Super Symmetric Cork Sniffers' PI

[VacuumVoodoo]

These are some of the odd conventions in the my SPICE.
"V1" next to the symbol of a battery identifies a voltage source.
V2 is signal generator. X1 and X2 are the triodes.

As to how it works:
It's a basic differential/paraphase PI with equal 150k plate resistors and a few twists.
I have separated adjustment of symmetry - the X3 trimpot - from B+ supply
so it doesn't disturb DC operating conditions.
Gain symmetry is achieved by feeding a very small fraction of out of phase signal
(relative to input V2 signal) into the X2 triode grid.
This weird arrangement is in fact a full frequency range POSITIVE feedback.
Only very small fraction of X1 triode output needs to be fed back
to result in gain change required for exact symmetry.

C6 capacitor works in a more "mysterious" way.
It provides positive feedback above the frequency at which
Miller effect begins to affect the bandwidth.
Because full amplitude signal is fed back in phase with the signal at X2 grid
it acts as bootstrap on the Miller capacitance effectively cancelling it.
You can say it performs a disappearing elephant trick.
Thus the over 150kHz bandwidth at 35db gain.

Regarding asymmetrical resistor and capacitor values -
positive feedback's effect on impedances inside the feedback loop
is inverse to the effect negative feedback would have had.
I admit it's a bit difficult to wrap one's head around this,
especially as my explanation perhaps isn't very pedagogical.

[ChrisM]

Why the different grid stoppers and grid leaks?

To set the gain equal it looks like you have just added a simple variable voltage divider after one triode's output. Almost like half a PPIMV. Seems like you could add this to any PI to get the gain the same/balanced.

I am guessing though that your design is better at balancing though because the gain stays constant/balanced across all frequencies. Correct?

[VacuumVoodoo]

Why the different grid stoppers and grid leaks?

To set the gain equal it looks like you have just added a simple variable voltage divider after one triode's output. Almost like half a PPIMV. Seems like you could add this to any PI to get the gain the same/balanced.

I am guessing though that your design is better at balancing though because the gain stays constant/balanced across all frequencies. Correct?

The apparent similarity to a PPIMV is one of these details where the devil resides.
That voltage divider is what provides positive feedback to the X2 triode... as I tried to explain above.
The frequency response graph shows both outputs overlayed so yes, this circuits balances at all frequencies.

A word of caution though.
This circuit depends on two separate positive feedback paths.
Capacitances used to balance the whole thing are very small.
This requires very clean and careful layout and airy lead dress.
Adjusting the trimpot and trimcap is best made with signal generator
and oscilloscope or you just might overcompensate
and have yourself a 100kHz oscillator.

[jazbo8]

Nice results! But to make it work on the bench, it would require closely matched halves of the tube. Similar level of balance may be achieved by using the split-load/cathodyne configuration which is inherently more balanced and use the other half of the tube for gain. The disadvantage is that the cathodye does not handle clipping as well as the LTP.

Jaz

[VacuumVoodoo]

Nice results! But to make it work on the bench, it would require closely matched halves of the tube.
Jaz

That's a common misconception. The differential PI, and it's derivative the LTPI, is inherently unbalanced.
Look at basic differential pair as the left triode connected as normal common/grounded cathode gain stage,
the right one connected as common/grounded grid gain stage.
The right triode receives it's input into the cathode from the left triodes' cathode.
This can be seen as a signal coming from a cathode follower,
so it is smaller than input source signal at left triodes grid.
Thus the output signal from the right side will be smaller than from left.
This difference will be bigger the smaller common cathode resistor is.
Even with very high cathode resistor, or an ideal current sink, and perfectly matched triodes
the outputs will differ in amplitude due to signal reduction in the left triode between grid and cathode.
The conclusion is that well matched triodes may satisfy ones need for balance in the Universe,
but they don't do a lot for the PI since the circuit itself is inherently unbalanced.

This was an attempt at a no math simplified explanation.

[jazbo8]

That's a common misconception. The differential PI, and it's derivative the LTPI, is inherently unbalanced.

Point well taken, so the close matching of the two halves is not required with the carefully selected parts in your circuit, this will come in handy! What happens when the PI is heavily overdriven, does it have unusual artifacts that we should be aware of?

