It is not a problem, since when negatively biased grids have huge resistance that we can ignore.
But nice well made records of classical, jazz, or similar live music, have big crest - factor, i.e. big difference between average levels and peak levels on loud passages. To be effective power amps must use output tubes on the edge of clipping. That means, on loud passages voltages on grids would hit significant non-linearity unbalancing Concertina.
There are 2 major solutions to the problem:
1. Use buffers after Concertina,
2. Select load lines for output tubes such a way that output tubes already saturate when control grid voltage did not approach zero volt yet.
3rd solution- do what I did for the RLD and have higher value grid-stoppers. Clips and recovers exceptionally cleanly.
kenpeter makes some cogent points and one, in particular, nearly gets to the heart of SY's error.
First though, Happy Birthday SY!
Second, the pedant that I am, I must mention that, as wonderful as the term "correlated impedances" is (well done kenpeter!), there are terms of art that already exist: differential and common-mode impedances. Also, what kenpeter calls "error" currents are, I think in this context, customarily called "test" currents.
Anyhow, when the test currents are anti-correlated, i.e., differential, there's really only one impedance one can speak of that the test sources "see". This is because the two equal and opposite test current sources are actually in series; all of the current that exits one source enters the other.
In fact, the two current sources can be replaced with a single source and the circuit response is identical (try it; I did just to verify that I wouldn't make a fool of myself 🙂 ) The impedance "seen" by the (single) test current source is the differential impedance, the impedance between the two outputs.
Now, kenpeter astutely observes that testing the output impedance of an amplifier "against a signal that is identical" will incorrectly lead you to the conclusion that the output impedance is infinite.
He's very close to striking the heart of the trap that SY fell into. It's an easy trap to fall into because it's subtle. But, the effort you make towards understanding it will be rewarded.
I've attached a problem that might illuminate the "trap". Check it out and post your answers if you're interested.
Last edited:
4th solution -- what I did in Pyramids: LTP buffer after Concertina, and LED in screen grids of output tubes that senses current and controls LDR input attenuator. Compression of peaks sounds less nasty than clipping.
Great informations , thats why you use cathode followers in your schematic in earlier post .
Actually, it was a LTP (differential cascade). Concertina and LTP complement each other well.
What about the concept of magnitude?
take two objects, one is a 15 pound rock and the other is a piece of lint. Now place them in the middle of a highway. In my view, I can ignore the piece of lint even though it can be proven to have an effect on the tire of a car I am driving if I hit it. of course someone could chime in and say the diameter of the tires on your car could be 1 micron at which point the rock becomes the rockies.... ahhh the beauty of scale.
again I trust your definitions and math. what I question is ignoring the scales of these relationships given the caveat of the identical load. everyone is aware of what happens when things become unbalanced.
so did I and I posted the pics a hundred or so posts back. we just seem to be missing each other on scale. It strikes me as funny that you mention engineering yet your points seem based on physics. (I view engineering as opinion based physics)
why would someone do a flawed test like that?
I think it is using a simplified model for avoiding concern about lint in the road causing a crash and all of the "but what if" arguments on either side completely miss the point and should be argued by physicists. Engineers on the other hand should be creative and actually take calculated chances based on understanding of the problem needing to be solved.
could you redraw that. I do not know what Q1 means.
If you calculate (or measure) the input capacitance, the -3dB frequency will be 1/(2piRC). As a rule, if you have a pentode output stage, the input C will be quite low, a few pF. For a triode stage, you need to consider the Miller capacitance.
As a practical matter, the load presented by a pair of output tubes is extremely close, differing by about Dave Slagle's piece of lint. If there's a serious enough mismatch of load to cause a problem with the phase splitter, then you've got gross problems with output stage balance.
No, actually, I don't think that you do. In fact, I'm quite certain that you don't.
Its just that "plate impedance" means to me, a certain change of current
into or out of the plate node (probably to GND) causes a change of voltage
to GND, or vice-versa. Note: Cause and effect are both at the same place!!!
Why an "equal impedance" illusion bothers me. Becuase the main cause of
changes measured at our plate were caused by arbitrary currents from the
cathode! Thats simply not the defintion of an impedance. I put the current
in this place over here, and measure the result over here somewhere else...
