Schade and CFB exactly equivalent

[smoking-amp]

Yes, #4 there gets the advantage of a low Z driver direct to g2.

I re-tried the 21HB5 in Schaded g2 drive, without the 120 Ohm grid stopper this time, and no difference. So it must be the low impedance of g2 is messing with the Schade divider network.

Here are some similar variations using the 6HJ5.

1) 6HJ5 with Schaded g2 drive (20 mA/div Vert, 50 V/div Horiz.)

2) 6HJ5 with just g2 drive, but through the same 22K Ohm resistor used in the above Schade R divider setup. (20 mA/div Vert., 50 V/div Horiz.)

Looks awful, so the low g2 Z is what is messing up the Schade divider. But amazing how well the Schaded g2 (above) is able to almost clean up most of the mess.

3) 6HJ5 with direct g2 low Z drive (B. pentode) (50 mA/div Vert., 50 V/div Horiz.)

So g2 will have to have a Mosfet follower to drive it for any Schade arrangements.
That would then give g2 Schade curves that are evenly spaced, and nearly straight lines.

What I wonder is how that would compare with the linearity of g1 Schading. One loses the internal Mu of 4 in loop gain in the g2 version, but the g2 B. pent. curves are much more linear spaced to begin with. Could be close. But the g2 version is harder to drive.

...........

 

[smoking-amp]

Case #4 above (post 99 ) can be configured similar to a g2/g1 drive idea that has been around the last couple of years. That idea was to proportion the AC drive to g2 and g1 as 1/Mu internal. That way each grid does half the work, and the AC swing on g2 can be halved then. That used a Mosfet follower to drive g2, and an attenuator off that (1/Mu internal) to drive g1.

Instead of using an attenuator for g1, we can put a Schade network in as follows (diagram). This can reduce the drive on g1 AND lower distortion via Schade Fdbk. Win-win. Maybe call it Schady g2/g1 drive??? One would want to optimise the Schade divider ratio to get something like 1/Mu internal reduction of the g2 drive signal voltage swing appearing on g1.

This may require a higher supply voltage for the Mosfet to get enough g2 bias. One could do the same iso. cap and bias R for g2, like used on g1, to reduce that. G2 only needs 1/2 the voltage swing here compared to standard g2 drive. One the other hand, the Mosfet gate input probably has a cap there already, so the Mosfet could just be run off B+ with modest current to get g2 bias up. (R3 can be a high resistance for the N channel driver case, since g2 pulls it down already.)

One could use a P channel Mosfet driver/follower as well (swap R3 and Q1 positions), for those inclined. (Schematic shown with the N channel driver version) (the P channel version would more benignly fail, via a Mosfet short, to a tube turned off)

Grid 1 of the tube would like to have a low R grid stopper too (not shown), to prevent parasitic Osc.

 

[SpreadSpectrum]

Do you have enough power supplies to rig this up and draw some curves? It looks interesting.

 

[Gary P]

Here is the FQP2N90 Mosfet in Schade again with 30 Ohms in the source lead, and 50 mA/div Vert., and 50 V/div Horiz. for certain.

OK, you got me thinking... I've been using cascoded MOSFETs for both input and output stages running at tube like operating points. In my application the output stage is run with 50K resistors for local feedback from the drains to the gates. The sources each have 5Ω resistors that are connected to a CCS that biases the stage. The input stage drives the output stage with current.

I was wondering how the 576 could be used to show what the curves would look like. Had a thought about directly driving the feedback resistor from the step generator running in current mode. Went and messed around with the 576 to see if the idea would work. It does... The step generator in current mode seems to have enough voltage compliance to do this task just fine.

IRF820B MOSFET with 50KΩ connected from the drain to the gate. 5Ω on the source, step generator connected to the gate/feedback resistor.

 

[SpreadSpectrum]

That's really interesting, Gary. Thanks for posting. That is some impressive linearity and curve slope.

So it looks like with that super low effective "plate resistance" from the feedback, your SS Tabor amp output impedance is probably dominated by transformer losses. Pretty cool.

I've got some experiments lined up and mosfet cascodes are definitely near the top of the list.

 

[smoking-amp]

"Do you have enough power supplies to rig this up and draw some curves? It looks interesting."

I'll trace some screen driven, Schaded g1, tubes soon. Need to recover from being traumatized by the taxman lately. I used up all my printer paper filling out forms. There's probably a form for that too. Final insult: "enclose your check with Form V, which cannot be found here or anywhere with the 2000 other forms you already filled out"

In the mean time, here is a "Schade" like driver proposed by Wavebourn some while back, using a P channel Mosfet hung from B+. (link provided) It should be adaptable to screen grid drive. It has the benefit of removing power supply ripple, which the typical "Schade" R Fdbk suffers from. (the usual Fdbk R references the feedback to B+ instead of ground) (Wavebourn's approach only has current sources between ground referenced and B+ referenced signals)

It also does not require the increased drive signal that is needed with a series R after a Mosfet follower. (for mimicking a current source drive)

One could also just connect the driver plate directly to the Mosfet drain, but that would need a buffer yet for screen drive. (Mosfet gate would be biased from B+ then).


http://www.diyaudio.com/forums/tubes-valves/205965-unusual-amps-64.html#post2919395

............