 
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
|
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
This is going to be kind of long...sorry. I'd be particularly interested in a response from John Curl, or anyone else that knowledgeable.In an effort (unsuccessful) to figure out the source of intermodulation distortion, I was drawing simple circuits and thinking about them. That led me to draw a simplistic single-ended amplifier.
Picture a zero-impedance power supply connected to a load. At the other end of the load is the amplifying device (either tube or transistor) in a normal common cathode/emitter configuration, and the cathode/emitter is tied to ground (the circuit uses base/grid bias, I guess -- remember, this is simplified).
OK, so we have a power supply, a load, and an amplifying device. One model for the circuit is that the amplifying device is a variable resistor, with the resistance controlled by the input (base/grid) terminal. The resistance of the amplifying device determines how the supply voltage divides between the load and the device itself.
With no feedback (simple, remember) the output voltage, i.e., the voltage across the load, is the supply voltage times the load resistance divided by (the load resistance plus the amplifying device resistance). Hence, the output voltage is NOT a linear function of the input voltage, even with a perfectly linear amplifying device.
The oddest part of looking at things this way is that the nonlinearity is greatest when the load resistance equals the amplifying device resistance, and is least when the load approaches infinity or zero!
There is a second problem: The load sees the resistance of the amplifying device as its source impedance, so the impedance of the amplifier varies with the instantaneous output voltage. This means that without correction, as the output voltage goes up, it does so less than linearly, AND the impedance of the amp rises as well.
The output of an amplifier in this too-simple model is a logarithmic function of the input voltage. To reiterate, the most startling fallout of this model is that perfectly linear devices distort pretty badly if used without feedback!
I started thinking about solutions. The first solution I thought of was the obvious: feedback. The problem with feedback is there is ALWAYS residual error. The next solution I think is a winner: Provide corrective circuitry that applies an inverse function of the distortion component.
As a side note, here's another question/idea: What if we used a second tube (or tube pair for push-pull) as a feedback/correction tube, in parallel with the main output tube, and mix the signals in the output transformer (I don't know what you'd do in transistor circuitry)? It seems some optimization would be possible. Also, the feedback/correction tube might be of a different type -- say, a high-current, low-gain type like a 6AS7/6080 along side a 6550 or 6C33C main output tube.
Overall, I think the best idea is preemptive circuitry. The "Here's how it doistorts so here's the correction for that" approach.
Love to hear y'alls' thoughts.
Lee
Follow Ups:
I'm not sure, but the concepts that you are thinking about have been on the agenda for many decades. There are problems.
First, tubes have a distortion mechanism, [Child's Law] that is different from solid state. Also, triode operation complicates the predicted result, because the change in Rp with current will tend to linearize the transfer function, but it will be effected by R(L). If I lost you already, then you are reaching too far. There are two concepts that are also used, besides negative feedback. These are feedforward, and pre-distortion. Both can be used, but negative feedback is usually the easiest to use.
Thanks, John, and yes, you did lose me some. I've hadn't heard of Child's Law.Is pre-distortion predictive complementary distortion?
Thanks for the thoughtful reply.
Lee
I think so.
John,What's your opinion of 4 terminal amps (like Kenwood tried years back) that try to include the wiring in a servo loop and thus assure constant voltage at the speaker terminals? Assuming, of course, that that's what you want.
Maybe I got it wrong... but there is no single device resembling an ideal potentiometer you described. To be exact, no single device in any practical amplifier wiring. They all sink/source current as a more or less linear function of input voltage or current, so whatever 'resistance' is seen by the load is a function of input, load, supply etc.> > What if we used a second tube (or tube pair for push-pull) as a feedback/correction tube, in parallel with the main output tube, and mix the signals in the output transformer (I don't know what you'd do in transistor circuitry)?
If both (different or same) tube are driven with opposite phases,
this sounds like a push-pull amp. I've seen a web page (don't have URL at hand) where a push-pull amp had identical output tubes but their inputs/outputs were unbalanced about 1:3. The author says it sounded much better than a symmetrical PP or an SE with the same components. If both tubes are in phase, yet unbalanced... well, if one is a triode and another a tetrode this may add up to something more linear than each. But isn't this what happens in an ultralinear-wired tetrode? Maybe it's worth fine-tuning the UL tap ratio to optimize rather than add a second tube?p.s. Actually, one can build the 'potentiometer' around an opamp or even DAC, but that's too far from the topic :))
This post is made possible by the generous support of people like you and our sponsors:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]
