|
Home
/ FAQ
/ News Classifieds / Events |
Audio Asylum Thread Printer |
Get a view of an entire thread on one page |
67.169.53.166
| '); } else { document.writeln(''); } } else { document.writeln(''); } } else { document.writeln(''); } } // End --> |
In Reply to: RE: I can't believe no one has commented on this post... posted by Henry Pasternack on October 27, 2009 at 14:32:01
where I take the "white box" or "clear box" testing paradigm and Henry takes the "black box" approach. But will yield same results where I am curious about the physics known inside and designed inside (but obviously don't know enough to be any expert there) to Henry's "so what how it does it, it just does it, take the simple external model."
Correct me if I am wrong.
-kurt
Thanks for the reply, but I don't think you really got the point of my long article. (IMHO it's a brilliant essay, but I never expected anyone to read or understand it, not then, not now -- though I would be delighted if they did.)
Here is a synopsis for the attention challenged: All the common triode circuit models are just abstractions of the tube that have no real bearing on the actual internal structure or operation. The internal NFB model is one such abstraction that does away with the series (Thevenin) or parallel (Norton) plate resistor and replaces it with an active control structure. It's no closer to the reality of how the tube works than the models it replaces. Furthermore, it gives no additional insight into externally observable behavior of the tube and its use in practical problems complicates circuit analysis. The model is based on circular logic and awkward assumptions. Finally, the underlying approach applied to other problems leads to feedback appearing in all kinds of devices where we don't normally think of it existing, like diodes, resistors, electric motors, and parachutes.
I'm not actually that committed to the argument, not like I was three years ago, but I couldn't resist Tre's invitation to dredge up a few thousand words of archived historical drivel.
-Henry
I do understand all your points and maybe I am not conveying this understanding at all to you. And your essay was brilliant, Henry, and I mean that. I know I was certainly only one of the few who understood but pondered it over and said little about it for there are lingering doubts about what is knowable and what is important to know. Ya know?
To access the POTENTIAL of its internal actions always requires external devices to complete the circuit that shows what that device is doing. If you don't have any way to predict what's inside as in Black Box.
For the parachute, it is in a NFB situation when the application is showing the device's POTENTIAL. If it were used above the moon, it would never exhibit its POTENTIAL for any application, just as the feather and hammer simultaneous drop hides the properties inherent to a drop in atmosphere. External atmosphere differentiates the feather and the hammer and shows it's otherwise hidden basic property against wind resistance. Seeing it only on the moon would all make us think they were the same.
You have to apply the device to see what the device's internal characteristics are. That can be done by black box analysis. Or if we do another analysis outside its normal application and peered inside instead to "measure" what's going on in there - the White Box case would tell us and predict to us what it will do in atmosphere and no atmosphere before we even begin applying it. Let's see: Broad light fabric attached to weight and we know F=ma and Force up will equal Force down. In resistive atmosphere Force up is real and we might even know it by calculation beforehand. In a vacuum Force up = 0 and nothing is even there to open it. So all forces are straight down to a parabolic descent.
I see it as a viewpoint of believing you can predict the outcome of distortion in pentodes with varying plate loads without plate curves, but more basic physics that concludes calculable plate curves. And then do for structure analysis for triodes as well. And someone has written a computer program that will take triode structures and predict the plate curves from that and some other pre-known properties about the metals and coatings and what the heater is expected to see, etc., etc. And then we have insight into what triode curves are based on and what pentode curves are based on. No need to build the circuit to predict the basic outcome.
