Suitable operating point for 6SN7 as a differential driver ?

[ray_moth]

In my PP amp (pleased see attachment), I'm using the 6SN7 as a differential driver for triode-mode EL34s. Each half of the 6SN7 has about 230v plate-to-cathode and is drawing about 4 mA. Cathode-to-ground voltage is about 40v. I've tried to avoid the low current, low plate voltage often seen in some early designs but I guess my 4mA per side is still pretty low.

I wonder if, in other people's experience, I would improve the sonics by reducing the common cathode resistor (currently 4.7k) and reducing the plate load resistors (currently 33k) to get a higher current?

[Geek]

Class A output stages draw no grid current and the grid-leak resistors are high enough that there would be little distortion.

Personally, I have run drivers anywhere from 10 to 80% Pa and found little sonic difference when driving class-A or AB1.

[SY]

There is a linearity benefit to be had by running the 6SN7s at a higher current, maybe double what you're doing. I would hesitate to go much lower on the plate resistors.

Why not use a current source in the tail rather than a resistor?

[kevinkr]

Using a current source to bias your differential driver circuit is an excellent idea, ring of two transistorized current sources work quite well for this.. Note that the current source should be floated off of a low voltage negative supply (-12V will work fine) which does not even need to be regulated, but should be reasonably well filtered. I have found that the bias voltage present on the cathodes of the 6SN7 is not enough for the current source to function properly with audio present on the cathodes - drives the current source to cut off on negative going peaks. Sounds very nasty, ask me how I know. Using a current source for bias will improve common mode rejection, but more importantly improve AC balance. I think something on the order of 6 - 7mA per section is a good compromise. Plate resistors should not be much less than 20K.

[ray_moth]

Thanks very much for your responses. I don't have the bits & pieces to make a CCS for the driver at the moment but I'm going to try reducing the 6SN7 plate resistors from 33k to 22k, reducing the tail resistor from 4.7k to 2.7k and removing the 440 resistors in each cathode. This should increase the current in each half of the 6SN7 to about 7mA, whihc might sound better.

I also want to insert 1M resistors between 6SN7 plate and 6SL7 plate, to improve DC balance and introduce a bit of negative feedback across the driver.

BTW, what sonic improvement might I notice if I do use a CCS in the tail of the driver? I can understand ther need for it in the tail of the splitter, to achieve balance.

[Geek]

Originally posted by ray_moth
BTW, what sonic improvement might I notice if I do use a CCS in the tail of the driver?

An unbelieveable improvement in transient response for starters...

[SY]

Dropping the tail resistor will increase imbalance in the diff pair. Even a simple one transistor current source (you don't have to get fancy) will outperform that resistor.

Use your 33K resistors, use a simple current source (15-20 ma for the pair), and get the B+ up and you'll have a terrific driver stage.

[ray_moth]

Right, you've convinced me! How about if I use a PN2222A as a current source? I can get a useful neg. voltage from the 50v 100mA winding on my tranny and use it at the same time for neg. bias on the EL34s. Please see attachments to this and the next post (fingers crossed - this often fails!)

[ray_moth]

The power supply:

[SY]

I haven't gone through and figured out how much the local DC feedback upsets the operating points of each stage to which its connected, so no comment there, other than "lotsa luck." You'll have to do a bunch of iterative calculations to get at this.

But the CS for the diff amp can be put together in a more stable and simple way. Run a resistor (R1) from one of the positive supplies to the base of the 2222. Run a resistor (R2) from the emitter of the 2222 to one of the stable negative rails. Then connect a forward-biased diode and a reverse biased 6.2V zener in series from the 2222 base to the far end of R2. (Sorry to be wordy, I'm not near my scanner at the moment)

Calculate R2 from R2 = 6.2/i, where i is the total current for the diff pair plus feedback. Calculate R1 from (Vpos-Vneg+6.8)/10i.

Other advice: breadboard like crazy before cutting metal and doing a final build.