Is Cathodyne splitter really short of drive for output stage?

[jjjtttggg]

I've read that the cathodyne aka split load aka concertina phase inverter suffers from inability to adequately drive the output stage.  I'm not sure this is really true, unless I'm missing something, at least for a 6V6 type amp like 5E3.  In the 5E3 output stage design, using AX7 nominal plate curves, it looks to me like the phase inverter should operate very nearly linearly over an input range (at the grid) of -2v or so up to 85volts or so, with both output phases entirely capable of that kind of +/-40v swing.  That P-P range should totally slam 6V6 output tubes.  Further, it seems like if even more headroom were needed, one could use the output screen supply voltage instead of the preamp supply.

I've only built one amp (modified 5E3), so I have no experience with any other PI circuit, but I just don't see in the math any indication that the cathodyne should limit an amp's ability to overdrive the output tubes.  Obviously an additional gain stage would be needed to get there, but since the cathodyne saves a valve compared with the alternatives, that doesn't seem to be a concern.

I'm aware of the potential for the output stage to clamp the cathode of the PI and spike the anode gain, so I'll need to account for that in my design.  I'm also aware that excessive positive grid voltage will overload the PI in an ugly way, but by then I think I'm already slamming the output tubes waaaay beyond the "sweet" spot, so I'm not terribly concerned about that.

So what am I missing?  I've read in lots of places that the main disadvantage of Cathodyne is that it lacks drive for the output, but I just don't see that it should.   If I'm right, I think I want to stick with it in my next amp, because it just seems like a more elegant solution than LTP or Paraphase.

Any and all thoughts appreciated!

-J

[Willabe]

I can't show you the math but my stock 5E3 clone distorts just fine that cathodyne PI driving the 2x6V6's.

Look at this Fender tweed 5E5A Pro, if the same PI can drive 2x6L6GB/5881's into distortion, and it does, than it will drive 2x6V6's, which need less signal to give full output, just fine.   

http://el34world.com/charts/Schematics/files/fender/Fender_pro_5e5a_schem.pdf

[HotBluePlates]

I've read that the cathodyne aka split load aka concertina phase inverter suffers from inability to adequately drive the output stage.  I'm not sure this is really true ... at least for a 6V6 type amp like 5E3.  ...

You got that 5E3, and you can turn up the volume enough to make the output tubes distort, right?  Then obviously for this amp/example, the split-load has all the drive needed and then some...

I've read that the cathodyne aka split load aka concertina phase inverter suffers from inability to adequately drive the output stage.  ...  I've read in lots of places that the main disadvantage of Cathodyne is that it lacks drive for the output ...

Be careful to distinguish between "voltage gain" and "output voltage swing" when looking at phase inverter circuits.  "Drive" is ill-defined and could be misinterpreted by readers (and maybe by writers).

No doubt about it, the split-load has a voltage gain which only approaches 1, so it has the least gain (by itself) of any of the phase inverter circuits.  This could be what some mean when they say it lacks "drive".  Practical split-load inverters are rarely used without additional gain stages before/after them, so this is somewhat immaterial in actual amps.

The split-load's challenge is that it needs to develop 2 output signals, moving in opposite directions, still leave voltage across the tube, and do all this from a single supply voltage.  Whether it can do this well cannot be known without specifying the supply voltage for the split-load stage and the needed peak output voltage (or at least the bias voltage of the output tubes).  Any claim of whether the split-load can/can't drive the output stage is meaningless without knowing those two voltages.  It
We'll use the Princeton Reverb as an example.

The split-load supply voltage is 240v, and the needed peak output voltage (per output) is equal to the 6V6 bias voltage at 34v peak.  So this means after the idle bias of the split-load is considered, it should be able to swing at least 34v in either direction, while still leaving voltage across the split-load's triode.

As biased, the split-load's plate is at 200v and its cathode is at 50v at idle.  If a signal is applied to the split-load's grid which reduces current through the triode, the limit is for the plate to move to the B+ voltage and the cathode to move to ground voltage.  Each can move 40-50v which satisfies the 34v peak output requirement in this direction.

What happens when the split-load triode draws more current, pulling the plate and cathode towards the same voltage?  For the plate, 200v - 34v = 166v and for the cathode, 50v + 34v = 84v.  This leaves 166v - 84v = 82v across the triode.  That is likely a good amount of plate-to-cathode voltage to support the plate current needed to swing the outputs in this direction.

This last step (plate swinging negative, cathode swinging positive) is the portion of signal swing most likely to be a limiting factor with a split-load inverter.  It is also the main spot you should be investigating to determine is the split-load has "enough drive".  Hopefully you see why this can't be evaluated without knowing the supply voltage and output tube bias.

