"...before the **** crows three times."
My main point, though, was that grid current draw will upset nearly any phase splitter balance. In practice, the load presented by output tubes does not include significant grid current draw unless the output tubes are gassy (replace them!), the output stage is biased to AB2 (you need a buffer, even with an LTP or paraphase!), or you're clipping (in which case the phase splitter imbalance is transient). And the differences in input capacitances for any tubes that have sufficient balance for acceptably low distortion are lint-grade minimal- yes, I suppose if you make your output stage PP pair one Chinese EL34 and one Mullard KT66, you might be able to measure some imbalance!
edit: Apparently literary quotations are not good enough for the autocensor, sorry about that. Substitute "rooster."
In Piyramid-7M I had degeneration resistors, in cathodes of triode LTP. In previous one, Pyramid-VII, I used pentode LTP driver.
...but in Pyramid-V that we measured together when you were younger and still here LTP was made of 6N6P and had no degeneration resistors. Nobody complained then, since according to our observations the first tube dominated in generating distortions.
...but in Pyramid-V that we measured together when you were younger and still here LTP was made of 6N6P and had no degeneration resistors. Nobody complained then, since according to our observations the first tube dominated in generating distortions.
I did mention the plate should have an effect 1/(Mu-1) at the cathode.
And there would be impedance consequences for bypass capping that.
Things that don't change: Fixed caps, fixed resistors, and 100% garaunteed
correlations, are all candidate elements for calculating an "impedance". If
any wiggle in an arbitrary manner, the arbitrary currents must be excluded.
For the wiggle was not caused by any stimulus at the test node.
If I am testing a node by injecting 1KHz current, and watching for voltage.
I am not allowed to inject a totally arbitrary 1KHz signal into another node
and falsely manipulate the local result to any impedance I might care to
fake it. If we close a feedback loop to make that other signal, then its no
longer arbitrary. It now has a local cause. It can be calculated.
An unguaranteed correlation that won't be there when you need it most,
has no relation to a local cause. Falls into the category of arbitrary things.
Like fair weather friends, they are not knowable or calculable.
In the quoted example of the cathode bypass cap. If it would always be
there, and we can depend upon it, we are allowed to figure it in. But if
you imply it shunts away arbitrary cathode driven output grid currents,
and somehow changes the plate impedance; Those currents should not
have been part of the plate equation to begin with. In that regard only,
nothing changes...
Last edited:
Again, an astute observation. If one wants to measure the impedance of a node (to ground), then one must inject a current into that node only, zeroing all other independent signal sources.
If, as is essentially the case with SY's claim, one insists on injecting a current into one node while extracting it from another node, the impedance measured is the impedance between the nodes, not the impedances of the two nodes to ground.
OK, I finally get you now (#429)...
Before, I could not pinpoint exactly what was your beef with Sy.
Thought the whole argument silly maths about the unknowable
ratio of a disconnected cause/effect... And it still is...
But if you stimulate both nodes equal and opposite AC coupled
currents of 100% garaunteed correlation, we have to ask if that
stimulous GND becomes unknowable, or insignificant, or both???
If it becomes insignificant, might not matter that its unknowable.
Possible that both of you are right, and arguing about nothing...
I mean, you can't see the difference between cathodyne GND,
and stimulous GND doing any arbitrary crazy thing. And if you
tied em' together and forced them equal, would look the same.
Its differential, and its SE to GND, and I don't know anymore...
We should be asking if a 100% garaunteed linear correlation of
driven loads is a realistic ideal to believe in, or not? If not, then
we don't need to be asking absurd questions and getting upset
about absurd answers...
I can give you a whole bag of absurd answers...
Before, I could not pinpoint exactly what was your beef with Sy.
Thought the whole argument silly maths about the unknowable
ratio of a disconnected cause/effect... And it still is...
But if you stimulate both nodes equal and opposite AC coupled
currents of 100% garaunteed correlation, we have to ask if that
stimulous GND becomes unknowable, or insignificant, or both???
If it becomes insignificant, might not matter that its unknowable.
Possible that both of you are right, and arguing about nothing...
I mean, you can't see the difference between cathodyne GND,
and stimulous GND doing any arbitrary crazy thing. And if you
tied em' together and forced them equal, would look the same.
Its differential, and its SE to GND, and I don't know anymore...
We should be asking if a 100% garaunteed linear correlation of
driven loads is a realistic ideal to believe in, or not? If not, then
we don't need to be asking absurd questions and getting upset
about absurd answers...
I can give you a whole bag of absurd answers...
Last edited:
Yes, that's the only way, under those conditions, to guarantee the "equal load" criterion. It's much easier when looking at this circuit to use the Thevenin definition of source impedance to understand how it will work in the sorts of circuits where it's used, which in fact DO inject equal and opposite currents during the signal swing.
Yes. Thevenin deals with two-node equivalent circuits only. So to determine the Thevenin equivalent of the node pair P & K, you could place a floating current source between the P & K, determine the difference between the P & K voltages, and divide by the current source current to get a floating Thevenin equivalent source of impedance slightly less than 2/gm.
The logical extension of this treatment is to then connect a floating load between the P & K. Now, the Cathodyne does not really drive a floating load, but the center of the floating load is 0V because the P and K voltages are equal and opposite. So you can tie that point to ground without affecting any voltages or currents in the circuit. And of course, that's what the Cathodyne really drives when nothing on the load side causes an imbalance, such as a grid current.
