Triodes and feedback theory
A few days ago jan.didden posted an article written by Jerald Graeme of Burr-Brown fame. The article is aimed at feedback in op-amps. Jerald Graeme and his crew have been prolific in the number of white papers and books they have written. Almost every document references the original electronics feedback paper written in 1934 by H S Black. You might want to take a look. https://ia802703.us.archive.org/24/items/bstj13-1-1/bstj13-1-1.pdf Feedback theory is now applied to medical research and many other disciplines beyond electronics. Triodes Too?
I went looking at Amazon to see if Graeme had written a book on this topic, in fact he has: Optimizing Op Amp Performance 1997. I purchased a “like new” copy for about $40. The single most interesting thing to me that Graeme had to say was this: “intuitive observations, guide the model derivation, and then comparison of the model’s transfer response with that of the circuit confirms the model’s validity”. How about that, sharing the process, and not just the results as if it was correct the first time. I think of it as intuitive deduction, starting with you have worked hard to learn, filling in the blanks with what you think best fits and then do reality testing in the lab, wash rinse and repeat. See page 30 heading 2.1.
In chapter 1 Graeme applies Black’s feedback model to the non-inverting input of the op-amp. Graeme develops the concepts including feedback factor represented by β, β+for positive feedback and β- negative.
In Chapter 2 Graeme applies and adapts Black’s feedback model to the inverting input of the op-amp. Chapter 2 is where the feedforward factor represented by α is introduced. α accounts for the inverted op-amp input, and attenuates the Vinput.
Too many words, more about triode non-inverting and inverting outputs later.
DT
A few days ago jan.didden posted an article written by Jerald Graeme of Burr-Brown fame. The article is aimed at feedback in op-amps. Jerald Graeme and his crew have been prolific in the number of white papers and books they have written. Almost every document references the original electronics feedback paper written in 1934 by H S Black. You might want to take a look. https://ia802703.us.archive.org/24/items/bstj13-1-1/bstj13-1-1.pdf Feedback theory is now applied to medical research and many other disciplines beyond electronics. Triodes Too?
I went looking at Amazon to see if Graeme had written a book on this topic, in fact he has: Optimizing Op Amp Performance 1997. I purchased a “like new” copy for about $40. The single most interesting thing to me that Graeme had to say was this: “intuitive observations, guide the model derivation, and then comparison of the model’s transfer response with that of the circuit confirms the model’s validity”. How about that, sharing the process, and not just the results as if it was correct the first time. I think of it as intuitive deduction, starting with you have worked hard to learn, filling in the blanks with what you think best fits and then do reality testing in the lab, wash rinse and repeat. See page 30 heading 2.1.
In chapter 1 Graeme applies Black’s feedback model to the non-inverting input of the op-amp. Graeme develops the concepts including feedback factor represented by β, β+for positive feedback and β- negative.
In Chapter 2 Graeme applies and adapts Black’s feedback model to the inverting input of the op-amp. Chapter 2 is where the feedforward factor represented by α is introduced. α accounts for the inverted op-amp input, and attenuates the Vinput.
Too many words, more about triode non-inverting and inverting outputs later.
DT
Yes very good book.
As to studying feedback/feedforward on a triode, the best way to start may be to model the triode as any other (opamp like graphic) block with two inputs and an output. All the known feedback equations can then be applied, and it is the open loop gain expression that locks it in.
Just like with opamps there are more inputs like the power supply connections / anode. Possibly these can me modeled as part of the open loop gain.
If this works, it would open up a huge knowledge base to the triode without a lot of reinventing the wheel.
Another book tip: David Mindell 'Between human and machine - Feedback, Control and Computing before Cybernetics'.
The title doesn't give it away but a large part of the book is devoted to the struggle to apply what Harry Black developed to everyday problems in feedback control. For instance, at the start of the war BuOrd (the US Navy ordnance development organisation) saw the lesson from the Blitzkrieg and the need for accurate anti aircraft fire control. A classical feedback problem. Yet the Navy was very much focused on manual control for big guns. After all, an enemy battle ship doesn't move that much while your shells are on their way towards it. You look were they land and adjust your elevation and azimuth (or train as the Navy called it) and fire again. Human in-the-loop feedback.
With aircraft it's much different. Those f*cking pilots refuse to fly level with constant speed. Then those Japanese Zero pilots added insult to injury by changing their altitude as well, all the time! By the time the shell arrived where you thought the plane was, it was already a mile away. Good firing directors can predict a lot from pilot behaviour, especially those who were former pilots, but basically it was a guessing game.
Interestingly, the process to smooth track history and extrapolate to a future position is EXACTLY the same as electronic filtering problems, too much smoothing and you miss the next manoeuvre, not enough smoothing and your aim point jerks all over the place.
Anyway, what I was getting at was that while at Bell Labs they were routinely building feedback amps, traditional Navy developers like at Sperry Gyroscope still say a servo amp as an open loop amp with the input-output error returned to the input. They did not realise it is a closed loop, and couldn't get their head around the stability problems they got! Only after Bell Labs and MIT got involved in the firing director problem did the two realise they were talking about the same thing! Then Bode and Nyquist chipped in and that was the start of our current body of knowledge on feedback and feedback stability. Fascinating!
Buy the book.
Jan
Last edited:
Hello,
In the Chapters 1 and 2 of the book Optimizing Op Amp Performance the op-amp is treated as the basic building block. Black’s feedback model was adapted to the 3 legged creatures with one output, one non-inverting input and one inverting input. On paper the op-amp conceptually looks pretty basic. Zoom in on the schematic of an op-amp made of discrete parts and it does not look nearly so simple after all. For example take a look at this op-amp made up of FET’s.
http://www.forsselltech.com/media/attachments/JFET_Opamp.PDF
I want to start more basic, perhaps a JFET or a vacuum triode. I like the idea of a triode better. I want to approach the feedback model in a fashion similar to triode noise modeling. Each source of noise is computed as if it is an Ein at a single input, the grid. With a model that assumes a single input we need to address the two potential triode outputs and program the feedback math (loops) so that feedback from these outputs to the single input matches actual circuit performance.
Assuming the grid is the single input what are the two potential outputs? We only have the anode and cathode to look at. Say we apply a DC power supply to the triode, positive voltage to the anode and negative to the cathode. Not so much voltage as to melt anything. Next we apply a AC voltage to the grid making sure that the grid is always negative in relation to the cathode, we do not want the grid to be a second anode. If we probe around with our oscilloscope probe we will only see AC voltage at the grid not at either the anode or the cathode. We will see DC voltage across the tube and there will be current flow through the tube. Now add a resistor between the B+ and the anode, also add a resistor between the cathode and negative terminal of the DC power supply. Try probing around again, we will find non-inverted AC voltage between the cathode and the cathode resistor and we will find inverted voltage between the anode and anode resistor. These are the 2 basic possible triode outputs, non-inverted and inverted. The feedback model will need to loop these voltages back or forward to the input and predict the behavior of a real circuit on the bench.
Jan, thanks for the book idea. I placed an order at Amazon. The fire control modeling you are speaking of is leading to next generation AI fighter planes with no pilot.
DT
In the Chapters 1 and 2 of the book Optimizing Op Amp Performance the op-amp is treated as the basic building block. Black’s feedback model was adapted to the 3 legged creatures with one output, one non-inverting input and one inverting input. On paper the op-amp conceptually looks pretty basic. Zoom in on the schematic of an op-amp made of discrete parts and it does not look nearly so simple after all. For example take a look at this op-amp made up of FET’s.
http://www.forsselltech.com/media/attachments/JFET_Opamp.PDF
I want to start more basic, perhaps a JFET or a vacuum triode. I like the idea of a triode better. I want to approach the feedback model in a fashion similar to triode noise modeling. Each source of noise is computed as if it is an Ein at a single input, the grid. With a model that assumes a single input we need to address the two potential triode outputs and program the feedback math (loops) so that feedback from these outputs to the single input matches actual circuit performance.
Assuming the grid is the single input what are the two potential outputs? We only have the anode and cathode to look at. Say we apply a DC power supply to the triode, positive voltage to the anode and negative to the cathode. Not so much voltage as to melt anything. Next we apply a AC voltage to the grid making sure that the grid is always negative in relation to the cathode, we do not want the grid to be a second anode. If we probe around with our oscilloscope probe we will only see AC voltage at the grid not at either the anode or the cathode. We will see DC voltage across the tube and there will be current flow through the tube. Now add a resistor between the B+ and the anode, also add a resistor between the cathode and negative terminal of the DC power supply. Try probing around again, we will find non-inverted AC voltage between the cathode and the cathode resistor and we will find inverted voltage between the anode and anode resistor. These are the 2 basic possible triode outputs, non-inverted and inverted. The feedback model will need to loop these voltages back or forward to the input and predict the behavior of a real circuit on the bench.
Jan, thanks for the book idea. I placed an order at Amazon. The fire control modeling you are speaking of is leading to next generation AI fighter planes with no pilot.
DT
This is the scientific method. See Karl Popper's book, "Conjectures and Refutations:
the Growth of Scientific Knowledge".
If you make a false assumption then you may draw false conclusions.DualTriode said:
Assuming conventional biasing, the normal valve input is grid-cathode (not grid) and the output is anode-cathode (not anode and cathode separately - the same signal current flows though both). Put the valve in a circuit, and then you might have circuit input from grid-negative supply and circuit output from cathode-negative supply (CF) but the valve sees this as exactly the same as the grounded cathode stage with circuit output from the anode-negative supply.
Good points. There never is 'a' input - input signal is ALWAYS between two nodes, just as output is always between two nodes. That is why a multimeter has two probes, you cannot measure a voltage on a point, only with respect to another point!
The reason why it is seldom looked at it like that is because it is assumed that the implicit reference to 'the' input is ground. In your triode, the grid is only 'the' input if you assume the cathode is grounded.
It is not by accident that 'voltage' is really 'potential diference' and difference needs two nodes.
Jan
The reason why it is seldom looked at it like that is because it is assumed that the implicit reference to 'the' input is ground. In your triode, the grid is only 'the' input if you assume the cathode is grounded.
It is not by accident that 'voltage' is really 'potential diference' and difference needs two nodes.
Jan
I have to say that there is a definite physical limitation in the comparison of feedback control of op amps to tube circuits. It generally only works perfectly in circuits that don't have to apply current. When I say this I'm speaking directly about integrated circuit op amps. IC op amps have huge voltage gain and then appropriate feedback to linearize it for the actual voltage gain actually needed. But the power dissipation requirements in any op amp that amplifies current won't allow IC op amps. There isn't any IC that I know of that will allow large current amplification to use in its feedback circuit.
This gets back to THE major advantage of valves over semiconductors in amplifiers for current drive to speaker transducers. Unlike most semiconductors there is soft limiting in valves as they reach the limit of the transfer characteristics in overdrive. You do not have the harsh audio character that comes from spray of higher harmonics as you come up against that limit that you have in semiconductors.
This is why there will always be this contest in sand amps to create the highest possible power output with the most number of semiconductors in parallel in the output stage. They just have to do that so that in the inadvertent case that a musical transient approaches a design limit there will be enough reserve power that it is unlikely to trigger that harsh sound. Of course it leads to numerous topologies avoid that either add complexity, cost, or heat and inefficiency.
So op amp theory really only has its place IN THE REAL WORLD in voltage amplification due to the cost and inefficiency of having super gain stages that amplify current. I think most semiconductor amplifier guys (not all) do not really understand this valve advantage that I can't see ever going away. At least in the final current amplification stage.
This gets back to THE major advantage of valves over semiconductors in amplifiers for current drive to speaker transducers. Unlike most semiconductors there is soft limiting in valves as they reach the limit of the transfer characteristics in overdrive. You do not have the harsh audio character that comes from spray of higher harmonics as you come up against that limit that you have in semiconductors.
This is why there will always be this contest in sand amps to create the highest possible power output with the most number of semiconductors in parallel in the output stage. They just have to do that so that in the inadvertent case that a musical transient approaches a design limit there will be enough reserve power that it is unlikely to trigger that harsh sound. Of course it leads to numerous topologies avoid that either add complexity, cost, or heat and inefficiency.
So op amp theory really only has its place IN THE REAL WORLD in voltage amplification due to the cost and inefficiency of having super gain stages that amplify current. I think most semiconductor amplifier guys (not all) do not really understand this valve advantage that I can't see ever going away. At least in the final current amplification stage.
LM3886? And a number of others...
The limit is usually the voltage limit of the power supply voltage, which the SS amp can pull very close to (and yes, it is abrupt.... but the SS can get very close to the rail, while a tube leaves lots and lots of supply voltage unused for that 'soft clipping'). Adding more devices in a SS amp won't let it pull any higher than the supply rail, it will only support more current without blowing up.
Heat and inefficiency? Seriously? Compared to a tube amp??
I made a class A, single ended SS amp recently, 60Watts output. Its steady state dissipation, and heat is nearly identical to that of a classic MC60 tube amp (push-pull and nearly class B)! Power tubes routinely burn more dissipation just keeping the filaments hot than most SS do to bias the whole amp.
Bwaslo,
All I would reply to you is that you may not know what you are talking about and do not know what you do not know. This is mostly because you have an orientation that precludes an open mind, i.e. confirmation bias. You're going into the weeds rather than recognize the intuitive facts staring you in the face. Why do you think musicians that use electronic amplification to reproduce their instrument's characteristic sound mostly use tube amps? (I'm thinking guitar amps here.) It's because there is a smooth transition from normal drive to overdrive conditions. They still haven't invented a guitar amp that plays as well at the limits of its power reproduction as a tube amp.
This fact is not limited to guitar amps. The soft clipping allows a much lower power hi fi tube amp to sound just as powerful as a much higher power transistor amp. It's not because of the tube sound. It's because if the same power level in a transistor amp hit the same short term transient level as the tube amp it will be immediately obvious that the transistor amp is clipping. Not so with the valve amp. These facts are starring you in the face. Where have you been?
All I would reply to you is that you may not know what you are talking about and do not know what you do not know. This is mostly because you have an orientation that precludes an open mind, i.e. confirmation bias. You're going into the weeds rather than recognize the intuitive facts staring you in the face. Why do you think musicians that use electronic amplification to reproduce their instrument's characteristic sound mostly use tube amps? (I'm thinking guitar amps here.) It's because there is a smooth transition from normal drive to overdrive conditions. They still haven't invented a guitar amp that plays as well at the limits of its power reproduction as a tube amp.
This fact is not limited to guitar amps. The soft clipping allows a much lower power hi fi tube amp to sound just as powerful as a much higher power transistor amp. It's not because of the tube sound. It's because if the same power level in a transistor amp hit the same short term transient level as the tube amp it will be immediately obvious that the transistor amp is clipping. Not so with the valve amp. These facts are starring you in the face. Where have you been?
I won't go into who is talking WAY beyond his understanding....
The reason that some (not even most) tube amp designs play louder than a SS amp of the same power rating is because those tube amps are rated where the distortion is some % value, but they can put out significantly more power at more % distortion. Which is another way of saying that their distortion gets higher than the manufacturer wants to admit to before the amp really reaches its max output voltage. That description doesn't seem as nice as "soft clipping" though. (Analog EE, here, too BTW). Whereas a typical SS amp will usually (because of the higher feedback it can use, before going unstable, than a transformer coupled tube amp could) maintains low distortion up to nearly the point where it runs out of supply rail. Which is another way of saying "hard clipping", which is an expression that implies that this is somehow a bad thing. Don't clip it, and not a problem. Clip the tube amp as hard, it will sound awful, too.
Don't you mean to CREATE their instrument's characteristic sound? Because an electric guitar, when you're listening only to the strings, doesn't even vaguely resemble the sound from the tube amp's speaker! (A former guitarist here, BTW). The amp is a creative instrument, not a accurate reproducer.
The reason that some (not even most) tube amp designs play louder than a SS amp of the same power rating is because those tube amps are rated where the distortion is some % value, but they can put out significantly more power at more % distortion. Which is another way of saying that their distortion gets higher than the manufacturer wants to admit to before the amp really reaches its max output voltage. That description doesn't seem as nice as "soft clipping" though. (Analog EE, here, too BTW). Whereas a typical SS amp will usually (because of the higher feedback it can use, before going unstable, than a transformer coupled tube amp could) maintains low distortion up to nearly the point where it runs out of supply rail. Which is another way of saying "hard clipping", which is an expression that implies that this is somehow a bad thing. Don't clip it, and not a problem. Clip the tube amp as hard, it will sound awful, too.
Hello Jan and All,
Okay okay okay. Voltage is a differential, a voltage source has a head and a tail and your Fluke DMM has both a red and a black probe.
Take a look at the Figures in Merlin’s new book; Merlin is very careful to show both positive end and negative end of each input and output. Also take a look at Graeme’s book, most often he only shows one end of an input or output. I suppose the exception is still the rule. Merlin is exceptional.
To quell the civil unrest let us assume that the black terminals of everything are connected in common; voltage sources, voltage outputs, voltage meters, most of the time that will be ground. Now you will tell me that will make the input voltage to the grid positive in relation to the cathode.
We have the ability to superimpose any DC bias that we want.
exeric and bwaslo please be mindful that this is about triodes and feedback, not soild state vs vacuum tube. We could be just as well be talking JFET and feedback. Solid state or vacuum tube it does not make any difference.
DT
Last edited:
Feedback is a circuit technique that can be applied to ANY circuit. The available literature gives masses of info how the limits of input impedance/current, gain, etc impact on the final result. A full treatise even includes forward propagation of the input signal through the feedback network to the output!
Saying that there is a physical limit to a circuit technique is barking nonsense.
It is true that the numbers and results from applying feedback to tubes are different from applying it to solid state or opamps, of course they are.
But the principles are always the same, the equations are the same, and the outcome is different because the input to the equations is different.
Jan
PS maybe a little tip: if you feel that Bill Waslo doesn't know what he is talking about, it must be because you are new here. Tsk tsk!
You may wish to work on your own credibility first. So far you're not making a whole lot of sense.
Last edited:
No, I will tell you that the input voltage to the tube is the difference between the signal voltage (against ground) at the grid and the cathode!
BTW Please do not mix bias voltage with signal voltage. I thought we were now interested in signal, not the static bias. Lets keep the discussion clean.
Jan
Last edited:
Jan,
The intent is to make the tail of everything common, reference the input to the same reference as the output and the tail end of the cathode resistor.
The notion of grid bias from the voltage generator will disappear as soon as we bias the grid with voltage across a cathode resistor.
For the discussion I did not want grid positive A2 operation, which would confuse everything.
DT
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.