Thanks,
Jaz

[VacuumVoodoo]

Well, everything depends on how sensitive your cork sniffing nose is
If you can smell a single rose petal in barrel of wine then you may want to
also include a DC balancing adjustment so that both plates will sit at he same potential.
It's easily done by connecting a 100 Ohm trimmer pot between cathodes.
Existing cathode resistor will then go to trimpot wiper.
This is a more handsome way than trimpot hanging on B+ as there won't be high voltage on it.
Generally, the gain and bandwidth balancing technique shown in this example
can be applied to a LTPI in the same way.
Both outputs clip at the same level on respective positive swings when overdriven.
Negative swings clip, or more correctly, saturate at different levels
- that's when the opposite triode is not conducting and is clipping on positive swing
so imbalances get emphasized..
It is of no consequence since on negative swing the power tube will be driven into cut-off anyway.
What needs to be investigated IRL is transition from cut-off and saturation to linear region in both directions.
These transitions introduce discontinuity in the signal.
Positive feedback circuits don't like this and may produce bursts of high frequency oscillation.
If this happens we may need to reduce the bandwidth closer to human range.

[David Root]

Here's a more conventional LTP with constant current sink, which the designer (Patrick Turner) says doesn't need balanced tubes and is dependent for balance only on the "equality of the resistance of the loads on each half of the LTP" ie how well you match the 33K plate resistors.

This is more mundane than Alex's very creative offering but should balance as well????

[martin manning]

The link won't open for me (?)

[VacuumVoodoo]

Your attachment is gone, David...
Current sink yes, there's one straight from an old book in the Tubewonder,
the ground end of the sink is connected to -70VDC bias source
so LTPI grids can be biased at true OV ground potential.

Sometimes it's fun to just check if ones wacky idea works.
High power tube amp for ultrasound imaging might benefit from it.
Technological throwback but at least it's of terrestrial origin,
not some alien stuff made from melted sand.
I misplaced my tin foil hat....

[pdf64]

My experience is that shared, unbypassed cathodes on a paraphase can result in ugly blips of oscillation when overdriven.
Split the cathode (and bypass them as desired) gets rid of it.
Pete

[jazbo8]

Mr. Turner's description of the LTP with CCS tail is around the middle of this page >> http://turneraudio.com.au/Integrated5050.htm

[jazbo8]

My experience is that shared, unbypassed cathodes on a paraphase can result in ugly blips of oscillation when overdriven.
Split the cathode (and bypass them as desired) gets rid of it.
Pete

Do you mean split the cathode resistors, bypass them then connect to a common resistor? Otherwise, there is no current sharing, no?

Jaz

[VacuumVoodoo]

My experience is that shared, unbypassed cathodes on a paraphase can result in ugly blips of oscillation when overdriven.
Split the cathode (and bypass them as desired) gets rid of it.
Pete

Your observation is correct.
However, the circuit is not really a paraphase PI as I pointed out earlier.
Perhaps I should have stressed this point stronger.
To be strict it is a basic differential pair, which classic paraphase is not.
Common cathode resistor, bypassed or not, must have been just another
penny saving attempt.
I used full BW weak positive feedback instead of a large cathode resistor
as in the classic Schmitt inverter, which when equipped with
bootstrapped self bias becomes a LTPI as we know it.
A strong positive feedback at high frequencies extends bandwidth by at least a decade.
Maybe sometime during summer I'll test it,
but no promise - I'm currently in a "testing in theory" mode.
That's why I invited TAGers to give it a try.

[pdf64]

the circuit is not really a paraphase PI as I pointed out earlier.
Perhaps I should have stressed this point stronger.
To be strict it is a basic differential pair, which classic paraphase is not.
Common cathode resistor, bypassed or not, must have been just another
penny saving attempt.
I used full BW weak positive feedback

Alex, apologies but I can't see the positive feedback that you mention as being key to its operation?
C6 excepted, it looks like a regular paraphase - what am I missing?
Could you devise a 'circuit theory of operation for dummies'?
Thanks.

Jaz, I was suggesting eliminating the current sharing between the cathodes, change to independant biasing / bypassing.
Pete