Am I crazy??

[SoulFetish]

Okay, this is weird. But it's late, so hopefully I'm just over tired and I'm overlooking something dumb.

This should be quick and easy.. but unfortunately i'm the one who's telling the story, so there goes any hope for "quick", but I'll try.....

I'm doing some testing on my output stage, right. I removed the DC coupled cathode followers, and wired it to be driven from the plates of the LTP (as used in more traditional configurations). I want to honestly ditch the driver stage if I don't think it adds anything tonally or in the response of the amp. I don't want to fall in love with an idea just because I'm invested in it. I'm ready to put the meat of the design on the chopping block and change it, or call it a failure and start something fresh.
(oh, refer to the schematic below. I drew it up quick but the values are pretty standard, so I didn't include them for time considerations)
Anyways, I was running a simple test tonight to set it up for using global feedback, and without feedback. With feedback engaged, I fed the input of the amp with a sine wave, and was monitoring the voltage driving the grid of the phase inverter. It was relatively arbitrary, but I adjusted the volume control so the voltage at the input of the LTP was 1VRMS. Nice clean sine wave. Normal right?
For testing purposes, I wanted to compensate the non-feedback PI drive voltage to be the same as with feedback, so I installed a 1M5 potentiometer in series with the 470k grid leak and kept all controls the same. So, when I removed the feedback. The voltage at the grid (test point #1) of the inverting triode was a little lower with the pot at minimum resistance? (basically out of circuit). I was like "what?"
Then I thought, "well, the negative feedback input is at the non-inverting grid, so maybe I need to test the signal at the non-inverting grid.
Here's where it gets weird: I confirmed that the feedback signal coming from the 8Ω tap was not open. So, while monitoring the signal (test point #2) at the non-inverting grid with feedback engaged, I pulled the feedback and the signal voltage cut in almost half!?? I did it again, and there is clearly a drop in voltage without the feedback signal.
So, the simple answer would be that the phase is reversed and I'm not using negative feedback at all and it's positive. Wouldn't that cause crazy oscillation?
And if the feedback was positive, why would the presence control works as it should with the global feedback connected.

What. the. hell. is going on?

[SoulFetish]

Would it be helpful to list the used component values on the schematic?

edit: I figured better to provide a complete picture of the amp under test.

[Dave H]

The feedback also goes to the tail which changes the cathode voltage. The signal on the grids is wrt ground. The input signal as far as the tube is concerned is its grid to cathode voltage. To test if the feedback is negative monitor the output then remove the feedback.

[SoulFetish]

yeah, yeah. Thats righ! In fact, the non-inverting grid is grounded as far as AC is concerned.
This makes sense now. Trying to do this test by measuring the grid to earth by voltages seems convoluted. I could just as easily measure the plate voltage and make the adjustment til those match.
So I was probably seeing the effect of the feedback bootstrapping tail, yeah?

[Chuck H]

Also, haven't you altered the phase (inverse) if you input at the other PI grid.

[Helmholtz]

So, while monitoring the signal (test point #2) at the non-inverting grid with feedback engaged, I pulled the feedback and the signal voltage cut in almost half!?? I did it again, and there is clearly a drop in voltage without the feedback signal.

Why does it surprize you that the signal at the non-inverting grid drops if you remove the NFB signal?
Don't forget that the 0.1µ cap is the coupling cap to the inverting grid. Signal at both its ends is the same.
So the inverting grid signal is identical to the signal across the 5.6k shunt resistor.
For normal operation without NFB, the inverting grid doesn't need a signal, as the triode is drven from the cathode signal.
But the difference in cathode currents causes a signal drop across the 5.6k resistor.
So the inv. grid signal without NFB is an artefact of the circuit.


I recommend to watch phase relationships between signals with a 2-channel scope using the PI input signal as reference.
This should allow you to see that connecting the NFB not only increases the inverting grid signal, but also reverses its phase (I think).

[pdf64]

yeah, yeah. Thats righ! In fact, the non-inverting grid is grounded as far as AC is concerned.
This makes sense now. Trying to do this test by measuring the grid to earth by voltages seems convoluted. I could just as easily measure the plate voltage and make the adjustment til those match.
So I was probably seeing the effect of the feedback bootstrapping tail, yeah?

I think it would be best to make the ‘feedback to open loop operation’ comparison based on the output level; as you’ve seen, bootstrapping affects the input impedance and hence the signal level there, and inside a feedback loop, all bets are off, as the signal includes an error correction component, so it can look and measure a bit weird.

[g1]

As a side note, can't speak about this particular case, but incorrect OT phase (positive feedback) does not always cause crazy oscillation.

[Helmholtz]

As a side note, can't speak about this particular case, but incorrect OT phase (positive feedback) does not always cause crazy oscillation.

