How do I add EFB to a UL connected output stage?

[kward]

The example schematic below shows a "normally" wired EFB regulator for a PP output stages for a stereo chassis configuration. The 380V from the point labeled A comes from the center taps of the output transformers. The 326V from the point labeled D feeds the screens of the output tubes, and the -42V from the point labeled E, is the bias voltage for the output stage.



But when the output stage is connected in Ultra Linear mode, there is no need to supply the screen voltage from the EFB regulator since it's obtained from the OPT's UL taps. How would the wiring of the EFB regulator need to change to support a UL connected output stage?

I can see how possibly either of these two answers below would be correct:

1. Keep everything the same, just don't connect the screens to point D. (feed the screens from the UL taps on the output transformers instead).
2. Since the mosfet is now not used, remove it from the circuit, and feed the 1M voltage sense resistor that is connected the base of the inverter transistor directly from point A. Of course that will slightly change the other resistor values in the EFB circuit since there would be more current now flowing through that 1M resistor than before, but that's okay (if this is the right approach) since I can redesign the rest of the EFB circuit to support the new voltages/currents.

Or is there a better method that I am not considering? Or perhaps is this not advised?

 

[thorpej]

The point of the original EFB circuit was to pin the bias point to the screen voltage so the bias gets hotter as the screen voltage sags. With the regulated screen voltage, it works the same, but of course never has to self-adjust because the screens are regulated.

Because the UL connected output stage has a constantly-fluctuating screen voltage (that is derived from the plate voltage + the AC feedback), seems like you'd pin it to the plate voltage so that the bias gets hotter when that sags instead.

Although, on the face if it, seems like you have a positive feedback loop in that situation -- B+ sags, bias gets hotter, tubes conduct more, B+ sags some more, and so on. It might be worth the experiment, though.

 

[Shadowdog]

Isn't the Dynaco SCA/ST-35 UL , so maybe something from Dave's EFB used on them could be incorporated.

 

[thorpej]

Isn't the Dynaco SCA/ST-35 UL , so maybe something from Dave's EFB used on them could be incorporated.

Indeed, I'd forgotten about that, and he does in fact take the LM337 voltage reference off of a voltage divider on the B+ rail. So that seems like the ticket, Kevin.

Eliminate the screen supply MOSFET and take the voltage sense directly off B+ (rather than the source of the now-absent MOSFET), adjust the rest of the bias circuit as needed for the desired bias voltage.

 

[kward]

Isn't the SCA-35 using the cathode EFB version? Whereas this is the "other" version of EFB. Maybe the solution is the same. I'll go read that article again on the ST/SCA-35.

 

[gadget73]

There are two versions of it as I understand. One is the grid bias, which can either run on a negative DC supply and feed the grids or it can replace a cathode resistor. Same net effect though.

The grid supply is what you have pictured. I think you'd reference it to the output transformer center tap since there is no screen supply.

 

[thorpej]

Exactly right, the cathode vs grid style is just about dealing with the specifics of how the circuit was originally built. But the principle of pinning it to a reference voltage is the same.

 

[thorpej]

Isn't the SCA-35 using the cathode EFB version? Whereas this is the "other" version of EFB. Maybe the solution is the same. I'll go read that article again on the ST/SCA-35.

This "other" version is really 2 different things, though, right? It's the "enhanced fixed bias, grid flavor" that pins the negative bias voltage to the screen voltage (in an inverse fashion -- as screen voltage drops, bias voltage becomes less negative). This pinning the bias (whichever flavor, grid or cathode) to a particular important reference voltage is the "enhanced" part. In theory, that could be deployed in a circuit *without* the other portion, which is a regulated screen supply.

As I recall, the origin of this screen regulator was Dave setting out to improve a a particular Fisher (or was it a Scott? I don't recall...) model that used a power resistor to provide the voltage drop needed for the screens, but of course that sags and leads to more distortion as the power output increases.

 

[Shadowdog]

I think the Dynaco's mentioned were his first EFB mod & were very similar as the circuits are the same around the output tubes where the mod was used for the bias.

 

[kward]

Yes, I agree with both of you now that the right solution is to tap the voltage that feeds the sense resistor from the plate in this scenario.

The other part of my question (that I didn't state very clearly in my opening post) is about the rate of sag, or said differently, the rate at which the grid bias voltage follows the voltage sag at the (plate or screen). I know in a pentode connected output stage, you want that sag relationship to be linear. Meaning, if the screen voltage drops by 10% of its quiescent value, you want the bias voltage to increase by 10% of the value you set to create that quiescent condition.

