• 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.

split push-pull output stage

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi all,

Just about to buy the output transformers when I had an idea.... :eek:

take a look at the attached output stage, any comments would be very appreciated. OPT's are a *bit* expensive to buy and try.

Basically the Idea is to split the B+ windings where they are normally connected in the transformer. This then allows separate control of the B+voltage in each half of the output stage.

and if thats then controlled with reference to some fixed voltage and a sensing resistor you can tweak the voltages so that the current balances in each half

Should work in a similar way to SY and Brian Becks modulated B+ in the Unity gain stage thread(s)

maybe.

I'm a bit unsure but if it works then this might also avoid amplifying the noise voltages present with a more normal servo bias system.

Also does bringing the windings out in this way affect the performance of the transformer, when compared to the normal way?

or should I just go buy the standard part?

Andy
 

Attachments

  • balanced.jpg
    balanced.jpg
    39.7 KB · Views: 428
Transformers are basically a summing device, what ever you do to them will work, or it won't. Your idea will work and you will reap some benefit from what you suggest.

Most OPT's will tolerate up to 10ma DC offset and some amplifier designers actually use this to flatten the permeability curve in the core, which flattens the distortion curve for frequencies under 400 Hz, with commercial core materials. Same applies to Nickel and amorphous core but it gets more difficult to implement and vastly more costly. Expect to pay around $130 US each for really first class PP iron with a commercial core.

Bud
 
In essence, you do have a common primary since they're on the same core. DC currents cancel, and from the standpoint of AC, the windings are connected together at a virtual center tap (at the ends of each winding)- the B+ is an AC ground. So this isn't parallel, it really still is push-pull.
 
Well I thought so, the Magnetic field still works the same? so far as the iron is concerned the wires are still connected - this is a question...

So the important thing is now - do I stuff up the transformer if I ask for this as a custom design (pennies permitting) as with all DIY there's the possibility that I might end up joining the windings anyway and doing the normal thing... I would like to try it though.

Incidentally, although I havent contacted Sowter yet I WILL use their trafo's on this amp, something like the U070, but with this winding if I can arrange it. Or specifically unless there are good reasons not to. i.e if I bottle it and go all conventional. Or it costs more than the national debt of a small country

In fairness, this amp has so much sand in it at the moment I'm hoping Virginie Ledoyen turns up ;) (the actress from 'the beach' y'see)

Andy
 
Most if not all transformers have two wire feed for the primary..
The two primary sections (Start of one, Finish of other) are brought out and soldered to one single lead wire for Center-Tap...
If you peel back the outer paper wrap, you will see this...seperate the flying leads and resolder a separate lead wire for the othe primary half...and your good to go..
DC offset does not flaten the permeablity curve.... A gap will do that, so that the inductance is flatter over wider operating range..
Your effective permeabilty, thus inductance, will go down but your flux density headroom will increase a bit...

Chris
 
This is an interesting design idea that I haven’t seen before. Has anyone seen this before? Of course, it is in fact push-pull as SY points out. If I understand your concept correctly, you’re adjusting and balancing the plate currents by slowly varying the plate voltages using the cathode resistor to sense currents (they could be very small resistors). By “slowly” I assume you intended your servo to be an integrator with a feedback cap. The usual servo would change grid voltage to keep cathode current at a set point. I suppose changing plate voltage would work too, although with less sensitivity. I’m not sure of the advantage. Can you explain? Still, it’s nice to see something new ( to me anyway)…
 
A common misconception about transformers is that the core is useful and does work. What is really going on is that the ferrous bit has formed a bounding box and the EMF emitted by the driver wires passing through that box is being reduced from infinite in size to a few cubic CM. This is essentially an antenna event.

The wires actually drive each other, they also charge the ferrous material, but that is both incidental and detrimental to their real purpose, to create a linear mag field and transform energy from one wire to another. The core influences events due to it's permitivity, essentially how fast a mag field layer, one for each core sheave in a stack, can be created within the inner perimeter of that sheave or laminations window.