Maybe not, but the presence control would not operate as expected.

[Chuck H]

Maybe not, but the presence control would not operate as expected.

Absolutely this^^^^^^^^^^^^^^^^

I expect the original design was in correct phase, but it was mentioned that during the testing the alternate PI input was used. This absolutely would invert the phase. I think there may be some dirt in the tests. For what it's worth it happens to me ALL THE TIME before I've managed to evaluate everything clearly. And I've been put on track by kind folks here at the forum many times.

[SoulFetish]

Right on, Chuck.
Sometimes you miss the connection between what your seeing with what you already know. Which is what I was presuming was the case here. DaveH’s first reply drew the link I needed.

Chuck, your my other EL84 brother here. You ever use any compensation across your OT primary?

[Chuck H]

Chuck, your my other EL84 brother here. You ever use any compensation across your OT primary?

You mean like a Zobel network (aka conjunctive filter) across the plate loads? All the time. In fact the only time I don't is when the power amp won't ever be pushed much beyond clipping. But I typically use less capacitance than usual. The resistance I use is has always been the suggested 1.25 to 1.5 x impedance @ frequency or a little higher. Not sure why EL84's always seem to show more tendency to "ringing" on the waveform when pushed hard compared to other tubes. There's surely a technical reason, but I don't know it. I do know that a Zobel network across the primary (using lower capacitance than usual) doesn't noticeably rob any power and seems to solve the problem.

IIRC one of the higher minds here suggested that two filters, each from the plate to the OT CT would be a better arrangement. I think it had to do with mitigating component failure due to voltage differential peaks. Which CAN get pretty extreme. I go to some trouble sourcing parts that are up to the task. Having had earlier experiences with failures in these circuits.

[SoulFetish]

IIRC one of the higher minds here suggested that two filters, each from the plate to the OT CT would be a better arrangement. I think it had to do with mitigating component failure due to voltage differential peaks. Which CAN get pretty extreme. I go to some trouble sourcing parts that are up to the task. Having had earlier experiences with failures in these circuits.

Yeah, that's the way I'm doing it. I've been getting some good reads over at Max Robinson's site. Here is the article he posted from Tim Smith which describes the formula for determining which values of RC to put across each side of the Primary.
https://www.angelfire.com/electronic...Smith-NFB.html


According to Smith's sources, these are the snubber values I should be using. I don't know how the hell I came up with the values for the RC network I'm using now, but they ain't right (2.7k & 2.2nF??).

[Chuck H]

The first time I implemented the Zobel I was under a time crunch for delivery and recognized the ringing on the scope analysis. So I did a little reading about typical values for this circuit and then used the scope to see what was happening. It's bound to be different values for different OT characteristics. I'm still straight across the plate leads using the ubiquitous 10k resistance value for an 8k primary Z. I didn't like the tonal changes with the typical cap values though so I started reducing that. I found that 1500pf would stop 90% of the ringing, absorb most of those pesky HF voltage spikes in the primary and was only barely audible. Like I said, I was under a time crunch so I used this blunt, brute force method. Just watch the scope and listen and change stuff till it works.

When I built a 4xEL84 version of my amp for a guy I arbitrarily used a 5k resistor and 3.3n cap. The amp sounds great and behaves on the scope but this was a lazy/clumsy thing to do since the miller capacitance using more tubes should change things beyond just halving the resistance and doubling the capacitance. Possibly or probably even eliminating the need for a Zobel on this amp since you don't see them used on 4xEL84 amps like they are in 2xEL84 models.

Subsequent experiments since I acquired ltspice mimic everything I saw building the 2xEL84 design like a mirror (kudos to Dave H for having good tube and transformer models which he kindly sent me). And tweaking values looks to me that it should be possible to increase resistance and further reduce capacitance. The key here being that, relative to the typical characteristics of transformer inductance and primary impedance we're targeting a fairly consistent frequency range. A larger value resistor would reduce the shunt ratio, but it also reduces the knee frequency for a given capacitor value so a lower value can be used. So this is how I'm "tuning" the circuit. Still using brute force rather than formulas but using the scope and my ears (and now ltspice modeling) to get whatever a given build needs.

I'm not trying to clone any circuits used by others. My only goal is to reduce ringing and spikes while keeping the audible consequence to a minimum.

[Helmholtz]

I found that compensating an amp for perfect stability, good square wave response and minimum ringing with a resistive load does not ensure stability with a guitar speaker.

With the Vibrolux Rev. (https://music-electronics-forum.com/...ibrolux-reverb) I tried a Zobel network as well as lead or lag NFB compensation.

Though results looked good with a resistive load, any of these measures increased instability with speakers.