G1=G2/mu Scaled Drive Strawman Design
The basic idea comes from smoking-amp in another thread:
The linearity is similar to screen drive but the voltage swing required is much less.
I adapted an amp design I've been developing to try out the concept. I haven't built it yet; still in the conceptual phase.
The basic topology is class AB, local current feedback loading a pentode+MOSFET CCS LTP gainstage and inverter with gyrators in the anode supply. Current feedback is derived across R122 and R125
The screen drive method borrows from some class A2 amps I've built, using the "mu output" of the gyrator for super-low drive impedance.
The scaled G1 drive voltage is generated by voltage divider R115, R116 from G2 to cathode. This is then buffered by MOSFET source follower Q106. There will be about 5V offset G2 positive WRT G1 in addition to the scaling this way (not sure of the effect...)
I show a stacked power supply arrangement with a third grid drive supply returned to the output cathodes. This keeps the grid current local to the output tube loop from grids to cathode. Other arrangements are certainly possible but I have had good results with this.
I guess the 6DQ6 gm to be around 3500-4000 wrt G2 using scaled drive, resulting in an effective anode resistance of 250-300 ohms in this circuit, hence the 6K6 OPT Zpri giving a damping factor of say (300+DCR) /1650 or about 5.
I think the sensitivity is about 2VRMS input for full sinewave output of about 50W. That can be tweaked by replacing all or part of R108 with diodes.
Comments or suggestions are welcome!
Interesting design. I see some anti-triode driver snuck in there too.
I did some further measurements for the 6HJ5 on the curve tracer using the resistor divider scaled drives. I found that the g2 was ramping up to 5 mA while the g1 was ramping up to 20 mA, with stepping from 0 to 55 V on g2 (1/4 that on g1). There was just a small amount of upward curvature in the current ramps versus voltage steps, probably the beginning of a diode curve (and a current spike around 0 V plate voltage). Would really need about twice that voltage range to fully test I think, but so far it looks as if the two grid inputs have near-enough resistive characteristics. So I am planning on just using a resistive divider from the g2 Mosfet follower drive down to the g1 for my first proto test ( g1 impedance is even conveniently scaled for the divider). A little droop in the g1/g2 ratio might actually further linearize the transfer anyway.
Probably can't use the simulator models for the output stage? At least I wouldn't expect the grid currents to be well modeled. But some scaled diode V to I may work for those.
For my 1st proto I'm thinking of using two 5 Watt video pentodes for drivers, with Schading feedbacks from the output plates to the driver cathodes. Since that ups the driver grid input voltage requirements, I will use some triodes (freebies in the video bottles) to make an LTP splitter up front.
I'm thinking of approaching the scaled drive concept from another viewpoint, but it will require a custom transformer. One comment - the control grid gets a follower, but the screen doesn't. You might consider running the follower into the screen instead, with a resistive divider to the control grids. I would think that being farther up on the drive feeding chain, the screens are more likely to pull current and require the services of a follower.
On the 6HJ5 (triode Mu 4.2) I found that the g1 draws 4X the current of the g2 grid (using 1/4 the voltage drive of g2). Probably some shielding of g2 occuring with the aligned grids, but one would also expect a positive g1 to "focus" electrons on g2 at some point. Maybe there will be some nasty "phase change" at a higher drive voltage than the +55V I have tested (on g2) so far.
I'm hoping that positive g1 causes the focusing focus point to occur short of the g2 wires, causing a crossover spray of electrons that largely misses g2. Have to get the proto together to find out. Certainly, g2 will take the brunt of the current when the plate V drops below 50 V (observed on the tracer, biggest current spikes on g2 at low plate V. Hmm, I'm going to re-check that again, not real sure). Would be nice if g2 could droop a bit under that abuse.
Michael does have the Mu follower to drive the g2.
I have been working on a circuit to drive both grids at once for a while. I have performed a few experiments along these lines and none resulted in anything useful. That hasn't stopped the process though, and if it rains tomorrow the switch might get thrown on my biggest step in this direction yet.
Years ago I found mention of a "high Mu triode" connection in the data sheet for a medium sized pentode (or tetrode) DH transmitting tube. It simply had G1 and G2 tied together. If my memory isn't faulty (highly possible) there were curves for this mode. I can not find that info anywhere now, but I have changed computers several times since then, and it might have been on paper.
