2

enter image description here

The figure shows the output stage of a vacuum-tube audio amplifier. While I was reading through this schematic, I noticed there is a some kind of feedback associated with R24. However, this feedback seems positive feedback to me.

The right side of R24 is connected to 'BLU' node on the output transformer. The 'BLU' node is also connected to the anode of push-pull output driver tube that is taking the flipped voltage polarity (due to differential nature) and outputs inverted signal. Therefore, I think there are two inverted stages and hence it should be positive feedback.

How does this feedback associated with R24 (or R27) work in this amplifier?

Here is the link for the full schematic (source):

enter image description here

Null
  • 7,448
  • 17
  • 36
  • 48
Emm386
  • 577
  • 2
  • 14
  • The negative feedback loop is probably outside of the shown part of this schematic. Show the full schematic. – Jens Aug 09 '22 at 00:30
  • @Jens I've just updated the post with a link showing the full schematic – Emm386 Aug 10 '22 at 01:11
  • Ah, thats better. As I supposed, the source of the negative feedback path is the small transformer winding BLK/WHT-WHT and this voltage is fed back to the cathode of the input stage. The gain divider looks like 2.7 kohm / 68 ohm, 32 dB. – Jens Aug 10 '22 at 01:35
  • @Jens, I appreciate your comment. I have two questions Q1: I think what you are referring is the global feedback network. Could you explain how this global feedback network is related to the local feedback above? Q2: 2.7kohm/68ohm is the closed loop gain of the global feedback. Again, I don't see how this is related to the local feedback above. – Emm386 Aug 10 '22 at 07:19
  • Yes, that is an interesting local positive feedback. It gives V3 more headroom, more dynamic range and a higher load impedance. 12BH7 already has good linearity, but this might still be improved by this circuit - I'm not sure about. – Jens Aug 10 '22 at 19:36

1 Answers1

2

What's hard to tell from the schematic, is the phase and ratio of the transformer windings.

The YEL and ORN windings are all common to the output transformer (BLU). The phasing, I think, is consistent: a dot should be placed at the top of each winding.

This amplifier is famous for using cathode feedback, with a specialized bifilar wound transformer.

Bifilar means, the two windings are made at the same time, in the same place, laid side by side; this gives minimal leakage inductance, between the windings: low enough that the output doesn't turn into a power oscillator at a few 100 kHz(!).

Since the windings are equal, every volt the plate pulls down by, is a volt pushed up at the cathode. This massively degenerates (i.e. applies local negative feedback) the voltage gain of the output stage (it's about 1), necessitating very high gain from the driver stage(s).

So already, we see that, although the identified loop is a positive feedback loop, the loop gain is in fact very small, and it's at least plausible that it's stable.

Note that +FB must overcome the existing gain first: for small amounts, gain and distortion rises, and this can even be useful by itself to secure greater -FB elsewhere; eventually, gain reaches infinity at some point, and a hysteresis loop begins to open. This is the (mono-)stability limit, and if there are reactive elements in the loop (namely C8/C9), oscillation ensues. If not, bistability results (the level latches into one side or the other of the hysteresis loop). So it must definitely be used sparingly, in something linear like an audio amplifier.

In this case, note that R24/R27 acts as a voltage divider, from output plates, to V3 plates, through V3 plate resistance to GND (or, to each other, in part). Probably, the total driver+output stage voltage gain is still less than in a more conventional design, hence the additional gain+phase splitter stage required (V2), plus a high gain front-end (V1). These are all direct coupled to minimize low-frequency phase shift (a big problem with a conventional AC-coupled three-stage amplifier, like a "Williamson" design).

Note also the 12AZ7 (V4) is bootstrapped: the cathode load resistor is split in two, and a large coupling capacitor feeds the tap from the output cathodes. This way, it only has to drive the output grids, as needed; just the little bit of loading from the resistors (R29, R30) is saved. They're even supplied from a separate winding (only independent, I think, because of the different DC voltage applied to it?), bootstrapping the plates as well.

I don't know these amplifiers in great detail, or how many they made with this design, to which the same analysis applies; likely you can find more authoritative resources from audiophile enthusiasts/historians/restorers. (Any watching: you are welcome to correct/expand this post!)

