Home |
Search |
Today's Posts |
#121
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
"Chris Hornbeck" wrote in message ...
Maybe what's not obvious is that the small-signal linear model is the crux of the biscuit. Large-signal non-linear extrapolations are way, way down the road in this kind of discussion. No non-linear large-signal models can be extrapolated from linear small-signal models(*). The twain don't meet. But what *should* be obvious from your many posts is that there is no linear model that includes feedback. Maybe a non-linear model might? Out of my field, unfortunately, so can't comment. I believe Byrns' point is that a three-halves power transconductance whose input is the sum of Vg and Vp / mu is one such model. There's no general way to represent the large-signal model in terms of an independent transconductance and admittance. But we don't need to solve the general case to disprove the feedback proposition. Because there are no cross-terms in the triode equation, if you set the condition that Ip is constant, the three-halves power exponent goes away and Vp becomes a linear function of Vg. Then gm has no dependence on Vp and rp has no dependence on Vg. This is why all the plate curves have the same shape. You can derive the Norton model assuming a nonlinear transconductance that is a one-half power function of Vp and a nonlinear admittance that is a negative one-half power function of Vg. The product of the two is a constant, equal to mu. The model meets the requirements that it has no feedback and also that it expresses the Child-Langmuir equation. This is particularly revealing since CCS operation is the exact condition under which the triode feedback proposition claims that feedback is at its maximum. It doesn't mean you can't model the triode as a transconductance wrapped in a feedback loop, but only that it isn't mandatory to do so. I know Byrns will hate this answer, because he hates everything I say as a matter of principle. One possible objection is that with constant current operation, this model gives no way to demonstrate externally that the transconductance and admittance correspond to Child-Langmuir. They could just as easily be perfectly linear, or any pair of functions whose product is constant. I would respond by arguing that this precisely proves my point. There is no external way to prove that there is negative feedback inside the triode, either. Any such assertion is just begging the question since it presumes the conclusion from the outset. As I keep saying, these models are just abstractions and you can come up with one to "prove" just about any internal representation of the tube. SPICE models, in particular, are intended purely for numerical simulation and are even further removed from reality than Norton and Thevenin. I think it's interesting that Byrns has little or no interest in SPICE models that take into account subtle triode distortions. A SPICE model that doesn't take into account subtle tube distortions is like a fashion model with a bad figure. The one doesn't fit the data, and the other doesn't fit her clothes. Both are ugly. -Henry |
#122
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
"Ian Iveson" wrote in message k...
Ipk = c * (u * Vgk + Vpk) ^ 1.5 This is a "feedback equation" only in your imagination. This is an example of what happens when you get so attached to an idea that it blinds you to any other ideas, as obvious as they may be. You can interpret this equation as a non-linear transconductance, Ip(Vg), with the addition of negative feedback, Vp / mu, from the plate. What none of the triode-feedback crowd wants to see is that you can just as well interpret it as a non-linear admittance, Ip(Vp), with the addition of a forward control mechanism, Vg * mu. You just don't need feedback to model the tube this way. That's because the alleged plate feedback is subsumed into the definition of the admittance. In other words, if you start with the premise that the triode is a pentode, you need to presume there is a fictitious screen grid supplying feedback from the plate to explain its behavior. But if you start with the premise that the triode is a diode, all you need to add is a control grid. I would argue, both physically and historically, that the statement "A triode is a diode with a control grid" makes a hell of a lot more sense than "A triode is a pentode with a fictitious screen grid connected to the plate." I think the diode interpretation follows much more naturally from the actual physical embodiment of the tube. Is there really anything more that needs to be said? -Henry |
#123
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
"Henry Pasternack" wrote in message ...
You can derive the Norton model assuming a nonlinear transconductance that is a one-half power function of Vp and a nonlinear admittance that is a negative one-half power function of Vg. Before Byrns and Jute get hysterical, let me say I inadvertently reversed Vp and Vg in that sentence. -Henry |
#124
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
In article ,
"Henry Pasternack" wrote: "Ian Iveson" wrote in message k... Ipk = c * (u * Vgk + Vpk) ^ 1.5 This is a "feedback equation" only in your imagination. This is an example of what happens when you get so attached to an idea that it blinds you to any other ideas, as obvious as they may be. You can interpret this equation as a non-linear transconductance, Ip(Vg), with the addition of negative feedback, Vp / mu, from the plate. What none of the triode-feedback crowd wants to see is that you can just as well interpret it as a non-linear admittance, Ip(Vp), with the addition of a forward control mechanism, Vg * mu. You just don't need feedback to model the tube this way. That's because the alleged plate feedback is subsumed into the definition of the admittance. Henry, I have little doubt that you are correct about this, I am just not smart enough to be able to figure out the equations that define the operation of your "non-linear admittance, Ip(Vp), with the addition of a forward control mechanism, Vg * mu". Could you please post the defining equations. Has anyone else out there been able to work out the actual details of the model Henry is suggesting? In other words, if you start with the premise that the triode is a pentode, you need to presume there is a fictitious screen grid supplying feedback from the plate to explain its behavior. "If you start with the premise that the triode is a pentode" why would "you need to presume there is a fictitious screen grid supplying feedback from the plate to explain its behavior"? Why not just assume that the actual screen grid of the pentode is connected to the plate, why would a "fictitious screen grid" be needed for this odd view? Actually I think you are twisting this view around backwards, the actual view is that the negative feedback in a triode can be defeated by inserting a screen grid in the triode, between the control grid and the plate, and connecting the screen grid to a source of fixed potential. But if you start with the premise that the triode is a diode, all you need to add is a control grid. I think that's exactly how most everyone believes you create a triode, I haven't heard anyone propose building a triode by adding a second screen grid to a pentode, what is your point with this idea? Regards, John Byrns |
#125
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
"John Byrns" wrote in message ...
"If you start with the premise that the triode is a pentode" why would "you need to presume there is a fictitious screen grid supplying feedback from the plate to explain its behavior"? Why not just assume that the actual screen grid of the pentode is connected to the plate, why would a "fictitious screen grid" be needed for this odd view? You're describing a real pentode strapped with external screen grid feedback. An actual triode has no screen grid so if feedback depends on a screen grid, it has to be fictitious. -Henry |
#126
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
Henry Pasternack wrote: "John Byrns" wrote in message ... "If you start with the premise that the triode is a pentode" why would "you need to presume there is a fictitious screen grid supplying feedback from the plate to explain its behavior"? Why not just assume that the actual screen grid of the pentode is connected to the plate, why would a "fictitious screen grid" be needed for this odd view? You're describing a real pentode strapped with external screen grid feedback. An actual triode has no screen grid so if feedback depends on a screen grid, it has to be fictitious. -Henry One can make a model where there is a screen in a triode and you have a shunt FB network between anode to the space charge and from grid to space charge, but there is no need. We surely all know that the anode voltage controls Ia in a diode, or in a triode with a fixed Eg with respect to Ek. Notice how in diode operation or operation of triode in diode mode that input signal is to the anode, and you cannot increase Ea without an increase in Ia. In the diode operation of the triode case where input is to the anode, we do not have to have a situation with always increasing Ia with a rise in Ea because we can make Eg more negative with respect to Ek and thus make Ia any value we like; Ia may constant with rising Ea, or even become less; ie, the anode resistance is negative but only if we hooked up an external positive current FB network to counter the negative voltage FB we are stuck with in the triode case. If one places a CCS as a cathode bias resistance in a triode, then any rise in Ea will give a constant Ia. And if Ea rises then Ek rises by an amount = Ea / µ where the µ is the amp factor of the triode, ie, after the NFB is taken into account for its operation. but during normal triode operation where the input is not to the ande, but to the grid, and output is from the anode instead, then the output voltage has an effect on Ia which opposes the effect of the grid. And the triode allows us to use a CCS as the anode load, something impossible with a diode configuration, and the triode then becomes a pure voltage gain device with no current change to the idle bias condition. It is then we are tempted to model the triode as a very high gain device such as a pentode were it not for the anode arm of electrostatic control of voltage division. Indeed we can model the triode as some kind of pentode but with the two arms of opposing or shunting electrostatic fields which sum at the space charge and either control Ia jointly, or control anode voltage without any Ia change. Once the ficticious pentode model is chosen for a tube like 6SN7, and with a shunt NFB network with equivalant resistors of say 200k grid to space charge and 2M space charge to anode, or 1 : 20, then it doesn't matter what the gain of the ficticious pentode is without the anode FB as long as its way above the triode gain. NFB makes variations in open loop gain tend to disappear and closed loop gain depends purely on shunt NFB resistance values, or in the real case of a triode, the electrostatic effects of anode and grid on the electrons at the space charge. Thse two interactive electrostatic effects are determined by realtive electrode distances and grid wire winding pitch, and when the recipe for a 6SN7 is accurately adhered to anywhere one always gets the same type of triode because its controlled by the same amount of internal NFB. I might also say that one should consider a load such as a transformer connected to a triode. The transformer would normally try to produce some harmonic distortion which does in fact appear at the anode of any type of driving device with a finite source resistance. Not only does the vacuum tube have some distortion generation of its own, but loads such as transformers also cause some distortion and the lower the source resistance feeding the transformer, the lower the distortion, so triodes with their own internal NFB are favoured unless we use external loop NFB to reduce the apparent source resistance of pentodes ( or beam tetrodes ). The distortion signal voltage isn't at the input grids, so the anode distortion signal will oppose its own production and where a +ve going Dn signal occurs, Ia increases to oppose it, and vice versa for a -ve going Dn signal. If that isn't NFB, I don't know what it is, because the same NFB active action occurs in shunt or series global NFB loops set up around many types of amplifiers to reduce the THD and reduce Rout. Power amps with pentode output tubes have about the same Rout with 15dB of global NFB as when simply triode connected with screens taken to anodes. One can vary the amount of internal FB of a multi grid tube by choosing the ratio of B+ to screen turns relative to B+ to anode turns in the OPT. About 40% screen taps allows enough internal NFB to exist to reduce THD of an EL34 etc to about the same as where triode connected, but not cause a loss of power range due to grid current limitations. But one may go further with such a connection and use say an extension of turns in the OPT to have say 50% MORE signal voltage of the same phase applied to the screen than in the triode connection. Ra and effective µ is further reduced below triode values as a direct result of increasing the local triodic NFB. There is a further limitation of output power and thus efficiency, so nobody does this commercially. Alternatively, the gain of a pentode or tetrode and the high Ra can be increased with positive voltage FB applied by reversing the usual UL connections. This usually makes an amp unstable, but when applied within boundaries of very well designed circuits and OPT, the PFB thus applied makes the open loop gain higher so that the global NFB applied becomes more effective, and THD and Rout is lower despite the PFB. Such measures along with positive current FB loops to make amp Rout low, or zero ohms, or even a negative resistance, ie, more output voltage with lower load, are not employed afaik in any modern amp because of the troble with stabilisation and the fact that such jiggery pokery with FB is frowned upon by nearly all audio cognescenti these days. Analysing the UltraLinear connection should lead ppl to the idea that there is NFB application possible within output tubes in a local circuit, and that it can be varied widely between almost no internal NFB at all with pure pentode tetrode connection, or up to triode connection where the screen has the same signal as the anode, and transmits its effect on behalf of the anode fully. Note that any multigrid could be made into a real triode but its anode would have to occupy the position of same distance as the screen is from the cathode grid structure for the same triode eyiypentfwhere oin s g a triode µ. This would make anodes need to be smaller dia, and perhaps more difficult to keep cool and so the bigger anodes of power pentodes and tetrodes is to some advantage. The multi grids are somewhat wild and have high untamed Rout and THD and are in need of control, hence about a total of 20dB of one or more loops of NFB of some kind is required to tame the multis. The most common form used since 1955 has been the UL NFB local balanced connection combined with global NFB. Patrick Turner. |
#127
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
In article ,
"Henry Pasternack" wrote: "Chris Hornbeck" wrote in message ... Maybe what's not obvious is that the small-signal linear model is the crux of the biscuit. Large-signal non-linear extrapolations are way, way down the road in this kind of discussion. No non-linear large-signal models can be extrapolated from linear small-signal models(*). The twain don't meet. But what *should* be obvious from your many posts is that there is no linear model that includes feedback. Maybe a non-linear model might? Out of my field, unfortunately, so can't comment. I believe Byrns' point is that a three-halves power transconductance whose input is the sum of Vg and Vp / mu is one such model. There's no general way to represent the large-signal model in terms of an independent transconductance and admittance. But we don't need to solve the general case to disprove the feedback proposition. Because there are no cross-terms in the triode equation, What exactly do you mean by a "cross-term"? I believe the triode equation I posted did have "cross-terms". These "cross-terms" are the result of the three halves power nonlinearity in the model, at least there are what I think of as "cross-terms". if you set the condition that Ip is constant, the three-halves power exponent goes away and Vp becomes a linear function of Vg. Then gm has no dependence on Vp and rp has no dependence on Vg. Agreed. This is why all the plate curves have the same shape. I don't see how you get here from there, and I don't agree that all the plate curves have the same shape. You can derive the Norton model assuming a nonlinear transconductance that is a one-half power function of Vp and a nonlinear admittance that is a negative one-half power function of Vg. I am not following where the one-half power and negative one-half power functions come from in your Norton model? Ignoring the problem with the "cross-terms" in the triode equation, wouldn't the nonlinear admittance at least have to be a three halves power function so that Child's diode law at least applied to the output when the grid voltage is set to zero? The product of the two is a constant, equal to mu. The model meets the requirements that it has no feedback and also that it expresses the Child-Langmuir equation. I agree that there is no feedback in your model, however I don't see how it in any way expresses the Child-Langmuir equation, unless I am missing something with your negative half power function? This is particularly revealing since CCS operation is the exact condition under which the triode feedback proposition claims that feedback is at its maximum. It doesn't mean you can't model the triode as a transconductance wrapped in a feedback loop, but only that it isn't mandatory to do so. I at least have never claimed that the feedback model is the only way to model the triode, I have repeatedly stated this, yet you persist in implying that I have said feedback is necessary to model the triode. What I have said is that you hadn't as yet presented your non feedback triode model. You have now done that above, although I find the half power function you used to be quaint, although that may simply be a refection of my ignorance. It is also interesting that your non feedback model does not even model non linearity of the triode as well as my feedback model does, given your earlier criticism of my model for its failure to accurately model the triodes nonlinearity. I know Byrns will hate this answer, because he hates everything I say as a matter of principle. One possible objection is that with constant current operation, this model gives no way to demonstrate externally that the transconductance and admittance correspond to Child-Langmuir. They could just as easily be perfectly linear, or any pair of functions whose product is constant. You said it not me! By the way how did you choose that half power function for your model, rather than a linear or say a three halves power function? I would respond by arguing that this precisely proves my point. There is no external way to prove that there is negative feedback inside the triode, either. And exactly where did I make the claim that you could tell by external observations if there was or wasn't feedback inside the triode. Go back and read what I have posted, this whole sub thread started when I said I had a triode model that incorporates negative feedback, that is a fair distance from saying feedback is the only way to model the triode. IIRC, and I may not, my statement was in response to a poster who was trying to say that the triode could not be modeled with feedback. Any such assertion is just begging the question since it presumes the conclusion from the outset. Say again, what conclusion are you talking about here? As I keep saying, these models are just abstractions and you can come up with one to "prove" just about any internal representation of the tube. I don't see how any model "proves" the internal operation of the tube? SPICE models, in particular, are intended purely for numerical simulation and are even further removed from reality than Norton and Thevenin. I would disagree, Norton and Thevenin are further removed from reality than even spice models, almost by definition. I think it's interesting that Byrns has little or no interest in SPICE models that take into account subtle triode distortions. Why is it necessary that I have an interest in these more accurate models, it's not even like your proposed model is more accurate than mine is? Remember that this sub thread started with my claim that I had a spice model that was based on feedback, I made no claims for its accuracy. A SPICE model that doesn't take into account subtle tube distortions is like a fashion model with a bad figure. What defines a "bad figure"? To a great extent a bad figure is in the eye of the beholder. The one doesn't fit the data, and the other doesn't fit her clothes. Both are ugly. How does a "bad figure" imply that her clothes don't fit, or that she is ugly? There is no guarantee that a woman who doesn't have a "bad figure" will automatically fit her clothes. In either case the clothes must be made to fit the figure as it exists, either good or "bad" whatever that is. Regards, John Byrns |
#128
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
In article ,
"Henry Pasternack" wrote: "John Byrns" wrote in message ... "If you start with the premise that the triode is a pentode" why would "you need to presume there is a fictitious screen grid supplying feedback from the plate to explain its behavior"? Why not just assume that the actual screen grid of the pentode is connected to the plate, why would a "fictitious screen grid" be needed for this odd view? You're describing a real pentode strapped with external screen grid feedback. An actual triode has no screen grid so if feedback depends on a screen grid, it has to be fictitious. The point you are getting confused on is that a screen grid is not added to a pentode to add feedback, a screen grid is added to a triode to defeat the internal feedback, or what ever it may actually be that is inside a triode which causes it to have a relatively low value of Rp. Regards, John Byrns |
#129
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
"John Byrns" wrote in message ...
The point you are getting confused on is that a screen grid is not added to a pentode to add feedback, a screen grid is added to a triode to defeat the internal feedback, or what ever it may actually be that is inside a triode which causes it to have a relatively low value of Rp. No, I'm not confused, but I am bored with this discussion. And I have other things to do. I will tell you that the one-half power exponent comes from taking the derivative of the Child-Langmuir equation. -Henry |
#130
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
In article ,
"Henry Pasternack" wrote: "John Byrns" wrote in message ... The point you are getting confused on is that a screen grid is not added to a pentode to add feedback, a screen grid is added to a triode to defeat the internal feedback, or what ever it may actually be that is inside a triode which causes it to have a relatively low value of Rp. No, I'm not confused, but I am bored with this discussion. And I have other things to do. I will tell you that the one-half power exponent comes from taking the derivative of the Child-Langmuir equation. That just confirms the fact that you are badly confused, I do however agree that this is a boring discussion, and we probably all have other things with which we could more profitably occupy our time. Regards, John Byrns |
#131
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
"John Byrns" wrote in message ...