At about the same time I saw a circuit for an amplifier using a pentode (807 I think) where G2 was directly driven and there was a resistor from G2 to G1. It was a push pull design using an interstage transformer. I breadboarded something similar using 6L6GC's and remember a whole bunch of power and a whole bunch of distortion. It sounded like some of my early amp experiments using the big germanium power transistors salvaged from 1960's car radios.
More recently (the Tube Sale at AES thread) I was melting tubes in screen drive mode (over 100 watts from a pair of 98 cent 6BQ6's) when I got the idea to tie G1 to G2 and drive them both with a PowerDrive circuit. The whole thing went bang on power up and there wasn't much left of the mosfets. I chalked the failure up to a wiring error (an unproven guess).
About a week later after some more successful other experiments I returned to the G1=G2 connection only to have it blow up again. This time I know it was wired right, so maybe oscillation was to blame. I gave up on this idea.
After reading the above mentioned thread and looking at the 6HJ5 curves I have decided to try it again. This time I am not going to tie the two grids together. I know that a mosfet buffer works to drive either G1 or G2 (done that on numerous tubes) and I know that a 6SN7 can drive two mosfets to the 200 V P-P level (done that to drive multiple tubes) so why can't I drive 2 mosfets from one 6SN7 and wire one mosfet to each grid of a power tube?
Since I tried some similar experiments back in the Adjustable Distributed Load thread, but never finished my experiments, these ideas should be investigated too. In fact it was the discovery of the breadboard for those experiments while closet cleaning that prompted this "lets blow it up one more time before tossing it" round of experiments.
I have several "HoneyDo" projects lined up for tomorrow, but they are all outdoors. Lets hope it rains, it rained most of today.
Looks like kind of "Anti-UL" concept. Nice implementation, BTW.
Here are some pics. These were taken on a timebase scope rather than the curve tracer, so are versus time rather than plate voltage as on the tracer. 11 stepped traces are sequenced, with every other one using an opposite 90 degree side of the power line sine wave voltage. There are actually two slightly different versions, since 11 does not divide out evenly by twos (rising and falling edges of 60 Hz may be reversed between sets) so some miss-match may be present between pics since 60 Hz sync was unknown on the scope. But general trend is obvious. ( I need to get the curve tracer set up to accept an external vertical signal so I can do regular grid current curves)
Current signals here was from a Tek AM503 DC current probe setup. Grid drives were 4 to 1. +55V max stepping on g2.
1) cathode current 50mA/div
2) plate current 50 mA/div
3) g1 current 10 mA/div
4) g2 current 10 mA/div
I'm puzzled by the cathode current and plate current traces, should be more similar I would think. Maybe I got the scale wrong on one. Well, at least one can see that the g2 current is the most spikey when plate voltage is low. Looks like g2 baseline current ramping is about half the g1 current ramp. Close to linear current ramps overall for g1 and g2, but some curvature upwards, probably a diode V/I curve. g2 looks more curved. g1 being closer to a straight ramp (resistive Z) is desirable as far as using just one Mosfet driver with a resistive divider for g1.
40K gate is quite a bit of external phase shift.
How does that compare with an internal mu?
Even a strapped triode...
For 300pF gate, I compute a 13KHz corner.
Though that doesn't take into account the
50K source follower load.
The resistive divider in the previous pics was approx. 300 Ohm from g2 to g1 and 100 Ohm from g1 to cathode. The 300 Ohm was adjusted until the voltages on the grids were 4 to 1. With 55V max stepping, that would be 137 mA thru the divider alone. But g1 is drawing about 80 mA max. Not a problem if it is a linear Z, since the attenuator was adjusted for a 4 to 1 result, but since it is not strictly linear, it may be causing the straigtening of the g1 current curve, since it would reduce the divider ratio as it draws excess (above the expected linear ramp) current. Probably g1 is just as curved as g2. I'll have to rig up a Mosfet g1 driver to test it on the tracer again.
Hmmm, maybe it will eventually be useful to put some extra resistance in the g2 leg too. The non-linear current draw by the grids is self linearizing that way (versus the internal 3/2 power transfer function).
In retrospect, what you found makes sense, as the control grid is closer to the cathode than the screen, so it should draw more current for a given positive drive. The place to look for freaky behavior will be at screen voltages of about 100V or so, though with the scaled G1/G2 drive 50V is as likely to drive the tube as hard as 100V G2 drive alone. I'd expect to only see such excursions at or near clipping for sufficiently frisky drivers.
|All times are GMT. The time now is 02:19 AM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.
Copyright ©1999-2016 diyAudio