Cathode follower question

[gyraf]

Hi Group,

Doing an output stage with rather limited anode voltage (+120V), I run two cathode followers in differential/push-pull mode coupled to each side of output transformer - see attached .jpg

There's always a resonance bump where coupling capacitor meets transformer primary inductance, which can be tamed by adding series resistance-with-capacitor-in-parallel, raising output impedance in sub-low frequency range.

In order to get this sorted in the best way my intuition tells me to get rid of the coupling capacitor. I really do like to minimize parts in signal path.

Both ends of output transformer sees approximately +32VDC (+/-10%, depending on tube) so there is a net DC voltage difference of a couple of volts at max. - but this gets evened out when omitting the coupling capacitor (by the rather low winding resistance of the transformer, some 30 Ohms).

Measures fine. Both frequency response (gets rid of a 6dB bump at 11Hz) and MOL/50Hz (1% @ +18dBu both with and without coupling capacitor)

The questions are:

Am I getting this wrong - I can't get to wrap my head around implications for DC working conditions in CF stage?

Any other potential problems that I should consider?

Jakob E.

 

[Harpo]

A gapped transformer ? I wouldn't assume both tube filaments coming up simultaniously, so the 32VDC wouldn't be common mode but differential for a short period of time.
I've never been a tube guy, so maybe disregard.

 

[ruffrecords]

Interesting topology. I also worry about what happens to dc conditions in circuits like this. It is the same problem as you get when paralleling two tubes.

If you think about this as a CF version of a regular push pull circuit, then if the transformer were in the anode circuit the HT would be fed to the centre tap. So how about connecting the centre tap of the CF transformer to 0V . Not sure how you would do the bias.

Cheers

Ian

 

[gyraf]

Even at non-even heat-up, current flowing in the primary is probably quite small, less than half the quiescent 6.5mA?

The transformer is silicon-iron "C"-core, and thus should not be overly sensitive to small DC-currents (if I get it right)

??

 

[AndreasS]

Hello Jacob,

pls. try connect the grids like the Blumlein garter:

http://www.tubecad.com/2005/May/21/BlumleinAutoBiasCircuit.png

Regards Andreas

 

[emrr]

I think you are fine removing the cap, especially if it is measuring fine.  I'm not seeing an obvious problem. 

Other options I'd look at if the primary center tap are available:

C/R from pri CT to ground along lines Ian suggested, could be better with symmetry introduced, resonances may improve slightly, or not.  Those problems lead to:

R96/R102 combined as one 150R from pri CT to R105/R106 combined as one 2K35R, rather than R96/R102 from cathode.  This might be terrible! 

 

[abbey road d enfer]

Am I getting this wrong - I can't get to wrap my head around implications for DC working conditions in CF stage?

Nothing wrong with this. Not any wronger than standard push-pull, for which proper operation depends on how well the halves are balanced. For optimal performance, you should provide a way of adjusting /balancing the operation point, either by balancing the cath resistors or the grid bias.

As someone mentioned, there is a possibility to connect the center tap to 0v, taking the cath res off, but then you have to implement grid bias.

 

[joaquins]

Why not to bias it from the center tap with a resistor? This way you have your AC at the transformer with DC opposed and the DC bias with that resistor, could be bypassed with a cap if that helps, but I'm not sure it's necessary. About 150Ω or 180Ω should do the DC bias, since would be like having the two 330Ω in parallel for a given DC current and you don't need the 4k7 resistors anymore since they were there to provide dual polarity movement from the idle point which will be taken care by the transformer now.

Is this a center tapped transformer or you have both primary windings (4 wires) independently? In that case you could just connect each winding in series with the 330Ω resistor for each tube which will be easier to balance the current, they will already work more balanced from start and even better if you add a balance pot, maybe 100Ω pot, wiper to ground, each end to one 330Ω resistor. This config will have higher output impedance but could be corrected bypassing with caps the resistor+pot biasing, in which case you still have the cap.

The easiest way of getting rid of the 11Hz bump is increasing the cap value quite a bit so the cut frequency goes to a much lower freq where it's no longer a problem.

JS

 

[PRR]

Hard to *know* without transformer parameters (with and without DC).

But if you move the transformer leads to the bottoms of the 330r resistors, you get some output impedance and thus some damping on your under-damped bass resonance. Output level is a little smaller, output impedance a little higher, but both may be acceptable. And it is zero-cost (not your main priority, but does mean easy testing).

 

[PRR]

> low winding resistance of the transformer, some 30 Ohms

What is nominal primary impedance? 30 Ohms suggests maybe 600 Ohms, not 10K as we would normally load a tube. 6mA into 600r is not a large output level (maybe +10dBm).

THD will be non-negligible, though reduced by push-pull action.

 

[joaquins]

I don't know ECC88, don't have the parameters in my head but a 12AU7 could drive 600Ω in a cathode follower configuration, of course having 660Ω in series would probably be a problem for LF response, and in push pull configuration each triode should drive a 300Ω equivalent load. Of course it will depend on what you load it with but still will probably need to be quite low to damp the transformer, at least adding a zobel or a resistor to load it.

In any case the load on the tubes would be quite similar to what they have now, even marginally higher impedance.

For damping the transformer without changing anything a resistor could be added in parallel with the cap, lower DC than shorting the cap out but more damping, something like the 312 output configuration... 10 cents more expensive than what PRR suggested but no loss in level and no desoldering needed from the original configuration.

JS

 

[gyraf]

Thanks for all your suggestions!

I'll go ahead with this topology then, and I'll experiment with the cross-coupled grid bias idea to see if that gets me even further.

Attached the response with and without the 10uF coupling capacitor. Being Polyester, I can't go to much higher values. And I don't really want to low-cut pre output driver for correction.

Output load is 2:1 into 600 Ohms (thus looking like 2K4 to the CF stage output). I see a MOL of +18dBu@1%, which is pretty much what I want and need.

Output transformer primary inductance is some 21H (calculated from 11Hz@10uF).


Jakob E.