That just confirms the fact that you are badly confused... Yes, and we all know that I'm a child molester. Talk is cheap, Byrns. Now, go away or I shall taunt you a second time. -Henry |
#132
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
In article ,
"Henry Pasternack" wrote: "John Byrns" wrote in message ... That just confirms the fact that you are badly confused... Yes, and we all know that I'm a child molester. Talk is cheap, Byrns. Now, go away or I shall taunt you a second time. Go for it. Regards, John Byrns |
#133
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
Andre Jute wrote: In an effort to be agreeable, I tried hard to give you negative feedback inside the tube as an explanation of the overwhelming superiority of triodes (or trioded pentodes) for audio reproduction, among other reasons because NFB is accessible to many who belong on RAT and is a genetic deformity of the silicon scum whose only purpose on RAT is dissension. NFB is what the silicon slime abuse to make their inadequate components sound passable, and what even tubies inspired by the age of sophisters and cost-accountants use to linearize pentodes. NFB thus has a base level of familiarity which gives it a head start in any black box model intended to explain something to diplomaed quarterwits among the silicon slime as well as the better-educated kibbitzers in my own camp. But fine, you want to reject my explanation, then you must offer a better reason to explain why triodes are such superior amplification devices to anything else, so much more pleasing to the ear, so much more accurate to the cultivated taste. Despite my cracks about the metaphysics of tubes, there *has* to be an electrical reason for the superiority of triodes. But sure, all kinds of input is welcome. Andre Jute There are now 136 messages in the thread I started with the post above. In all that bandwidth, all that shouting, all that bad temper, no one has taken up the challenge to offer a better reason [than internal negative feedback] to explain why triodes are such superior amplification devices to anything else, so much more pleasing to the ear, so much more accurate to the cultivated taste All that we have seen is that Patrick Turner and I have solidified our faith in our transconductance model substantially and that John Byrns, arriving late to the thread (a wise man to absent himself from this boring round about the mulberry bush), has further emphasized the native appeal of the transconductance model graphically. None of us deny there may be another explanation. We just say that NFB inside the triode seems a logical explanation, and the easiest to assume since we are already familiar with the mechanism and math, and because it so readily relates the triode to the less attractive pentode. That's a whole bunch of persuasive practical reasons already. We don't need faith for our explanation, just the knowledge that it works for us. Against that, you mantra-chanting gaggle of naysayers offer us what precisely? Some basic principles that Plodnick, having regaled us with unlikely tales of being Terman's annointed successor, can't even copy out right and refuses to transform into any useful form, probably because he doesn't know how. That's hardly a productive use of so much bandwidth and time. The depth of the supporting arguments for our position, and the experience backing it, is also very striking when compared to the thin, unnourishing broth of half-digested and, in some cases halfbaked or even plain perverse, theories you're throwing against our solid bulwark. Andre Jute Habit is the nursery of errors. -- Victor Hugo |
#134
Posted to rec.audio.tubes
|
|||
|
|||
Does anyone know what happened to the audio asylum?
|
#135
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
Andre Jute wrote:
You, Andre, I will answer. Patrick can go **** himself, since he's determined to answer my technical discussion of a subject with mindless insults and unsupported criticisms. Phil wrote: Patrick Turner wrote: Try reading a few more books about triodes rather than seeking some NFB de-bunking premise from Henry. With all due respsect to Henry, he ain't the world's authority on vacuum tube theory. He could be the devil himself, but that's irrelevant, along with his knowledge, when it comes to the quality of his arguments on this matter. You don't "disprove" a theory by saying the author is stupid or evil. You ignore the characteristics of the author altogether, and focus on the accuracy, or lack thereof, of the theory itself. Reasoning 101. Unless of course the party in question has a track record of deliberately lying. In the case of Henry Pasternack, often referred to RAT as Pompass Plodnick, Google archives show a long history of Pasternack lying on professional and other matters for the sake of "winning" some argument, and of Pasternack committing other unscientific and immoral acts, for which his only excuse is, once the further lies are stripped away, in Pasternack's own words, "my zeal to flame Andre". In this particular case Pasternack came here hoping to have a big fight with me but I merely patronized him a little and sent him on his way with a flea in his ear; he was stuck with Patrick. Because Pasternack came for me first, and because of his history of lying on professional matters for personal gratification, and because Pasternack's first post on the subject (to Chris Hornbeck) was ambiguous, we don't actually know whether he really believes what he says now, or whether his hatred of me has once more painted him into another corner which he will now try to justify with a berm of math. Those new to RAT who wish to see earlier examples of Pasternack lying on professional matters for personal gratification should look up the case where Pasternack told a newbie not to listen to me when I advised a primary impedance on his output transformer 2*Rp or higher; Pasternack told him instead to choose an output impedance equal to the plate resistance. Read that again. Pasternack surely knew that the primary impedance should be twice or more the plate resistance but he lied about it to a newbie "in my zeal to flame Andre". There are hundreds of further posts in Pasternack which tried to justify his Zo = Rp stance but eventually John Byrns nailed Pasternack's hide to wall. Other examples are plentiful, and I have already in this thread given references to a URL that proves Pasternack's contempt for the scientific method. Furthermore, Pasternack in the throes of his hatred will commit totally immoral acts. He ran with Michael LaFever's Magnequest Scum, who flooded the single driver conference with graphic homosexual pornography sent in my name in an effort to drive me out; they were not there before I came, they were not there after I left. Some of that was traced to Pasternack's server. Even worse is the case of the two little girls of an Italian engineer who built one of my designs. He accused Pasternack of sending graphic homosexual filth to his computer, where his two little girls saw it. He didn't even know who Pasternack was when he traced the filth to Pasternack. Pasternack's Magnequest Scum associate Bob Chernofsky said on the Joenet (there's an archieve at Harvard if you want to look it up; search for Sound List) that Pasternack did it because he was bored with not being able to get at me directly. Do you really want to hold this scumbag Pasternack up to us as an impartial fount of engineering wisdom? I'm no fan of the Magnequest gangster tactics -- remember my all out fights with "the nuke-man" years ago -- but if Henry is right on the issue of NFB in triodes, then he's right. Now, I'm not saying he definitely is right, but I am saying that we should look at his arguments and his reasoning, alone, since it is intellectually invalid to claim that "since X is evil, therefore his claim that 2 + 2 = 4 must be incorrect." Get real, Phil. We know better, and it is up to newbies like you to inform themselves before they goof up dumb opinions. Andre Jute Stop bleating. Please, please, please give me the Silence of the Lambs. PS Do I need to explain that we shall know by your response to my sharp remarks how steady your judgement is? Do I need to explain that you don't refute scientifically valid reasoning by "cutting your opponent to shreds?" That proves nothing, ever. Now, it certainly is an indication of potential problems, but that just means you examine the reasoning carefully, which we were doing already. Patrick wanted to "refute" my point that he could not prove that NFB is responsible for a triode's low Rp, by saying that I could not prove that NFB was not responsible. Yes! That's what I've been trying to say for a long time now! I can't prove either one, and neither can he, and as far as I know, neither can you. But tell me, unless some example appears that requires us to the NFB model to obtain accurate predictions or good designs, wouldn't the simpler resistance + battery model lead to fewer errors and less confusion? Don't you believe that triodes with an output Rp of 800 ohms and no "standard" local or global feedback sound better than a pentode with an output Rp of 800 ohms from some type of feedback? If the pentode simply got its 800 ohms of Rp by adding the same amount of feedback as the triode already has, then why does the triode sound better? If you have no explanation, then don't you run the risk of convincing some people by your use of the triode NFB model that they might as well use a pentode with feedback, "since it will sound the same as a triode with its internal feedback?" Do you see what I mean? The triode NFB model, in addition to being just as unprovable as the resistance model, and more difficult to understand and therefore use, has a tendency to mislead people into believing that "pentodes sound the same as triodes, since they both use NFB to get low Rp." They don't sound the same, and for all you know, it's because NFB screws up the sound, and triodes do *not* have internal NFB! All disadvantages, and no advantages, as far as I can see. Unless, of course, you simply want to *argue* that "pentodes are too just as good sounding as triodes." Now, Andre, surely you of all people wouldn't want to lie to and mislead audio lovers, like some other people you occasionally mention, right? But by all means, the moment you prove that the resistance model is flawed, and that only the NFB model can correctly model triode behavior in the audio realm, I will say we should start using it. So far, however, the only difference I can see is that the NFB model fails to predict the sonic difference between triodes and pentodes, which is an indication -- although not a proof -- that the NFB model is the one which is flawed. Make sense? Phil |
#136
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
"Phil" wrote in message ...
You, Andre, I will answer. Patrick can go **** himself, since he's determined to answer my technical discussion of a subject with mindless insults and unsupported criticisms. I politely but firmly demand that you do not respond to or repost standerous statements made about me. If you have any question about this, please contact me in person so we can discuss. -Henry |
#137
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
Phil wrote: An attempt at reason... generally successful to a reasonable reader. Phil: I would remind you of the adage about micturation windwards. Peter Wieck Wyncote, PA |
#138
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
Phil wrote: Andre Jute wrote: You, Andre, I will answer. Patrick can go **** himself, since he's determined to answer my technical discussion of a subject with mindless insults and unsupported criticisms. Unfortunately Phil suffers from such sensitivity that he is offended when challenged intellectually. The only way you can really learn is to put your ego in the dustbin, and assume you are wrong until you have proved otherwise, and you simply ain't proved anything to me yet, so i have no intention to **** myself as you suggest. Phil wrote: Patrick Turner wrote: Try reading a few more books about triodes rather than seeking some NFB de-bunking premise from Henry. With all due respsect to Henry, he ain't the world's authority on vacuum tube theory. He could be the devil himself, but that's irrelevant, along with his knowledge, when it comes to the quality of his arguments on this matter. You don't "disprove" a theory by saying the author is stupid or evil. You ignore the characteristics of the author altogether, and focus on the accuracy, or lack thereof, of the theory itself. Reasoning 101. Unless of course the party in question has a track record of deliberately lying. In the case of Henry Pasternack, often referred to RAT as Pompass Plodnick, Google archives show a long history of Pasternack lying on professional and other matters for the sake of "winning" some argument, and of Pasternack committing other unscientific and immoral acts, for which his only excuse is, once the further lies are stripped away, in Pasternack's own words, "my zeal to flame Andre". In this particular case Pasternack came here hoping to have a big fight with me but I merely patronized him a little and sent him on his way with a flea in his ear; he was stuck with Patrick. Because Pasternack came for me first, and because of his history of lying on professional matters for personal gratification, and because Pasternack's first post on the subject (to Chris Hornbeck) was ambiguous, we don't actually know whether he really believes what he says now, or whether his hatred of me has once more painted him into another corner which he will now try to justify with a berm of math. Those new to RAT who wish to see earlier examples of Pasternack lying on professional matters for personal gratification should look up the case where Pasternack told a newbie not to listen to me when I advised a primary impedance on his output transformer 2*Rp or higher; Pasternack told him instead to choose an output impedance equal to the plate resistance. Read that again. Pasternack surely knew that the primary impedance should be twice or more the plate resistance but he lied about it to a newbie "in my zeal to flame Andre". There are hundreds of further posts in Pasternack which tried to justify his Zo = Rp stance but eventually John Byrns nailed Pasternack's hide to wall. Other examples are plentiful, and I have already in this thread given references to a URL that proves Pasternack's contempt for the scientific method. Furthermore, Pasternack in the throes of his hatred will commit totally immoral acts. He ran with Michael LaFever's Magnequest Scum, who flooded the single driver conference with graphic homosexual pornography sent in my name in an effort to drive me out; they were not there before I came, they were not there after I left. Some of that was traced to Pasternack's server. Even worse is the case of the two little girls of an Italian engineer who built one of my designs. He accused Pasternack of sending graphic homosexual filth to his computer, where his two little girls saw it. He didn't even know who Pasternack was when he traced the filth to Pasternack. Pasternack's Magnequest Scum associate Bob Chernofsky said on the Joenet (there's an archieve at Harvard if you want to look it up; search for Sound List) that Pasternack did it because he was bored with not being able to get at me directly. Do you really want to hold this scumbag Pasternack up to us as an impartial fount of engineering wisdom? I'm no fan of the Magnequest gangster tactics -- remember my all out fights with "the nuke-man" years ago -- but if Henry is right on the issue of NFB in triodes, then he's right. Now, I'm not saying he definitely is right, but I am saying that we should look at his arguments and his reasoning, alone, since it is intellectually invalid to claim that "since X is evil, therefore his claim that 2 + 2 = 4 must be incorrect." Get real, Phil. We know better, and it is up to newbies like you to inform themselves before they goof up dumb opinions. Andre Jute Stop bleating. Please, please, please give me the Silence of the Lambs. PS Do I need to explain that we shall know by your response to my sharp remarks how steady your judgement is? Do I need to explain that you don't refute scientifically valid reasoning by "cutting your opponent to shreds?" That proves nothing, ever. Now, it certainly is an indication of potential problems, but that just means you examine the reasoning carefully, which we were doing already. Patrick wanted to "refute" my point that he could not prove that NFB is responsible for a triode's low Rp, by saying that I could not prove that NFB was not responsible. Yes! That's what I've been trying to say for a long time now! I can't prove either one, and neither can he, and as far as I know, neither can you. But tell me, unless some example appears that requires us to the NFB model to obtain accurate predictions or good designs, wouldn't the simpler resistance + battery model lead to fewer errors and less confusion? Don't you believe that triodes with an output Rp of 800 ohms and no "standard" local or global feedback sound better than a pentode with an output Rp of 800 ohms from some type of feedback? If the pentode simply got its 800 ohms of Rp by adding the same amount of feedback as the triode already has, then why does the triode sound better? If you have no explanation, then don't you run the risk of convincing some people by your use of the triode NFB model that they might as well use a pentode with feedback, "since it will sound the same as a triode with its internal feedback?" Do you see what I mean? The triode NFB model, in addition to being just as unprovable as the resistance model, and more difficult to understand and therefore use, has a tendency to mislead people into believing that "pentodes sound the same as triodes, since they both use NFB to get low Rp." They don't sound the same, and for all you know, it's because NFB screws up the sound, and triodes do *not* have internal NFB! All disadvantages, and no advantages, as far as I can see. Unless, of course, you simply want to *argue* that "pentodes are too just as good sounding as triodes." Now, Andre, surely you of all people wouldn't want to lie to and mislead audio lovers, like some other people you occasionally mention, right? But by all means, the moment you prove that the resistance model is flawed, and that only the NFB model can correctly model triode behavior in the audio realm, I will say we should start using it. So far, however, the only difference I can see is that the NFB model fails to predict the sonic difference between triodes and pentodes, which is an indication -- although not a proof -- that the NFB model is the one which is flawed. Make sense? Phil Sorry Phil, but to me you make no sense. This is of course a public forum, and to start any post by telling me to go **** myself won't bring a positive response from me unless YOU can PROOVE there is no NFB in a triode. This is something you have failed to achieve. I don't care about being told to **** myself, and i wish the group to know this, and to know that you cannot seem to understand a triode IMHO. The superiority of triodes for audio amps is due to the NFB that is within them. Its very difficult to get a pentode amp to measure AND sound as well, although many people would fail to identify a pentode amp versus a triode amp in a blind AB test if both were designed to have the same Rout and power ceiling and bandwidth. For valid amplifier comparisons, the playing field must be level. Patrick Turner. |
#139
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
In article ,