If you're working from a raw design, you'll need to start with the supply voltage & output tube bias then ask, "Will there be enough plate-to-cathode voltage (given my desired load resistors) when there is 4x the bias voltage subtracted from the supply voltage?"  And then you need to subtract a bit more voltage to ensure the tube can swing the needed output without running into limits or linearity issues.

Returning to that Princeton, 34v * 4 = 136 and 240v - 136v = 104v.  And from our ealier example, another 20v or so of available voltage is there as a buffer between the output requirement and tube capability.

You could also reality-check this by looking at a loadline for the Princeton's 12AX7 for 112kΩ (both 56kΩ loads are in series) and check that it will pass (240v - 82v)/112kΩ = 158v/112kΩ = 1.41mA at 82v on the plate at or below 0v on the grid.

[jjasilli]

My take on this topic is that the cathodyne PI originated at a time before overdrive tone or any sort of distortion was considered desirable.  In a guitar amp with a simple preamp circuit (few gain stages), after insertion loss in the tonestack, a cathodyne PI might not output enough voltage to overdrive the power tubes, per Hotblue's post above.  The result could be a clean amp with little to no overdrive anywhere in the signal chain.  As Hotblue points out this is not truly the fault of the cathodyne, but it got a bad rap for "cleanliness" anyway.  A split-tail PI can provide more than 1:1 gain, so a weak signal into the split-tail PI can overdrive the power tubes.


OTOH, I think the PI itself can be overdriven with a strong input signal.  To my ears a cathodyne has a rather raw sounding overdrive tone; a split-tail sounds more smooth.  Hence mods to the Princeton PI like the Paul C mod and Stokes mod for higher plate supply voltage and/or re-bias for more or smoother overdrive.

[jjjtttggg]

Thank you all for your excellent responses.  I'm happy to learn that I'm thinking correctly. I'm thinking I want to build an amp where phase inverter does not overdrive at all, but simply passes any pre-amp distortion through unchanged, with enough headroom to overdrive the output tubes whether that be with that unmodified but already distorted large magnitude preamp signal, or simply a clean but large peak to peak preamp signal.

Straightforward enough ... conceptually anyway!

[HotBluePlates]

...  In a guitar amp with a simple preamp circuit (few gain stages), after insertion loss in the tonestack, a cathodyne PI might not output enough voltage to overdrive the power tubes ...

Note that this scenario is not an issue with the split-load inverter, but only bad design.

Gain of the split-load tends towards Mu/(Mu+2), or a gain of 0.98 for a 12AX7.  If we're looking at the Princeton Reverb from before, this means the signal into the split-load needs to be at least 34v/0.98 = ~34.7v peak.

But I always consider the gain stage prior to the split-load as a part of the overall output stage (phase inverter, output tubes, output transformer), especially because the feedback loop almost always includes this stage.

And since that 12AX7 gain stage has a voltage amplification of ~60, input to that stage only needs to be ~0.6v peak (before feedback) to fully drive the output tubes.

If the amp's preamp, tone stack, Volume control, etc imposes too much loss for a typical guitar signal to deliver this small peak signal to the input of the stage ahead of the split-load... Well, that's just bad design in my book.

The above should also highlight why a good amp design always proceeds from output stage backwards to the input jack, to ensure a later stage's drive needs are met by a preceding stage.

[PRR]

> I've read

You can read ANYthing on the interwebs.

Hemingway said something about "a good BS detector". Extra-strength when reading the internet. (Now someone will say they 'read' that Hemingway is an internet expert...)

As a rule of thumb: if nothing "dumb" is done, and cathodyne is fed B+ similar to final's G2 voltage, and final's Mu(g2) is significantly more than say 5, it will overdrive the final fine.

"Most" audio power pentodes are Mu(g2) of 10, unless it is 18 (EL84). It's like they knew what you were going to do!

You can fall short driving 2A3 (Mu=4.2 triode) with cathodyne. Even if you include the 2A3's ~~62V self-bias in the feed to the cathodyne, it just about works, and cathodyne bias is fussy.

Putting loss between driver and final is "dumb".

Putting significant NFB local to the final (UL, cathode-taps) reduces final's useful gain and prompts re-evaluation. If Mu(g2)=10 then you cut final gain in half with NFB, you may fall short.

Some quick-design amps run the cathodyne at very reduced voltages, or poor choice of bias or resistor values. Direct-coupled from a previous stage, a cathodyne is liable to have not-much plate-cathode drop. But you can screw-up any topology.

Would I BS you? Assume I would and evaluate for yerself.

[clyde]

...  In a guitar amp with a simple preamp circuit (few gain stages), after insertion loss in the tonestack, a cathodyne PI might not output enough voltage to overdrive the power tubes ...

Note that this scenario is not an issue with the split-load inverter, but only bad design.

Gain of the split-load tends towards Mu/(Mu+2), or a gain of 0.98 for a 12AX7.  If we're looking at the Princeton Reverb from before, this means the signal into the split-load needs to be at least 34v/0.98 = ~34.7v peak.