How the coil is constructed, how much of the "capacitance" couples across a dielectric barrier from one charged wire to an uncharged wire and how much does not, how much of the cubic bounding box volume will support coherent, field event layers, and how fast the core will accept and release those field events are what determines how accurate a transformation occurs. This core release event is called remanence and is the period of time that the field event is being maintained by the ferrous sheave, by that ferrous bit remaining a pseudo permanent magnet and holding on to it's polarization into a bipolar magnetic event.

When you have two driving wires, they will couple to a driven wire and both will influence the eventual "signal" that driven wire exhibits. If one driver has less coherent coupling, either in flux generated B Field or static moment generated E Field / D Field events, you will have a distorted summation of the two driven "signals", but, it will still be a summation.

When you introduce other signal operants to the driver coils, they will have a modifying effect upon the ability of the wires to perform their antenna event and the likely hood of you predicting what you will think of that modification is ZERO. So, just try it.

Be prepared to use signal modifying components of differing materials and construction and take notes and post your findings please, as the rest of us would rather you did this, so we do not have to. Not that we could duplicate your scenario in a thousand years of trying, but your information will provide hints and be welcome for that.

Bud
 
Wheeww. That’s more than I can handle! Maybe you’re making all this too hard. I think the important thing is that once these servos converge, the “DC” current through one primary winding is forced to be exactly the same as the “DC” current through the other primary winding. If the two windings have the same number of turns and balanced DCR values, then the “DC” voltage drop across each winding will be identical too. If they don’t have balanced DCRs, then there will be a slight voltage imbalance, the same as in regularly configured OPTs – but it doesn’t matter since it’s the amp-turns that count – and that is balanced by the servos. Then, it’s up to AC signal (which the servo is insensitive to) to make the usual flux changes needed to pass the signal through the OPT. I don’t see a fundamental problem here.
 
Modulating the plate voltage, by servo, in order to aquire DC balance??? Is that what this is about???
For pentodes this will require a MASSIVE plate change...big enough to distrupt the puch-pull operation and changing the headroom....lack of symmetry ... The rail voltage nees to be locked in and the same for halves of Push-Pull....
For pentodes i have used adjustments on the screen voltage to balance the DC currents.... DC current drift is slow in tubes, therefore you can use a very slow time constant in the servo control loop...
If your dealing with TRIODES, then you would adjust/control the heater supply to control the DC plate current balance... the thermal time constant of the heater wires is plenty longer than the DC current offset drift time...
Personally I would not want an active servo control loop on at all times...even with a time constant way lower than a Hz....
Tubes don't drift around that quickly, if they do then you should consider better tubes... A servo that is in sleep mode need only turn on once per hour for example for a brief moment to re-calibrate the balance..then go back into sleep mode....

Chris
 
cerem,

You're right that this won't work with pentodes - the sensitivity (of plate current to plate voltage) is way too low. And yes, in that case varying the screens is an option. But he's showing triodes here. I don't see this concept as fundamentally flawed vis-à-vis a servo for the grid voltages. The heater voltage changes would work and I’ve seen that done before, but there you might be changing other parameters too much, upsetting symmetry. The servo has a long history in grid driving, but it's not for everyone, either on principle or due to the suspected "sound" it might introduce. I happen to think that with some care, servos can work fine. I would be more tempted to drive grids with a servo, but this idea is an interesting variant worth tossing around.
 
Brian Beck said:
Wheeww. That’s more than I can handle! Maybe you’re making all this too hard. I think the important thing is that once these servos converge, the “DC” current through one primary winding is forced to be exactly the same as the “DC” current through the other primary winding. If the two windings have the same number of turns and balanced DCR values, then the “DC” voltage drop across each winding will be identical too. If they don’t have balanced DCRs, then there will be a slight voltage imbalance, the same as in regularly configured OPTs – but it doesn’t matter since it’s the amp-turns that count – and that is balanced by the servos. Then, it’s up to AC signal (which the servo is insensitive to) to make the usual flux changes needed to pass the signal through the OPT. I don’t see a fundamental problem here.