Phil wrote: Do I need to explain that you don't refute scientifically valid reasoning by "cutting your opponent to shreds?" That proves nothing, ever. Now, it certainly is an indication of potential problems, but that just means you examine the reasoning carefully, which we were doing already. Patrick wanted to "refute" my point that he could not prove that NFB is responsible for a triode's low Rp, by saying that I could not prove that NFB was not responsible. Yes! That's what I've been trying to say for a long time now! I can't prove either one, and neither can he, and as far as I know, neither can you. But tell me, unless some example appears that requires us to the NFB model to obtain accurate predictions or good designs, wouldn't the simpler resistance + battery model lead to fewer errors and less confusion? While it doesn't prove anything, doesn't the way the electrostatic fields from the grid (input), and the plate (output), interact within the triode to determine the plate current suggestive of negative feedback? Henry Pasternack criticized my feedback based triode model on the grounds that it does not completely capture the fine details of the distortion produced by triodes. Your simpler resistance + battery triode model is even further removed from accurately modeling triode distortions. Even the half power gm + admittance model that Henry himself proposed doesn't do as well at modeling triode distortion as my simple feedback model does. Not only that but unlike my model, Henry's model doesn't even begin to attempt to model the way the grid voltage in a real triode affects Rp and the way the plate voltage affects Gm. Don't you believe that triodes with an output Rp of 800 ohms and no "standard" local or global feedback sound better than a pentode with an output Rp of 800 ohms from some type of feedback? If the pentode simply got its 800 ohms of Rp by adding the same amount of feedback as the triode already has, then why does the triode sound better? If you have no explanation, then don't you run the risk of convincing some people by your use of the triode NFB model that they might as well use a pentode with feedback, "since it will sound the same as a triode with its internal feedback?" Do you see what I mean? How do you know that triodes actually sound better? Has anyone ever actually done head to head listening tests between triode and pentode amplifiers that are otherwise identical except for the configuration of the output stage? I kind of doubt these tests have ever been done, and without them you can't truthfully say that triodes actually sound better than pentodes. For comparative triode vs. pentode listening tests to be valid, the amplifiers being compared must be completely identical except for the configuration of the output stage, this implies the following, among other things: 1.) The triode and pentode tubes used must be as nearly identical in capability as possible. 2.) The feedback applied to the pentode must be local feedback encompassing only the output stage, not overall loop feedback as is typically used in pentode amplifiers. 3.) The input sensitivity and admittance of the pentode with local feedback must be the same as the triode to allow the use of the same driver stage for both amplifiers. 4.) The input admittance requirement rules out using shunt feedback, is that the right name, from plate to grid. Cathode feedback seems the most likely to meet the requirements. Two commercial amplifiers using cathode feedback are the QUAD II and McIntosh "unity coupled" series of amplifiers, but neither of these two amplifiers exactly meets the requirements. The McIntosh amplifier has more feedback in the output stage than is required to simulate the performance of a triode, and the output stage is considerably less sensitive than a triode, requiring a driver stage with considerably more voltage output capability. The QUAD design probably comes close to the required output stage sensitivity and amount of feedback. Unfortunately the QUAD design operates the output tubes in the "Ultralinear" mode, and not as true pentodes. We need a design somewhere between the QUAD and McIntosh, incorporating the lower local feedback of the QUAD with the true pentode operation of the McIntosh. Until you have done comparative listening tests using a pair of amplifiers designed along the lines I have described above, there is no way you can make the claim that triodes sound better than pentodes. The triode NFB model, in addition to being just as unprovable as the resistance model, and more difficult to understand and therefore use, has a tendency to mislead people into believing that "pentodes sound the same as triodes, since they both use NFB to get low Rp." All you have to do to know the resistance model is wrong is to look at some triode curves, also you have not yet shown that it is not true that "pentodes sound the same as triodes, since they both use NFB to get low Rp." They don't sound the same, and for all you know, it's because NFB screws up the sound, and triodes do *not* have internal NFB! What evidence do you have that pentodes with local feedback don't sound the same as triodes? Have you done the required listening tests with a pair of otherwise identical amplifiers as I described above? Until you have done this test you can't legitimately make the claim that "they don't sound the same". All disadvantages, and no advantages, as far as I can see. Unless, of course, you simply want to *argue* that "pentodes are too just as good sounding as triodes." The advantage of pentodes is their greater efficiency, if triodes and pentodes sound the same, why use triodes with their lower efficiency? Now, Andre, surely you of all people wouldn't want to lie to and mislead audio lovers, like some other people you occasionally mention, right? But by all means, the moment you prove that the resistance model is flawed, and that only the NFB model can correctly model triode behavior in the audio realm, I will say we should start using it. The simple resistance model is clearly flawed in that it doesn't capture any of the triodes distortion mechanisms. Even Henry Pasternack's vaguely defined half power Gm & admittance model fails to model triode nonlinearity as well as my simple feedback model does. So far, however, the only difference I can see is that the NFB model fails to predict the sonic difference between triodes and pentodes, which is an indication -- although not a proof -- that the NFB model is the one which is flawed. Make sense? No, you are starting from the assumption that a triode sounds better than a pentode with local negative feedback in otherwise identical amplifiers, you must first demonstrate this to be true. Then even if it is true that triodes sound better, how does that disprove the validity of the NFB model and prove the validity of the "resistance" model? The "resistance" model is clearly flawed because it can't even come close to reproducing triode curves as measured on a curve tracer. Regards, John Byrns |
#140
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
Phil wrote:
You, Andre, I will answer. Patrick can go **** himself, since he's determined to answer my technical discussion of a subject with mindless insults and unsupported criticisms.wrote: [Big snip about the tiresome Pasternack] Patrick wanted to "refute" my point that he could not prove that NFB is responsible for a triode's low Rp, by saying that I could not prove that NFB was not responsible. Yes! That's what I've been trying to say for a long time now! I can't prove either one, and neither can he, and as far as I know, neither can you. I believed you first time. I also believe Patrick. I don't claim to be able to prove the case either way. But the purpose of a model is to reflect reality. The transconductance model, which includes the NFB in mu, in my opinion comes closest. Patrick believes this so firmly that it is an article of faith with him. So what? He's got a lot of bench experience to back up his faith. But tell me, unless some example appears that requires us to the NFB model to obtain accurate predictions or good designs, wouldn't the simpler resistance + battery model lead to fewer errors and less confusion? Don't you believe that triodes with an output Rp of 800 ohms and no "standard" local or global feedback sound better than a pentode with an output Rp of 800 ohms from some type of feedback? Why? For what? Some mystical reason? I think you've mistaken my attitude, Phil. I talk up a storm, and will speculate with anyone on any subject and crack some jokes too along the way, but when it comes down to spending money I'm an ultra-conservative designer, and I stick rigidly to the scientific method. If the pentode simply got its 800 ohms of Rp by adding the same amount of feedback as the triode already has, then why does the triode sound better? If you have no explanation, then don't you run the risk of convincing some people by your use of the triode NFB model that they might as well use a pentode with feedback, "since it will sound the same as a triode with its internal feedback?" Do you see what I mean? No, I don't. And I'm not sure you know what you mean either. You're thinking with the rush of adrenaline Patrick has sent to your head. I have spent ten years explaining why a zero negative feedback amp (actually a zero or low negative feedback class A1 triode or trioded pentode amp) sounds so good. The explanation deals very strongly with subliminal effects of adverse *proportions* of NFB residual artifacts, rather than the total THD. That is, I explain it in terms of the quality of the amp's silence, not just its amount. Thus part of my explanation of the superiority of triodes over pentodes is that pentodes (in common applications, not in my amps) require a level of feedback that drives the proportion of the residual noise which is particularly objectionable to the ear into the stratosphere, while the feedback in the triode actually reduces the objectionable proportion below the threshold. It's a very subtle thing to understand but there is a very great deal of anecdotal evidence to support this viewpoint. Furthermore, we have worked practical examples here on RAT to demonstrate that the means by which I achieve the silence (high voltage, high current, high load) work exceedingly well on all the parameter I consider important -- look up the tables John Byrns worked up when those idiots Bob Chernofsky and Michael LaFevre challenged me to a design contest; search for "Bubbaland 300B". The triode NFB model, in addition to being just as unprovable as the resistance model, and more difficult to understand and therefore use, has a tendency to mislead people into believing that "pentodes sound the same as triodes, since they both use NFB to get low Rp." People who cannot discriminate differentially between the effects of higher harmonic residuals of NFB are not my problem. I have explained it to them in both words and math over the years. They can look up my writings on the subject and educate themselves. Your argument of "ease of understanding and use" holds no water. The added complication of my view rewards us for the additional effort with overwhelming advantages for practical application in the design of good amps. It is because of this understanding of mine that my amps sound so much better than some designed by people who cannot be bothered to put their minds in gear. They don't sound the same, Right. and for all you know, it's because NFB screws up the sound, Absolutely. But you have to go one step further and ask which part of the NFB screws up the pentode sound and how, and that brings you back to my view of differential damage due to higher harmonic residuals. and triodes do *not* have internal NFB! It's still possible, I suppose, but there is too much that your view then leaves unexplained. All disadvantages, and no advantages, as far as I can see. You're just wrong, man. All advantages is what I see. Your model breaks Einstein's rule that the explanation for a thing should not be simpler than is necessary to explain it. An "explanation" that leaves so much unexplained as your resistance model is no explanation at all, merely a framework for raising questions unanswerable in terms of the "explanation", which questions then become the proof of the ineptness of the "explanation". Unless, of course, you simply want to *argue* that "pentodes are too just as good sounding as triodes." Now, Andre, surely you of all people wouldn't want to lie to and mislead audio lovers, like some other people you occasionally mention, right? That is a polemical argument that has nothing to do with the science here. However, to bring it back to the science, it is in fact my contention, made many times here on RAT, backed by ABX tests conducted with professional classical musicians as my test panels, that trioded pentodes in Class A1 with zero negative feedback sound "more natural", i.e. closer to what the musicians expect to hear when playing their own recordings, than anything else including directly heated real triodes (usually 300B). That again tends to support my twin claims that NFB is dangerous to the sound and that there is some NFB inside triodes. But by all means, the moment you prove that the resistance model is flawed, and that only the NFB model can correctly model triode behavior in the audio realm, I will say we should start using it. Why should I prove your model false? Read some Popper, pal. All that is required here is for me to ascertain that my model is scientifically valid and to have discovered in practice that it is more use to me in designing amps than your model. Competing models is a disease of the Pasternack type of thinking, of trying to win arguments rather than to learn something new. As for you supporting my model or not, why should I care? I don't even care whether Patrick or John support my model; I know that if I write to them for technical help on another matter, they will help me regardless of whether I support their view. (That's another strike against Pasternack, incidentally: he begs openly for support for his views, and publicly persecutes those who do not support him. That is patently unprofessional and unscientific behaviour on Pasternack's part.) So far, however, the only difference I can see is that the NFB model fails to predict the sonic difference between triodes and pentodes, which is an indication -- although not a proof -- that the NFB model is the one which is flawed. Make sense? No. In my reinterpretation of the residuals of NFB, the proportional division of THD between the harmonics is predictive of the sonic quality of an amp, as the bare round THD number is not. Thus my model predicts quality, and supports my experimental findings. Phil All in all, Phil, I would say that you have about half a foot to stand on (nobody can disprove your model but you haven't described any true utility for it whereas we know our transconductance model has sound practical utility), and I can understand why Patrick was getting short with you. Don't let it sour your relationship with him. Patrick, though sometimes infuriatingly detailed, is a very helpful, friendly guy. Regards, Andre Jute Visit Jute on Amps at http://members.lycos.co.uk/fiultra/ "wonderfully well written and reasoned information for the tube audio constructor" John Broskie TubeCAD & GlassWare "an unbelievably comprehensive web site containing vital gems of wisdom" Stuart Perry Hi-Fi News & Record Review |
#141
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
John Byrns wrote: In article , Phil wrote: Do I need to explain that you don't refute scientifically valid reasoning by "cutting your opponent to shreds?" That proves nothing, ever. Now, it certainly is an indication of potential problems, but that just means you examine the reasoning carefully, which we were doing already. Patrick wanted to "refute" my point that he could not prove that NFB is responsible for a triode's low Rp, by saying that I could not prove that NFB was not responsible. Yes! That's what I've been trying to say for a long time now! I can't prove either one, and neither can he, and as far as I know, neither can you. But tell me, unless some example appears that requires us to the NFB model to obtain accurate predictions or good designs, wouldn't the simpler resistance + battery model lead to fewer errors and less confusion? While it doesn't prove anything, doesn't the way the electrostatic fields from the grid (input), and the plate (output), interact within the triode to determine the plate current suggestive of negative feedback? Henry Pasternack criticized my feedback based triode model on the grounds that it does not completely capture the fine details of the distortion produced by triodes. Your simpler resistance + battery triode model is even further removed from accurately modeling triode distortions. Even the half power gm + admittance model that Henry himself proposed doesn't do as well at modeling triode distortion as my simple feedback model does. Not only that but unlike my model, Henry's model doesn't even begin to attempt to model the way the grid voltage in a real triode affects Rp and the way the plate voltage affects Gm. Don't you believe that triodes with an output Rp of 800 ohms and no "standard" local or global feedback sound better than a pentode with an output Rp of 800 ohms from some type of feedback? If the pentode simply got its 800 ohms of Rp by adding the same amount of feedback as the triode already has, then why does the triode sound better? If you have no explanation, then don't you run the risk of convincing some people by your use of the triode NFB model that they might as well use a pentode with feedback, "since it will sound the same as a triode with its internal feedback?" Do you see what I mean? How do you know that triodes actually sound better? Has anyone ever actually done head to head listening tests between triode and pentode amplifiers that are otherwise identical except for the configuration of the output stage? I kind of doubt these tests have ever been done, and without them you can't truthfully say that triodes actually sound better than pentodes. For comparative triode vs. pentode listening tests to be valid, the amplifiers being compared must be completely identical except for the configuration of the output stage, this implies the following, among other things: 1.) The triode and pentode tubes used must be as nearly identical in capability as possible. This is difficult for comparisons to be fair. A real triode such as the 300B is capable of 8 watts. A suitable tube which produces 9 watts in SE pentode mode is an EL34. if 24 watts is the Pda for the pentode and efficiency is 40%, then a typical max is 9.6 watts. But the EL34 cannot be safely used to obtain 8 watts in triode. The KT66 or humble 6L6 could also be used in beam tetrode mode to get 8 watts SE. 2.) The feedback applied to the pentode must be local feedback encompassing only the output stage, not overall loop feedback as is typically used in pentode amplifiers. But how much NFB? How many dB? If one is comparing a 300B to an EL34 pentode with local NFB then the amount of local NFB must be sufficient to reduce the effective Ra of the EL34 to the same 800 ohms as the 300B. Ra' with local series voltage NFB = Ra / ( 1 + [ µ x ß ] ), where Ra is the pentode/beam Ra as per the data for the Iaq, and µ is the pentode/beam amplification factor, or data gm x Ra, and ß is the fraction of the output fed back. You lazy *******s out there can work out that if you know Ra' is to equal 800 ohms, and Ra for EL34 = 12,000 ohms in pentode, and gm = 0.01A/V. then ß can be found. I refuse to spoon feed all of you, and you must be able to work out these utterly basic figures to be able to understand amplifiers. The condition for true pentode/beam operation with NFB is that the screen be bypassed fully to the cathode, so if the CFB connection is used from an OPT to get the series voltage NFB, the screen isn't bypassed to OV like it is ina Quad II amp. 3.) The input sensitivity and admittance of the pentode with local feedback must be the same as the triode to allow the use of the same driver stage for both amplifiers. To compare the triode / pentode/beam tubes, the drive voltage need only be a LINEAR drive voltage of low Rout from any type of source. One wouldn't want the 2H distortion of a triode driver stage cancelling the 2H of the output stage. The amount of 2H generated in driver stages can be say 2% at 40Vrms, and this is about what is needed to completely cancel 2H of a pentode stage with local NFB. See my SE35 pages. But loads on the pentode stage change the pentode/beam 2H quantity drastically, with triode like 2H at low RLs and then reducing to zero at some critical load value, then 2H increases again as RL rises but the phase of the 2H is opposite to below the critical RL value. 4.) The input admittance requirement rules out using shunt feedback, is that the right name, from plate to grid. I don't think shunt NFB can be outlawed if the source resistance of the driver amp used for a test is blamelessly low, as it would be if the driver stage was say an EL84 in triode with a 6CG7 input stage and with its own NFB loop to reduce its Rout of about 2k down to say 500ohms, so that the low input resistance of the shunt NFB network does not have an interative relationship with the driver stage. Cathode feedback seems the most likely to meet the requirements. Two commercial amplifiers using cathode feedback are the QUAD II and McIntosh "unity coupled" series of amplifiers, but neither of these two amplifiers exactly meets the requirements. Lincoln Walsh was another brand. The McIntosh amplifier has more feedback in the output stage than is required to simulate the performance of a triode, and the output stage is considerably less sensitive than a triode, requiring a driver stage with considerably more voltage output capability. But at least the signal applied to the screen is the same as the cathode, so the McI is a case of real beam operation with NFB. But it is substantially a class B circuit, and indeed no comparison can be made between McI circuitry and pure class A triode op without loop NFB. The QUAD design probably comes close to the required output stage sensitivity and amount of feedback. Unfortunately the QUAD design operates the output tubes in the "Ultralinear" mode, and not as true pentodes. We need a design somewhere between the QUAD and McIntosh, incorporating the lower local feedback of the QUAD with the true pentode operation of the McIntosh. Your quite correct John about the Quad. It uses a combination of CFB applied around the grid AND the screen circuit. I do the same with my 8585 circuit, and it measures/sounds well. Where the screen remains bypassed to the cathode the tube is in pentode/beam mode. Above I have offered enough info in a formula for those really interested to work out how much CFB must be used to render a given pentode or beam tube to have an equivalent Ra to a given triode. I will be surprised if anyone takes over from my lead and works our the exact op conditions for an EL34 to behave like a 300B, then correctly publishes them here. I fear everyone is asleep. Until you have done comparative listening tests using a pair of amplifiers designed along the lines I have described above, there is no way you can make the claim that triodes sound better than pentodes. The consensus or predjudice is that triodes are king. I suggest that since SE amps are also in such high esteem these days, that comparisons be made using SE sample amps. Using PP clouds the issue because then there is added confusion and bickering between guys who are for or against PP circuits which usually have much less THD if the circuit works in class A. The triode NFB model, in addition to being just as unprovable as the resistance model, and more difficult to understand and therefore use, has a tendency to mislead people into believing that "pentodes sound the same as triodes, since they both use NFB to get low Rp." All you have to do to know the resistance model is wrong is to look at some triode curves, also you have not yet shown that it is not true that "pentodes sound the same as triodes, since they both use NFB to get low Rp." They don't sound the same, and for all you know, it's because NFB screws up the sound, and triodes do *not* have internal NFB! What evidence do you have that pentodes with local feedback don't sound the same as triodes? Have you done the required listening tests with a pair of otherwise identical amplifiers as I described above? Until you have done this test you can't legitimately make the claim that "they don't sound the same". All disadvantages, and no advantages, as far as I can see. Unless, of course, you simply want to *argue* that "pentodes are too just as good sounding as triodes." The advantage of pentodes is their greater efficiency, if triodes and pentodes sound the same, why use triodes with their lower efficiency? Yes, we all know triodes have effciencies about 1/2 that of pentodes or beam tubes. The efficiency never constitues a sonic advantage or disadvantage. Triodes mean a higher power bill, but it may be worth it because the spectral purity of the triode is better than a pentode with NFB if the Ra of the pentode is made equal to the triode. But why stop here? Is a well made Williamson with trioded output tubes AND with Williamson's specified 20dB of global NFB really superlative? Does it sound ****ed up by the NFB? Would it not be better to connect the two KT66 output tubes in beam tetrode mode while keeping the loop NFB AMOUNT = 20dB ? I think the triode connection of KT66 constitutes a considerable amount of local NFB, and with the rated class A load of 10k a-a it may be equal to 15dB if we were to consider the KT66 triode as a kind of beam tube as a beam tube with local NFB. So with the KT66 set up as beam tubes, then we'd have to apply 35db of global NFB to force the beam tubes to operate as the triodes with 20dB global NFB. But it would be