Page 1 of 1

#1 Neat idea for push-pull cathode auto-bias

Posted: Sat Jul 30, 2016 8:12 pm
by Thermionic Idler
I've been doing some research on biasing methods for push pull driver and output stages.

The cathode bias methods seem to be divided into two main camps - camp 1 has a single shared cathode resistor, camp 2 has a resistor for each cathode with a corresponding cathode bypass capacitor to ground (the Leak TL12 is an example of this along with many others).

Each has their pros and cons - the former has the benefit of no capacitors in the signal path - but the valves have to be perfectly matched, and stay that way. With the latter of course, there is less of a requirement for perfect matching - but you have two capacitors, usually electrolytic because of the large values required, in the signal path.

I came across a post from a guy on diyaudio.com who made an interesting point about the latter separate bias method. You don't actually need to have two caps to ground - all you need to do is cross connect a single non-polarised capacitor from one cathode to the other.

However he didn't make any comment about what value would be appropriate, and I wasn't able to find any example schematics of that connection method - they all had the two caps going to ground.

So I reasoned when examining the stage as a current loop - the current in the conventional arrangement passes through two bypass caps connected in series across ground, usually of 100uF value. So if you were to replace that with a single cap, the value would theoretically need to be 50uF.

Swapping out two 100uF caps for a single 50uF (or nearest) sounds like a pretty good idea to me. You can buy Obbligato polyprops in 47uF values for one thing, so then you're no longer pushing the signal through electrolytics.

I ran up a couple of Spice models to check my figures were right and I've linked them below. The context is a direct-coupled driver stage which is why all the voltages are raised some way above ground and the cathode resistors are high values.

Conventional circuit on the right, cross coupled equivalent on the left. As you can see, the frequency sweep result is identical for both.

Anyway, I thought I'd put this out there. Also would be interested if anyone can think of any 'gotchas' with this.

Image

Image

#2 Re: Neat idea for push-pull cathode auto-bias

Posted: Sat Jul 30, 2016 9:57 pm
by Nick
The normal reason to use separate cathode parts is to handle the stage going into class B, not sure just how well the shared cap would deal with that.

#3 Re: Neat idea for push-pull cathode auto-bias

Posted: Sat Jul 30, 2016 11:44 pm
by Mike H
FWIW, when I was dreaming up my EL84 p-p amp, I worked on the OP stage first in LTspice computer simulation.

I started with a single CCS for two cathodes joined. It did not handle large signals at all well, with lots of visible distortion.

I then merely put a large value cap across the CCS to ground and then it could do large output swings with a visibly much cleaner and symetrical looking signal. Bear in mind no negative feedback at all at this point, just driving it with non-inverted and inverted p-p input signals.

My final configuration was separate CCS for each valve cathode, thus keeping each ones bias current the same, and bypassed to ground also with separate electrolytic caps.

The last fly in the ointment was the tendency for the capacitors to charge up in class B operation, where one valve is cut off and the other wants to conduct a current that is greater than the sum of both valves' quiescent biased value. As each CCS uses a depletion mode MOSFET, I addressed this issue by connecting drain to gate of each MOSFET with a Zener diode whose value is 2 - 3 Volts higher than the normal cathode Voltage. So if the capacitors charge up in class B and exceed this threshold the MOSFET's are turned on harder to compensate by delivering the extra current draw. This also simulates in LTspice quite well. The problem witb the caps charging up is it has altered the grid bias when the stage comes back into class A, as well as varying the Va for each o/p valve.

HTH

#4 Re: Neat idea for push-pull cathode auto-bias

Posted: Sun Jul 31, 2016 10:46 am
by Thermionic Idler
Nick wrote:The normal reason to use separate cathode parts is to handle the stage going into class B, not sure just how well the shared cap would deal with that.
Good point, I wasn't sure either so I thought I'd have a quick play with that this morning.

Firstly I redrew the circuit as the way I'd elevated everything by 150V was playing merry hell when the thing started drawing grid current, so I simply referenced the cathodes to -150 instead and dropped B+ to 270.

Image

So what we have here is 20V input to trioded KT66's on 270V B+ at 440Hz, 130ms into the transient.

Here's what's happening with the grid current, the cross connected version is drawing less in the sim…

Image

And here is the current measurement across the two load resistors and also the current draw through the two B+ voltage supplies - which seems to imply less distortion happening on the cross connected version unless I'm missing something?

Image

#5 Re: Neat idea for push-pull cathode auto-bias

Posted: Sun Jul 31, 2016 11:56 am
by pre65
You may well be aware of this, but read note 1 on this data sheet. :)

http://tdsl.duncanamps.com/pdf/kt66.pdf

#6 Re: Neat idea for push-pull cathode auto-bias

Posted: Sun Jul 31, 2016 1:10 pm
by Thermionic Idler
Oh yes, I should include that stopper resistor as that'll have a big influence on the grid current - at the moment the voltage sources have an output impedance of zero. Will do that and rerun the comparison later on.

