then hit where you choose to ground it.
Sorry, due my limitation English skills cannot be more clear like that...

[catalin gramada]

I don't see any difference between two circuits as time cathode resistor(s) are bypassed so mostly original version take advantage from economy of components...As time you use individual cathode resistors the shunt nfb resistor in the bottom to ground rules no need to use a "floating" secondary, but even so the corect wiring as time you use the global nfb is exactly how it was presented in my sketch no matter if common OT is grounded or "floating" = meant the common secondary should be grounded quite where you grounded the cathode of the stage where nfb is applied not ground at the speaker jack. So you keep the OT common isolated, the speaker jack isolated and run the nfb twisted wires pair (send and return) directly over 10 ohm resistor. Put the 10 ohm in the bottom to ground and...voila! You have a grounded secondary. Then will be very easy to choose and see any diferences just moving the 10ohm resistor -together with nfb wires from OT- in top or bottom. Anything else should be keeping isolated.