difficult to apply 35db global NFB around a W amp because stability would be threatened, or that is to say the stability margin is almost zero, and we want a margin of at least 15dB, achievable with triodes, and with a decent OPT and with a well conceived critical damping network. Now, Andre, surely you of all people wouldn't want to lie to and mislead audio lovers, like some other people you occasionally mention, right? But by all means, the moment you prove that the resistance model is flawed, and that only the NFB model can correctly model triode behavior in the audio realm, I will say we should start using it. The simple resistance model is clearly flawed in that it doesn't capture any of the triodes distortion mechanisms. Even Henry Pasternack's vaguely defined half power Gm & admittance model fails to model triode nonlinearity as well as my simple feedback model does. The battery + R model of the triode or the low Rout generator + R model of the triode is very useful for basic circuit voltage gain analysis. I have more about all this at my website. But it explains nothing of the NFB action in the triode, even though the R element representing the dynamic Ra is the RESULT of the NFB action. So far, however, the only difference I can see is that the NFB model fails to predict the sonic difference between triodes and pentodes, which is an indication -- although not a proof -- that the NFB model is the one which is flawed. Make sense? As we have tried to point out, a lot depends on how comparisons are made. A raw pentode without any NFB is a horrid amplifier with many harmonics in the spectral performance and with Ra about 10 times higher than a triode so when connected to almost any speaker with much variation in Z the response is horribly non-flat. Also, since transformer distortion depends on the driving source resistance, problems from the iron make pentode sound worse. Its great for a guitar amp though; the more THD the better. The rock music industry depends on distortion for its existance. No, you are starting from the assumption that a triode sounds better than a pentode with local negative feedback in otherwise identical amplifiers, you must first demonstrate this to be true. Then even if it is true that triodes sound better, how does that disprove the validity of the NFB model and prove the validity of the "resistance" model? The "resistance" model is clearly flawed because it can't even come close to reproducing triode curves as measured on a curve tracer. You can't compare a resistor to a triode. A resistor isn't an amp with 4 terminals, of which 2 are commoned at 0V. A resistor has only 2 terminals and its input signal is applied across it, and its "output" is the current though it. A triode has its input applied to its grid, and its output is at its anode and load connection. The output interacts with input in a feedback circuit, something that is impossible in a resistor, so WTF does anyone try to liken a triode's amplifier action to a resistor? Ppl against NFB in triodes just don't want to accept the NFB was designed by the God Of Triodes and put there into every triode by Him and allowed to be discovered and used for about 44 years before the invention of an alternative which was commercially successful, ie, a transistor. Some types of transistors sound better than others, and the amount and method of NFB affects the sound. But that's a whole other lot of arguments. Argy bargy battles about triodes versus pentodes have raged since 1935, and as diversity in devices was expanded human agreement over circuit correctness and sonic performance is now equal to ( zero / infinity ) %. Patrick Turner. Regards, John Byrns |
#142
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
Henry Pasternack wrote:
"Phil" wrote in message ... You, Andre, I will answer. Patrick can go **** himself, since he's determined to answer my technical discussion of a subject with mindless insults and unsupported criticisms. I politely but firmly demand that you do not respond to or repost standerous statements made about me. If you have any question about this, please contact me in person so we can discuss. -Henry That's fair enough. I felt uncomfortable about responding to this anyway, as I see no point to (1) bringing up ancient history, (2) including you in it, since I was never aware of your involvement in anything evil, and (3) using the technique of "cutting your opponent to shreds" given the apparent honesty and intelligence of your responses. I wasn't exactly sure what to do about it, mind you, but your suggestion seems like the right thing to do. Phil |
#143
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
In article ,
Patrick Turner wrote: John Byrns wrote: How do you know that triodes actually sound better? Has anyone ever actually done head to head listening tests between triode and pentode amplifiers that are otherwise identical except for the configuration of the output stage? I kind of doubt these tests have ever been done, and without them you can't truthfully say that triodes actually sound better than pentodes. For comparative triode vs. pentode listening tests to be valid, the amplifiers being compared must be completely identical except for the configuration of the output stage, this implies the following, among other things: 1.) The triode and pentode tubes used must be as nearly identical in capability as possible. This is difficult for comparisons to be fair. A real triode such as the 300B is capable of 8 watts. A suitable tube which produces 9 watts in SE pentode mode is an EL34. if 24 watts is the Pda for the pentode and efficiency is 40%, then a typical max is 9.6 watts. But the EL34 cannot be safely used to obtain 8 watts in triode. The KT66 or humble 6L6 could also be used in beam tetrode mode to get 8 watts SE. You are talking about configuring a pentode as a triode by connecting the screen grid to the plate, creating a triode with a virtual plate at the location of the screen grid, when you do that you no longer have a pentode, you have effectively turned the pentode into a real triode with its inherent internal negative feedback. My idea was to use a pentode with the screen grid held at a constant voltage relative to the cathode so that the tube actually operates as a pentode, and then providing external local negative feedback around the pentode to approximate the negative feedback that occurs within a real triode. This means the available power output would be similar to that obtainable with the tube operating as a pentode, not the lower power output that is available with the pentode strapped as a triode by connecting the screen grid to the plate. 2.) The feedback applied to the pentode must be local feedback encompassing only the output stage, not overall loop feedback as is typically used in pentode amplifiers. But how much NFB? How many dB? There are probably several ways to decide how much negative feedback to use, but what I had in mind was to apply enough negative feedback to bring Rp of the simulated triode down to the same value as the Rp of the real triode it is being compared with, as you describe below. The triode and pentode would also be chosen so that when enough negative feedback is applied to the pentode to bring its Rp down to the same Rp as the triode, the u of the simulated triode would be the same as the u of the actual triode it is being compared with. If a 300B is used as the comparison triode, an EL34 with the correct amount of negative feedback applied to make the Rps equal, would probably have a u that is too high. A 6L6 would probably give about the right values of Rp and u, but would probably fall a little short on power output capability as compared with the 300B. One might try to match a 2A3 rather than the 300B, but again the obvious pentodes don't quite fit the requirements, so a more extensive search would probably be necessary to find a suitable pair of tubes. If one is comparing a 300B to an EL34 pentode with local NFB then the amount of local NFB must be sufficient to reduce the effective Ra of the EL34 to the same 800 ohms as the 300B. Ra' with local series voltage NFB = Ra / ( 1 + [ µ x ß ] ), where Ra is the pentode/beam Ra as per the data for the Iaq, and µ is the pentode/beam amplification factor, or data gm x Ra, and ß is the fraction of the output fed back. You lazy *******s out there can work out that if you know Ra' is to equal 800 ohms, and Ra for EL34 = 12,000 ohms in pentode, and gm = 0.01A/V. then ß can be found. I refuse to spoon feed all of you, and you must be able to work out these utterly basic figures to be able to understand amplifiers. I don't believe this is a problem for most here, with one obvious exception who seems loath to commit himself to any concrete calculations. We can simplify this calculation by using another form of this formula where Rp' = (1/(ß * Gm)) || Rp, then applying the law of tens that an old vacuum tube era engineer taught me when I was an impressionable youth. The law of tens says that we can ignore numbers in a calculation that are either more than ten times greater than the primary quantity we are interested in, or tens times smaller, depending on the nature of the calculation. In this case Rp' is the parallel combination of Rp and 1/(ß * GM). In this case Rp at 12,000 Ohms is more than ten times greater than Rp' at 800 Ohms, so we can ignore Rp, and simplify the equation to Rp' = 1/(ß * Gm). If we want Rp' to be 800 Ohms and Gm equals 10 mA/V, then ß must equal 0.125. The condition for true pentode/beam operation with NFB is that the screen be bypassed fully to the cathode, so if the CFB connection is used from an OPT to get the series voltage NFB, the screen isn't bypassed to OV like it is ina Quad II amp. This condition is easily met by providing an additional winding on the output transformer with the same number of turns as the cathode feedback winding. The screen is then feed through this extra winding which maintains the correct relation between the screen and cathode potentials. 3.) The input sensitivity and admittance of the pentode with local feedback must be the same as the triode to allow the use of the same driver stage for both amplifiers. To compare the triode / pentode/beam tubes, the drive voltage need only be a LINEAR drive voltage of low Rout from any type of source. One wouldn't want the 2H distortion of a triode driver stage cancelling the 2H of the output stage. The amount of 2H generated in driver stages can be say 2% at 40Vrms, and this is about what is needed to completely cancel 2H of a pentode stage with local NFB. See my SE35 pages. But loads on the pentode stage change the pentode/beam 2H quantity drastically, with triode like 2H at low RLs and then reducing to zero at some critical load value, then 2H increases again as RL rises but the phase of the 2H is opposite to below the critical RL value. You raise an interesting issue here, you seem to be suggesting that some of the triode sound may be due to synergy between the triode output stage and the driver stage. Considering the number of different driver configurations used in modern triode amplifiers, I wonder how much of a factor this effect actually is? Further investigation of this effect seems needed. 4.) The input admittance requirement rules out using shunt feedback, is that the right name, from plate to grid. I don't think shunt NFB can be outlawed if the source resistance of the driver amp used for a test is blamelessly low, as it would be if the driver stage was say an EL84 in triode with a 6CG7 input stage and with its own NFB loop to reduce its Rout of about 2k down to say 500ohms, so that the low input resistance of the shunt NFB network does not have an interative relationship with the driver stage. That is presumably another valid approach, but the thing I don't like about it is that it doesn't take a standard triode amplifier circuit as its basis. It does have the considerable virtue of eliminating the need for a complex custom output transformer in the simulated triode pentode amplifier. It is also interesting in that shunt feedback is the method of operating the 6L6 as a triode that was suggested by RCA in their original paper on the design of the 6L6. A simple cathode follower might also serve as a way of driving the shunt feedback network for the less demanding among us. Cathode feedback seems the most likely to meet the requirements. Two commercial amplifiers using cathode feedback are the QUAD II and McIntosh "unity coupled" series of amplifiers, but neither of these two amplifiers exactly meets the requirements. Lincoln Walsh was another brand. There were many other brands, Electrovoice was another, but the QUAD and McIntosh designs are the most commonly known. The McIntosh amplifier has more feedback in the output stage than is required to simulate the performance of a triode, and the output stage is considerably less sensitive than a triode, requiring a driver stage with considerably more voltage output capability. But at least the signal applied to the screen is the same as the cathode, so the McI is a case of real beam operation with NFB. But it is substantially a class B circuit, and indeed no comparison can be made between McI circuitry and pure class A triode op without loop NFB. The QUAD design probably comes close to the required output stage sensitivity and amount of feedback. Unfortunately the QUAD design operates the output tubes in the "Ultralinear" mode, and not as true pentodes. We need a design somewhere between the QUAD and McIntosh, incorporating the lower local feedback of the QUAD with the true pentode operation of the McIntosh. Your quite correct John about the Quad. It uses a combination of CFB applied around the grid AND the screen circuit. I do the same with my 8585 circuit, and it measures/sounds well. Where the screen remains bypassed to the cathode the tube is in pentode/beam mode. Above I have offered enough info in a formula for those really interested to work out how much CFB must be used to render a given pentode or beam tube to have an equivalent Ra to a given triode. I will be surprised if anyone takes over from my lead and works our the exact op conditions for an EL34 to behave like a 300B, then correctly publishes them here. I fear everyone is asleep. They only wakeup when you say something controversial, or seriously blaspheme triodes. The u of the simulated triode will be too high if you use an EL34 with only enough cathode feedback to bring Rp down to 800 Ohms. The 6L6 should give nearly the correct values of Rp and u if a ß of 0.25 is used. Regards, John Byrns |
#144
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
Phil wrote: Henry Pasternack wrote: "Phil" wrote in message ... You, Andre, I will answer. Patrick can go **** himself, since he's determined to answer my technical discussion of a subject with mindless insults and unsupported criticisms. I politely but firmly demand that you do not respond to or repost standerous statements made about me. If you have any question about this, please contact me in person so we can discuss. -Henry That's fair enough. I felt uncomfortable about responding to this anyway, as I see no point to (1) bringing up ancient history, (2) including you in it, since I was never aware of your involvement in anything evil, and (3) using the technique of "cutting your opponent to shreds" given the apparent honesty and intelligence of your responses. I wasn't exactly sure what to do about it, mind you, but your suggestion seems like the right thing to do. Phil Absolutely amazing. Let's take Phil's hypocrisy step by step: That's fair enough. I felt uncomfortable about responding to this anyway, But you don't feel uncomfortable condemning me by implication and name, because my name is mentioned above? You're a hypocrite, Phil. And you don't feel uncomfortable condemning Patrick, an innocent third party, by implication and name, because *his* name is also mentioned above? You're a hypocrite, Phil. as I see no point to (1) bringing up ancient history, I see. The fact that Henry Pasternack consistently, for several years on end, committed vicious and immoral acts, and approved publicly of such acts committed by his associates in the Magnequest Scum, are irrelevant while Pasternack agrees with you, eh Phil? You're a hypocrite, Phil. (2) including you in it, since I was never aware of your involvement in anything evil, The evidence of Pasternack's vicious immorality and consistent lies on professional matters for personal reasons was clear on the newsgroups, still stands on the newsgroups, are referred to by me in enough detail to check them on the Google archive and the Harvard Sound List archive, and you, Phil, have deliberately chosen not to look, to blind yourself to the fact that Pasternack lies on professional matters. You're a hypocrite, Phil. and (3) using the technique of "cutting your opponent to shreds" given the apparent honesty and intelligence of your responses. English translation: Phil says, "The only people who are sincere, honest and intelligent are those who agree with me." You're a hypocrite, Phil. The phrase "cutting your opponent to shreds" is itself a dishonest polemical device. Of course honest people will cut a dishonest, vicious piece of scum like Pasternack to shreds. You're not only a hypocrite, Phil, you are dishonest in argument, as we saw in your abuse of Patrick, and again in this post of yours. Crows of a dishonest feather flock together. You belong with Pasternack. I wasn't exactly sure what to do about it, mind you, but your suggestion seems like the right thing to do. English translation: Phil says, "I, Phil, see electronics as a popularity contest. If Pasternack agrees with me he is always right, regardless of his history of lying on electronics, regardless of his proven history of making electronic statements in furtherance of his personal vendettas." Below my signature is the evidence of Pasternack's professional unreliabity and personal viciousness that you refuse to consider. That is only a sample, of course; a hundred times that much lies in the archives. You're a hypocrite, Phil, and you're dishonest. I had hopes for you. Too bad. Flick. Andre Jute Visit Jute on Amps at http://members.lycos.co.uk/fiultra/ "wonderfully well written and reasoned information for the tube audio constructor" John Broskie TubeCAD & GlassWare "an unbelievably comprehensive web site containing vital gems of wisdom" Stuart Perry Hi-Fi News & Record Review Phil wrote: Patrick Turner wrote: Try reading a few more books about triodes rather than seeking some NFB de-bunking premise from Henry. With all due respsect to Henry, he ain't the world's authority on vacuum tube theory. He could be the devil himself, but that's irrelevant, along with his knowledge, when it comes to the quality of his arguments on this matter. You don't "disprove" a theory by saying the author is stupid or evil. You ignore the characteristics of the author altogether, and focus on the accuracy, or lack thereof, of the theory itself. Reasoning 101. Unless of course the party in question has a track record of deliberately lying. In the case of Henry Pasternack, often referred to RAT as Pompass Plodnick, Google archives show a long history of Pasternack lying on professional and other matters for the sake of "winning" some argument, and of Pasternack committing other unscientific and immoral acts, for which his only excuse is, once the further lies are stripped away, in Pasternack's own words, "my zeal to flame Andre". In this particular case Pasternack came here hoping to have a big fight with me but I merely patronized him a little and sent him on his way with a flea in his ear; he was stuck with Patrick. Because Pasternack came for me first, and because of his history of lying on professional matters for personal gratification, and because Pasternack's first post on the subject (to Chris Hornbeck) was ambiguous, we don't actually know whether he really believes what he says now, or whether his hatred of me has once more painted him into another corner which he will now try to justify with a berm of math. Those new to RAT who wish to see earlier examples of Pasternack lying on professional matters for personal gratification should look up the case where Pasternack told a newbie not to listen to me when I advised a primary impedance on his output transformer 2*Rp or higher; Pasternack told him instead to choose an output impedance equal to the plate resistance. Read that again. Pasternack surely knew that the primary impedance should be twice or more the plate resistance but he lied about it to a newbie "in my zeal to flame Andre". There are hundreds of further posts in Pasternack which tried to justify his Zo = Rp stance but eventually John Byrns nailed Pasternack's hide to wall. Other examples are plentiful, and I have already in this thread given references to a URL that proves Pasternack's contempt for the scientific method. Furthermore, Pasternack in the throes of his hatred will commit totally immoral acts. He ran with Michael LaFever's Magnequest Scum, who flooded the single driver conference with graphic homosexual pornography sent in my name in an effort to drive me out; they were not there before I came, they were not there after I left. Some of that was traced to Pasternack's server. Even worse is the case of the two little girls of an Italian engineer who built one of my designs. He accused Pasternack of sending graphic homosexual filth to his computer, where his two little girls saw it. He didn't even know who Pasternack was when he traced the filth to Pasternack. Pasternack's Magnequest Scum associate Bob Chernofsky said on the Joenet (there's an archieve at Harvard if you want to look it up; search for Sound List) that Pasternack did it because he was bored with not being able to get at me directly. Do you really want to hold this scumbag Pasternack up to us as an impartial fount of engineering wisdom? Get real, Phil. We know better, and it is up to newbies like you to inform themselves before they goof up dumb opinions. Andre Jute Stop bleating. Please, please, please give me the Silence of the Lambs. PS Do I need to explain that we shall know by your response to my sharp remarks how steady your judgement is? |
#145
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
Andrew Jute McCoy, rather than address substance exuded: Absolutely amazing. Let's take Phil's hypocrisy step by step: (after which the usual drivel oozes forth) And, about your fantasy on Peter Drucker? You lie as actual humans breath. Peter Wieck Wyncote, PA |
#146
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
John Byrns wrote:
In article , Patrick Turner wrote: John Byrns wrote: How do you know that triodes actually sound better? Has anyone ever actually done head to head listening tests between triode and pentode amplifiers that are otherwise identical except for the configuration of the output stage? I kind of doubt these tests have ever been done, and without them you can't truthfully say that triodes actually sound better than pentodes. For comparative triode vs. pentode listening tests to be valid, the amplifiers being compared must be completely identical except for the configuration of the output stage, this implies the following, among other things: 1.) The triode and pentode tubes used must be as nearly identical in capability as possible. This is difficult for comparisons to be fair. A real triode such as the 300B is capable of 8 watts. A suitable tube which produces 9 watts in SE pentode mode is an EL34. if 24 watts is the Pda for the pentode and efficiency is 40%, then a typical max is 9.6 watts. But the EL34 cannot be safely used to obtain 8 watts in triode. The KT66 or humble 6L6 could also be used in beam tetrode mode to get 8 watts SE. You are talking about configuring a pentode as a triode by connecting the screen grid to the plate, creating a triode with a virtual plate at the location of the screen grid, when you do that you no longer have a pentode, you have effectively turned the pentode into a real triode with its inherent internal negative feedback. My idea was to use a pentode with the screen grid held at a constant voltage relative to the cathode so that the tube actually operates as a pentode, and then providing external local negative feedback around the pentode to approximate the negative feedback that occurs within a real triode. This means the available power output would be similar to that obtainable with the tube operating as a pentode, not the lower power output that is available with the pentode strapped as a triode by connecting the screen grid to the plate. 2.) The feedback applied to the pentode must be local feedback encompassing only the output stage, not overall loop feedback as is typically used in pentode amplifiers. But how much NFB? How many dB? There are probably several ways to decide how much negative feedback to use, but what I had in mind was to apply enough negative feedback to bring Rp of the simulated triode down to the same value as the Rp of the real triode it is being compared with, as you describe below. The triode and pentode would also be chosen so that when enough negative feedback is applied to the pentode to bring its Rp down to the same Rp as the triode, the u of the simulated triode would be the same as the u of the actual triode it is being compared with. If a 300B is used as the comparison triode, an EL34 with the correct amount of negative feedback applied to make the Rps equal, would probably have a u that is too high. A 6L6 would probably give about the right values of Rp and u, but would probably fall a little short on power output capability as compared with the 300B. One might try to match a 2A3 rather than the 300B, but again the obvious pentodes don't quite fit the requirements, so a more extensive search would probably be necessary to find a suitable pair of tubes. If one is comparing a 300B to an EL34 pentode with local NFB then the amount of local NFB must be sufficient to reduce the effective Ra of the EL34 to the same 800 ohms as the 300B. Ra' with local series voltage NFB = Ra / ( 1 + [ µ x ß ] ), where Ra is the pentode/beam Ra as per the data for the Iaq, and µ is the pentode/beam amplification factor, or data gm x Ra, and ß is the fraction of the output fed back. You lazy *******s out there can work out that if you know Ra' is to equal 800 ohms, and Ra for EL34 = 12,000 ohms in pentode, and gm = 0.01A/V. then ß can be found. I refuse to spoon feed all of you, and you must be able to work out these utterly basic figures to be able to understand amplifiers. I don't believe this is a problem for most here, with one obvious exception who seems loath to commit himself to any concrete calculations. We can simplify this calculation by using another form of this formula where Rp' = (1/(ß * Gm)) || Rp, then applying the law of tens that an old vacuum tube era engineer taught me when I was an impressionable youth. The law of tens says that we can ignore numbers in a calculation that are either more than ten times greater than the primary quantity we are interested in, or tens times smaller, depending on the nature of the calculation. In this case Rp' is the parallel combination of Rp and 1/(ß * GM). In this case Rp at 12,000 Ohms is more than ten times greater than Rp' at 800 Ohms, so we can ignore Rp, and simplify the equation to Rp' = 1/(ß * Gm). If we want Rp' to be 800 Ohms and Gm equals 10 mA/V, then ß must equal 0.125. The condition for true pentode/beam operation with NFB is that the screen be bypassed fully to the cathode, so if the CFB connection is used from an OPT to get the series voltage NFB, the screen isn't bypassed to OV like it is ina Quad II amp. This condition is easily met by providing an additional winding on the output transformer with the same number of turns as the cathode feedback winding. The screen is then feed through this extra winding which maintains the correct relation between the screen and cathode potentials. 3.) The input sensitivity and admittance of the pentode with local feedback must be the same as the triode to allow the use of the same driver stage for both amplifiers. To compare the triode / pentode/beam tubes, the drive voltage need only be a LINEAR drive voltage of low Rout from any type of source. One wouldn't want the 2H distortion of a triode driver stage cancelling the 2H of the output stage. The amount of 2H generated in driver stages can be say 2% at 40Vrms, and this is about what is needed to completely cancel 2H of a pentode stage with local NFB. See my SE35 pages. But loads on the pentode stage change the pentode/beam 2H quantity drastically, with triode like 2H at low RLs and then reducing to zero at some critical load value, then 2H increases again as RL rises but the phase of the 2H is opposite to below the critical RL value. You raise an interesting issue here, you seem to be suggesting that some of the triode sound may be due to synergy between the triode output stage and the driver stage. Considering the number of different driver configurations used in modern triode amplifiers, I wonder how much of a factor this effect actually is? Further investigation of this effect seems needed. 4.) The input admittance requirement rules out using shunt feedback, is that the right name, from plate to grid. I don't think shunt NFB can be outlawed if the source resistance of the driver amp used for a test is blamelessly low, as it would be if the driver stage was say an EL84 in triode with a 6CG7 input stage and with its own NFB loop to reduce its Rout of about 2k down to say 500ohms, so that the low input resistance of the shunt NFB network does not have an interative relationship with the driver stage. That is presumably another valid approach, but the thing I don't like about it is that it doesn't take a standard triode amplifier circuit as its basis. It does have the considerable virtue of eliminating the need for a complex custom output transformer in the simulated triode pentode amplifier. It is also interesting in that shunt feedback is the method of operating the 6L6 as a triode that was suggested by RCA in their original paper on the design of the 6L6. A simple cathode follower might also serve as a way of driving the shunt feedback network for the less demanding among us. Cathode feedback seems the most likely to meet the requirements. Two commercial amplifiers using cathode feedback are the QUAD II and McIntosh "unity coupled" series of amplifiers, but neither of these two amplifiers exactly meets the requirements. Lincoln Walsh was another brand. There were many other brands, Electrovoice was another, but the QUAD and McIntosh designs are the most commonly known. The McIntosh amplifier has more feedback in the output stage than is required to simulate the performance of a triode, and the output stage is considerably less sensitive than a triode, requiring a driver stage with considerably more voltage output capability. But at least the signal applied to the screen is the same as the cathode, so the McI is a case of real beam operation with NFB. But it is substantially a class B circuit, and indeed no comparison can be made between McI circuitry and pure class A triode op without loop NFB. The QUAD design probably comes close to the required output stage sensitivity and amount of feedback. Unfortunately the QUAD design operates the output tubes in the "Ultralinear" mode, and not as true pentodes. We need a design somewhere between the QUAD and McIntosh, incorporating the lower local feedback of the QUAD with the true pentode operation of the McIntosh. Your quite correct John about the Quad. It uses a combination of CFB applied around the grid AND the screen circuit. I do the same with my 8585 circuit, and it measures/sounds well. Where the screen remains bypassed to the cathode the tube is in pentode/beam mode. Above I have offered enough info in a formula for those really interested to work out how much CFB must be used to render a given pentode or beam tube to have an equivalent Ra to a given triode. I will be surprised if anyone takes over from my lead and works our the exact op conditions for an EL34 to behave like a 300B, then correctly publishes them here. I fear everyone is asleep. They only wakeup when you say something controversial, or seriously blaspheme triodes. The u of the simulated triode will be too high if you use an EL34 with only enough cathode feedback to bring Rp down to 800 Ohms. The 6L6 should give nearly the correct values of Rp and u if a ß of 0.25 is used. Regards, John Byrns I'm reading this part of the thread with great pleasure; it reminds me of the glory days of RAT when high theory was everyday meat on the table. There's some confusion here about which tubes we're talking about. Perhaps I am to blame for it, in phrasing the thread headline too bluntly as "triode superiority". But the results of tests and my actual opinion, which comes as no surprise to you, is not the unqualified "triodes are superior" but that "ZNFB or low NFB Class A1 triodes and trioded pentodes are superior on the ear of qualified listeners to any other form of reproduced music". My best amp ever, in the ears of my test groups of professional performers, is PP EL34 Class A running ZNFB or less than 6dB of NFB; I usually test it against expensive 300B gear either of my own decising or commercially available and *behind the curtain* it usually wins. Next, the problem with pentode sound is not so much the pentode itself but the amount of NFB required to get it to deliver its power with any pretense to quality, the NFB creating recombinant harmonic artifacts which are very disturbing to qualified listeners. Thus, if we use any NFB at all in the levelling process, as in the CFB example, the question is opened up of the recombinant harmonic effects of NFB *which we are then adding to only one of the contestants*, the pentode, to disadvantage the pentode yet again. Andre Jute Visit Jute on Amps at http://members.lycos.co.uk/fiultra/ "wonderfully well written and reasoned information for the tube audio constructor" John Broskie TubeCAD & GlassWare "an unbelievably comprehensive web site containing vital gems of wisdom" Stuart Perry Hi-Fi News & Record Review |
#147
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
In article .com,
"Andre Jute" wrote: I'm reading this part of the thread with great pleasure; it reminds me of the glory days of RAT when high theory was everyday meat on the table. There's some confusion here about which tubes we're talking about. Perhaps I am to blame for it, in phrasing the thread headline too bluntly as "triode superiority". But the results of tests and my actual opinion, which comes as no surprise to you, is not the unqualified "triodes are superior" but that "ZNFB or low NFB Class A1 triodes and trioded pentodes are superior on the ear of qualified listeners to any other form of reproduced music". My best amp ever, in the ears of my test groups of professional performers, is PP EL34 Class A running ZNFB or less than 6dB of NFB; I usually test it against expensive 300B gear either of my own decising or commercially available and *behind the curtain* it usually wins. Next, the problem with pentode sound is not so much the pentode itself but the amount of NFB required to get it to deliver its power with any pretense to quality, the NFB creating recombinant harmonic artifacts which are very disturbing to qualified listeners. Thus, if we use any NFB at all in the levelling process, as in the CFB example, the question is opened up of the recombinant harmonic effects of NFB *which we are then adding to only one of the contestants*, the pentode, to disadvantage the pentode yet again. Hi Andre, Please cut me a little slack if I have missed something, as I wasn't in on this thread from the beginning. I don't understand your apparent thesis that the triode is superior, if indeed your "PP EL34 Class A running ZNFB" is your best amp ever, as judged by your "test groups of professional performers"? Are the EL34s in this amp triode connected with the screen grids strapped to the plates? If this is the case, the implication is that an EL34 with the screen tied to the plate makes a better triode than the 300B used in several of your other amp designs? You also seem to be saying that the negative feedback that is inherent in the operation of a triode is not responsible for its superior sound, do you have any clues yet as to exactly what it is that is responsible for the superior sound of the triode and the triode connected pentode? Regards, John Byrns |
#148
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
John Byrns wrote:
In article .com, "Andre Jute" wrote: I'm reading this part of the thread with great pleasure; it reminds me of the glory days of RAT when high theory was everyday meat on the table. There's some confusion here about which tubes we're talking about. Perhaps I am to blame for it, in phrasing the thread headline too bluntly as "triode superiority". But the results of tests and my actual opinion, which comes as no surprise to you, is not the unqualified "triodes are superior" but that "ZNFB or low NFB Class A1 triodes and trioded pentodes are superior on the ear of qualified listeners to any other form of reproduced music". My best amp ever, in the ears of my test groups of professional performers, is PP EL34 Class A running ZNFB or less than 6dB of NFB; I usually test it against expensive 300B gear either of my own decising or commercially available and *behind the curtain* it usually wins. Next, the problem with pentode sound is not so much the pentode itself but the amount of NFB required to get it to deliver its power with any pretense to quality, the NFB creating recombinant harmonic artifacts which are very disturbing to qualified listeners. Thus, if we use any NFB at all in the levelling process, as in the CFB example, the question is opened up of the recombinant harmonic effects of NFB *which we are then adding to only one of the contestants*, the pentode, to disadvantage the pentode yet again. Hi Andre, Please cut me a little slack if I have missed something, as I wasn't in on this thread from the beginning. The necessary information is spread over many threads and my netsite. I don't understand your apparent thesis that the triode is superior, if indeed your "PP EL34 Class A running ZNFB" is your best amp ever, as judged by your "test groups of professional performers"? The headline is a provocative outgrowth of a bad-tempered set of threads on whether a triode has native or built-in negative feedback. I have always said that best amp I ever built was the trioded EL34 PP. In fact I spelt it out earlier in this thread as well, but nobody took the slightest notice. Are the EL34s in this amp triode connected with the screen grids strapped to the plates? Yes. If this is the case, the implication is that an EL34 with the screen tied to the plate makes a better triode than the 300B used in several of your other amp designs? Yes. The comparison is ZNFB or very low NFB triode strapped EL34 PP against SE ZNFB 300B. In level matched tests with a third, usually solid state placebo amp also behind the curtain, a sophisticated listening panel will choose the PP EL34 pseudo-triode amp every time. (The SS amp commonly gets no votes.) There are complications, in that when you double up the tubes on each, to compare PSE and PPP, all other factors equal, except that now you can use better speakers, specifically ESL63, the results get a bit closer simply because the speakers offer more resolution to a test panel accustomed to listening to details, and that big broadcasting DHT like 845 and SV572-xx whip little EL34, but for practical, level-matched tests at the entry level -- not meaning money, for all these amps are expensive, but the entry level of very high resolution -- sure as hell triode strapped PP ZNFB EL34 hold the upper hand. You have to be in PP 300B at umpteen-eleven times the price, with a lot of high-ticket high-carriage iron on board, before the natural triodes make as good a showing as the humble, inexpensive, ultra-versatile EL34; by then it is clearly no longer a fair test (but then, in terms of high street hi-fi prices even the base EL34/300B test has only a tenuous connection to what most people would regard as reality). I give you the additional information because we need to be clear that what I have proved isn't that PP amps are better than SE, merely that the EL34 is a stonking good tube, especially at the price. The test only incidentally pitches a PP against an SE amp; I had long before this discovered that SE or PP matters not when both are made with triodes running in Class A and you have a very light or absent hand on the NFB; what I set out to prove in these later tests with the PP trioded EL34 was something about Class A sound, and something about the composition of distortion, and something about the lowest imperceptible level of loop or anyway additional NFB. You also seem to be saying that the negative feedback that is inherent in the operation of a triode is not responsible for its superior sound, On the contrary, it seems to me very likely that the negative feedback inside a triode is responsible for the quality of its sound. To me it is a very persuasive argument that to bring a pentode back even partway towards the silence of a triode you have to use the NFB that was taken away to make it a pentode! However, it is clear to me that external forms of NFB, as in loop feedback, UL, CFB, etc, act differently to the native or internal NFB of a triode. This is easily seen in comparative harmonic output analysis of an EL34 amp with a pentode/UL/triode switch and a pot or switch to alter the amount of loop NFB, as in my T113 "Triple Threat" PP EL34. do you have any clues yet as to exactly what it is that is responsible for the superior sound of the triode and the triode connected pentode? Sure. The difference in sound between a triode (natural or screentied) and a pentode is caused by the different makeup of the harmonics. Regardless of the absolute level of distortion, the triode, however made, has a lower proportion of odd and higher harmonics than the pentode. It is known that very low percentages of odd and higher harmonics can cause uneasiness in listeners. It is strong supporting information that the difference between a good 300B amp and a bad one can be read in the proportion of 3rd harmonic left on the bad amp (as you yourself showed in a table when you compared my HVHCHL "Hedonist" to the distrastrous BobC/LaFevre/Magnequest "Bubbaland" amp roundabout 1998 -- that was a big clue on the way to my present understanding -- thanks!). The easiest way to prove the importance of the proportions inside the THD to yourself is to put a pentode/UL/triode switch and an NFB pot on a pentode amp and to study the makeup of the distortion in each mode at various levels of NFB including zero NFB. Of course, it took me a bit longer than a single sentence, because I had to eliminate all the common factors between the various topologies (I wasted years on AC balance in PP amps) before I isolated the odd factor and could examine it closely. There is no other parameter where the change is so striking and so inexplicable. Even at zero NFB, the third and higher harmonics fall through the floor when you turn the triode switch on a pentode amp. Of course, before you even get there, you notice that in the trioded pentode, the THD is lower than in the other two modes (pentode, UL) at any (low) level of NFB. It is difficult not to conclude that the effective removal of the pentode's screen grid adds additional NFB which works in some internal way. It is equally difficult to see (and hear) the different result of this internal NFB and not to conclude that it works differently from external feedback. Thus my fear that, if you use any form of external feedback to make the triode and the pentode "electrically equivalent", you will disadvantage the pentode because the makeup of its harmonics will be adverse *and will be heard by the panel*. I don't actually see any alternative way which won't turn the pentode into a triode, which defeats the purpose of a straight-up challenge, so I think you and Patrick are on a wild goose chase. But that has never stopped us speculating before! Regards, John Byrns These, incidentally, are the same test results I interpret as meaning that there is no such thing as an SE sound, that what we hear which so appeals is a Class A ZNFB or very low NFB sonic signature. People just confuse Class A with SE because only a tiny, tiny minority even of the ultrafidelista have ever heard pure Class A PP amp. HTH. Andre Jute Visit Jute on Amps at http://members.lycos.co.uk/fiultra/ "wonderfully well written and reasoned information for the tube audio constructor" John Broskie TubeCAD & GlassWare "an unbelievably comprehensive web site containing vital gems of wisdom" Stuart Perry Hi-Fi News & Record Review |
#149
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
Andrew Jute McCoy blathered and lied: The headline is a provocative outgrowth of a bad-tempered set of threads on whether a triode has native or built-in negative feedback. I have always said that best amp I ever built was the trioded EL34 PP. In fact I spelt it out earlier in this thread as well, but nobody took the slightest notice. It built such an amp exactly as it visited Peter Drucker in his NYC offices... And exactly as Mr. Drucker was an Hungarian Refugee as Mr. McCoy stated so emphatically. Mr. McCoy lies as actual human beings breath. Peter Wieck Wyncote, PA |
#150
Posted to rec.audio.tubes
|
|||
|
|||
Henry Pasternack's Norton triode model???
In article ,
"Henry Pasternack" wrote: I believe Byrns' point is that a three-halves power transconductance whose input is the sum of Vg and Vp / mu is one such model. There's no general way to represent the large-signal model in terms of an independent transconductance and admittance. But we don't need to solve the general case to disprove the feedback proposition. Because there are no cross-terms in the triode equation, Henry blows it here because the triode equation I posted, Ipk = k * (Vgk + Vpk/u) ^ 1.5 actually does include "cross-terms", although Henry would like us to believe it doesn't. if you set the condition that Ip is constant, the three-halves power exponent goes away and Vp becomes a linear function of Vg. Then gm has no dependence on Vp and rp has no dependence on Vg. This is why all the plate curves have the same shape. This of course is not true, all the plate curves in a real triode, as well as in my model, do not have the same shape. You can derive the Norton model assuming a nonlinear transconductance that is a one-half power function of Vp and a nonlinear admittance that is a negative one-half power function of Vg. The product of the two is a constant, equal to mu. The model meets the requirements that it has no feedback and also that it expresses the Child-Langmuir equation. To correct a typo "Henry Pasternack" then wrote: You can derive the Norton model assuming a nonlinear transconductance that is a one-half power function of Vp and a nonlinear admittance that is a negative one-half power function of Vg. Before Byrns and Jute get hysterical, let me say I inadvertently reversed Vp and Vg in that sentence. OK, after correcting that typo has anyone been able to make sense of Henry's Norton model? As near as I can tell, for the plate circuit to express the Child-Langmuir equation, it is necessary for Henry's nonlinear admittance to be a positive one-half power function of Vp, not a negative one-half power function of Vp as Henry said, is this another typo? Regards, John Byrns |
#151
Posted to rec.audio.tubes
|
|||
|
|||
Henry Pasternack's Norton triode model???
On Wed, 11 Oct 2006 01:53:36 GMT, John Byrns
wrote: OK, after correcting that typo has anyone been able to make sense of Henry's Norton model? As near as I can tell, for the plate circuit to express the Child-Langmuir equation, it is necessary for Henry's nonlinear admittance to be a positive one-half power function of Vp, not a negative one-half power function of Vp as Henry said, is this another typo? Algebraic models have no resonance for me. A model that gets me all hot-n-bothered must start (or be referential to) a physical model. My personal physical model of triode behavior is of a transconductance "engine" with a mu multiplier, just because it's kinda-sorta like the actual machinery. The cathode makes an electron cloud, the grid turns that into a transconductance device, geometry gives a mu. Like that. Child's equations are pre-War approximations, not revealed Truth. They have no physical basis. For me, this numbers bantering is both wrong-headed and boring. Excuse my French, but it's totally unrelated to the issue of choosing a defendable model. But, FWIW, your SPICE model seems, to me, totally disconnected from physical reality. Educate me? OTOOOH, what a wonderful discussion. And an even more wonderful digression from... well, you know, all the other stuff here sometimes. Much thanks, as always, Chris Hornbeck |
#152
Posted to rec.audio.tubes
|
|||
|
|||
Henry Pasternack's Norton triode model???
On Tue, 10 Oct 2006 23:30:24 -0500, flipper wrote:
My personal physical model of triode behavior is of a transconductance "engine" with a mu multiplier, just because it's kinda-sorta like the actual machinery. The cathode makes an electron cloud, the grid turns that into a transconductance device, geometry gives a mu. Like that. Add that it's also a function of the plate-cathode potential and ya got it. The tranconductance engine is (mostly, small-signal, linear- model) independent of anode voltage, other than that there *is* one (an anode pulling the electrons). The next "block diagram in the SPICE model" should be geometry; lack thereof is my beef with overly simplistic, pseudo-mathematical modeling. It's tempting to try to do double-entry bookkeeping by imagining the transconductance engine to be modulated by anode voltage, because there is some small effect there, but it's misleading. It may crudely represent the large-signal non-linear model, but I believe it poorly represents the real world, small- or large-signal. And, for me worse, adds no information. Of course that's two 'inputs' but who's counting? This is a part of your religion, but I remain agnostic. All good fortune, Chris Hornbeck |
#153
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
John Byrns wrote: In article , Patrick Turner wrote: John Byrns wrote: How do you know that triodes actually sound better? Has anyone ever actually done head to head listening tests between triode and pentode amplifiers that are otherwise identical except for the configuration of the output stage? I kind of doubt these tests have ever been done, and without them you can't truthfully say that triodes actually sound better than pentodes. For comparative triode vs. pentode listening tests to be valid, the amplifiers being compared must be completely identical except for the configuration of the output stage, this implies the following, among other things: 1.) The triode and pentode tubes used must be as nearly identical in capability as possible. This is difficult for comparisons to be fair. A real triode such as the 300B is capable of 8 watts. A suitable tube which produces 9 watts in SE pentode mode is an EL34. if 24 watts is the Pda for the pentode and efficiency is 40%, then a typical max is 9.6 watts. But the EL34 cannot be safely used to obtain 8 watts in triode. The KT66 or humble 6L6 could also be used in beam tetrode mode to get 8 watts SE. You are talking about configuring a pentode as a triode by connecting the screen grid to the plate, creating a triode with a virtual plate at the location of the screen grid, when you do that you no longer have a pentode, you have effectively turned the pentode into a real triode with its inherent internal negative feedback. My idea was to use a pentode with the screen grid held at a constant voltage relative to the cathode so that the tube actually operates as a pentode, and then providing external local negative feedback around the pentode to approximate the negative feedback that occurs within a real triode. This means the available power output would be similar to that obtainable with the tube operating as a pentode, not the lower power output that is available with the pentode strapped as a triode by connecting the screen grid to the plate. If you'd read further on you would have seen that I have dealt with the issue of keeping the EL34 in pentode by bypassing its screen to the cathode, so it woking in pure pentode mode, albeit with NFB in some external loop. 2.) The feedback applied to the pentode must be local feedback encompassing only the output stage, not overall loop feedback as is typically used in pentode amplifiers. But how much NFB? How many dB? There are probably several ways to decide how much negative feedback to use, but what I had in mind was to apply enough negative feedback to bring Rp of the simulated triode down to the same value as the Rp of the real triode it is being compared with, as you describe below. The triode and pentode would also be chosen so that when enough negative feedback is applied to the pentode to bring its Rp down to the same Rp as the triode, the u of the simulated triode would be the same as the u of the actual triode it is being compared with. This may never or rarely ever be the case. If a 300B is used as the comparison triode, an EL34 with the correct amount of negative feedback applied to make the Rps equal, would probably have a u that is too high. A 6L6 would probably give about the right values of Rp and u, but would probably fall a little short on power output capability as compared with the 300B. One might try to match a 2A3 rather than the 300B, but again the obvious pentodes don't quite fit the requirements, so a more extensive search would probably be necessary to find a suitable pair of tubes. Its very easy to work out what the effective µ will be for a pentode with say local CFB. The EL34 has Ra = 12k, and gm = 0.011A/V, so its pentode µ = 132. If we wanted an effective Ra = 800 ohms = the 300B then The effective Ra' with series voltatge NFB = 800 = Ra / ( 1 + [ µ x ß ] ) = 12,000 / ( 1 + [ 132 x ß] ). So 1 = 15 / ( 1 + [ 132 x ß] )...... divide both sides by 800 So 1 + 132ß = 15, -1 = 14, So ß = 14 / 132 = 0.106. Therefore if you have 300 turns in the cathode and 2,700 turns in the anode winding for a total of 3,000 turns of primary in an OPT, there is 10% of the total power producing signal fed back in series with the grid input signal. So let us suppose there was 135Vrms across the anode winding, there will be 15Vrms across the cathode winding. Without a load Vak / Vgk = pentode µ, and in this case the pentode µ = 132, so Vgk = 150V / 132 = 1.136vrms. There is 15V at k, so total input voltage with no load is 15 + 1.136 = 16.13Vrms, and the effective µ = 150Vrms / 16.136V = 9.29, which is much higher than the µ for a 300B, and its due to the higher initial gm of the pentode. Under the no load conditions cited the amount of NFB used is 20 log x voltage gain or in this case µ reduction, which is 1/14.21 = 23dB of applied NFB. This is a very similar amount of NFB that is applied when the screen is simply strapped to the anode, and where Ra becomes 1,250 ohms. However the local CFB network from the OPT is a linear network unlike the trioding case where the applied FB is via a voltage field effect following a 3/2 rule. Where the screen is bypassed to 0V instead of to the cathode, the effective Ra is more difficult to calculate but it is still low, but its always more than the pure pentode with FB case. The effect of such local NFB on pentode distortion is to reduce it but not alter the spectral content. Triode connection results in a far lesser number of disastrous harmonics to start with. Hence Triode connection is still favoured over pentode connection with NFB. Bypassing screen to 0V and using CFB results in the pentode or beam tube having triode like gains AND more triode like spectra since some of the applied NFB between screen and cathode as well as between grid and cathode. So the Quad II method, ie, Acoustical connection is an effective one at extracting the best performance from a multigrid tube which give triode like Ra, adequate gain, and triode like THD, but with nearly maximum power available in pentode or beam. I find that the Acoustical is more effective than having 50% UL screen taps a la Leak. If one is comparing a 300B to an EL34 pentode with local NFB then the amount of local NFB must be sufficient to reduce the effective Ra of the EL34 to the same 800 ohms as the 300B. Ra' with local series voltage NFB = Ra / ( 1 + [ µ x ß ] ), where Ra is the pentode/beam Ra as per the data for the Iaq, and µ is the pentode/beam amplification factor, or data gm x Ra, and ß is the fraction of the output fed back. You lazy *******s out there can work out that if you know Ra' is to equal 800 ohms, and Ra for EL34 = 12,000 ohms in pentode, and gm = 0.01A/V. then ß can be found. I refuse to spoon feed all of you, and you must be able to work out these utterly basic figures to be able to understand amplifiers. I don't believe this is a problem for most here, with one obvious exception who seems loath to commit himself to any concrete calculations. Who on earth would that be? We can simplify this calculation by using another form of this formula where Rp' = (1/(ß * Gm)) || Rp, then applying the law of tens that an old vacuum tube era engineer taught me when I was an impressionable youth. The law of tens says that we can ignore numbers in a calculation that are either more than ten times greater than the primary quantity we are interested in, or tens times smaller, depending on the nature of the calculation. In this case Rp' is the parallel combination of Rp and 1/(ß * GM). In this case Rp at 12,000 Ohms is more than ten times greater than Rp' at 800 Ohms, so we can ignore Rp, and simplify the equation to Rp' = 1/(ß * Gm). If we want Rp' to be 800 Ohms and Gm equals 10 mA/V, then ß must equal 0.125. I didn't use that lazy old engineer's method of 90% accuracy will do. I got about the same result as you did though. The condition for true pentode/beam operation with NFB is that the screen be bypassed fully to the cathode, so if the CFB connection is used from an OPT to get the series voltage NFB, the screen isn't bypassed to OV like it is ina Quad II amp. This condition is easily met by providing an additional winding on the output transformer with the same number of turns as the cathode feedback winding. The screen is then feed through this extra winding which maintains the correct relation between the screen and cathode potentials. There is no need for the screen winding, and having resistances from the B+ to the screens and electros which are connected between screen and cathode is OK, since the screen supply is OK at say 300V while Ea is at 350V, and the screen dropping R are not a low value so they don't load down the cathode windings whose output soure resistance is 1/100 of the effective Ra' of the pentode. In a PP amp there can be 10% taps along the anode windings to cross couple the screens to get the same effect as a screen winding. But why such efforts when the far simpler Acoustical method is so effective, and harmonically better? 3.) The input sensitivity and admittance of the pentode with local feedback must be the same as the triode to allow the use of the same driver stage for both amplifiers. To compare the triode / pentode/beam tubes, the drive voltage need only be a LINEAR drive voltage of low Rout from any type of source. One wouldn't want the 2H distortion of a triode driver stage cancelling the 2H of the output stage. The amount of 2H generated in driver stages can be say 2% at 40Vrms, and this is about what is needed to completely cancel 2H of a pentode stage with local NFB. See my SE35 pages. But loads on the pentode stage change the pentode/beam 2H quantity drastically, with triode like 2H at low RLs and then reducing to zero at some critical load value, then 2H increases again as RL rises but the phase of the 2H is opposite to below the critical RL value. You raise an interesting issue here, you seem to be suggesting that some of the triode sound may be due to synergy between the triode output stage and the driver stage. Considering the number of different driver configurations used in modern triode amplifiers, I wonder how much of a factor this effect actually is? Further investigation of this effect seems needed. Some triode drivers produce almost as much but rarely ever more than the triode output stages they drive, so a 300B needing 50Vrms grid drive from say a 6V6 in triode is where the 5% mainly 2H is partially cancelled by say 2.5% in the 6V6 THD, also mainly 2H. So you get 2.5% THD overall, plus some added 4H, and other garbage, but these second order products are low down. The 2H of the input stage in a typical 3 stage SET amp is additive to the THD of the output tube, so its important to use a CCS load on the input triode to minimise its THD and thus reduce the overall THD. Its possible to get much more effective THD cancellation with output stages configured with pentode of beam tubes with CFB, as in the case of my SE35 which is fully documented at my website. 4.) The input admittance requirement rules out using shunt feedback, is that the right name, from plate to grid. I don't think shunt NFB can be outlawed if the source resistance of the driver amp used for a test is blamelessly low, as it would be if the driver stage was say an EL84 in triode with a 6CG7 input stage and with its own NFB loop to reduce its Rout of about 2k down to say 500ohms, so that the low input resistance of the shunt NFB network does not have an interative relationship with the driver stage. That is presumably another valid approach, but the thing I don't like about it is that it doesn't take a standard triode amplifier circuit as its basis. It does have the considerable virtue of eliminating the need for a complex custom output transformer in the simulated triode pentode amplifier. It is also interesting in that shunt feedback is the method of operating the 6L6 as a triode that was suggested by RCA in their original paper on the design of the 6L6. A simple cathode follower might also serve as a way of driving the shunt feedback network for the less demanding among us. Pages 333 and 334 of RDH4 give methods of shunt FB exploiting the high Ra of a driving tube as part of the shunt NFB. Also see my schematics pages on shunt NFB amps at http://www.turneraudio.com.au/miscel...hematics2.html Cathode feedback seems the most likely to meet the requirements. Two commercial amplifiers using cathode feedback are the QUAD II and McIntosh "unity coupled" series of amplifiers, but neither of these two amplifiers exactly meets the requirements. Lincoln Walsh was another brand. There were many other brands, Electrovoice was another, but the QUAD and McIntosh designs are the most commonly known. The McIntosh amplifier has more feedback in the output stage than is required to simulate the performance of a triode, and the output stage is considerably less sensitive than a triode, requiring a driver stage with considerably more voltage output capability. But at least the signal applied to the screen is the same as the cathode, so the McI is a case of real beam operation with NFB. But it is substantially a class B circuit, and indeed no comparison can be made between McI circuitry and pure class A triode op without loop NFB. The QUAD design probably comes close to the required output stage sensitivity and amount of feedback. Unfortunately the QUAD design operates the output tubes in the "Ultralinear" mode, and not as true pentodes. We need a design somewhere between the QUAD and McIntosh, incorporating the lower local feedback of the QUAD with the true pentode operation of the McIntosh. Your quite correct John about the Quad. It uses a combination of CFB applied around the grid AND the screen circuit. I do the same with my 8585 circuit, and it measures/sounds well. Where the screen remains bypassed to the cathode the tube is in pentode/beam mode. Above I have offered enough info in a formula for those really interested to work out how much CFB must be used to render a given pentode or beam tube to have an equivalent Ra to a given triode. I will be surprised if anyone takes over from my lead and works our the exact op conditions for an EL34 to behave like a 300B, then correctly publishes them here. I fear everyone is asleep. They only wakeup when you say something controversial, or seriously blaspheme triodes. The u of the simulated triode will be too high if you use an EL34 with only enough cathode feedback to bring Rp down to 800 Ohms. The 6L6 should give nearly the correct values of Rp and u if a ß of 0.25 is used. We just caculated the EL34, and BTW, KT88 etc also respond well to CFB. µ' ( ie, effective amplification factor after FB is applied ) for EL34 to get Ra = 800 ohms is around 9, and not too high, and could be around 5, which would allow not too high a drive voltage, and get Ra' a lot lower than 800 ohms. I am using ß on my 300 watters at 20%, and with screens bypassed to 0V, and the measured results are the best I have seen. Its also sounds well. See my pages with full schematics and explanations. You are right about 6L6 or KT66. Even trioded, the Ra drops to only 1.6k. Later tubes such as EL34 and KT88/6550 KT90 have more gm and lower ra to start with so NFB is all the more effective. I have not tried the 13E1 with CFB but I suggest it would respond exceptionally well; QE208 is another fabulous beam tetrode which i think is even better...... Patrick Turner. Regards, John Byrns |
#154
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
