• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Gain stage with g2-feedback.

A common way to do a gain stage is to combine a common cathode circuit with a cathode follower. The common cathode block may be a triode strapped pentode. Something like this:

[IMGDEAD]http://www.veiset.net/tech/P01.gif[/IMGDEAD]


Since the cathode of the CF is close to the plate voltage of the pentode it is temping to connect the screen grid to the CF’s cathode. This will result in feedback from the CF to the input stage.

[IMGDEAD]http://www.veiset.net/tech/P02.gif[/IMGDEAD]


This will at least do two things: reduce the output impedance and linearize the CF.

Here are some measurements of the two circuits. Blue curve is without and red curve with g2-feedback. The output impedance is 690 ohm for the plain CF and 280 ohm with g2-feedback.
Most of the distortion belongs to the common cathode block, so it might be worthwhile to play with the working point for that stage.

[IMGDEAD]http://www.veiset.net/tech/THD_g2_feedback.gif[/IMGDEAD]


Jan E Veiset
 
Nice idea and interesting results, Jan!

Would be even nicer with some empirical additional data about the change in stage amplification factor, although this probably could be computed backwards from the changes in Zout and THD, together with the coefficients of the tube being used in the CF position, and whatelse.

So hey, please also add the A figures, if the circuit still is on the bench and it doesn´t take you too much effort to measure four AC peak-peak voltages :)

Regards,

Tom
 
The gain is quite identical in the two circuits, in the first circuit the feedback is local from plate to g2 and in the other the feedback loop is extended to include the CF. With the same plate resistor (56k) A=11 when 6SJ7 runs as a triode and A=10.5 with the extended feedback loop.
I'm not sure why the gain is a little bit lower in the second case. Since the gain of the CF is less than unity there will be less feedback to g2 versus triode mode and that should have given me more gain, not less. :xeye:

Jan E Veiset
 

Shoog

Member
2002-08-15 10:16 pm
Eire
I was thinking of trying this with my FVP5 clone which has a super linear cathode follower second stage. I got into discussion with Joe Rasmussen about it and he was strongly of the opinion that the Common Cathode stage as a triode with a high R load it would sound better than a UL configuration.
He put me off trying the idea, but I may just give it a go when I do my next major rebuild.

One thing that might be an issue is that having the anode and the screen at the same voltage will increase the screen current which puts an additional none linear load on the common cathode. I thought about using a resistive voltage divider to avoid this and reduce the screen current to insignificant levels.

Shoog
 
jane said:
I'm not sure why the gain is a little bit lower in the second case. Since the gain of the CF is less than unity there will be less feedback to g2 versus triode mode and that should have given me more gain, not less. :xeye:

I think it finally sunk in. When the screen is not connected to the anode I will lose some transconductance and the gain will decrease.

Jan E
 
Thanks for your comment pedroskova :)

Some additional information about the circuit: The input stage runs without decoupling of the cathode. This will bring in some degenerative feedback in this stage. If we decouple the cathode the gain will increase and the loop feedback will also increase. When the cathode is decoupled it is easier to calculate the effect of g2 feedback. With g2 feedback the gain of the 6SJ7 is close to triode operation, but rp will be a bit higher singe g2 is not connected to the anode, so this is just a rough estimate:
Ap (pentode mode) = gm x Rl = 1.6mA/V x 56k = 90x
At (triode mode) = mu x Rl / (Rl+rp) = 19 x 56 / (56+8) = ~16x

Ap / At = 90 / 16 = 5.5x, if we adjust this number by the gain of the CF (~0.9) expected effect of g2 feedback should be 5.5 x 0.9 = ~5x

This indicates that the output resistance (with decoupled cathode in the input stage) should be about 5x lower than without feedback. Rout without feedback was 690R and 690R/5 = ~140 ohm. Rout with decoupled Rk was measured to be 150 ohm.

Jan E Veiset
 

Shoog

Member
2002-08-15 10:16 pm
Eire
Sorry my misunderstanding,

I think at this stage it would be well worth actually building the circuit and seeing if it performs in reality. On paper it looks like this idea should produce a worthwhile improvement, but the proof of the pudding is in the eating.

I for one will be very interested in such a test.

Shoog
 
I think at this stage it would be well worth actually building the circuit and seeing if it performs in reality. On paper it looks like this idea should produce a worthwhile improvement, but the proof of the pudding is in the eating.

Here is the pudding: :)

24MAR08B.jpg


The results I'm writing about in this thread are a mixture of theory and practical measurements.

The circuit performs about equal to a triode connected pentode. The gain and distortion will be about the same as for a triode connected pentode. If you find a well suited pentode for triode connection and a good operating point you'll have good gain stage. The benefits of connecting the screen grid to the cathode of the CF are that you will include the CF in the feedback loop which will linearize the CF and decrease the output resistance. I'll suppose the feedback loop also will increase the PSRR of the circuit a bit (I haven't measured that yet).

Jan E Veiset
 
Hi Jan,

If you replace the 56K plate load resistor with a CCS, the loop gain will go way up. Which should linearize the gain better and lower the output Z. Putting a voltage dropping zener (maybe 50V) between the 6SR7 cathode and the screen grid should linearize the gain too, by decreasing the screen grid current (by moving the screen voltage below the plate voltage).

Instead of the CCS, one could also try 2 load resistors in series for the 6SJ7 and a bootstrapping cap. from the 6SR7 cathode to their junction. (simulating a CCS load for the 6SJ7)

Don
 
smoking-amp said:
Instead of the CCS, one could also try 2 load resistors in series for the 6SJ7 and a bootstrapping cap. from the 6SR7 cathode to their junction. (simulating a CCS load for the 6SJ7)

I did some measurements with bootstrapping today (27+27k plate resistors and a bootstrap cap). This linearized the input stage and the distortion dropped with a factor of ~2x. However, the output impedance increased due to the positive feedback from the bootstrapping. This effect will be less pronounced for a common cathode stage with decoupled cathode.

THD Zout Circuit description
1.3% 790 6SJ7 Triode connected + CF
0.6% 1250 Bootstrapped 6SJ7 Triode + CF
1.0% 280 g2 feedback from CF
0.4% 800 Bootstrap + g2 FB from CF

THD measured at Vout = 10V rms (28V pp).
THD is approximately proportional to output level.

Jan E Veiset
 
After building (and testing) the prototype, I though it would be a shame to put this idea back in the drawer so I decided to make a "living room" preamp out of it...
I found the distortion to be lowest when running the input stage at low current. At Ik=~1mA and a corresponding plate resistor the THD figures shows:
0.05% at 1V rms output.
0.5% at 10V rms output.
Output impedance with EBC33 is ~200ohm.
Frequency response: 10Hz – 100kHz (+0/-3dB, 50k load)
S/N: ~78dB ref 1Vrms output.

The preamp sounds very similar to a standard common cathode / CF preamp that I’ve got but the measurements are a bit better (half the distortion, third the Zo, etc).

I’ve got a helping hand from Mapletree Audio, Canada, with the chassis.

At the work bench:
[IMGDEAD]http://www.veiset.net/hyperlinear/25MAI08A.jpg[/IMGDEAD]

On a sunny afternoon:
[IMGDEAD]http://www.veiset.net/hyperlinear/28MAI08A.jpg[/IMGDEAD]


Jan E Veiset