Berning output buffer

[Bill*B]

Can't get that David Berning output buffer out of my mind. I understand the autobias - but the circuit itself - as the transistor is a p channel, does that make it a push-pull affair, rather like a quasi White follower?

[SY]

The p-FET acts like the mirror of a vacuum tube, or alternatively, you can think of the tube acting like the n-FET in a complementary source follower stage.

And it works very well, albeit with a relatively high output impedance.

[Bill*B]

I doubt that the tube and FET have the same "curves" i.e. response to input signal. Does the FET mirror the tube because it could swing more current that the tube, but the tube limits the amount of current passed on the the FET, so that they are forced to be equal? Please don't despair at my ignorance - I'm not fluent with solid state and am struggling to understand this. Regards, Bill.

[SY]

No sweat, dude!

The FET and tube do have different curves, but that's stuffed down a bit because they're both so heavily degenerated. The match is still not perfect, but the idea was to have the buffer be equally capable of sourcing and sinking current. The distortion performance is OK (I measured about 0.1% at 2VRMS out), but a few orders of magnitude worse than the preamp I currently use.

[Bill*B]

Thanks, Sy! I hadn't considered feedback. If you were going to duplicate this circuit, what tube/transistor combo would you select?

[SY]

The best version of it I have tried used an ECC88 and a J271. The 2K resistors become 470 ohm, the B+ is 120V.

[PRR]

> a push-pull affair, rather like a quasi White follower?

The output stage?

More like the common sand-state loudspeaker amp output: NPN and PNP complementary symmetry.

Not exact symmetric complements, but....

> I doubt that the tube and FET have the same "curves" i.e. response to input signal.

Both are working as cathode followers. Gain is 0.9 to 0.98, so the "curve" is very slight. It isn't the same on both sides, but similar since both triode and FET are field-effect resistors. And the two curves go in opposite directions, so they tend to cancel. Not exact, but better than a triode against a resistor, and with hope of peak output current equal or higher than idle current.

The input stage is a cathode follower driving a grounded-grid stage. It is not push-pull like a long-tail pair: even-order distortion is not cancelled. It is however low because the 1K5 resistor swamps the variation in Cathode and Source resistances.

[Bill*B]

Thanks, guys. You've been a great help. Bill.

[SY]

I should make it clear that the distortion figure I quoted was for the buffer only- the gain stage has worse distortion performance. At 3VRMS out, the gain stage has about 0.5% distortion, with 2nd HD at -41dB, 3rd HD at -69dB, 4th HD at -87dB.

If you bypass the cathode resistor, at the same output, the 2nd HD increases to -34dB, 3rd drops to -81dB, 4th stays about the same at -85dB. That's probably a more musical distribution, but the 1% THD offends my Inner Engineer.

The advantage of the gain stage topology is bandwidth- the Miller Effect is eliminated.

[Bill*B]

Sy, with the 6DJ8/J271 combo at 120V B+, did you still keep B- at around -20V?

Berning shows a linesatage buffer on his web site of the same general configuration, but using a 12AX7 at what looks to be pretty low current, and minus the op amp. This is in the "P-1" preamp support section. I assume it was an early effort.

PRR, Sy, any comments on this style as compared to a CF with CCS load? Again, my thanks for letting me tap your experience.