It seems that the new SVT like to eat their 12AU7 Drivers

[jazbo8]

The cathode followers in this design have two purposes as far as I know. First, they stop the high capacitance of the power tube grids and wiring from forming a pole that might destabilise the feedback loop. Second, they help prevent thermal runaway by providing a low impedance sink for gas current coming out of the power tube grids.

I don't think it is intended to be a Class-AB2 design that runs the power tubes into grid conduction.

Those are good reasons, but why are the high plate-to-cathode and high cathode-to-heater voltages required? For example, what if the B+/plate for the CF is lower to say 150V, and the B- is lowered to -100V (with the appropriate changes made to the bias network of course), perhaps it would be a bit less linear than the higher-voltage case as R.G. suggested, but it should still be enough to meet the above conditions, no?

[bob p]

Good question. There is an answer. I think that everyone here is just waiting for someone else to draw the load lines.

[es345]

Those are good reasons, but why are the high plate-to-cathode and high cathode-to-heater voltages required? For example, what if the B+/plate for the CF is lower to say 150V, and the B- is lowered to -100V (with the appropriate changes made to the bias network of course), perhaps it would be a bit less linear than the higher-voltage case as R.G. suggested, but it should still be enough to meet the above conditions, no?

Regarding the plate voltage I don't have a reasoning but an interesting observation:

If you look to older SVT schematics of the powerstage and the one here in this thread: in all schematics the plate supply of the follower stage is coupled to the screens of the power tubes therefore having the same voltage. In the powerstage with 6146B/ 12BH7 that voltage was 220V, in the older powerstage with 6550/12BH7 it was 350V, now in the one here in the thread it is 380V. So perhaps no recalculation of the design has been done while changing the supply voltage.

Regarding the negative supply:

The bias at the control grid is -47V. This means at full power you need to be able to supply about -94V at the control grid. The follower should not be in total switch off at this moment, otherwise you can get nasty transient effect . With -180V you are on the safe side.
There is a second aspect which also require a high negative voltage : to get -47V at the cathode of the follower stage you need a bias less than -47V at the grid of that triode. I don't know how much it is for a 12AU7. For the 12BH7 in the SVT schematic -77V is mentioned which result in a negative peak of -124V.

[jazbo8]

Thanks for sharing the observations, I have not done a design or try to analyze the power tube's screen to the driver's plate, but this "E-Linear" (I don't think Ampeg ever called it that... as the name came later) connection is a form of LNFB, which lowers output distortion and increases the damping factor (good for bass).

SVT's -180V negative bias supply is kinda reasonable compared with Dumble SSS' -328V (really?!) And your calculations showed that such high negative voltage wasn't necessary (for the 6L6's), if the high differential voltage is killing the tubes, then lowering the plate and the negative supply voltage should be consider a good fix as long as there is no sonic impact.

[es345]

I assume that LNFB means local negative Feedback. In the old schematics there is no Feedback between the Screens and the plate of the Driver, ist just the same supply.

[es345]

for reference here are the 2 SVT schematics, unfortunately low quality only

[Steve Conner]

The minimum plate voltage required for the 12AU7 CF in this circuit will be set by how much current you want to deliver to the power tube grids. If you ask for too much, the CF itself will draw grid current and charge up its coupling capacitor, causing blocking distortion.

I had a look at the tube curves for the 12AU7. At a plate voltage of ~300V, the maximum possible output current with Vgk=0 is off the scale, something like 50mA. So, I think you could easily reduce the plate voltage of the CF with no ill effects. At 150V you can have 20mA and at 100V, 15mA.

The 12BH7 is considerably more potent, it will crank out 40mA at 150V and 25mA at 100V.

Note that these figures are the plate-to-cathode voltage, you must subtract the amount of positive voltage you want on the power tube grids. But as I said, I don't think the SVT was ever designed for serious Class AB2 operation. The 6146B originally used in it is rated for an average control grid current of only about 5mA.

[es345]

I share that view that it is not a design for AB2. A clear indication for that is the gridstopper value of 47K in front of the power tubes(R29,34,43...).

[jazbo8]

I assume that LNFB means local negative Feedback. In the old schematics there is no Feedback between the Screens and the plate of the Driver, ist just the same supply.

Yes, local negative feedback, not 100% sure, but I think it goes like this: change in power tube plate voltage -> change in power tube screen voltage -> change in CF plate voltage -> change in CF cathode voltage -> change in power tube grid voltage.

[es345]

The term "negative feedback" usually is used for feedback loops within signal path. What you are describing is a possible path through the supply path, not along a signal path. This means DC change effects. And here I don't see a reasonable impact along the path you mentioned. A small change (caused by e.g. SAG) at the CF plate will have basically zero effect at the cathode (it follows the grid). And still no effect for the signal.

[jazbo8]

The term "negative feedback" usually is used for feedback loops within signal path. What you are describing is a possible path through the supply path, not along a signal path. This means DC change effects. And here I don't see a reasonable impact along the path you mentioned. A small change (caused by e.g. SAG) at the CF plate will have basically zero effect at the cathode (it follows the grid). And still no effect for the signal.

Yes, I may have mixed things up a bit, I was thinking about the E-linear connection where the plate of the driver is also connected to the power tube's screen grid, although I forgot that E-linear uses an UL connection... In addition, the small screen grid resistor does not really sag the screen voltage much, so not much change in Vg2 occurs... thus little if any LNFB.

[loudthud]

The term "negative feedback" usually is used for feedback loops within signal path. What you are describing is a possible path through the supply path, not along a signal path. This means DC change effects. And here I don't see a reasonable impact along the path you mentioned. A small change (caused by e.g. SAG) at the CF plate will have basically zero effect at the cathode (it follows the grid). And still no effect for the signal.

Looking at the plate curves for a pentode, plate voltage has very little effect on cathode (or plate) current until plate voltage drops very low. The lines run parallel to the X axis. In a triode however, if the grid voltage is held constant, plate current will reduce when plate voltage is reduced.

[jazbo8]

I guess the change in the screen supply from sag is small enough to be ignored in this case... In any case, I love to see some scope shots at various settings to see just how this rather unusual PPICF works.

[es345]

Looking at the plate curves for a pentode, plate voltage has very little effect on cathode (or plate) current until plate voltage drops very low. The lines run parallel to the X axis. In a triode however, if the grid voltage is held constant, plate current will reduce when plate voltage is reduced.

Fully agreed. Now let us have a look to this specific cathode follower circuit. Plate voltage is 380V, cathode -45V, cathode resistor 47K connected to -180V resulting in ~2,9 mA cathode current. The cathode resistor is a local feedback element for AC and DC, due to the high value you can treat the 47K also as a non ideal current source. Looking to the 12AU7 spec: 2,9mA, plate to cathode 425V leads to about -26V between grid and cathode. Now let us reduce the plate e.g. by 125V (32% change wrt ground). A view to the spec shows that for the same current we will need now about -16V grid to cathode. So the cathode voltage will be about -56V resulting in a drop of about 7% of the cathode current. So a 32% drop at plate supply results in ~7% current drop only. (This estimation is is mathematically not fully correct but good enough to show the trend)

The assumed voltage drop at full power from 380V to 255V (32%) might be to high. If the supply voltage drop would be e.g. 10% then the current change woul be less than 2% , thats what I have called "basically no difference"