Yeah, right. If that's what Thevenin had in mind, I'm not impressed.
Let probable correlation of cathode error to plate error fool us to think we
have measured cause and effect in the same place, but it is not true! This
is ungaraunteed correlation that will not be there when you need it most.
Correlation is a very useful illusion, but its not an impedance.
into or out of the plate node (probably to GND) causes a change of voltage
to GND, or vice-versa. Note: Cause and effect are both at the same place!!!
Why an "equal impedance" illusion bothers me. Becuase the main cause of
changes measured at our plate were caused by arbitrary currents from the
cathode! Thats simply not the defintion of an impedance. I put the current
in this place over here, and measure the result over here somewhere else...
Yeah, right. If that's what Thevenin had in mind, I'm not impressed.
Let probable correlation of cathode error to plate error fool us to think we
have measured cause and effect in the same place, but it is not true! This
is ungaraunteed correlation that will not be there when you need it most.
Correlation is a very useful illusion, but its not an impedance.
If you put a global loop around an amp, such that you measure cause
and looped effect at the same plate , THAT might be an impedance
lowering or equalizing mechanism...
But in the open loop situation, accidental correlation of two errors at
different locations having mutual effect at the plate is not impedance.
"S**t happens" is not an impedance. Don't need to cipher no maths...
Unconnected cause and effect is unworthy of any equation...
and looped effect at the same plate , THAT might be an impedance
lowering or equalizing mechanism...
But in the open loop situation, accidental correlation of two errors at
different locations having mutual effect at the plate is not impedance.
"S**t happens" is not an impedance. Don't need to cipher no maths...
Unconnected cause and effect is unworthy of any equation...
Last edited:
Again, kenpeter, your observations are astute. You're zeroing in on the difference between driving point impedance and trans-impedance. You are most definitely "getting it".
Last edited:
Quote:
Now, kenpeter astutely observes that testing the output impedance of an amplifier
"against a signal that is identical" will incorrectly lead you to the conclusion that
the output impedance is infinite.
Slagel> why would someone do a flawed test like that?
I don't know why any reasonable person would... But same broken logic has lead
us to believe an equally loaded concertina might have equal impedance. And this
misrepresentation of correlation as impedance has never before raised alarms???
How many people had to be sleeping for how many decades for this to happen?
I remember being taught the same nonsense, and never questioned it. My bad...
Now, kenpeter astutely observes that testing the output impedance of an amplifier
"against a signal that is identical" will incorrectly lead you to the conclusion that
the output impedance is infinite.
Slagel> why would someone do a flawed test like that?
I don't know why any reasonable person would... But same broken logic has lead
us to believe an equally loaded concertina might have equal impedance. And this
misrepresentation of correlation as impedance has never before raised alarms???
How many people had to be sleeping for how many decades for this to happen?
I remember being taught the same nonsense, and never questioned it. My bad...
Last edited:
Lets take two nodes of Dave's lint, identical in all regards except:
One has a locally caused relationship we might call an "impedance".
The other a remotely caused arbitraryship we can call an "accident".
It doesn't matter they measure the same. Accident does not have
an impedance. The pink elephant seen in the mirror does not weigh
three tons. Flapping wings at the right time don't lessen it's weight.
If not also a real elephant, same as the one standing on your head,
it weighs whatever the glass weighs; the rest is illusion.
Last edited:
Okay, now for a practical question: How to design a concertina with direct coupling from the previous triode?
See my topic related to EL84 pushpull amp. I made a e88cc version and the final one uses 6n6p for both preamp/pi stages. The ideea is to find the HT/anode load balance so that the anode voltage of the preamp comes a few volts (bias voltage for concertina) lower than concertina cathode. A good practice is to split the HT 1/3 on each concertina resistor and tube. But I know that it is recommended to go between 1/4 and 1/3 for resistors. Let's say that your HT is 300V, then anode/cathode resistors should drop between 75-100V each so that across the tube you have 150-200V. Draw load lines and find the best balance for both stages.
Is the correlation accidental?
Lets just look at the plate impedance of the circuit as if it were a plate follower. Does the output impedance remain the same if we place a capacitor across the 22K resistor in the cathode?
the way I read your argument, the cap will not change the output impedance.
maybe the term we should be using to make everyone happy is "apparent output impedance"
dave