Just in case anyone claims my sims didn't match my black box diagram. Here is what happens when the grid of a tube loads either output. (Note the frequency scale now goes to 200K) While there are indeed differences, they are nowhere near as drastic as the results if I followed Chris's advice.
Dave, from what you wrote, it seemed (strangely) that the LCR input impedance was a flat 6Kohms. I thought this odd and asked you to confirm. In the absence of a response, I had to assume that that was strange, but true.
You asked me to provide a drive impedance of 6K for each load. I ask you to look at my sims and tell me that the drive impedance for each of your loads is not 6K. This is what you asked for. This is what I gave you.
You see, you asked me the wrong question.
What you needed was a requirement of a differential source impedance of 12K for the pair of loads. Had you asked for that, I would have given you the matched solution you arrived at.
But I took this as a golden opportunity to make my point.
Again, please look at my sims and tell me I did not give you a circuit which drives each 6K load with 6K.
And again, the conclusion is clear – Cdyne P-gnd and K-gnd loads are different.
You asked me to provide a drive impedance of 6K for each load. I ask you to look at my sims and tell me that the drive impedance for each of your loads is not 6K. This is what you asked for. This is what I gave you.
You see, you asked me the wrong question.
What you needed was a requirement of a differential source impedance of 12K for the pair of loads. Had you asked for that, I would have given you the matched solution you arrived at.
But I took this as a golden opportunity to make my point.
Again, please look at my sims and tell me I did not give you a circuit which drives each 6K load with 6K.
And again, the conclusion is clear – Cdyne P-gnd and K-gnd loads are different.
Last edited:
I gave you a circuit and the simple requirement to select two resistor values.
I didn't ask you to prove anything. For my application your conclusion that different valued resistors were needed is incorrect and now I am out of work because I just had 1000 circuit boards stuffed based on your advice.
Maybe Single ended and Differential are the proper terms to use? I'll accept you have proven the single ended plate and cathode impedances are different. I just question their value to a balanced circuit. I do accept that one the circuit becomes unbalanced, then we need to consider both the single ended and differential current loops but from day 1 I have been clear that I am not interested in the unbalanced case.
dave
I didn't ask you to prove anything. For my application your conclusion that different valued resistors were needed is incorrect and now I am out of work because I just had 1000 circuit boards stuffed based on your advice.
Maybe Single ended and Differential are the proper terms to use? I'll accept you have proven the single ended plate and cathode impedances are different. I just question their value to a balanced circuit. I do accept that one the circuit becomes unbalanced, then we need to consider both the single ended and differential current loops but from day 1 I have been clear that I am not interested in the unbalanced case.
dave
I drew up a circuit and stated that in order for the LCR to give proper frequency response it must be driven from a 6K source.
Then I asked for the value of series resistance required to make each behave identically.
It is obvious that the means you chose to analyze the circuit and select the resistors gave the wrong results when put into practice.
dave
Then I asked for the value of series resistance required to make each behave identically.
It is obvious that the means you chose to analyze the circuit and select the resistors gave the wrong results when put into practice.
dave
Dave, it is not true that I chose how to analyze the circuit. I had no choice in analyzing the circuit. From your posts:
748: "It is also required that the 6K LCR be driven from exactly a 6K source."
762: "source Z driving the LCR must indeed be 6K."
I know I gave you results other than what you wanted. I simply gave you what you demanded. I even asked you to be certain that that is what you wanted!
And since, forgive me, I did not believe this was a real job, it was a great opportunity to demonstrate that the design required different series resistances to present different source impedances. And that the Cdyne Zouts are different.
And wasn't that real purpose of this exercise? To determine Cdyne operating characteristics?
748: "It is also required that the 6K LCR be driven from exactly a 6K source."
762: "source Z driving the LCR must indeed be 6K."
I know I gave you results other than what you wanted. I simply gave you what you demanded. I even asked you to be certain that that is what you wanted!
And since, forgive me, I did not believe this was a real job, it was a great opportunity to demonstrate that the design required different series resistances to present different source impedances. And that the Cdyne Zouts are different.