An analogy is High Frequency Structure Simulator software, or HFSS. It allows you to draw a metal and dielectric full structure of however complicated you can get to in practical terms (hint, simplest geometries only, please). Then you can run the calculations and see what the EM fields do when sourced at one end through various frequency sweeps or just one point. This has helped many designers make some weird microwave accessories to better perfection, such as couplers. You now have a White Box to add into another application surrounding it and predict its basic performance, such as return loss, insertion loss, coupled loss, etc. Very helpful to add to further powerful microwave simulators of full circuits. But the circuits are only good as the designed couplers and those are based on known physical modeling and then making them as close as the model. So again, we see beforehand what kind of "in circuit" "NFB" will take place as we vary the circuit. Seeing it in action after the fact and modeling only after the build and measurement of the device and ignoring the structure simulator is a valid course of action and viewpoint on how you want to see that Black Box model. Just ignore the insides that do have loops and feedback all in there and how it became optimized by a more fundamental physical model. Or know that it was only this good because it "Clear Box" peered inside and did do that modeling work for a more precise design and a more predictable Black Box output spec.
So this "new wave" is going on now. It's not enough to design a resistor by over processing with spirals as it will increase costs beyond the competitors and diminish performance. It can be designed and built by more stringent first pass controls and predictions from physical structural models. And then it really does matter that you know exactly how much NFB the device will cause in the circuit and what that is "internally" the "internal NFB" is, for lack of a better term.
Now we need to differentiate those terms, like maybe the expected resistance, inductance, and capacitance along with what distortions are in it from nonlinear dielectrics in use from the device itself and what it will expect to do in a normal simulator of a full circuit it's used in.
This is why I cannot dismiss the internals in knowledge and its own internal circuit that might have that wrongful by a tad "internal NFB" when placed in the circuit that shows what its properties are when applied for its intended use.
There is just a new lower level of simulation that's happening, and has happened for many years actually. Parts need optimization by computer, too.
Or am I still way off base here? I don't think so. Maybe I am not understood what I am trying to say. NFB requires more than one part and it needs stimulus, too. A device with "internal NFB" is assumed that the property of that poor term will come out in the circuit (environment) it was intended to be in.
-kurt
Interesting points, Kurt. This discussion has probably gotten too esoteric to be of interest to most readers, so I'll try not to go on too long.
I agree that "white box" models based on direct internal testing or knowledge of device physics can help us do more accurate simulations and hopefully create better designs.
I'm a bit hung up on whether the "triode internal NFB" model is really a white box model, or if it's just another black box model based on externally observable behavior that doesn't really tell us anything useful about how the triode works internally.
The paper that was discussed on r.a.t., as I recall, wasn't derived from the physics equations (i.e., Child-Langmuir). I do know if you crack open a genuine triode and go looking for that hidden screen grid to do white box testing, you won't find it.
So, let's not pretend the triode has a fictitious internally connected screen grid. Instead, let's treat it as a test probe that we use, in fact or in a thought experiment, to disconnect the hypothetical feedback loop inside the tube. I can't claim that this is really all that different from opening up a feedback amplifier and using a test probe to short out an internal feedback node for just the same purpose. Except, of course, in the case of the amplifier the internal feedback node is a tangible internal structure, whereas in the triode it's harder to visualize. But to a mathematician, I suppose, it's clear enough.
Getting back to the introduction to my long essay, I can't argue that there is no way to model internal NFB in the triode. I just don't see what you gain from it versus, say, more detailed physics-based models. I know that internal negative feedback is a good way to model grid-plate capacitance, for instance, but I've never seen a practical problem where the kind of feedback model we're talking about here made the solution easier.
I know that Mark Kelly, whom I respect as a math genius with a strong interest in triode modeling, looked at this question and came away unimpressed. Though I've had some strenuous arguments, notably with one fellow from Australia, nobody's ever framed it in such a way that I said to myself, "Holy moly, now I see da light!"
In the end, I'm waiting for people to take this triode internal NFB thing and start solving problems with it that nobody's been able to solve by conventional methods. Or taking previously hard problems and making them simple. That will convince me.
-Henry
After reading thourgh lots of the triode-has-internal-FB writings it seems for the most part to be a 'proof' that the pentode is bad.
cheers,
Douglas
Friend, I would not hurt thee for the world...but thou art standing where I am about to shoot.
Post a Followup: