After soliciting advice in an earlier thread for different ways to add negative feedback to a simple 12ax7 amplification stage, I think I now know how to better direct my questioning.
I have attached sketches of two different circuits. The first one is the known-good single triode stage. It is directly-coupled to a rudimentary electrostatic speaker of unspecified impedance (hence the high DC offset). But for the fact that it requires a bit more drive voltage than my smartphone can provide, it does everything I need it to.
In the spirit of eliminating the separate drive amplifier (a little solid-state thing) and utilizing the other triode in the 12ax7 envelope, I would like to graft on an input stage adding a little bit of gain and hopefully allowing for some negative feedback from the output in order to linearize the whole thing.
The second sketch is my proposed circuit for a direct-coupled current feedback amplifier (sans component values) that I would like to implement. I have been unable to find a guide for calculating the appropriate values for the desired operating points, so I'm hoping someone here can point me in the right direction. I put together a prototype trying to account for the interaction of the additional components, but the results deviated from my estimates pretty woefully.
I am very flexible on the values of the input stage and feedback components, but I would like to preserve the behavior of the output stage (if not the actual component values).
By my calculations, the known-good plate circuit represents a resistance of roughly 110k with no signal (250V / 2.3mA). For simplicity's sake, I decided to use a feedback resistor of 110k (I assume this is effectively in parallel with the plate circuit). Since this would presumably draw the same amount of current as the plate circuit, I halved the load resistor to 11k to preserve the 50V drop (50V / 4.6mA). This feedback resistor was then coupled to ground through the input stage's 470R cathode resistor in order to generate a bias voltage for that stage of approximately 1V [470R * (250V / 110,470R)] and set the gain at about 235 (admittedly more than I need).
I am clearly overlooking some factors since my experimental results are pretty far off of the calculated values. Any advice would be greatly appreciated.
-Nick
I have attached sketches of two different circuits. The first one is the known-good single triode stage. It is directly-coupled to a rudimentary electrostatic speaker of unspecified impedance (hence the high DC offset). But for the fact that it requires a bit more drive voltage than my smartphone can provide, it does everything I need it to.
In the spirit of eliminating the separate drive amplifier (a little solid-state thing) and utilizing the other triode in the 12ax7 envelope, I would like to graft on an input stage adding a little bit of gain and hopefully allowing for some negative feedback from the output in order to linearize the whole thing.
The second sketch is my proposed circuit for a direct-coupled current feedback amplifier (sans component values) that I would like to implement. I have been unable to find a guide for calculating the appropriate values for the desired operating points, so I'm hoping someone here can point me in the right direction. I put together a prototype trying to account for the interaction of the additional components, but the results deviated from my estimates pretty woefully.
I am very flexible on the values of the input stage and feedback components, but I would like to preserve the behavior of the output stage (if not the actual component values).
By my calculations, the known-good plate circuit represents a resistance of roughly 110k with no signal (250V / 2.3mA). For simplicity's sake, I decided to use a feedback resistor of 110k (I assume this is effectively in parallel with the plate circuit). Since this would presumably draw the same amount of current as the plate circuit, I halved the load resistor to 11k to preserve the 50V drop (50V / 4.6mA). This feedback resistor was then coupled to ground through the input stage's 470R cathode resistor in order to generate a bias voltage for that stage of approximately 1V [470R * (250V / 110,470R)] and set the gain at about 235 (admittedly more than I need).
I am clearly overlooking some factors since my experimental results are pretty far off of the calculated values. Any advice would be greatly appreciated.
-Nick
Attachments
I don't think you are going to get any gain out of the second stage because the cathode resistor will need to be fairly large to bias the stage correctly in this direct coupled arrangement, which will also kill all of the gain available.
I don't really have any suggestions on what to do differently because I don't know what you are trying to accomplish yet. What are the requirements--what does the stage need to do?
I don't really have any suggestions on what to do differently because I don't know what you are trying to accomplish yet. What are the requirements--what does the stage need to do?
He can use a bypass capacitor across the second tube's cathode resistor to bring back the gain.
This will be tricky to get right, though. Do you only want a 50V peak swing?
Last edited:
Thanks for the responses. I am using a 12ax7 because that is one of the few tubes I had on hand to experiment with. The fact that it works astonishingly well leads me to believe that I'm doing something right — I just thought it could be improved by adding another gain stage and implementing some negative feedback.
jerluwoo — The diagram you have attached is very much along the lines of what I would like to implement as a front end. The circuit with the 22k load resistor should effectively be though of as the "power stage". I'm not opposed to trying a similar design based around a 12au7 (or similar — the minimalism of the dual triode envelope appeals to me for some reason) and I really just want to wring a little more performance out of something that is already working quite well. More than anything, I'd like to understand how the components are interacting in the direct-coupled current feedback arrangement I sketched. I'm not asking anyone to do my homework for me, I'd just like to know where I might be able to find the information.
rayma — I omitted the power stage's bypass capacitor from the drawing since my immediate concern is simply understanding how the circuit is operating during quiescent conditions. It could absolutely be added later on — a question I was saving for later was whether it was beneficial to the negative feedback scheme to get as much gain as possible from the last stage. Yes, something approaching 50V peak-to-peak would suffice (I realize there will be losses as the rail voltage is approached — I could always compensate by slightly lowering the bias point to 240V or so).
kward — I think my response to rayma addressed the bypass capacitor issue. As stated in my original post, I am using this circuit to provide a 250V bias and +/- 50V peak-to-peak signal swing into a rudimentary electrostatic panel of unknown (but high) impedance. The rock-simple single triode stage I sketched actually accomplishes this quite well, I just felt it could be improved slightly and I'm curious about the workings of the direct-coupled current feedback arrangement.
jerluwoo — The diagram you have attached is very much along the lines of what I would like to implement as a front end. The circuit with the 22k load resistor should effectively be though of as the "power stage". I'm not opposed to trying a similar design based around a 12au7 (or similar — the minimalism of the dual triode envelope appeals to me for some reason) and I really just want to wring a little more performance out of something that is already working quite well. More than anything, I'd like to understand how the components are interacting in the direct-coupled current feedback arrangement I sketched. I'm not asking anyone to do my homework for me, I'd just like to know where I might be able to find the information.
rayma — I omitted the power stage's bypass capacitor from the drawing since my immediate concern is simply understanding how the circuit is operating during quiescent conditions. It could absolutely be added later on — a question I was saving for later was whether it was beneficial to the negative feedback scheme to get as much gain as possible from the last stage. Yes, something approaching 50V peak-to-peak would suffice (I realize there will be losses as the rail voltage is approached — I could always compensate by slightly lowering the bias point to 240V or so).
kward — I think my response to rayma addressed the bypass capacitor issue. As stated in my original post, I am using this circuit to provide a 250V bias and +/- 50V peak-to-peak signal swing into a rudimentary electrostatic panel of unknown (but high) impedance. The rock-simple single triode stage I sketched actually accomplishes this quite well, I just felt it could be improved slightly and I'm curious about the workings of the direct-coupled current feedback arrangement.
How to design valve guitar amplifiers is a good place to start. My advice is to use ac coupling. I know in the other thread they told horror stories of instability etc. etc. ,but, it's not such a huge issue that the choice of capacitor values can't generally solve. In all honesty, you do not need the second stage to achieve your output goal of +/- 50v peak. The circuit I posted has a gain of ~60 which will get your desired output level with 1v input. Without feedback and if your speaker doesn't actually load it down as you think, the distortion should be under 1%. However an electrostatic speaker is in essence is a variable capacitor by nature, increasing the loading on the gain stage as frequency rises, and varying as the membrane vibrates moving toward and away from the stator. The second triode would be better served as a cathode follower to buffer the voltage stage.
Last edited:
This was discussed at length in the previous thread. I tried using two lm317s in series as constant current sources, but the result was far inferior to the 22k resistor loaded circuit in my initial post in this thread in terms of both gain and distortion. Most of the proposed topologies using the other triode as an active load weren't workable since they require that the output be half the supply voltage.
I've attached a sketch including some measurements that I made after basically circuit-bending my initial design to try to and get close to the desired operating points. I don't like operating on the fly like this, but I was trying to make sense of things. The original second-stage cathode resistor of 680R should've pulled about 2.3 mA through the tube, with roughly the same amount effectively running in parallel to ground through the 110k feedback resistor. I had to raise the cathode resistor to 1k to even get 1.66 mA through the tube. The feedback path seems to be pulling far more current than it ought to.
Maybe my readings will make sense to someone.
Attachments
Shouldn't. Should be able to set the voltage drop across the triode acting as the active plate load to pretty much anywhere you want. Ex: To replace the 47K resistor in your latest diagram, you would need something like a 137 ohm cathode resistor, and connect the grid to the bottom side of that cathode resistor. Doing so will drop 106V across the tube and essentially "do" the same thing as the 47K plate load resistor at quiescent conditions, in addition to keeping current draw somewhat constant at AC conditions. Whether or not this would make a well performing stage is a different question, but it should be possible to meet the DC conditions easily.
Maybe what you didn't account for is that adding the 100K feedback resistor will drop more voltage across the 11K plate load resistor than intended (because of the DC path to ground that is provided), which will shift the tube's quiescent operating point. Again, I'm not sure where exactly your calculations differ from experimental results, just some ideas.
Overall, my impression is that you are running the 12AX7's too hot. Something to consider is starting with AC coupling for the feedback loop, to break the dependency on needing to meet both AC and DC conditions simultaneously. With a 12AX7, it seems better to run the stages in the neighborhood of 1 mA of quiescent current, give or take. My opinion of course.
http://www.blueguitar.org/new/text/threads/from_mark/SBench_Loadlines_Series.pdf This is about the best explanation on choosing tube operating points you can find. Use of the datasheet and plate curves is a must for working with tubes. Trying to cobble together a multistage circuit with feedback without knowing how to set up a single stage correctly is pointless.
I take slight umbrage at the suggestion that I don't know how to design a single stage correctly. My single-triode circuit — simple as it may be — functions the same on paper as it does in real life at DC and it can deliver a relatively clean signal (free of clipping, at least) into the unconventional load I am using. It is pulling double-duty providing both the static DC bias for the electrostatic panel and the necessary signal voltage to drive it. I am very pleased with its performance and simplicity, I just wanted to expand on the idea a little bit. I used the datasheet and plate curves to set the operating points and load line exactly as I wanted them to be — operating well-within the performance envelope of the tube. I've gotten some helpful suggestions along the way, but telling me that the working circuit I built isn't working doesn't get me any closer to answering the actual question I asked. I am slowly coming to a better understanding of how all of the currents are interacting within the direct-coupled arrangement, so I will just soldier on on my own.
Sorry dude. This place can be pretty awesome for kicking around ideas sometimes. Unfortunately, some seem to misunderstand that this hobby isn't about making the best amplifier possible. It is about learning, exploring ideas, and having fun along the way. You write your own requirements.
Your question deserves a serious answer even if your approach is suboptimal (and I'm not saying that it is, I know nothing about driving these speakers), because answering it will help you learn.
Your question deserves a serious answer even if your approach is suboptimal (and I'm not saying that it is, I know nothing about driving these speakers), because answering it will help you learn.
I attempted to account for this by changing the load resistor from 22k in the single-triode circuit to 11k in the feedback arrangement. I tried to explain myself in the initial post, but I've attached a before-and-after sketch of this change. It's helpful for my understanding to take it step-by-step like this. In the final circuit, the cathode resistor of the input stage would be in series with this feedback resistor, but its small value would be incidental to this particular part of the calculation. The feedback resistor could always be changed, but 110k was a convenient value (roughly equivalent to resistance presented by the tube with 250v @ 2.2mA going through it — and a factor of ten larger than the load resistor for good measure).
One of the things that was causing me problems was that I assumed the voltage at the junction of the feedback resistor and cathode resistor would set the operating point of the first stage all by itself. I think I need to consider the effect of both currents going through this fixed resistance — the current from the feedback resistor and the current the input triode will actually draw on its own as a result of the cathode bias.
SpreadSpectrum — ¯\_(ツ)_/¯
Attachments
Last edited:
In my experience, things don't work out perfectly like you would expect from the data sheets. You always have to make allowances for tubes to behave a little bit different from their published characteristics. Treat your calculations as a ballpark and add some trimmer pots in there to get it tweaked.
That said, there is probably a lot that could be done to this circuit to optimize it to decrease distortion. Performance may be acceptable to you but could be improved. Keep that in mind for later.
If I were doing this I would probably AC couple the feedback which would give you a lot of freedom to adjust DC conditions and set up the stages to give less distortion.
For now, can you draw two diagrams, one with calculated DC values and one with measured DC values so that I can see the difference and possibly point out where you are going wrong? I'm still unclear on what the difference between your calculations and measurements are.
That said, there is probably a lot that could be done to this circuit to optimize it to decrease distortion. Performance may be acceptable to you but could be improved. Keep that in mind for later.
If I were doing this I would probably AC couple the feedback which would give you a lot of freedom to adjust DC conditions and set up the stages to give less distortion.
For now, can you draw two diagrams, one with calculated DC values and one with measured DC values so that I can see the difference and possibly point out where you are going wrong? I'm still unclear on what the difference between your calculations and measurements are.
I have attached a sketch of the known-incorrect initial values that I tried with the operating points I anticipated. This is the best I can do since I didn't keep a record of the voltages present with the original values. The first post in this thread contains an sketch of the actual values and voltages after I made some component changes on the fly to try to approximate the intended operating points.
Forget for a moment that this was ever designed to amplify anything — for now I just want to be able to select components that will give the desired operating conditions at DC.
I think the crux of my problem surrounds the operation of the cathode resistor in the input stage. My initial — erroneous — thinking was that the voltage divider formed by the feedback resistor and the first stage's cathode resistor would result in roughly 1v at the cathode (470r * (250v / 110,470r)). I assumed — erroneously — that 1v through the 470r cathode resistor would draw approximately 2.2mA through the input triode circuit dropping about 100v across the 47k plate resistor. I think there is a complex interaction with the feedback current and the input stage's own current that is impacting the bias of the first stage. That's where I'm at. Thanks for pursuing an understanding of this with me.
Attachments
GoatGuy — Thanks for posting this schematic. I've reposted it here for clarity. I assume this is from a simulator and thus basically correct. If so, I think I may have identified a critical flaw in my reasoning. My Ohm's Law may be a bit rusty — please feel free to correct me at any point along the way.
1.) The second stage is dropping 56v through a 33k resistor — dictating a current flow through that resistor of 1.7mA.
2.) This 1.7mA represents the collective current being drawn through this plate resistor by both the second stage triode AND the feedback resistor (which is essentially grounded at DC).
3.) The 330k feedback resistor is connected to ground through the 470r cathode resistor of the input stage, dropping 169v for a current draw of .51mA.
4.) The second stage is dropping 1.35v across the 470r cathode resistor — dictating a current flow through both the tube and cathode resistor of 2.87mA.
How can this be? Shouldn't the current through the second stage's 33k plate resistor have to equal the current through the feedback resistor (.51mA) in addition to the current through the second stage's triode (2.87mA)? I would expect a total of 3.38mA through the 33k resistor (which should force a voltage drop of 125.4v instead of 56v).
Attachments
You forgot to account for the additional current that the feedback resistor will put into the first stage cathode resistor. For it to have 1V across it, it needs to be half its resistance value. You have ~2.1mA from the triode plate current and ~2.2mA from the feedback resistor for a total of 4.3mA. Ohm's law will show that you will get about 2V across a 470 Ohm resistor, which will throw off your plate voltage in your first stage. That's why the 220 Ohm resistor worked in the end.
The node on the second stage plate on this schematic appears to violate Kirchoff's current law. I'm guessing there is a typo or something.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.