SSS to negativ bias voltage???

[martin manning]

The output impedance of the cathode follower driver is much lower than the typical fixed bias arrangement where any grid current demanded by the power tubes must be pulled through the 220k bias feed. For the cathode follower it's around 500 ohms. 12AX7 can't supply much current, but the difference is audible.

[soeves]

Ok! But Even the drivertube must pull the electrons through the 220 kohm resistance, just like in the typical fixed bias setup. So how come it makes the bias voltage on output tube more stable?

[martin manning]

This situation is different. If the load tries to pull the cathode voltage down, the grid becomes more positive wrt the cathode. This causes current flow through the tube to increase, counteracting the cathode voltage drop.

[Richard1001]

Ok! But Even the drivertube must pull the electrons through the 220 kohm resistance, just like in the typical fixed bias setup. So how come it makes the bias voltage on output tube more stable?

The difference is that the traditional bias supply has a high impedance connecting a fixed bias voltage to the grid through 220k resistors in series. Very little current draw from the power tube grids will pull the voltage up very easy.

The powertube grids draw almost no current when idle, but with a large signal voltage on the grids, the gridcurrent draw will vary. Also, with high frequency signals (specially high frequency's found in distorted signals) the input capacity will be a higher load for a conventional long tail phase inverter.

A cathode follower has very low output impedance and a very high input impedance. The driver tube will actively maintain the bias voltage that is set by the reference voltage. The reference voltage connected to the driver tube grids will see a very high impedance keeping the voltage constant regardless of the signal applied. The cathode will follow this reference voltage with a low impedance. It can do this because it can vary the current draw through the tube itself which translates to the right dropping voltage over the 220k resistor. The amount of current draw of the power tube grids has very little effect on the bias voltage.

Also, with high frequency signals applied, the low impedance of the driver tube can handle the input capacity of the powertubes more easy, giving a higher bandwidth.

But although the output impedance is very low, ofcourse there is a limit to the maximum current the driver can deliver. This limit to the maximum current will also limit the signal bandwidth. An ECC83 is not a very powerful tube for a driver and the 220k cathode resistor limits the maximum current (together with the supplied voltage) But even this relative low current is still a whole lot more than a conventional high impedance bias circuit can supply.

Lowering the supply voltage will limit the performance of the driver, so in the original design the tube is set to the very limit of what it can handle. For some (not all) current production tubes this is sometimes to much.

[martin manning]

Lowering the supply voltage will limit the performance of the driver, so in the original design the tube is set to the very limit of what it can handle.

I don't see any performance issues (voltage swing, bandwidth) with running the drivers using a bipolar supply at +/- 160V. It's a much better solution IMO.

[GAStan]

Lowering the supply voltage will limit the performance of the driver, so in the original design the tube is set to the very limit of what it can handle.

I don't see any performance issues (voltage swing, bandwidth) with running the drivers using a bipolar supply at +/- 160V. It's a much better solution IMO.

FWIW my SSS was originally build with the same setup you have. I have since converted it to the bipolar supply Martin is suggesting with good results. I am unable to detect any tonal or "feel" differences between the two. The conversion is not difficult for someone able to build a SSS, and is reversible.

[rootz]

Ok! But Even the drivertube must pull the electrons through the 220 kohm resistance, just like in the typical fixed bias setup. So how come it makes the bias voltage on output tube more stable?

But although the output impedance is very low, ofcourse there is a limit to the maximum current the driver can deliver. This limit to the maximum current will also limit the signal bandwidth. An ECC83 is not a very powerful tube for a driver and the 220k cathode resistor limits the maximum current (together with the supplied voltage) But even this relative low current is still a whole lot more than a conventional high impedance bias circuit can supply.

Lowering the supply voltage will limit the performance of the driver, so in the original design the tube is set to the very limit of what it can handle. For some (not all) current production tubes this is sometimes to much.

Is it and does it?
First of: the 12ax7 does not limit bandwidth. The output impedance in conjunction to the Miller capacitance of the grids of the output tubes do. I totally agree that it is a limited tube if you shoot at Class AB2 operation.
When providing grid current to the power tubes, the current comes from the anode side of the driver, so the 220k is essentially out of the equation. No limits there.
Lowering the positive voltage does not limit the performance in any meaningful way IMHO, if you keep certain safe margins. Vak needs certain margins to keep gm up. But as seen in Martin's power supply and SSS005 this does not need to be the full screen voltage at all. In SSS005 the plate node for the driver is actually pretty high impedance. This of course limits the AB2 capabilities, but so does the 12ax7. It seems class AB2 operation was never Dumble's goal in itself, otherwise he probably would have designed it better. He certainly was competent enough to do that!