Andre Jute wrote: John Byrns wrote: In article , Patrick Turner wrote: John Byrns wrote: How do you know that triodes actually sound better? Has anyone ever actually done head to head listening tests between triode and pentode amplifiers that are otherwise identical except for the configuration of the output stage? I kind of doubt these tests have ever been done, and without them you can't truthfully say that triodes actually sound better than pentodes. For comparative triode vs. pentode listening tests to be valid, the amplifiers being compared must be completely identical except for the configuration of the output stage, this implies the following, among other things: 1.) The triode and pentode tubes used must be as nearly identical in capability as possible. This is difficult for comparisons to be fair. A real triode such as the 300B is capable of 8 watts. A suitable tube which produces 9 watts in SE pentode mode is an EL34. if 24 watts is the Pda for the pentode and efficiency is 40%, then a typical max is 9.6 watts. But the EL34 cannot be safely used to obtain 8 watts in triode. The KT66 or humble 6L6 could also be used in beam tetrode mode to get 8 watts SE. You are talking about configuring a pentode as a triode by connecting the screen grid to the plate, creating a triode with a virtual plate at the location of the screen grid, when you do that you no longer have a pentode, you have effectively turned the pentode into a real triode with its inherent internal negative feedback. My idea was to use a pentode with the screen grid held at a constant voltage relative to the cathode so that the tube actually operates as a pentode, and then providing external local negative feedback around the pentode to approximate the negative feedback that occurs within a real triode. This means the available power output would be similar to that obtainable with the tube operating as a pentode, not the lower power output that is available with the pentode strapped as a triode by connecting the screen grid to the plate. 2.) The feedback applied to the pentode must be local feedback encompassing only the output stage, not overall loop feedback as is typically used in pentode amplifiers. But how much NFB? How many dB? There are probably several ways to decide how much negative feedback to use, but what I had in mind was to apply enough negative feedback to bring Rp of the simulated triode down to the same value as the Rp of the real triode it is being compared with, as you describe below. The triode and pentode would also be chosen so that when enough negative feedback is applied to the pentode to bring its Rp down to the same Rp as the triode, the u of the simulated triode would be the same as the u of the actual triode it is being compared with. If a 300B is used as the comparison triode, an EL34 with the correct amount of negative feedback applied to make the Rps equal, would probably have a u that is too high. A 6L6 would probably give about the right values of Rp and u, but would probably fall a little short on power output capability as compared with the 300B. One might try to match a 2A3 rather than the 300B, but again the obvious pentodes don't quite fit the requirements, so a more extensive search would probably be necessary to find a suitable pair of tubes. If one is comparing a 300B to an EL34 pentode with local NFB then the amount of local NFB must be sufficient to reduce the effective Ra of the EL34 to the same 800 ohms as the 300B. Ra' with local series voltage NFB = Ra / ( 1 + [ µ x ß ] ), where Ra is the pentode/beam Ra as per the data for the Iaq, and µ is the pentode/beam amplification factor, or data gm x Ra, and ß is the fraction of the output fed back. You lazy *******s out there can work out that if you know Ra' is to equal 800 ohms, and Ra for EL34 = 12,000 ohms in pentode, and gm = 0.01A/V. then ß can be found. I refuse to spoon feed all of you, and you must be able to work out these utterly basic figures to be able to understand amplifiers. I don't believe this is a problem for most here, with one obvious exception who seems loath to commit himself to any concrete calculations. We can simplify this calculation by using another form of this formula where Rp' = (1/(ß * Gm)) || Rp, then applying the law of tens that an old vacuum tube era engineer taught me when I was an impressionable youth. The law of tens says that we can ignore numbers in a calculation that are either more than ten times greater than the primary quantity we are interested in, or tens times smaller, depending on the nature of the calculation. In this case Rp' is the parallel combination of Rp and 1/(ß * GM). In this case Rp at 12,000 Ohms is more than ten times greater than Rp' at 800 Ohms, so we can ignore Rp, and simplify the equation to Rp' = 1/(ß * Gm). If we want Rp' to be 800 Ohms and Gm equals 10 mA/V, then ß must equal 0.125. The condition for true pentode/beam operation with NFB is that the screen be bypassed fully to the cathode, so if the CFB connection is used from an OPT to get the series voltage NFB, the screen isn't bypassed to OV like it is ina Quad II amp. This condition is easily met by providing an additional winding on the output transformer with the same number of turns as the cathode feedback winding. The screen is then feed through this extra winding which maintains the correct relation between the screen and cathode potentials. 3.) The input sensitivity and admittance of the pentode with local feedback must be the same as the triode to allow the use of the same driver stage for both amplifiers. To compare the triode / pentode/beam tubes, the drive voltage need only be a LINEAR drive voltage of low Rout from any type of source. One wouldn't want the 2H distortion of a triode driver stage cancelling the 2H of the output stage. The amount of 2H generated in driver stages can be say 2% at 40Vrms, and this is about what is needed to completely cancel 2H of a pentode stage with local NFB. See my SE35 pages. But loads on the pentode stage change the pentode/beam 2H quantity drastically, with triode like 2H at low RLs and then reducing to zero at some critical load value, then 2H increases again as RL rises but the phase of the 2H is opposite to below the critical RL value. You raise an interesting issue here, you seem to be suggesting that some of the triode sound may be due to synergy between the triode output stage and the driver stage. Considering the number of different driver configurations used in modern triode amplifiers, I wonder how much of a factor this effect actually is? Further investigation of this effect seems needed. 4.) The input admittance requirement rules out using shunt feedback, is that the right name, from plate to grid. I don't think shunt NFB can be outlawed if the source resistance of the driver amp used for a test is blamelessly low, as it would be if the driver stage was say an EL84 in triode with a 6CG7 input stage and with its own NFB loop to reduce its Rout of about 2k down to say 500ohms, so that the low input resistance of the shunt NFB network does not have an interative relationship with the driver stage. That is presumably another valid approach, but the thing I don't like about it is that it doesn't take a standard triode amplifier circuit as its basis. It does have the considerable virtue of eliminating the need for a complex custom output transformer in the simulated triode pentode amplifier. It is also interesting in that shunt feedback is the method of operating the 6L6 as a triode that was suggested by RCA in their original paper on the design of the 6L6. A simple cathode follower might also serve as a way of driving the shunt feedback network for the less demanding among us. Cathode feedback seems the most likely to meet the requirements. Two commercial amplifiers using cathode feedback are the QUAD II and McIntosh "unity coupled" series of amplifiers, but neither of these two amplifiers exactly meets the requirements. Lincoln Walsh was another brand. There were many other brands, Electrovoice was another, but the QUAD and McIntosh designs are the most commonly known. The McIntosh amplifier has more feedback in the output stage than is required to simulate the performance of a triode, and the output stage is considerably less sensitive than a triode, requiring a driver stage with considerably more voltage output capability. But at least the signal applied to the screen is the same as the cathode, so the McI is a case of real beam operation with NFB. But it is substantially a class B circuit, and indeed no comparison can be made between McI circuitry and pure class A triode op without loop NFB. The QUAD design probably comes close to the required output stage sensitivity and amount of feedback. Unfortunately the QUAD design operates the output tubes in the "Ultralinear" mode, and not as true pentodes. We need a design somewhere between the QUAD and McIntosh, incorporating the lower local feedback of the QUAD with the true pentode operation of the McIntosh. Your quite correct John about the Quad. It uses a combination of CFB applied around the grid AND the screen circuit. I do the same with my 8585 circuit, and it measures/sounds well. Where the screen remains bypassed to the cathode the tube is in pentode/beam mode. Above I have offered enough info in a formula for those really interested to work out how much CFB must be used to render a given pentode or beam tube to have an equivalent Ra to a given triode. I will be surprised if anyone takes over from my lead and works our the exact op conditions for an EL34 to behave like a 300B, then correctly publishes them here. I fear everyone is asleep. They only wakeup when you say something controversial, or seriously blaspheme triodes. The u of the simulated triode will be too high if you use an EL34 with only enough cathode feedback to bring Rp down to 800 Ohms. The 6L6 should give nearly the correct values of Rp and u if a ß of 0.25 is used. Regards, John Byrns I'm reading this part of the thread with great pleasure; it reminds me of the glory days of RAT when high theory was everyday meat on the table. There's some confusion here about which tubes we're talking about. Perhaps I am to blame for it, in phrasing the thread headline too bluntly as "triode superiority". But the results of tests and my actual opinion, which comes as no surprise to you, is not the unqualified "triodes are superior" but that "ZNFB or low NFB Class A1 triodes and trioded pentodes are superior on the ear of qualified listeners to any other form of reproduced music". My best amp ever, in the ears of my test groups of professional performers, is PP EL34 Class A running ZNFB or less than 6dB of NFB; I usually test it against expensive 300B gear either of my own decising or commercially available and *behind the curtain* it usually wins. Quite a lot of ppl dislike multigrids connected in triode but in my mind EL34, KT88, 6550, KT90 make great triodes when triode connected. And they measure often better than the data curves suggest they would because so often the gear used in 1955 to get those very data curves were not all that accurate. Next, the problem with pentode sound is not so much the pentode itself but the amount of NFB required to get it to deliver its power with any pretense to quality, the NFB creating recombinant harmonic artifacts which are very disturbing to qualified listeners. The "pentode itself" is a wild thing, and also the beam tetrode, and these create a large number of odd and even numbered harmonics some of which change phase relative to fundemental depending on the load. PP operation in heavy class A is the only way to get decent pentode class A performance without using the huge total amounts of NFB found in EAR509, McIntosh etc, and its the huge amount of NFB needed to suppress the re-combinant artifacts arrising from too little FB around a poor performer to begin with, because in the two brands mentioned the output tubes were nearly in class B bias conditions. McIntosh got 50 watts from a pair of 6L6 in 1949. ppl later were able to extract 80 watts from a pair of 807, 100 watts from EL34, and 140 watts from KT88, and all to no avail with hi-fi apps without NFB the THD was horrendous. But with for pure class A the power output is up to 45% of the total anode dissipation of the output tubes and the harmonics are at least not entirely evil up to 10% of the total class A power and all people actually use on most days/evenings at home, so hence a pair of 6550 in beam mode in class A were quite a good proposition, giving 32 watts of class A with THD under 3% if each tube was biased at 36 watts, and not much worse than the class A1 triode connected maximum of 16 watts. At 2 watts which connection prodices the least THD in heavy class A1? Hence the Rout of the amp is the issue, rather than abominable distortion products. UL operation was a godsend to amp makers struggling to argue that beam or pentode amps were as good as triodes, which seemed to get the thumbs up amoung the cognescenti as well as convince the engineers that although triodes were inefficient, they did measure rather well as Williamson had proved. UL reduced pentode and beam THD to less than triode levels, the spectra was nicer, and you didn't need a huge amount of global NFB to reduce Rout. Peter Walker's Acoustical also got a message across to the cognescenti; his tricks didn't sound too bad. Walker patented his connection, and nobody else could use it, or I am sure they would have. I use it in my amps; it is a very nice way to hook up output tubes. BUT, and there is always a but, or several buts, the drive amp is very important in all amps using low gain output stages. Thus, if we use any NFB at all in the levelling process, as in the CFB example, the question is opened up of the recombinant harmonic effects of NFB *which we are then adding to only one of the contestants*, the pentode, to disadvantage the pentode yet again. The Main reason for NFB around pentodes and beam tubes is to reduce their Ra relative to RL to get at least a triode damping factor or better. To avoid the worst of ecombinant NFB artifacts the NFB should be "gentrified" and in practice this means using the UL connection or the Acoustical which avoids the necessity to keep the screen at the same signal potential as the cathode. Peter Walker did know what he was doing when he offered the world his Quad II design and anyone who has fooled around with iron, wire, and multigrids will realise that although triode connection is extremely unlikely to be sonically poor, the Acoustical multigrid connection is not any poorer, measures pretty well, and produces lots more ceiling without an extra pair of KT66. I have even placed KT90 into Quad II amps with a revised PS and slight mod of the input driving pair of EF86 pentodes and got some great results. http://www.turneraudio.com.au/quad2powerampmods.html Along with pentode or beam output tubes, let us definately NOT forget little input pentodes...... Patrick Turner. Andre Jute Visit Jute on Amps at http://members.lycos.co.uk/fiultra/ "wonderfully well written and reasoned information for the tube audio constructor" John Broskie TubeCAD & GlassWare "an unbelievably comprehensive web site containing vital gems of wisdom" Stuart Perry Hi-Fi News & Record Review |
#155
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
Andre Jute wrote: John Byrns wrote: In article .com, "Andre Jute" wrote: I'm reading this part of the thread with great pleasure; it reminds me of the glory days of RAT when high theory was everyday meat on the table. There's some confusion here about which tubes we're talking about. Perhaps I am to blame for it, in phrasing the thread headline too bluntly as "triode superiority". But the results of tests and my actual opinion, which comes as no surprise to you, is not the unqualified "triodes are superior" but that "ZNFB or low NFB Class A1 triodes and trioded pentodes are superior on the ear of qualified listeners to any other form of reproduced music". My best amp ever, in the ears of my test groups of professional performers, is PP EL34 Class A running ZNFB or less than 6dB of NFB; I usually test it against expensive 300B gear either of my own decising or commercially available and *behind the curtain* it usually wins. Next, the problem with pentode sound is not so much the pentode itself but the amount of NFB required to get it to deliver its power with any pretense to quality, the NFB creating recombinant harmonic artifacts which are very disturbing to qualified listeners. Thus, if we use any NFB at all in the levelling process, as in the CFB example, the question is opened up of the recombinant harmonic effects of NFB *which we are then adding to only one of the contestants*, the pentode, to disadvantage the pentode yet again. Hi Andre, Please cut me a little slack if I have missed something, as I wasn't in on this thread from the beginning. The necessary information is spread over many threads and my netsite. I don't understand your apparent thesis that the triode is superior, if indeed your "PP EL34 Class A running ZNFB" is your best amp ever, as judged by your "test groups of professional performers"? The headline is a provocative outgrowth of a bad-tempered set of threads on whether a triode has native or built-in negative feedback. I have always said that best amp I ever built was the trioded EL34 PP. In fact I spelt it out earlier in this thread as well, but nobody took the slightest notice. Are the EL34s in this amp triode connected with the screen grids strapped to the plates? Yes. If this is the case, the implication is that an EL34 with the screen tied to the plate makes a better triode than the 300B used in several of your other amp designs? Yes. The comparison is ZNFB or very low NFB triode strapped EL34 PP against SE ZNFB 300B. In level matched tests with a third, usually solid state placebo amp also behind the curtain, a sophisticated listening panel will choose the PP EL34 pseudo-triode amp every time. (The SS amp commonly gets no votes.) There are complications, in that when you double up the tubes on each, to compare PSE and PPP, all other factors equal, except that now you can use better speakers, specifically ESL63, the results get a bit closer simply because the speakers offer more resolution to a test panel accustomed to listening to details, and that big broadcasting DHT like 845 and SV572-xx whip little EL34, but for practical, level-matched tests at the entry level -- not meaning money, for all these amps are expensive, but the entry level of very high resolution -- sure as hell triode strapped PP ZNFB EL34 hold the upper hand. You have to be in PP 300B at umpteen-eleven times the price, with a lot of high-ticket high-carriage iron on board, before the natural triodes make as good a showing as the humble, inexpensive, ultra-versatile EL34; by then it is clearly no longer a fair test (but then, in terms of high street hi-fi prices even the base EL34/300B test has only a tenuous connection to what most people would regard as reality). I have heard music via a pair of made in HK 18watt Vincent PP amps using two 300B which to me did all the business really well. it wasn't an expensive amp in1995, when I used to attend the NSW Audiophile Society's meetings to see what other folks got up to. For PP triodes and 18 watts you need about 400V and 150mA, and an OPT with about 8k : 8 ohms, and your'e away. The same PS could be used for a pair of EL34, and get less power, or a pair of 6550 in triode for maybe slightly more power. But the iron need only be about the same and not cost a lot' the cost is mainly in the higher costs of the 300B and perhaps in its needed pair of 5V filament supplies. I personally don't see any point to using EL34 in triode when for such a small extra outlay 6550 or KT88 in triode could be used and which would have lower Ra-a and more power which is very much worth that slight extra outlay. I give you the additional information because we need to be clear that what I have proved isn't that PP amps are better than SE, merely that the EL34 is a stonking good tube, especially at the price. Also don't forget the EL34 replacement called the 6CA7, a beam tetrode, also an underated tube. I had to do a minor repair on a 'THD" branded SE guitar amp last week which had a Russian 6L6GC running with 45 watts of anode dissipation with the anode only a tiny bit red in one place. It was fine at 35 watts. EL34 will glow red with less Pda than this 6L6GC. So I it seems the humble 6L6GC now made with its actual higher pda capability well above the original 22 watt rating could produce good triode power in a PP circuit. The test only incidentally pitches a PP against an SE amp; I had long before this discovered that SE or PP matters not when both are made with triodes running in Class A and you have a very light or absent hand on the NFB; what I set out to prove in these later tests with the PP trioded EL34 was something about Class A sound, and something about the composition of distortion, and something about the lowest imperceptible level of loop or anyway additional NFB. I tried PP EL34 in triode in a pair of Quad II amps for 13 watts with 6db of global NFB and the owner said it lacked the sparkling dynamics of a 10 watt amp using 6GW8 in class AB1 and about 17dB of GNFB. Only when i went to KT88 in triode with 12 dB of GNFB did the owner say that the Quads then sounded as good as the UL amp and that the bonus was a slighty higher ceiling because the KT88 in triode with Ea at 400V and fixed bias the amp gave 20 watts AB1. There is lots about modding Quad II at http://www.turneraudio.com.au/quad2powerampmods.html I conclude from my observations and from what customers tell me that 12dB of NFB is fine in a triode PP amp. Nobody I know can tell the difference between a normal triode amp with 12db of GNFB and the same amp with its screens taken to UL taps and with the FB left the same. You also seem to be saying that the negative feedback that is inherent in the operation of a triode is not responsible for its superior sound, On the contrary, it seems to me very likely that the negative feedback inside a triode is responsible for the quality of its sound. To me it is a very persuasive argument that to bring a pentode back even partway towards the silence of a triode you have to use the NFB that was taken away to make it a pentode! Well, yes, but impossible.... However, it is clear to me that external forms of NFB, as in loop feedback, UL, CFB, etc, act differently to the native or internal NFB of a triode. This is easily seen in comparative harmonic output analysis of an EL34 amp with a pentode/UL/triode switch and a pot or switch to alter the amount of loop NFB, as in my T113 "Triple Threat" PP EL34. Well indeed there IS a difference in FB transfer between external and triode internal transfer. Local or global NFB loops are normally employed in linear mechanisms; ie the CFB of an OPT with CFB is where the cathode voltage is magnetically locked to the energy elsewhere generated. The fed back voltage is exactly the same in spectral content at the cathode as at the anode' only the phase is opposite. The local CFB with screen fully bypassed to cathode acts to reduce all the spectra of the pentode without NFB, and most pentodes suffer double digit THD at clipping when tested to gain the published data for SE use; 13% THD is about what is quoted, and a horrible mix compared to the 5% of the trioded case, with most only 2H. Note that these figures are for the rated load use stated for the tube; Loading means a heck of a lot with output tube distortions. Unloaded, the pentode is abysmally horrible. But when CFB or external GNFB is added, the higher the load, the higher the gain so for a given value of ß, the greater tha amount of NFB applied, so although the THD rises as load increases with no NFB, it is limited in rise when a given amount of NFB is applied. The triode THD reduces to almost nothing when load value is very high, indicating its inmternal FB is maximally effective, similar to the case od the unloaded pentode with a given amount of CFB as i calculated in a previous case for the EL34. When loaded though, the gain of the triode goes lower, and the internally applied amount of NFB becomes lower but still present, and the transfer of the NFB becomes maximally affected by the nonlinear 3/2 voltage effect on Ia. A similar thing occurs in UL, but intermediate between triode and pentode, and somewhat unique, ie, different to either triode or pentode. A UL amp with 43% taps and EL34 may make a maximum of say 32 watts, and in triode its 14 watts, but the UL amp will have no more THD than the triode amp at 5 watts, and be spectrally as clean. Trouble is the Rout. The UL amp has Rout about = RL, and GNFB must be used to further reduce Rout. In the past many makers competed with each other to get the highest possible amount of GNFB applied around their UL amps to be able to say they had 0.03% THD at 45 watts. 