Going back to the floating current source to obtain a Thevenin equivalent circuit, imagine that you replaced the single source with two sources in series, each one passing identical currents in the same direction. There is no change in circuit operation. Since the P & K voltages are equal and opposite, and the current sources are identical, the potential at the junction of the current sources is 0V. So you can ground that point if you wish. No net current flows into or out of ground, so doing so in no way changes circuit operation. Thus, this means of testing is equivalent to using the floating current source for the Thevenin two-terminal equivalent circuit.
Finally, if you do such a test on the bench or in a simulation, you can obtain the result that the P-K impedance is about 2/gm. But if you instead apply Kirchoff's Law and derive the equations, you'll be able to see what portions of Vp and Vk are due to ip and what ones are due to ik. Then you can determine the P-gnd and K-gnd impedances by applying the definition of impedance between node A and B in a linear circuit : the change in the voltage between nodes A and B due to and divided by the current flowing into A and out of B. And you'll be able to see that the impedance between the P & gnd is much higher than that between the K and gnd. You'll see that the Plate (to ground) impedance is nowhere near half of 2/gm, and that the Cathode to ground impedance is also somewhat larger than half of 2/gm.
Uh, whatever. I don't understand half that stuff...
What I'm pondering right now that it doesn't matter if test current(s) are
GND ref'd or not. What matters is that resulting voltage(s) are GND ref'd.
And they are. But they can also be ref'd to each other, so differential isn't
wrong either.
What I'm pondering right now that it doesn't matter if test current(s) are
GND ref'd or not. What matters is that resulting voltage(s) are GND ref'd.
And they are. But they can also be ref'd to each other, so differential isn't
wrong either.
Last edited:
kenpeter, the impedance measured by a test current source is defined precisely as the voltage across the test source divided by the current through the test source, assuming all other independent sources are zeroed. This is the essence of an ohm-meter. (By the way, this is why, in a simulator, one usually sets the test source current to 1. Then the magnitude of the voltage equals the magnitude of the impedance since, to get the impedance in ohms, you divide the voltage by 1 Amp.)
So, if one end of the test current source is connected to ground, the relevant voltage is referenced to ground (think of connecting the black lead of the ohm-meter to ground).
If neither end of the test current source is connected to ground, the relevant voltage is not referenced to ground (think of connecting the ohm-meter across two nodes that aren't grounds).
So, if one end of the test current source is connected to ground, the relevant voltage is referenced to ground (think of connecting the black lead of the ohm-meter to ground).
If neither end of the test current source is connected to ground, the relevant voltage is not referenced to ground (think of connecting the ohm-meter across two nodes that aren't grounds).
For any who might be interested, Jan Didden just posted my letter on Linear Audio's Letters to the Editor page: Linear Audio | Letters . To summarize, I agree with all of Mr. Vogel's math discussing the Cathodyne (some of which vindicates what Alfred and I have said) but point out differences between what Mr. Yaniger and Mr. Vogel are saying. This is strange, because Mr. Vogel says that he agrees with Mr. Yaniger!
I'm hoping Mr. Vogel can clear this up.
I'm hoping Mr. Vogel can clear this up.
I've had it happen to me in a couple of ways.
one way was my b+ was too low.
another my resistor didn't match (anode-cathode)
used too small of resistor (wattage wise) and thier resistance drifted non linearly as the circuit heated up (temperature coefficient got in the way)
tied too many output tubes, had to use a common cathode buffer amp between the phase splitter and 8- 6550's .
one way was my b+ was too low.
another my resistor didn't match (anode-cathode)
used too small of resistor (wattage wise) and thier resistance drifted non linearly as the circuit heated up (temperature coefficient got in the way)
tied too many output tubes, had to use a common cathode buffer amp between the phase splitter and 8- 6550's .
other things can cause this too.
an old tube that is nonlenar. usually cased by sitting too long and the tube elements has to cook off the ionization. If this the case try subbing with a new tube and let that old nos tube cook for a while in a different line stage.
paralleled inductance of the resistors causing a change in inter-electrode capacitance causing gain skewing. good quality resistors (non inductive wire wounds work to your advantage in this circuit) work best for this circuit.
an old tube that is nonlenar. usually cased by sitting too long and the tube elements has to cook off the ionization. If this the case try subbing with a new tube and let that old nos tube cook for a while in a different line stage.
paralleled inductance of the resistors causing a change in inter-electrode capacitance causing gain skewing. good quality resistors (non inductive wire wounds work to your advantage in this circuit) work best for this circuit.
Chris, your current-mirror example in your LTE elegantly demolishes the heart of SY's argument.
With your simple example circuit, a circuit so simple that no algebra is required to see that the source impedances are vastly different, a circuit so simple that no algebra is required to see that the outputs give equal voltages across identical loads, there can be no rational doubt: SY's measurements do not imply equal and low source impedances.
With your simple example circuit, a circuit so simple that no algebra is required to see that the source impedances are vastly different, a circuit so simple that no algebra is required to see that the outputs give equal voltages across identical loads, there can be no rational doubt: SY's measurements do not imply equal and low source impedances.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.