Adding NFB to power tube cathode?

[jeff]

 I have a little SE I want to try adding NFB to. It's basically gain gain gain cathode follower tone vol power tube. If I add the NFB to the gain stage's cathode it'll be before the volume and tone so adjusting the volume and tone will change the NFB.

Do I do something like this?

What would be good values to try?

EDIT by sluckey... I cropped your pic.

[sluckey]

The easiest way to add NFB (AKA degenerative feedback) to your power tube cathode is simply eliminate the cathode bypass cap. Nothing else to do.

If you want to do it your way, don't connect the grid circuit to the feedback resistors. Just connect the tone and grid circuit to ground as usual. Look at the blackface champ circuit to get some resistor values to use as a starting place for the feedback divider. Don't forget, you 'may' have to switch polarity of the feedback.

[kagliostro]

As to decide the best resistors values

you can use trimmers

and than fixed resistors

Kagliostro

[jeff]

"If you want to do it your way, don't connect the grid circuit to the feedback resistors. Just connect the tone and grid circuit to ground as usual."

Thanks good catch.

"Don't forget, you 'may' have to switch polarity of the feedback."

Oh yeah thanks again.

"Look at the blackface champ circuit to get some resistor values to use as a starting place for the feedback divider.  "

In the champ the NFB goes to the cathode of a gain stage so I imagine the signal going to the cathode of a power tube has to be much bigger and the Champ values won't give the same result.



 

[jjasilli]

Another alternative:  note the cap between signal grid pin & ground pin on the VibroChamp AB764: http://www.guitarscanada.com/amp-tech-section/29808-fender_vibrochamp_ab764-pdf.html

[PRR]

The voltage-gain from power tube grid-cathode to speaker is very low, often not even unity. You can't get a lot of negative voltage feedback (controls speaker) any simple way. The attached image is usually the best you can do.

As Sluckey says: unbypassed cathode resistor gives simple NFB, but _current_ feedback. This leaves your speaker cone "less controlled". It lowers THD in this stage, but you need more drive voltage. With that large-loss tonestack in front, the stage before the tonestack has to make HUGE output (and THD) to push enough signal through TS to EL84 and get all the EL84's power.

Obviously both experiments are easy to try.

[VMS]

This page has a test and results of the circuit that PRR suggested:

http://livinginthepast-audioweb.co.uk/feedbackwindings.php

[jeff]

But does this mean you'll have DC going through your speaker/secondary? Is that OK?

I'm not sure about this and maybe I'm way off but I'll ask in the name of learning.

I think I understand how a speaker works. An AC signal through the coil makes an electromagnet that attracts/repells the speaker to the speakers magnet, and that moves the cone.

If there is a DC through the voice coil are you making an electromagnet that pumps your speaker out a little. If it is a little pre pumped them the will travel be "biased"? Maybe that's not the right term but, in other words, the travel of the cone will not be equal on the positive swing compared to the nevative swing? Is there a chance of damaging the speaker if it bottoms out?

The tube is biased at 44mA, so 44mA is going through the 8ohm speaker coil and the resistance of the OT. I'm not sure how it is divided but there is a voltage on the speaker, right? Is that bad?

[HotBluePlates]

But does this mean you'll have DC going through your ... speaker?

D.c. voltage present? Yes. D.C. current? Almost none. Read on.

But does this mean you'll have DC going through your ... secondary?

Yes.

Measure the resistance of your speaker (unconnected to the amp), and the secondary. The speaker might be 4-6 ohms (for an 8 ohm speaker), the secondary close to as low as your meter will read (probably ~0.2 ohms for a good digital meter, without a zero offset).

That means the connection to the secondary looks like ground for all practical purposes. *If* there is any d.c. present, the secondary itself is so low-resistance that almost all current that does flow will flow through the secondary itself rather than through the speaker.

Further, since the secondary looks essentially like ground, the feedback resistor to the tube cathode forms a voltage divider with the secondary resistance. That divides down the voltage present at the cathode, leaving almost no d.c. across the secondary.

It's safe to do (and has been done for many decades in many, many amps). You should choose a feedback resistor value that's quite a few times larger than the cathode resistor in question. If you furhter stack the cathode resistor on to of the shunt feedback resistor, as in PRR's example, you will reduce the d.c. present at the secondary to probably below the limits of measurability with your equipment.

The lack of sufficient voltage gain from the output tube cathode to the speaker will probably make the feedback ineffective unless you move back to the last stage with actual voltage gain (likely the one before the cathode follower).