[Richard1001]

Ok! But Even the drivertube must pull the electrons through the 220 kohm resistance, just like in the typical fixed bias setup. So how come it makes the bias voltage on output tube more stable?

But although the output impedance is very low, ofcourse there is a limit to the maximum current the driver can deliver. This limit to the maximum current will also limit the signal bandwidth. An ECC83 is not a very powerful tube for a driver and the 220k cathode resistor limits the maximum current (together with the supplied voltage) But even this relative low current is still a whole lot more than a conventional high impedance bias circuit can supply.

Lowering the supply voltage will limit the performance of the driver, so in the original design the tube is set to the very limit of what it can handle. For some (not all) current production tubes this is sometimes to much.

Is it and does it?
First of: the 12ax7 does not limit bandwidth. The output impedance in conjunction to the Miller capacitance of the grids of the output tubes do. I totally agree that it is a limited tube if you shoot at Class AB2 operation.
When providing grid current to the power tubes, the current comes from the anode side of the driver, so the 220k is essentially out of the equation. No limits there.
Lowering the positive voltage does not limit the performance in any meaningful way IMHO, if you keep certain safe margins. Vak needs certain margins to keep gm up. But as seen in Martin's power supply and SSS005 this does not need to be the full screen voltage at all. In SSS005 the plate node for the driver is actually pretty high impedance. This of course limits the AB2 capabilities, but so does the 12ax7. It seems class AB2 operation was never Dumble's goal in itself, otherwise he probably would have designed it better. He certainly was competent enough to do that!

Hmm...

There is a difference between the DC bias voltage and the AC signal. The AC signal is superimposed on the DC bias. You always want the AC signal to swing evenly around the constant DC bias voltage center. You don't want this center drifting.

And Mr Dumble did not design this driver circuit. It was copied from a McIntosh audio amplifier.

[martin manning]

And Mr Dumble did not design this driver circuit. It was copied from a McIntosh audio amplifier.

It is similar to the McIntosh MC-30/40/60/75 driver stage, but only with respect to the biasing. The McIntosh power stage is an unusual arrangement where the plates and screens of the power tubes are connected to opposite ends of the OT primary coil. Each driver's plate is connected to the screen of the tube it is driving, and all of the B+ comes through the OT primary CT. There are two versions of the bias supply, an earlier one that has the negative voltage derived from the HV secondary (like SSS 002), and a later one that uses a bias tap on the HV secondary. The later one uses a single half-wave rectifier, and the negative voltage is -150V. I guess we could say that Dumble was "inspired" by the McIntosh driver. In any event, they decided to go to a "less negative" voltage for the drivers, I would guess for reliability reasons.

[rootz]

But although the output impedance is very low, ofcourse there is a limit to the maximum current the driver can deliver. This limit to the maximum current will also limit the signal bandwidth. An ECC83 is not a very powerful tube for a driver and the 220k cathode resistor limits the maximum current (together with the supplied voltage) But even this relative low current is still a whole lot more than a conventional high impedance bias circuit can supply.

Lowering the supply voltage will limit the performance of the driver, so in the original design the tube is set to the very limit of what it can handle. For some (not all) current production tubes this is sometimes to much.

Is it and does it?
First of: the 12ax7 does not limit bandwidth. The output impedance in conjunction to the Miller capacitance of the grids of the output tubes do. I totally agree that it is a limited tube if you shoot at Class AB2 operation.
When providing grid current to the power tubes, the current comes from the anode side of the driver, so the 220k is essentially out of the equation. No limits there.
Lowering the positive voltage does not limit the performance in any meaningful way IMHO, if you keep certain safe margins. Vak needs certain margins to keep gm up. But as seen in Martin's power supply and SSS005 this does not need to be the full screen voltage at all. In SSS005 the plate node for the driver is actually pretty high impedance. This of course limits the AB2 capabilities, but so does the 12ax7. It seems class AB2 operation was never Dumble's goal in itself, otherwise he probably would have designed it better. He certainly was competent enough to do that!

Hmm...

There is a difference between the DC bias voltage and the AC signal. The AC signal is superimposed on the DC bias. You always want the AC signal to swing evenly around the constant DC bias voltage center. You don't want this center drifting.

And Mr Dumble did not design this driver circuit. It was copied from a McIntosh audio amplifier.

Agreed that the bias point should not shift. But in reality it simply doesn't in a meaningful in either the SSS005 power supply or Martin's solution.

And then again, both 002 and 005 benefit (or should I say: need) a protection diode against negative voltage at either CF driver plate node or screen node.