Design questions, driver stage for SE.

[jane]

Design questions, driver stage for SE.

Alternative A:
The pentode in the input stage is used for amplification. The anode load is a self-biased cathode follower (mu-follower). Reflected anode load for the pentode is ~300k, the output resistance from the pentode is approx 230k (300k||1M). This connection is used for shunt feedback.
The output stage is fed from the cathode of the mu-follower, approx output resistance is ~3k with a max voltage swing of +/-80Vp.



Alternative B:
A CCS replaces the anode load for the input pentode. Output resistance is ~1M. This connection will both be used for shunt feedback and feed a DC-coupled cathode follower for the output stage. Output resistance from the CF is ~400 ohm and max voltage swing is about -150/+100Vp.



In alternative B the amplification is higher, the output voltage swing is greater and the output resistance from the CF is lower.
The catch is the CCS. This will make the DC-working point for the pentode very dependent on pentode’s parameters and thermal drift in the CCS. A small variation in one of these parameters will alter the working point +/-100V.
One way to stabilize the working point can be to shunt a resistor (e.g. 500k) in parallel with the pentode.

Common conditions: Po in triode mode is ~20W, in UL greater than 30W. Zo without feedback is ~2 ohm in triode mode, ~6 ohm in UL and distortion is ca 6-8%. Shunt feedback is 10-14dB that will lower the distortion and Zo with a factor of 3-5x.
Input sensitivity for full tilt with feedback is ca 1Vrms.

Any suggestions what way to go?

Jan E Veiset
NO-6600

[jane]

The catch is the CCS. This will make the DC-working point for the pentode very dependent on pentode’s parameters and thermal drift in the CCS. A small variation in one of these parameters will alter the working point +/-100V.

One way to stabilize the DC working point can be to apply screen grid DC feedback to the input stage. How well that will work, I really don't know...
Maybe I’m starting to over complicate--I usually like to keep things simple. Anyway, here is the draft:



Jan E Veiset
NO-6600

[Eli Duttman]

Jan,

Pentodes are very effective current sinks. That makes them DIFFICULT to use with an active load. "Fighting" starts between the pentode and the load.

I suggest that you resistively load the 'SJ7 and DC couple it to a buffering cathode follower.

Pete Millett has a circuit for a low/high dual B+ rail supply from a single CT rectifier winding. PM's design could be set up lower voltage/high current and higher voltage/low current. That would allow you to use a LARGE load resistor for high gain from the 'SJ7. Regulating the higher B+ for low noise is also a possibility.

[rdf]

Hi Jan. You've probably already considered it but a possible option (haven't tried it yet) is to replace the pentode CCS in Version B with a choke. Its much lower DC impedance would better stabilize the static operating point. A variation of Version B with choke is one of the things lined up for measurement when I get back to town. I also want to play around with different resistive output loads to try minimizing higher distortion components.

BTW, the screen feedback technique you show is similar to that used in the Mullard 3-3, except in the latter case the screen is held above the plate and the whole thing runs at very low currents and voltages (0.2 ma and ~35 vdc from memory.)

[jane]

Originally posted by Eli Duttman
I suggest that you resistively load the 'SJ7 and DC couple it to a buffering cathode follower.

Hi Eli,
That’s an easy and good way to keep things simple, but the shunt feedback will like to see a high resistant driver. Let’s say we put in a 68k resistor: The anode current (after trimming the cathode resistor) will then be ~3 mA, which according to the data sheet is design-center, and the amplification will be something like gm x R = ~100.
OK, if we apply 14dB shunt feedback, Rfb need to be ~100k and the peak current in the feedback path will be more than 3mA. Hmm..

Jan E Veiset
NO-6600

[jane]

Originally posted by rdf
You've probably already considered it but a possible option (haven't tried it yet) is to replace the pentode CCS in Version B with a choke.

That's a very good (but expensive) idea!
That will give me the high output impedance that I'm looking for and definitively high voltage swing! A choke can be one way to go.

BTW, the screen feedback technique you show is similar to that used in the Mullard 3-3

I'm just born 50 years to late - I should have been working for Mullard in the 50s.

Jan E Veiset
NO-6600

[rdf]

If you're interested Hammond has a suitable inexpensive choke for benching, the 156C. 150H, ~3.5K ohms, 8 ma max., $12. In testing, a solid state CCS like the IXYS has more benign high-order distortion components but the Hammond is certainly good enough for proof of concept stuff. And they work a pip as grid bias supply smoothing chokes if you decide on a different plate load solution.

[jane]

Originally posted by rdf
Hammond has a suitable inexpensive choke for benching, the 156C.

Thanks for the suggestion.
I have never thought about the Hammond's filter chokes as plate chokes, but that is an inexpensive get and probably as good as anything else. 150H is a nice value: in series with a resistor for voltage drop the impedance will stay quite high even at low frequencies and the shunt feedback will help me linearize the LF response.



Jan E Veiset
NO-6600

[rdf]

Originally posted by jane
.... in series with a resistor for voltage drop the impedance will stay quite high even at low frequencies and the shunt feedback will help me linearize the LF response.

Good point, all of my playing around has been with low impedance triodes or triode-connected pentodes. The 156C is self-resonant around 3 kHz, which calculates in the range of 2.5 Meg max impedance. I was going to shunt it with a high value R, chosen by measurement to mimize harmonics above the second, to limit the variation seen above that by the driver. I didn't give sufficient thought to low frequency loading. Looks like a combination might be the better approach.

BTW, have you considered LED biasing for the pentode? Two reds are right in the ballpark and can result in as little as 10 ohms in the cathode circuit.

[jane]

I ended up with the mu-follower.
The amp is very powerful and clean sounding. The bass response is very impressive, no sign of OPT saturation at 20Hz/25W.



Some key figures:
Bandwidth (Po=10W): Triode: <10Hz-70kHz (-3dB), UL: <10Hz-75kHz (-3dB)
Max power output rms: Triode: 24W, UL: 45W
Distortion (Po=10W): Triode: 0.6%, UL: 0.3%
Damping factor: Triode: 6.9, UL: 5.3

Schematics:


Power Supply:


Internal wiring:


Cathode resistors and decoupling caps:


Distortion pattern 400Hz 10W UL:


Square wave 10kHz / 10W Po


Sowter make some very heavy transformers:


Jan E Veiset
NO-6600