Hello, I am trying to get a pnp transistor to carry an ac signal linearly from a high positive voltage source to a low negative voltage output.
I tried this
View attachment 98993
but this is what I got out of it
View attachment 98994
The pnp transistor was specifically designed for audio http://www.farnell.com/datasheets/866302.pdf
What could be causing the distortion in the 0-6k frequency range?

As you can see from prior posts, and you already know, this is an unusual circuit to design. I thought about it for a few minutes. Consider this. Suppose you have your existing 150 VDC supply with some AC signal on top of it. Connect that through two resistors to a negative supply, like -150 VDC. Now you have a resistor divider. You want the upper resistor to drop a nominal 174 VDC, and the lower resistor to drop 126 VDC, to get a nominal -24 VDC. It's just a resistor divider. Naturally, you'll use large resistor values which you can calculate. Your AC signal will be reduced by the ratio of the resistors, of course, but perhaps you can make up the loss with the next stage of amplification. This is a very brute force approach, and it wastes a lot of energy, but it's a place to start. How does that strike you?

 

These two statements seem to contradict each other.

The Tubelab guy, a few members on diy-audio, and think the tubecad guy as well stated that transistor followers don't mess with the sound, mainly transistors are claimed to be fine as long as you don't use them for gain.
It's hard to beat the experience of the tubelab and the tubecad guys, so I figured it was worth looking into. Sadly their circuits involved turning a positive voltage into a positive voltage, not a positive voltage into a negative voltage which leaves me hanging.

I'm not sure what you are trying to say with the servo, the servo just takes the voltage from the output of the amp and uses it to adjust the grid voltage of the top tube on the WCF. If you think of both WCF tubes as resistors, the servo is just adjusting the resistance of the top tube so the DC is 0 between them.

Coin master, The transistor in the your link in an NPN transistor, not PNP as shown in your schematic. I'm not sure which you intended.
As you can see from prior posts, and you already know, this is an unusual circuit to design. I thought about it for a few minutes. Consider this. Suppose you have your existing 150 VDC supply with some AC signal on top of it. Connect that through two resistors to a negative supply, like -150 VDC. Now you have a resistor divider. You want the upper resistor to drop a nominal 174 VDC, and the lower resistor to drop 126 VDC, to get a nominal -24 VDC. It's just a resistor divider. Naturally, you'll use large resistor values which you can calculate. Your AC signal will be reduced by the ratio of the resistors, of course, but perhaps you can make up the loss with the next stage of amplification. This is a very brute force approach, and it wastes a lot of energy, but it's a place to start. How does that strike you?

Yeah, that's originally where I got the idea, I started playing around with different types of configurations with tubes, transistors, and resistors in order to make some sort of divider. Unfortunately pure resistors of high value screw up the frequency response for some reason so it can't be that. My experience in electronics is limited so maybe there's something I haven't thought of yet.

 

It's just a resistor divider.

Good idea. It takes out the distortion of the transistor and is no worse for the AC voltage loss.

I'm not sure what you are trying to say with the servo,

I'm trying to say, "Which part are you calling a Servo?" I don't see any motors in that drawing and I don't see what is powering the motor.

 

I thought AC and DC were separate, why do I need more DC to create more AC? In the sim and on my oscilloscope the DC swing is a tiny fraction of a volt and the DC on the output mimicks the DC on the base. What am I not understanding?

In general, the AC and DC are sort of separate, but once you put a transistor in there, and you want it to run in the linear range, then the combination of the two DC and AC voltages becomes important, because a large AC swing can put the transistor into it's nonlinear operating range, where it is either saturated or cut off.

 

I'm trying to say, "Which part are you calling a Servo?" I don't see any motors in that drawing and I don't see what is powering the motor.

DC offset servo, as in the opamp that adjusts dc offset.

 

Unfortunately pure resistors of high value screw up the frequency response for some reason so it can't be that. My experience in electronics is limited so maybe there's something I haven't thought of yet.

The only thing I can think of that would cause a resistor to affect the frequency response is a wire wound resistor, which is as much of an inductor as a resistor. Or perhaps the capacitance from one end of the resistor to the other becomes more significant than I realize for large resistors. 1 Mohm resistors would draw about 150 ma, but perhaps that's too much for your source, or it may be too much equivalent resistance for the next stage.

Also, by "servo" I think you really mean voltage follower. We often refer to servos as short for servo motors.

Incidently, are you aware that transistors, especially large ones, have inherent capacitance at their inputs and outputs?

 

DC offset servo, as in the opamp that adjusts dc offset.

Oh. I just call them op-amps. Why do you have 20K pots with nothing connected to their ends?

 

may be too much equivalent resistance for the next stage.

It's about the capacitance of the next tube. I have used 100K grid stoppers to cause the Miller effect to quench an oscillation.

 

The only thing I can think of that would cause a resistor to affect the frequency response is a wire wound resistor, which is as much of an inductor as a resistor. Or perhaps the capacitance from one end of the resistor to the other becomes more significant than I realize for large resistors. 1 Mohm resistors would draw about 150 ma, but perhaps that's too much for your source, or it may be too much equivalent resistance for the next stage.

My claim on frequency response is soley on what LTspice is telling me. It really depends on the values of the two resistors but using pure resistor dividers causes all kinds of funky behavior with the phase degrees and frequency response. Then again, LTspice lied to me about my last circuit in a few different ways so I don't know whether I can trust it.

Maybe a pure resistor solution isn't a terrible idea after all if I try low value resistors with huge power ratings. I haven't physically tried pure resistors yet.

 

Interesting. Theoretically, resistors have no affect on the phase of an AC signal. I've never seen a simulator have that effect. But resistors in combination with active elements, like transistors, diodes, and tubes, can have phase effects, and maybe that's what LTSpice was showing you.

 

Oh wait, I just realized the divider is changing the input voltage and therefore the plate voltage of the input stage. That is the cause of the frequency response issue. Unfortunately this means that if I use a dc offset servo to somehow have variable divider resistance so the output of the amp remains at 0vdc I will also be varying the input tube operating point. I suppose I could use another servo to maintain 160v on the input stage.
Damn, dc offset servos are starting to become my best friend, almost every circuit I've come up with so far requires them to function.

Or maybe I can use a buffer, if I am to believe that source followers don't affect the sound. I'd rather have that as a last resort though.

The problem still remains what sort of component should I use to form the bottom portion of the divider. I need an electronically controlled variable resistance of some kind.

 

Nevermind, even after manually correcting the plate voltage the frequency response and phase is still off. Also I forgot buffers need a coupling cap on the input. Oh well, seems the plan has unraveled on me, it was a good try I guess.

 

Now, I've heard from reputable people that transistor followers don't have any sonic impact on the signal

At the golden-ears-audio-crazies philosophical level, transistors *always* have a sonic impact. That's why y'all still are ga-ga over empty-state devices (bottles, jars, valves, whatever). But I digress...

The term "follower" usually means one of two things in solid state circuit design: (transistor) emitter follower or (opamp) voltage follower. While the two have significant differences at the circuit level, in both cases the output voltage is essentially identical to the input (unity gain, zero phase) with a lower source impedance. Neither is useful as a DC level shifter, although the opamp voltage follower can be modified for this over a limited range.

Your PNP circuit suggests something different, a transconductance stage. This is a voltage input / current source output stage, and is common in current sense circuits where the result from a sensor on a power supply line needs to be level-shifted so it is ground-referenced. One approach to this is to use two transistors in a current mirror configuration, but getting audiophile-level distortion performance from an open-loop translation stage will be very difficult.

ak