But I always consider the gain stage prior to the split-load as a part of the overall output stage (phase inverter, output tubes, output transformer), especially because the feedback loop almost always includes this stage.

And since that 12AX7 gain stage has a voltage amplification of ~60, input to that stage only needs to be ~0.6v peak (before feedback) to fully drive the output tubes.

If the amp's preamp, tone stack, Volume control, etc imposes too much loss for a typical guitar signal to deliver this small peak signal to the input of the stage ahead of the split-load... Well, that's just bad design in my book.

The above should also highlight why a good amp design always proceeds from output stage backwards to the input jack, to ensure a later stage's drive needs are met by a preceding stage.

How would a cathode biased output stage be calculated?  Thank-you for this enlightening treatise. 

[Ritchie200]

Thank you all for your excellent responses.  I'm happy to learn that I'm thinking correctly. I'm thinking I want to build an amp where phase inverter does not overdrive at all, but simply passes any pre-amp distortion through unchanged, with enough headroom to overdrive the output tubes whether that be with that unmodified but already distorted large magnitude preamp signal, or simply a clean but large peak to peak preamp signal.

Straightforward enough ... conceptually anyway!

Elegant solution - I like that!  I know a perfect cathodyne amp for you!


Jim

[HotBluePlates]

... How would a cathode biased output stage be calculated?  ...

Voltage across the cathode resistor.

Select whatever resistance is needed to idle at 100% with the plate/screen voltage you have.  The grid is effectively at 0v because it is grounded through a grid-leak resistor (and there's no grid current, so no voltage drop).  So the voltage across the cathode resistor is the total grid-to-cathode voltage.

You might measure 21v at the cathode, and could call that "-21v grid-to-cathode".  Either way, you know you need a preamp/driver/phase inverter capable of at least 21v peak with any input signal and likely control settings.  You might figure your way output-to-input while assuming the Volume control is half-to-3/4-up (or wherever you want the end of clean headroom to begin).

If you were asking, "How to I design the output stage," that's a whole other topic in itself.

[jjjtttggg]

... I know a perfect cathodyne amp for you!


Jim

I'm interested!  What amp are you thinking of?

-J

[sluckey]

I bet it's a real porker! OinK OinK!   

[Ritchie200]

HEY!  Don't talk about my girl like that!  She aint heavy, she's my weapon!

jjjtttggg, don't let sluckey sway you from the promised land.  Here ya go...

http://el34world.com/Forum/index.php?topic=17548.0

Jim

[HotBluePlates]

jjjtttggg, don't let sluckey sway you from the promised land.  Here ya go...

I've read that the cathodyne aka split load aka concertina phase inverter suffers from inability to adequately drive the output stage.  ...  I've read in lots of places that the main disadvantage of Cathodyne is that it lacks drive for the output ...

... No doubt about it, the split-load has a voltage gain which only approaches 1, so it has the least gain (by itself) of any of the phase inverter circuits.  This could be what some mean when they say it lacks "drive".  Practical split-load inverters are rarely used without additional gain stages before/after them, so this is somewhat immaterial in actual amps. ...

And notice that in the Major, the added gain stage I mentioned is a push-pull gain stage between the split-load inverter and the output tube grids.

[jjjtttggg]

Interesting design.  I hadn't seen that one!  I've never thought of putting the extra gain needed after the cathodyne instead of before.  I'm thinking the motivation for that would be greater P-P voltage range at the output valve grids (at the cost of an extra valve), but maybe there's another reason I'm not seeing.

I'm planning on something that won't blow the windows out of my house, though , (ie. 2 x 6V6, or 2 x EL84 output stage), so by the calculations I mentioned above (I think) I've headroom enough at the cathodyne outputs to live with a single gain stage upstream.

I have some other questions about my amp design options, but they're not related to cathodyne/phase inverter, so I'll start as a new thread.

Thanks for sharing!
-J

[PRR]

> I've never thought of putting the extra gain needed after the cathodyne instead of before.

If you grew up in the 1950s, you'd recognize it as the "Williamson". Williamson was driving high voltage triodes as power tubes, so he needed some swing.

[jjjtttggg]

Love it PRR...

The "Williamson"...  No end to the learning it seems. :)

Thanks for the info.

-J

[Ritchie200]

The 6v6 and el84 may not give you enough dB at the band practice with a busy drummer - if that is what you have plans for.  Plus, your clean headroom will be long gone pushing any volume with those two toobs.  If it's a bedroom or for recording, you will be good.  However.....

Don't give up on the Major yet!  Did you read that entire thread?  The two tube Tubenit version would be roughly 50 watts of glorious UL KT88 booming headroom with a master and ppimv to give you a little grunt as needed at whatever volume you want.  Come to the KT88 dark side...

Jim