#7 Re: Neat idea for push-pull cathode auto-bias

Posted: Sun Jul 31, 2016 4:53 pm
by Nick
Well, yes, in the case of the cross connected version, you have a shared cathode resistor at AC so I would expect lower distortion. In the case of no linking cap you don’t have any cancellation between the two sides and as they are not linked through the load they are not connected in any way so nothing like a differential pair.

I would be looking at a linked inductive load, things may will be different in that condition than a purely resistive load.

#8 Re: Neat idea for push-pull cathode auto-bias

Posted: Tue Aug 02, 2016 12:15 pm
by Thermionic Idler
Agreed, I think I'll Spice up the full circuit I have in mind, then that'll be a more realistic model. When I added 10k stoppers to what I had above, the results were much closer although I think the cross coupled one still had the edge.

The holdup at the moment is me working out accurate models for the Lundahl interstage / output transformers.

The motivation behind all this is the presence in my spares box of 4x unused 47uF Obbligato caps...

#9 Re: Neat idea for push-pull cathode auto-bias

Posted: Tue Aug 02, 2016 8:56 pm
by izzy wizzy
A Tabor Clone on DIYAudio did something similar to this http://www.diyaudio.com/forums/tubes-va ... ished.html

cheers,

Stephen

#10 Re: Neat idea for push-pull cathode auto-bias

Posted: Tue Aug 02, 2016 10:40 pm
by Mike H
Thermionic Idler wrote:Agreed, I think I'll Spice up the full circuit I have in mind, then that'll be a more realistic model.
Also needs an actual output transformer, at the mo the plots don't look correct to me, without one, cos 2 separate anode resistors cannot emulate a p-p OPT. I'm assuming it's an output stage?

#11 Re: Neat idea for push-pull cathode auto-bias

Posted: Thu Aug 04, 2016 10:54 pm
by Thermionic Idler
izzy wizzy wrote:A Tabor Clone on DIYAudio did something similar to this http://www.diyaudio.com/forums/tubes-va ... ished.html

cheers,

Stephen
Ah yes - that's along the lines of a CCS circuit I saw in a Tubecad article where he had the caps back to back with a large polarising resistor.
Mike H wrote:
Thermionic Idler wrote:Agreed, I think I'll Spice up the full circuit I have in mind, then that'll be a more realistic model.
Also needs an actual output transformer, at the mo the plots don't look correct to me, without one, cos 2 separate anode resistors cannot emulate a p-p OPT. I'm assuming it's an output stage?
Interestingly enough it's actually for a driver stage, so non-class A performance may well be academic unless it goes into serious overload.

Here's the context - basically, I'm revisiting the balanced 300B push-pull circuit designs of Kevin Carter that Stephen above gave me a heads-up on some time back when I was investigating the Lynn Olson Karna (thanks Stephen). Most have a 6SN7 differential input with chokes on the anodes, DC coupled to 6BX7 drivers, through an interstage to the pair of 300B's.

When comparing the 6BX7 and KT66 data, I noticed that the KT66, when triode connected, has near identical figures to the 6BX7 for plate resistance, amplification factor and transconductance at 250V plate voltage. Just requires a little more current (60mA vs 42mA). The KT66 is also a popular model being manufactured by several companies and is very linear in triode-mode. By comparison, the 6BX7 is not being made any more, MJ asserts that finding low-distortion examples is difficult because it was never intended to be an audio valve, and rarely do two triodes in the same envelope match. (I have an OCD block on only using half of a double triode).

I posted to the K&K audio asylum board asking whether my KT66 driver idea had legs, and Kevin replied saying he'd used an EL34 in a similar manner which had worked very well. So I really think I'm on to something with this.

The original circuit had a shared cathode resistor or CCS on the driver pair though, and the data sheet for the KT66 stated that separate cathode bias resistors were "essential". At first, I thought it shouldn't be an issue if the valves were well matched - but of course in the real world valves do drift over time, so I started weighing up the pros and cons of making my life easier with auto-biasing. That's what eventually led me to the posts about the cross connection idea, and I realised I could use my nice Obbligato caps that I couldn't fit in the OTL's chassis.

Really, for distortion modelling I need to be doing this for the 300B output stage, the original model was built to verify that I was correct about the capacitor value w.r.t. frequency response. I just ended up using the same for testing distortion behaviour.

That all said, I have now modelled the Lundahl interstage so here's the latest circuit:

Image

And here is a comparison of the current through the interstage dummy loads R5 and R6 under overload conditions, sine wave input:

Image

It's quite interesting comparing the voltage/current behaviour at various points in the circuit.

#12 Re: Neat idea for push-pull cathode auto-bias

Posted: Thu Aug 04, 2016 11:09 pm
by Cressy Snr
Looks interesting Dave

#13 Re: Neat idea for push-pull cathode auto-bias

Posted: Fri Aug 05, 2016 7:28 pm
by Mike H
Thermionic Idler wrote: Interestingly enough it's actually for a driver stage, so non-class A performance may well be academic unless it goes into serious overload.
Oh is it? OK then. :D

You could have a centre-tapped choke for it instead then, as exampled in The GEC book of amplifier circuits, BUT then when I added it to an o/p stage (with OPT) it was very unstable.