Tim Williams
  • 22,874
  • 1
  • 20
  • 71
  • Tim, really appreciate such detailed comments. I have a question: I thought the impedance looking into the transformer is large as it transforms the impedance of 16Ω or 8Ω to much larger impedance by turn ratio of the transformer. Therefore, I've been thinking that this push-pull stage gain is fairly large. If this is true, then the positive FB loop gain is going to be large too. Another thing I noticed is that if suppressor grid of V8 or V9 can also work as a gain tap, then it forms negative feedback with the R24/R27. Do you have any thoughts about this idea? – Emm386 Aug 10 '22 at 20:37
  • 1
    @Emm386 You aren't understanding the circuit: a tube only sees voltages with respect to its electrodes. Specifically the cathode. Consider the cathode voltage, and then consider the plate and screen voltages with respect to it. There are windings connected to all of them. (I think you meant screen, not suppressor; KT88s are "Kinkless Tetrodes" and don't even have one.) – Tim Williams Aug 10 '22 at 20:45
  • Hm... yeah I guess I am not understanding this output drive stages properly. It is interesting structure tho, as the same AC voltage gets fed to both control grid and cathode of KT 88 (why?). I guess what you are saying is that the cathode of KT 88 is coupled to the anode plate of Kt 88 via transformer. can I ask you why one would do this kind of structure? what would be the benefits out of doing this so? – Emm386 Aug 10 '22 at 22:01
  • 1
    The cathode feedback provides negative feedback ("degeneration"). Reducing the output stage from, say 10-20 gain, down to about 1. Reducing its distortion by as much, and its output resistance by even more. – Tim Williams Aug 10 '22 at 22:34
  • Hi Tim, thanks again for the quick response. The follow up question I have is about this degeneration network. Isn't having a resistor in the cathode good enough for degeneration purpose? this amplifier has a transformer-looking structure (and I guess this YEL-YEL/WHT is magnetically coupled to M-216A output transformer?). On top of this, the cathode takes the same AC input voltage (scaled amplitude) as the grid one. Is there any circuit/theory that I can read to understand this structure better? – Emm386 Aug 11 '22 at 02:10
  • 1
    @Emm386 Good enough -- for what? A typical degeneration resistor takes up, whatever, 5, 10% of the drive voltage, and is based on load current. Note that current feedback raises the output impedance even further, and it's high to begin with, with tetrodes/pentodes. Low Zo is desirable for driving loudspeakers. Shunt feedback can be used to reduce Zo, and can apply more NFB, but requires more drive power as well. Whereas, feeding back half the output voltage, maybe 10X the grid voltage is applied as NFB -- far more than you could ever hope from a cathode resistor alone. – Tim Williams Aug 11 '22 at 02:46
  • so in summary, I can identify at least three feedback structure associated with the output driver + R24/R27. (1): C10 and C11 serve as bootstrapping to further increase the input impedance looking into the grid of KT88. (2) YEL/YEL-WHT transformer is coupled to the BLU/BLU-WHT magnetically and this forms a so-called unity-coupled feedback ( a type of cathode feedback). (3) R24/R27 serve as positive feedback to have more gain for V3 stage. am I missing something here? – Emm386 Aug 11 '22 at 05:51
  • 1
    And the orange winding, again bootstrapping. That covers it, yes. – Tim Williams Aug 11 '22 at 14:31
  • Tim, really appreciate all your comment. I learned lots of new things including cathode feedback idea and usage of bootstrapping in conjunction with this feedback. But I have one last question: What is the purpose of the orange winding? you mentioned that this is for plate bootstrapping, but I just don't understand the purpose of plate bootstrapping (and 12AZ7 is not even output driver). Could you explain me the purpose of plate bootstrapping using another coupled transformer? – Emm386 Aug 11 '22 at 18:59
  • 1
    Again, look at the voltage drop (plate to cathode). What would it be without the winding? Note the effect of plate resistance on the cathode follower: what would its gain be, without? And what would its supply voltage have to be, to cover the required voltage swing? – Tim Williams Aug 11 '22 at 20:03
  • If I understand correctly, the higher the plate resistance is, the lower the gain is. To cover the required voltage swing, the supply voltage must be higher. After reading through the schematic once again, I think the plate bootstrapping of V4 reduces the plate resistance. This will reduce the voltage swing drop at the cathode of V4. Then I have two questions: ORN - ORN/WHT center tap takes unregulated voltage from the center tap of the power supply transformer. isn't this supposed to be GND? how come this provides 205 V? question two: can't we just connect higher voltage to V4 anode? – Emm386 Aug 12 '22 at 00:35
  • 1
    Actually, after some consideration, I think the ORN-ORN/WHT transformer is trying to keep the voltage between anode and cathode constant over grid signal swing. It is bootstrapping, but not positive feedback. I think this is bootstrapping you often see from switched S/H circuit to minimize the distortion. In other words, the anode voltage goes up as grid voltage goes up. similarly, anode voltage goes down as grid voltage goes down. Therefore, the linearity of V4 stage will be improved by a negative feedback. But I would also love to hear your opinion. – Emm386 Aug 12 '22 at 03:36