LTP phase inverter design frustrations & weird voltages

[thelonious]

Hi all,

I am in the process of finishing my first amp design/rebuild. [Insert standard disclaimer here about how long this post is and what dumb things I will say because I am a n00b]. 

I started with an old Bell 5630 PA amplifier and did a recap job on it to see what it sounded like and make sure it was in working condition before I started modding. After I got it working, I read Merlin's preamp design book and did a ton of research online in this forum and others. Then I started modding things to see what would happen, and you know how that bug goes... four months later, there isn't much of the original left; it's a new amp. The layout is now 12AX7 > 5879 > 12AX7 FX loop > Moonlight tone control > 12AV7 Cathodyne PI > 5881 P-P output stage (cathode biased). I am fairly happy with the sound, especially considering I probably should have started with a little SE amp but instead got in way over my head with this monster project.

Now to the point of this post: I have tried switching in a LTP phase inverter, and no matter what I try, I get the same weird result. It acts kind of like a gate. There's extremely low output until the input signal increases to a certain level (in the latest failed design that was input level of ~35Vp-p), at which point the output suddenly spikes to massively loud with awful distortion. It's as if some kind of resistance were being overcome.

Here are the steps I have gone through, all with the same strange result described above:
1) Try dropping in a 12AX7 LTP design from Merlin's book without changing anything (lazy, I know, but it was a place to start). Problem begins.
2) Search forums to find out more about the LTP. Try different part values based on that research. No change.
3) Replace PI tube socket in case it had issues; also, try a different 12AX7. No change.
4) Put 68k grid stopper on each power tube grid just in case. Also put bigger grid stopper on PI. No change.
5) Replace LTP with cathodyne PI. Suddenly amp works properly and sounds great!
6) Rebuild LTP outside the amp (on tupperware, actually) using original tube socket and last used values, and sub that into the amp. Same problem as before.
7) Replace 12AX7 with a 12AV7 without changing part values - no change in problem.
8) Calculate load lines for 12AV7 and change part values accordingly. No change in problem.
9) Give up and go back to using cathodyne. Works great, sounds great again.

So---any ideas on what in the world I am doing wrong with the LTP? I'm fine with sticking with the cathodyne, I just don't want to be defeated by the LTP. 

I've attached schems of the two LTP designs I used in steps 1 and 8. The latter includes the screwed up voltages I tested on the step 8 design. Do those symptoms on a LTP (acting like a gate, then huge ugly output spike, weird voltages) ring any bells with you guys? Is there some basic mistake I am making in my ignorance that would cause this? Different part values I should try? Or, if this post is useless without schematics and voltages for the whole amp, just say so and I will draw them out over the next couple days and upload them.

Thanks in advance for your help. I've learned a ton in this forum and greatly appreciate the knowledge you guys share here.

[sluckey]

Cg2 should really be connected directly between grid 2 and ground but that's not causing your symptoms. And you must have a cap on the input to prevent upsetting the bias on the LTP. Do you?

[Colas LeGrippa]

LIKE sLUCKEY SAID, I am pretty sure you have forgotten a coupling cap somewhere. You must have a cap before the input and after the output, if not, eaxactly what you describe will happen.

Colas

[thelonious]

Ok, that makes perfect sense. I know I have an input cap on the current build with the cathodyne, I'll bet I forgot an input coupling cap on some of the early LTP builds. Also, I tried a bunch of different tone stacks, and in some of them I used vintage caps from the original amp. Those caps tested ok for value, but they easily could have been leaky and letting DC through to upset the LTP.

Thanks for the tip on Cg2!

Tonight I'll try the LTP again, making sure I have good caps on the input and outputs. If it's still misbehaving, maybe some of my wiring is bad and I have voltage leaking somewhere it shouldn't be. I'll report back.

[thelonious]

Finally got it working. I always thought coupling caps were to keep DC voltage from the previous stage off the grid of the following stage, so the arrangement I had was gain stage > coupling cap > tone control > volume pot > PI. What I never thought about was that the tone control and volume pot were affecting how the PI's grid was biased because I didn't also have an input coupling cap. The cathodyne worked anyway, but the LTP's balance was too messed up. All I needed to do was add a coupling cap immediately before the grid stopper - so you guys were totally right. Thank you for your help!

But that brings up the question, how do I know when I need two coupling caps between stages and when I don't? I thought the output coupling cap of the previous stage would also function as the input coupling cap of the next stage, if that makes sense. But now I'm wondering if every time I have a volume/gain pot between stages, I need both an output coupling cap before the pot, and an input coupling cap after the pot?

[sluckey]

how do I know when I need two coupling caps between stages and when I don't?

Look at the dc bias voltage on the grid. If it ain't zero volts, as in the case of the LTP, you probably need a grid coupling cap to prevent any tone pots, volume pots, or other resistors giving that dc bias voltage another path to ground. The typical LTP uses a bootstrap bias circuit, ie, grid resistor is returned to a resistor that is also connected to the cathode circuit. There will usually be a fairly high dc voltage at the junction of the grid resistor and the cathode 'tail' resistor. Without getting too deep, this bias arrangement results in a verrry high input impedance, even though the grid resistor may only be 1meg. The impedance is so high that even your 10meg input resistance DMM will load the real grid voltage and throw the actual bias voltage off. A 1meg volume pot loads the grid circuit even more, causing the bias to be way too low.

Other circuits use this bootstrapping bias too. For example, look at this Princeton split-load PI. The grid is bootstrapped to the cathode just like the LTP PI. Notice that there is 73.5v at the bottom side of the 1M grid resistor? And that there is 75v on the cathode? That 73.5 volts is also on the grid and that means that the cathode is 1.5v positive in respect to the grid. Or, you could say the grid is 1.5 volts negative with respect to the cathode. Sounds like typical bias for a small tube. Now look at what happens if you put a 1meg resistor (like a volume pot) from grid to ground. That 73.5v at the bottom of the original 1meg grid resistor now has a path to ground thru the added 1meg resistor. Current flows and now 1/2 that 73.5 volts will drop across each of those 1meg resistors. And you end up with 36.75v on the grid. Now the grid is negative 38.25v (75-36.75=38.25) in respect to the cathode. The tube will be nearly completely biased off and will really choke any signal applied to it. So, you can't just hang any resistance on a bootstrapped grid without messing up the bias. I chose the Princeton because the circuitry is much simpler, but the same principle holds for the LTP. The coupling cap in the Princeton is doing double duty. It's blocking plate voltage from the previous plate, and at the same time, it is preventing the previous stage from upsetting the high impedance bootstrapped grid bias circuit. If you just had to have a volume control between those stages, you would likely need another cap.

To summarize, if the grid voltage ain't zero, you will probably need a grid coupling cap.

[thelonious]

Thanks for the informative explanation! That completely answered my question and gave me new things to read up on...