Indeed, BudP I watched that go SOOO far over my head it was like watching a Shuttle Launch. And I used to think that the electrical engineering part of my course was run by a transormer sadist. Honestly I need to study the reply. I think you answered exactly what I was asking with theory in spades. Honestly it's like being in Dr Whalley's class again... :boggled:

BrianBeck. the IDEA is to avoid noise introduced through the marginally more normal output servo which equally can be arranged to get p-p balance. BUT I figured if you can alter the shape of things in the pre-amp stage it OUGHT to maybe perhaps, possibly scale up. It seemed like an elegant solution to that problem. Driving the scheme here is to me not so easy to me but thats a different story... I Just want to know at this point if its worth persuing. If I spend £400+ on OPT's I'd like to think they were pretty OK. Regardless of experimental configurations.

Cerrem, thats the Idea, But I dont see how it will mess up the P-P operation. I'm going for balance - with triodes (or UL :eek:) the whole thing is to equalise the magenetising currents in each half. Am I being real dumb here? & missing the importance of the voltages? (s'possible, actually more than possible)
I disagree with the thought of contolling heater voltage to alter the characteristics of the tube though - YES I am aware that this may not be completely logical but it seems wrong to me, there are plenty of published documents saying that its BAD when you run Vf low, or high...
OTOH I like the Idea of the servo that sleeps. Worth some investigation. How do you 'fix' the value you arrive at though...

So Thanks all, I think Its worth me seeing if I can get these OPT's in the way that I want, I can try then, even if I end up doing this in the normal way. 'course any results and you'll know first.

Andy
 
Andy,

OK, just to play devil's advocate, what noise are you wanting to reduce? Can you be more specific? I'm not aware of noise generally being a problem with a well-implemented grid-driving servo. Bias drift if there is a transition to Class B, is a problem, but there are solutions for that too (in either implementation).
 
Brian

before I need to go sleep, I want to say that "plan B" IS the bias servo's. I have no problem with those at all, Maxim have published some info on very low noise regulators which ought to make it pretty much equivalent to battery bias. (6nV * Mu in a couple hundred volts?)

Basically I want to try this, If it works great, If not then servo's re the way. I will need a maintenance free output stage. I dont have the kind of patience to play with bias settings regularly, I'd forget and cook something

Andy
 
With servos, I will tend to add an additional low-pass filter after the silicon, at perhaps an order magnitude higher frequency than the servo corner. This gets rid of any noise and tiny opamp signal outputs above, say, 10Hz. You don't want to get the two poles too close, because this will raise the Q of the resultant second-order filter, but 10:1 or even more is probably OK. I just don’t think that noise needs to be a problem with either type of servo. I’m still seeing only disadvantages to the implementation (but not fundamental flaws) of the B+ servo, relative to standard grid-driving servos. A B+ servo is possible, but I don’t know why you’d bother when the grid servo would be so much easier. I’m open to advantages of the B+ servo that I haven’t thought of yet, though, and that’s why asked. And I would not want to discourage experimentation here. Let us know how it works.

Have some sleep, Andy - you've earned it!
 
Brian this was just one of those ideas. I want to give myself some flexibilty. I'm a bit noise paranoid, the idea of amplifying grid noise is... well, enough to have me looking at this. Like so many things in audio I just want to give it a go :smash:

I'll admit this is not a properly thought through thing, If it had been I'd be reporting how well it worked, or not reporting how badly it sucked. For instance How far the voltages need to be off to compensate for tube mismatch - got no idea, Do you have cathode bias or a fixed negative voltage - No idea!

Also got some of these "ideas" for class AB in Bias servos, but you can do that with a normal transformer :)

Cheers

Andy
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.