And wasn't that real purpose of this exercise? To determine Cdyne operating characteristics?
The request was to give the proper value resistors to make the circuit behave as desired. (flat frequency response when driven from an inverse RIAA filter.)
I made it clear that in order to meet the goal, the 6K LCR must be driven by a 6K source.
If you gave me exactly what I asked for why don't the frequency response plots fufill the initial criteria?
I have two possible answers. The simulation is wrong or the analysis on how to choose the resistor values is wrong.
For me it was to test the theories here with an actual application and ultimately for me the correct answer is the one that gives the correct results.
I trust spice.
Last edited:
This is the sum total of what I knew about your LCR modules at the time I presented you with the solution. Clearly, your LCR modules somehow presented a contant 6K load to their drivers.
And indeed, in the solution I presented you, that is exactly what was accomplished - both 6K loads were driven by 6K source impedances. Please tell me how you can deny that from the simulation results?
Because, suddenly, after I provide you with the solution, you present me with a schematic that contradicts your first quote above! But if your LCRs really did have a 6K input impedance as you said, there is no reason that the inverse RIAA network would not have given you the desired results.
Bottom line:
(1) I followed your specs assiduously. Tell me where I did not?
(2) Even if I messed things up, which I certainly did not, the sim proves undeniably that in order to drive 6K loads with 6K source impedances, you need different value series resistors, And that implies the Cathodyne impedances were unbalanced to begin with.
Dude... here is the input impedance of a 7K30 LCR. trust me that the theoretical 6K is similar.
I would have suggested using a 600R LCR like the tango but that could have been problematic and couldn't work since the single ended plate impedance most tubes is higher than that.
In response to (1)
your circuit looks nothing like the circuit I suggested. When I inserted your results into the circuit I suggested, the results were wrong. Sure you proved something, it just didn't apply to this situation.
(2)
see my response to #1.
the only thing that I can do to explain the discrepancy we are having it you are using a single ended analysis to predict behavior of a differential circuit.
One mea culpa here... The reason I was silent for a few days was because I had sim results that didn't match my predictions. I was ready to eat crow and in fact I thought I had finally found a model that busted SY's boundary conditions and wondered why the two identical filters provided different responses. I was glad I didn't post this when I realized there is a substantial difference according to spice between a .0118 cap in the + leg and a .0118u cap in the - leg.
dave
I would have suggested using a 600R LCR like the tango but that could have been problematic and couldn't work since the single ended plate impedance most tubes is higher than that.
In response to (1)
your circuit looks nothing like the circuit I suggested. When I inserted your results into the circuit I suggested, the results were wrong. Sure you proved something, it just didn't apply to this situation.
(2)
see my response to #1.
the only thing that I can do to explain the discrepancy we are having it you are using a single ended analysis to predict behavior of a differential circuit.
One mea culpa here... The reason I was silent for a few days was because I had sim results that didn't match my predictions. I was ready to eat crow and in fact I thought I had finally found a model that busted SY's boundary conditions and wondered why the two identical filters provided different responses. I was glad I didn't post this when I realized there is a substantial difference according to spice between a .0118 cap in the + leg and a .0118u cap in the - leg.
dave
Obviously I had to choose numbers for my application with an actual tube and see the results.
Also it is obvious that different tubes have different characteristics so I made a guess at what the single ended plate and cathode impedances would be of a 6DJ8 under the circuit operating conditions I chose. If you would like me to solve for each independently under single ended conditions like you did, I will.
this doesn't circumvent the fact that your method gives the wrong answer to the initial problem since I have already shown that the only way to get matched "flat" response is with the pair of 5900 ohm resistors. I simply took 3K as a guess for the "plate impedance" of the Dj8 and subtracted it to get different values of the two resistors.