I'm wondering in a UL connected output stage if this linear relationship still holds. I think the answer is yes, but I don't know how to prove it with math or with theory. I had forgotten about the ST-35 using (the cathode version of) EFB in a UL configuration. So I think the example of the ST-35 answers this question, since it does offer a linear following relationship, even in UL model. Therefore I think the answer is yes, the relationship is linear between plate/screen sag and bias voltage change, regardless of Pentode or UL.

(and for that matter, if you had a triode output stage, the same answer probably still applies).

 

[Shadowdog]

That's got to be it right! Dave is brilliant with his circuits I must say!

 

[gadget73]

I'm thinking the DC portion of the screen supply in a UL amp isn't quite proportional to the DC portion of the plate supply. Assuming a 50% tap point and a 10 volt drop at the plate across the transformer, you're going to get a 5 volt drop at the screen. Basically the transformer coil acts like a big wirewound resistor in DC terms.

Add in the AC though and that goes out the window to some degree so I don't know how much or if it truly matters. Considering there is no DC-only point to tap off the screens in a UL amp I guess it doesn't matter all that much with regards to how the EFB works. If you had a dual wound transformer with a dedicated screen winding that could be fed from any DC point your heart desired it would be a different story.

 

[kward]

Yeah, I get your logic there. Well, I'm gonna try it. Proof will be in the pudding, so to speak, comparing distortion output at max power before vs. after.

 

[gadget73]

I wired up a cathode version in the Madison-Fielding amp I was tinkering with. I forget the specifics without having my notes in front of me but it made a considerable improvement in power and distortion, especially when comparing single channel vs both channel. It more or less tracks Dave's results in the original Tronola article on the SCA-35. The M-F is a pentode output vs UL but the amp is overall pretty similar. It also bottoms out the distortion at an incredibly low 23 ma per pair of tubes. The only down side, if you want to call it that, is distortion comes on a lot more like a solid state amp than a typical cathode bias amp. Its clean right to the point where it isn't rather than gradually getting worse over a wider range of power.

 

[thorpej]

So, what are you building??

 

[kward]

I've got two projects going simultaneously right now. The first is my attempt at another bench power supply that I have discussed briefly in other threads. But this question is concerning an update to the KT120 Mk II amp I built last year--adding an EFB output stage regulator to the UL connected output stage. This isn't my amp any longer but the new owner asked if I could retrofit that change in.

 

[kward]

It resembles more of an Eico HF-60 with KT120 tubes, but with Dynaco output iron. I coined the Mk II designation only to distinguish it from my first attempt several years before that. I didn't really realize it bore resemblance to the HF-60 until after I finished it.

 

[kward]

The new owner doesn't really focus on the details of how tube amps work. Our discussion was more along the lines of what can be done to "improve" the amp: sound quality, performance, etc. We talked through some changes in the power supply, how feedback amps work in general, and about output stages and their distortion causing characteristics as well as methods of reducing distortion. EFB is extremely attractive as a solution there. It's a no-brainer so to speak to try it, because it is inexpensive (~$10 for parts), all of the parts run cool, and the solution is not invasive (can be easily reverted if the owner doesn't like it).

 

[thorpej]

I've got two projects going simultaneously right now. The first is my attempt at another bench power supply that I have discussed briefly in other threads. But this question is concerning an update to the KT120 Mk II amp I built last year--adding an EFB output stage regulator to the UL connected output stage. This isn't my amp any longer but the new owner asked if I could retrofit that change in.

I'm looking forward to reading about Bench Supply Mk 2! I finally have a workbench in my garage again, so I'm finally able to make progress on may build of Bench Supply Mk 1.

 

[kward]

Here's what I think the EFB circuit looks like. Mosfet wasn't needed so it's been removed. I have available -146V DC source, 475V B+ (from center taps), and I need -63V bias output which I will feed to the bias trimmer pots. So this should do it. I'm using a (perhaps odd size) 680K sense resistor so that I can use standard E24 resistor values everywhere else.

The only real problem I see is the amp was built with four individual bias adjust circuits, one for each tube. It will be a little harder to adjust for lowest distortion because it will mean adjusting bias in small increments for each tube in the Push Pull pair (without disturbing DC balance between the two tubes) until I find the bias settings that produce lowest distortion. I think this can be done, it's just going to take more patience to hone in on the bias settings per tube that will be needed.