30db of applied GNFB was not uncommon in top brands using well made OPT. Meanwhile all the makers with lesser quality OPTs had to settle for no more than the standard 20dB of GNFB and 0.3% THD at clipping for class AB1. Now 17db is a lot of GNFB around any tube amp. In actual fact, few could really tell the difference between samples of all the above amp recipes at average normal listening levels of 86db, or 1/2 a watt average to each of two speakers. Nobody is going to win any prizes for applying more than just enough GNFB around any tube amp these days. Sanity has prevailed; we have done away with the uneccessary. A UL amp with 50% taps and in class A PP with 6550 typically makes 1% THD of mainly only 3H at about 30 watts and without GNFB. At 3 watts, output voltage is 1/3 of clipping, and THD is about 0.2%, and at 1 watt its THD is about 0.11%, and THD is already so low that it isn't of any concern and would not be discernible by 99% of the population. However, the Rout would seriously colour the sound, depending on the vaguaries of speaker Z so if 15dB of NFb were applied, the typical Rout without NFB of about 5 ohms if the secondary winding was a match for 5 ohms is reduced to about 0.7 ohms. All is well if the OPT has wide BW to start with and the driver amp is suitably set up. people knock UL amps with 6550, but it has been my experience to seriously dent all the egos of other makers of amps with triodes and low GNFB with samples of mine in demonstrations with the audiophile club, and while using 1/2 the output tube count. I have nothing against trying to use zero GNFB and rely on just the natural low Ra of the triode. But I would say that then one is compelled to use wide BW OPTs, lest the OPT unecessarily act as a filter, so don't try an old Leak Point 1 TL12 without its GNB in triode; the OPT has 50mH of leakage inductance! And one wants low winding loss OPTs, and with a nice high voltage ratio to give the output tubes a higher than normal AB1 loading. So when thinking of EL34 in triode I'd be thinking of 10k : 6 ohms OPT ratios, using Ea = 425V, Ia = at least 50mA. The Ra-a of the triode = 2.5k, and this is transformed to 1.5 ohms by the OPT Z ratio of 1,666:1. Then one adds the effective winding resistance of say 0.4 ohms at the sec and you still get Rout at about 1.9 ohms. This is one reason why not to use EL34 in triode when 6550, KT88, or KT90 will all give you 1.5 ohms under the same conditions. But the larger tubes allow a higher safe Ea at the same Ia, and so RLa-a can be higher, say 12k : 6 ohms, or 2,000 : 1 so the Ra-a of 1,800 ohms is reduced to only 0.9 ohms, and plus Rw, Rout = 1.3 ohms. 300B will do slightly better since the Ra is slightly lower than any of the larger octal power tubes in triode. The benefit of having each output tube seeing a class A load of 6k each is that the THD also is quite low, and negligible at ordinary listening levels. do you have any clues yet as to exactly what it is that is responsible for the superior sound of the triode and the triode connected pentode? Sure. The difference in sound between a triode (natural or screentied) and a pentode is caused by the different makeup of the harmonics. Especially the IMD caused by the varying gain with load changes. In the pentode, these are high, in the triode low, and its a function of the Ra.... Regardless of the absolute level of distortion, the triode, however made, has a lower proportion of odd and higher harmonics than the pentode. It is known that very low percentages of odd and higher harmonics can cause uneasiness in listeners. This very true, and when i built a pair of SET 4 watt amps using 2A3 they sounded cleaner than the same power from an EL84 in pentode with global NFB. However, I do have a pair of SE amps with EL84 with about 10% CFB and slight NFB and these are remarkably triode like, and very listenable. I think the 2A3 gives the best 4 watts possible out of all devices in the universe. It is strong supporting information that the difference between a good 300B amp and a bad one can be read in the proportion of 3rd harmonic left on the bad amp (as you yourself showed in a table when you compared my HVHCHL "Hedonist" to the distrastrous BobC/LaFevre/Magnequest "Bubbaland" amp roundabout 1998 -- that was a big clue on the way to my present understanding -- thanks!). The easiest way to prove the importance of the proportions inside the THD to yourself is to put a pentode/UL/triode switch and an NFB pot on a pentode amp and to study the makeup of the distortion in each mode at various levels of NFB including zero NFB. Of course, it took me a bit longer than a single sentence, because I had to eliminate all the common factors between the various topologies (I wasted years on AC balance in PP amps) before I isolated the odd factor and could examine it closely. There is no other parameter where the change is so striking and so inexplicable. Even at zero NFB, the third and higher harmonics fall through the floor when you turn the triode switch on a pentode amp. Well, switching that switch from pentode to triode is connecting the triode FB. Ra falls heaps, the iron distortions and intermods are much suppressed, and sure, the THD reduces along with IMD. Some ppl may not initially like the triode sound if the pentode with FB gave increased bass and treble due a speakers higher Z below 100Hz and above 500Hz, and this was the case in old radios, where the FB was omitted not only to save having to have another tube for the extra open loop gain, but to allow the high pentode Ra to compensate the miserable F response from the AF detector of the radio, typically 150hz to 2 kHz. But with hi-fi speakers powered by current sources, which is what a pentode amo is without its GNFB, the outcome is a bit tricky indeed. Of course, before you even get there, you notice that in the trioded pentode, the THD is lower than in the other two modes (pentode, UL) at any (low) level of NFB. It is difficult not to conclude that the effective removal of the pentode's screen grid adds additional NFB which works in some internal way. It is equally difficult to see (and hear) the different result of this internal NFB and not to conclude that it works differently from external feedback. I have posulated that the triode FB works along paths which act on a 3/2 voltage to current rule. But where very small changes level occur, the changes in THD are virtually linear because one acts along what is such a slight amount of bend in the Ra curves. When plotting the THD for a given class A amp on linear voltage output versus THD % scales the THD rises slightly faster than a straight line. So a given amp may make 5% at 20Vrms into a load, and 2% at 10V. One might draw a curve through the 0.0V and the 10v and 20V points and find that's the average for several samples of that amp. But wide variations in THD occur between samples; tubes are not always matched. In Quad II samples i have worked on with older tubes it isn't unusual to find 5 times the rated data THD. Swapping positions of the KT66 and or the EF86 can drastically reduce the THD. Usually they measure a whole lot better when all R&C are replaced, but still its worth final swapping of known healthy tubes to get the THD to measure low; and voila, usually the sound is blameless while every effort has been made to accomodate the imperfections of the items within the system. Class AB usually gives a slowly rising THD profile at first which is virtually the same as full class A, but when the threshold of AB commences the THD rises much more rapidly, and graphs of THD with kinks showing the THD increases rapidly with AB action. pentode amps are worst in this regard because their gain is prportional to load,, since gain = gm x RL, and RL reduces to half the class A load ( or to 1/4 RLa-a ) when one tube cuts off. Triode gm tends to rise and Ra get lower with increasing anode current so the AB transition produces less odd order harmonics; principally less 5H. The principle reason why PP triode output tubes in class A are so linear up to clipping where 1% THD is not uncommon is that the decline in gm with lessening Ia is about equal to the increase in gm with increasing Ia. The actual load seen by one tube of a PP pair affects the load seen by the other, and in fact the loads are curved lines for each tube. The PP amplifier is so inherently linear and majestically simple that it has stayed in favour for all these years............ But its hard to get a PNP transistor to act just like its NPN companion. They just don't come in twin pairs; more like distant cousins. Thus my fear that, if you use any form of external feedback to make the triode and the pentode "electrically equivalent", you will disadvantage the pentode because the makeup of its harmonics will be adverse *and will be heard by the panel*. This is the fundemental problem with comparisons of pentode with loop FB and triodes, arranged for the same Ra. However, I have used CFB in my SE35, and with screens bypassed to 0V and with Eg2 much lower than Ea to minimise Ig2, and got only about 1.5% THD without GNFB at 35 watts from a quad of 6CA7. With such a small amount of 2H at the output stage, it was easy to cancel it by the triode driver stage without setting up the driver with an adverse load to deliberately create 2H that will cancel in the output stage. The amp is designed to give maximal THD cancelations in the 4 to 6 ohm load range and the benefits seem real sonically during the tests I have done with SEUL amps using 13EI for a comparison. I only had to apply a small amount of GNFB in the SE35 to "complete the picture". The owner now uses NOS EL34 instead of the Sovtek or EH6CA7, and perhaps the harmonics produced are even more benign than they were when supplied with the Russian 6CA7. I don't actually see any alternative way which won't turn the pentode into a triode, which defeats the purpose of a straight-up challenge, so I think you and Patrick are on a wild goose chase. But that has never stopped us speculating before! One can only compare two different things to be more aware how each works. I sometimes wonder what the folks on different planets in the many universes out there thought when they discovered thermionic emission in vacuum tubes, and farnarcled around with 2 electrodes, then 3 electrodes, then 4, then 5, then 7 then 9.............. And we can't advise those who have not yet discovered tubes, and them out there can't advise us of what to look out for next. We are profoundly alone so far, scratching at the surface of the very hugely unknown............. Patrick Turner. Regards, John Byrns These, incidentally, are the same test results I interpret as meaning that there is no such thing as an SE sound, that what we hear which so appeals is a Class A ZNFB or very low NFB sonic signature. People just confuse Class A with SE because only a tiny, tiny minority even of the ultrafidelista have ever heard pure Class A PP amp. HTH. Andre Jute Visit Jute on Amps at http://members.lycos.co.uk/fiultra/ "wonderfully well written and reasoned information for the tube audio constructor" John Broskie TubeCAD & GlassWare "an unbelievably comprehensive web site containing vital gems of wisdom" Stuart Perry Hi-Fi News & Record Review |
#156
Posted to rec.audio.tubes
|
|||
|
|||
Henry Pasternack's Norton triode model???
In article ,
Chris Hornbeck wrote: On Wed, 11 Oct 2006 01:53:36 GMT, John Byrns wrote: OK, after correcting that typo has anyone been able to make sense of Henry's Norton model? As near as I can tell, for the plate circuit to express the Child-Langmuir equation, it is necessary for Henry's nonlinear admittance to be a positive one-half power function of Vp, not a negative one-half power function of Vp as Henry said, is this another typo? Algebraic models have no resonance for me. A model that gets me all hot-n-bothered must start (or be referential to) a physical model. My personal physical model of triode behavior is of a transconductance "engine" with a mu multiplier, just because it's kinda-sorta like the actual machinery. The cathode makes an electron cloud, the grid turns that into a transconductance device, geometry gives a mu. Like that. How is your "transconductance engine with a mu multiplier" anymore "kinda-sorta like the actual machinery" than is my model? In reality your model is a subset of my model, it is simply my model with the 3/2 power nonlinearity and the Vp feedback removed. The addition of the 3/2 power nonlinearity and Vp feedback actually make my model more like the actual machinery, than is your model, by incorporating some of the inherent distortion producing mechanisms that exist in the real triode machinery, and by providing an approximation of the large signal behavior of the triode machinery. Child's equations are pre-War approximations, not revealed Truth. They have no physical basis. Yes, my model is cribbed from prewar vacuum tube engineering texts, and while it is an approximation, it is clearly a much closer approximation to the actual triode machinery than is your simple linear model. I would say my model contains considerably more "revealed Truth" than does a simple linear small signal model like yours. These same old pre-War texts also tie the physical geometry into all this theory, although it has been years since I have looked at this old vacuum tube theory, and I never paid much attention to the details of exactly how the geometry ties into the theory. I simply can't understand upon what basis you claim any physical basis for a liner model like yours when a triode is anything but linear? For me, this numbers bantering is both wrong-headed and boring. Excuse my French, but it's totally unrelated to the issue of choosing a defendable model. What "numbers bantering are you talking about, and what is the proper procedure for choosing a "defendable model"? Perhaps if you can give me an example of how you would defend your model, I could put up a defense of my model. But, FWIW, your SPICE model seems, to me, totally disconnected from physical reality. Educate me? As I said above it is cribbed from pre-War text books, a period when technology had reached a pinnacle that has remained unchallenged ever since. However far it may be removed from reality, it is clearly more closely connected to reality than is your simple linear model which is simply a degenerate subset of my model, less closely connected to reality. Regards, John Byrns. |
#157
Posted to rec.audio.tubes
|
|||
|
|||
Explanation still required for triode superiority
"Andre Jute" said:
These, incidentally, are the same test results I interpret as meaning that there is no such thing as an SE sound, that what we hear which so appeals is a Class A ZNFB or very low NFB sonic signature. People just confuse Class A with SE because only a tiny, tiny minority even of the ultrafidelista have ever heard pure Class A PP amp. Well, I have, since that's about the only topology I use when building tube amps. I compared triode strapped JJ KT88s against NOS Philips XF4 EL34s, also triode strapped in the same circuit, with somewhat reduced supply voltage to not overstress the EL34s. For me, the KT88s won hands down, even with the lower anode voltage. I posted the circuit diagram some days ago in the "Step 2 for Fisher Iron" thread. You can still find it here, if you're interested: http://img508.imageshack.us/img508/1751/wkschemod0.jpg Unless one wants a SE amp, it doesn't get any simpler than that, I'm afraid. -- "Due knot trussed yore spell chequer two fined awl miss steaks." |
#158
Posted to rec.audio.tubes
|
|||
|
|||
Henry Pasternack's Norton triode model???
"Chris Hornbeck" wrote in message ... : On Wed, 11 Oct 2006 01:53:36 GMT, John Byrns : wrote: : : OK, after correcting that typo has anyone been able to make sense of : Henry's Norton model? As near as I can tell, for the plate circuit to : express the Child-Langmuir equation, it is necessary for Henry's : nonlinear admittance to be a positive one-half power function of Vp, not : a negative one-half power function of Vp as Henry said, is this another : typo? : : Algebraic models have no resonance for me. A model that gets me : all hot-n-bothered must start (or be referential to) a physical : model. : : My personal physical model of triode behavior is of a transconductance : "engine" with a mu multiplier, just because it's kinda-sorta like : the actual machinery. The cathode makes an electron cloud, the : grid turns that into a transconductance device, geometry gives : a mu. Like that. : : Child's equations are pre-War approximations, not revealed Truth. : They have no physical basis. err, of course they do. Terman gives an almost too simple deduction of plate voltage vs current. it goes like this: suppose you have a tube running I amps at V volts across. this means that the field set up by the space current is just large enough to offset the plate's electrostatic field (for that type of tube, at that current). now move the plate to 4V. to offset that, you need 4 times as many electrons (charge). we know that the kinetic energy of the electrons - supplied by the plate to cathode field - is now 4 times as large, so (1/2 mv^2) the electron speed is doubled. so it takes 4*2= 8 times as many electrons per time unit (current) to balance the new plate voltage. so Ia=k*Vak^3/2, with k some geometrically determined constant. Rudy physical evidently : : For me, this numbers bantering is both wrong-headed and boring. : Excuse my French, but it's totally unrelated to the issue of : choosing a defendable model. : : But, FWIW, your SPICE model seems, to me, totally disconnected : from physical reality. Educate me? : : : : OTOOOH, what a wonderful discussion. And an even more wonderful : digression from... well, you know, all the other stuff here : sometimes. : : Much thanks, as always, : : Chris Hornbeck |
#159
Posted to rec.audio.tubes
|
|||
|
|||
Henry Pasternack's Norton triode model???
"Ruud Broens" wrote in message ... ....so Ia=k*Vak^3/2, with k some geometrically determined : constant. : better formulated as: so, by induction, we know Ia is proportional to Vak^3/2, the multiplier k being dependent on geometry and cathode emission : Rudy : physical evidently |
#160
Posted to rec.audio.tubes
|
|||
|
|||
Henry Pasternack's Norton triode model???
On Wed, 11 Oct 2006 17:11:42 GMT, John Byrns
wrote: The cathode makes an electron cloud, the grid turns that into a transconductance device, geometry gives a mu. Like that. How is your "transconductance engine with a mu multiplier" anymore "kinda-sorta like the actual machinery" than is my model? The nonlinearities in typical triodes originate mostly in the transconductance engine. I'm not suggesting a linear model. My concerns with proposed algebraic models come from their conceptual framework. They seem like elaborate inversions designed to "prove" a pre-determined outcome. For example, any realistic model would have the transconductance device "first", because it literally is first. Some time delay and some further geometric effects follow. Much thanks, as always, Chris Hornbeck |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
What are they Teaching | Audio Opinions |