Incorperating a NFB loop into a paraphase inverter

[anson]

Hi everybody,

I've been tinkering around with the idea of changing my LTP to a paraphase inverter. The current output stage configuration that I have now is a LTP powering four EL-34's, class AB, with NFB, nothing too different from many other amps. But for adding a global NFB to a paraphase, how would I do that? I've looked at Supro amps to see how they did it, but it looks like they didn't bother with any NFB circuitry. If I put the NFB lead coming from the speaker tap directly to the input of the paraphase inverter, I'll have to use someting like a 20Meg series resistor to drop the NFB voltage down, assuming I use a 1M grid bias resistor on the phase inverter. That's the only thing I can think of at the moment. Any other ideas??
Anson

[Bruce / Mission Amps]

I haven't done it but, if you are using the one with two cathode biasing resistors, lift one of the resistors from ground (preferably the one that is being driven from the previous stage) with a 100 ohm resistor and insert the NFB voltage at the new junction.
Start with a NFB feeding resistor of at least 4k7 and see what happens.
Other wise, look at the NFB from a 5B6 Bassman and see if that simple NFB setup would work.... which is essential the same idea.

[anson]

I just had a look at that schematic, and I see how it makes sense. I'll give it a go and see how it works.

Thanks!!!
Anson

[anson]

Well, I tried the paraphase inverter, using 220k plate resistors and 2.2k cathode resistores, unbypassed. The sound was pretty good-I liked it. The distortion that was due to the nature of the paraphase was pretty enjoyable, for both clean and dirty tones. The NFB system wourked out good too-I used a 6.8k ohm series and a 190ohm shun resistor. To supply the grid signal for the second (paraphase) stage, I initally tried tapping off of the anode resistor of the firrst stage (at the junction of a 220k resistor, and a 25k pot, to help dial in a signal for good balance)--a split load. Then I used a coupling cap to feed the grid of the second stage, which was referenced to ground via a 1M resistor. the problem i got was after the amp was turned on for about a minute, two of the el34's would oscilate at a rate of about 2 hz, so the sound would go from full volume (allowed by the pre PI master volume, that is) to dead quiet at a rate of 2 hz. And the plates of those two tubes turned from red to normal at that same rad. Of course I didn't let this happen long before I shut the amp down. Any ideas of what this could be? It was actually a very cool sounding vibrato effect. Too bad it would toast my tubes before long!
I changed the second stage input from a split load, to the more traditional design tapping off of the fixed bias grid resistor, keeeping the coupling cap. and things worked better, but there was always a significant amound of hum. I tried the floating paraphase to, and that was even hummier. I thought my wiring was wrong, then re-did everything. Still hummed. Gettting tired of the hum, I've since changed everything back to a LTP, but am using the 6.8k series and 190ohm shunt resitors, which give the amp a better fell IMO then the 100k and 5k pair i had before. Anyboody do similar experiemnts?

[Steve Conner]

Here is how I added NFB to a floating paraphase inverter.
http://scopeboy.com/poweramp.gif

The oscillation/vibrato effect you got is low-frequency instability. It's liable to happen whenever you have three or more similar RC (or LR) time constants inside a feedback loop. They each contribute some phase shift, and if it ever all adds up to 180 degrees, it turns the negative feedback into positive feedback, making the thing an oscillator.

If you think about it, it's similar to the classic trem oscillator circuit, which has three time constants providing 60 degrees of phase shift each. You also have three: two coupling capacitors and an output transformer.

One fix is to make sure that the three time constants are all different by at least a factor of 3. But remember that the OPT's time constant gets shorter when it saturates at high power, or because of mismatched tubes. You can end up with a circuit that's only conditionally stable: it idles fine, but gets kicked into oscillation by loud notes.

[Joey Voltage]

Very True! The more phase shifts you have with in the loop, the more potential for disaster you have. the other thing that maybe can be done is to find the highest roll off for the lows, and the lowest roll off for the highs and shift that one constant far away from the others, (or conversely the other two up enough) so that it looks like you are only dealing with single pole filters.

[anson]

Thanks for the input! I didn't think of a paraphase as having multiple stages in itself, but I now see that it does. Actually, a LTP has multiple stages too, but they are DC coupled via the cathode, so I guess that would be why it's less susceptable to this kind of instability that i was experiencing.

Steve, actually my amp was oscillating either with NFB or withoutt NFB (i have a NFB on/off switch). I used .047u coupling caps and 470k grid resistors feeding the quad of el-34's, and a .022u coupling cap and a 1M grid resistoor for the paraphase resistor. Should oscillation still result with no global NFB?

the other thing that maybe can be done is to find the highest roll off for the lows, and the lowest roll off for the highs and shift that one constant far away from the others, (or conversely the other two up enough) so that it looks like you are only dealing with single pole filters.

...Sorry i'm not quite understanding you here. By this do you mean to make the two PI stages and the output stage have the same(or very similar) roll-off??


After reverting back to a floating paraphase and having it hum alot and not figuring out why, i reverted to a LTP but used a 6.8k tail resistor and a 190 ohm NFB shunt, with a 6.8k NFB series resistor. Then I put on 150k and a 220k plate resistors. This was after countess experiemnts with other values, of course! I Don't know how it looks on a scope, but it's sounds way more interesting to me than the previuos marshallesque way I had things set up before.

[ThermionicScott]

Should oscillation still result with no global NFB?

Nope. Sounds like you have some layout/circuit debugging to do before you can add the NFB. Do you have grid- and screen-stopper resistors installed? Got a pic of your build?

- Scott