I'm reasonably sure if I follow your sim model using the esource with a 3K resistor on top as the "ideal" tube" the same thing will happen. In fact, I'll bet you a GI-Joe with kung-Fu grip that it does.
dave
Also it is obvious that different tubes have different characteristics so I made a guess at what the single ended plate and cathode impedances would be of a 6DJ8 under the circuit operating conditions I chose. If you would like me to solve for each independently under single ended conditions like you did, I will.
this doesn't circumvent the fact that your method gives the wrong answer to the initial problem since I have already shown that the only way to get matched "flat" response is with the pair of 5900 ohm resistors. I simply took 3K as a guess for the "plate impedance" of the Dj8 and subtracted it to get different values of the two resistors.
I'm reasonably sure if I follow your sim model using the esource with a 3K resistor on top as the "ideal" tube" the same thing will happen. In fact, I'll bet you a GI-Joe with kung-Fu grip that it does.
dave
the file is here for all to enjoy
the tube symbols and models are from stephie's (nee steve's) changes so you have two choices.
follow the instructions in the stickies here
or insert your own working tube symbol and model in.
dave
Last edited:
Dave, you didn't look at my sim in detail any more than I looked at yours! Both of us are guilty!
Instead of 3K, try 272.026 on the top, and 5810.488 on the bottom (yeah, I know, the resolutions are ridiculous. Humor me.)
And please accommodate my admitted limitations by employing instead of a 6DJ8 a model of a triode where mu = 30 and rp = 3K. And set Rp = Rk = 6K.
And please e-mail me your sim file (refer to the PM I sent you) with the reverse RIAA and LCR models.
Thanks.
Instead of 3K, try 272.026 on the top, and 5810.488 on the bottom (yeah, I know, the resolutions are ridiculous. Humor me.)
And please accommodate my admitted limitations by employing instead of a 6DJ8 a model of a triode where mu = 30 and rp = 3K. And set Rp = Rk = 6K.
And please e-mail me your sim file (refer to the PM I sent you) with the reverse RIAA and LCR models.
Thanks.
Thank you, Dave.
It is the viewpoint of the simulator that matters, not that of our intuitions, right?
Your comments on resolution are spot on.
Give me some time like you took to examine this, OK? Meanwhile, you might want to fool around with the design I recommended. Or not, and see what I come up with.
It is the viewpoint of the simulator that matters, not that of our intuitions, right?
Your comments on resolution are spot on.
Give me some time like you took to examine this, OK? Meanwhile, you might want to fool around with the design I recommended. Or not, and see what I come up with.
Reminder how you can make concertina with impedance match and gain.
This works extremely well as long as no DC conducts into output grids.
Even with 20mA rushing down the splitter, forget about driving into A2.
It is caps, not Z imbalance, that forbids concertina to drive grid current.
The other caveat is when voltage drop across PNP approaches zero, not
going to split this clipping event evenly. So there is a diode to make
sure it clips in a sensible way, well before anything weird can happen
across the PNP.
R4 across VBE is a 3.3mA constant current source for concertina plate.
The triode loadline is held to constant current, for best possible Mu.
We try to squeeze an extra +50 and -50 gains from this concertina!
The only genuine advantage this circuit seems to offer....
Locally balanced Z's turn out to be a red herring of no utility.
And these Z's would be balanced by the global loop anyway.
My point only that it was pointless. Equal Z's prove no better
or worse than unequal Z's...
You may need Duncan amps' Triode and 6L6 models.
The high voltage PNP transistor model is built in....
This works extremely well as long as no DC conducts into output grids.
Even with 20mA rushing down the splitter, forget about driving into A2.
It is caps, not Z imbalance, that forbids concertina to drive grid current.
The other caveat is when voltage drop across PNP approaches zero, not
going to split this clipping event evenly. So there is a diode to make
sure it clips in a sensible way, well before anything weird can happen
across the PNP.
R4 across VBE is a 3.3mA constant current source for concertina plate.
The triode loadline is held to constant current, for best possible Mu.
We try to squeeze an extra +50 and -50 gains from this concertina!
The only genuine advantage this circuit seems to offer....
Locally balanced Z's turn out to be a red herring of no utility.
And these Z's would be balanced by the global loop anyway.
My point only that it was pointless. Equal Z's prove no better
or worse than unequal Z's...
You may need Duncan amps' Triode and 6L6 models.
The high voltage PNP transistor model is built in....
Attachments
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.