[PRR]

DC cathode current will split between speaker and OT secondary in the ratio of DC resistances. DC resistance of "8 ohm" speaker is 6 ohms. For decent efficiency, DC resistance of secondary must be much less than 8 ohm load. Perhaps 1 ohm, maybe less. Therefore the current splits 1/7th in speaker and 6/7th in OT secondary. Maybe 7mA.

Peak audio current in speaker is about 1 Amp, 1,000mA. A 500mA DC current will cause heat and push the cone off-center enough to matter. A 7mA current is "nothing"; causes less offset than typical factory tolerance.

You have DC current in the primary. Many more turns. The SE OT is designed for plate current in primary. Another little bit in secondary does no harm.

In push-pull form this was once a popular mod for Dynaco power amps. The improvement is small but it is free. No resistors necessary: there is so little forward gain that you don't want _any_ return path loss.

> move back to the last stage with actual voltage gain

Except he's adopted the notion of putting a tone-stack in front of the final tube. In most audio work, this would be "poor engineering". It has been popular for guitar amps where bad is sometimes good. But I'm not sure it lends itself to other good ideas like local NFB from the speaker.

[HotBluePlates]

Agreed. The tone circuit may not work as expected.

But I figured I'd let him discover for himself what it sounds like, and that the negative is probably better applied after a tone stack (or around a much different type of tone circuit).

[PRR]

> I figured I'd let him discover for himself

I read that he probably already understood the issue?

> If I add the NFB to the gain stage's cathode it'll be before the volume and tone so adjusting the volume and tone will change the NFB.

[jeff]

 I just noticed in the schematic that the cathode is connected to a 16ohm tap and the speaker to a 8. I also realised the OT I'm using has a 16 ohm tap. I suppose by using tha 16 ohm tap I'll get a little more NFB.

 I'll simply wire the 16 ohm tap wire to a switch, and select to ground the cathdoe resistor or connect it to the tap.

 I like testing things out with an A/B switch. Hearing it one way then rewiring it to hear it the other way just doesn't work for me. I'm never sure if it sounds better/different or if my mind makes me think I hear something better/different because I am expecting to hear it and I think I know what I'm supposed to be hearing.

"...putting a tone-stack in front of the final tube. In most audio work, this would be "poor engineering". "

Would this be considered "poor engineering" because of the loss in the tone stack means you don't get the maximum gain to the power tube, or is there another reason?

I'm gonna try this out first and get back to ya
Thanks for the help guys

[HotBluePlates]

I suppose by using tha 16 ohm tap I'll get a little more NFB.

Under normal circumstances (feedback from speaker to phase inverter, or stage before), then you're right.

Given a typical EL84 amp, your bias (voltage across the cathode resistor) can't be but 8-12v. Your amp's output power is ~4-5w. Pretend the speaker is a resistor. 5w*16 ohms = 80. SqRt 80 = 8.94v on the 16 ohm tap, at full power output. On the 8 ohm tap, it's only 6.32v.

To get your amp to make full output power, you need an voltage applied to the grid approaching the value of the bias (say, 9v). The output tube has, when this voltage is applied, a bigger voltage and current swing, as measured at the plate. The OT matches the tube plate to the speaker load, but does this by stepping down the voltage swing, and stepping up the current. We've already calculated how much a.c. voltage needs to be applied to a 16 and 8 ohm resistor to obtain the full output power of your amp.

If you think in terms of voltage only, from the EL84 grid to the speaker, 8-12vac (whatever matches your EL84 bias) has to be applied to arrive at 8.94vac (5w, 16 ohms) or 6.32v (5w, 8 ohms). If we were calculating gain, this is either almost no voltage gain, or an attenuation.

A feedback loop only works if there is excess gain to throw away. For example, say you had a gain of 40 (8v in, 320v out), and you wanted to use feedback and could accept a closed-loop gain of 20 (8v in, 160v out). The gain thrown away in the feedback is actually being used to reduce distortion or (if derived and applied the right way) to control speaker flap, reduce output impedance, increase input impedance, etc.

Since there is no-gain or even a loss from EL84 cathode to speaker, and since you will need a voltage divider to isolate the tube cathode from the speaker, you will get no useful feedback. For this injection point only, it's not enough (16 ohm tap) or really not enough (8 ohm tap). The same problem exists at the cathode follower output (voltage loss from this point to speaker, due to tone circuit then OT) and the cathode follower input (voltage loss due to cathode follower, tone circuit and OT).

Just trying to help you see what you might have already realized: that you have to have excess gain available to "throw away" in a feedback loop, in order for the NFB to actually do any work. The same probelm potentially arises when there's marginal gain thrown away in a loop, and we overdrive the input to the loop... eventually the input signal overpowers the amount of feedback available, and the feedback abruptly stops working.

[jeff]

I get what you're saying about not having as much of a signal.

 NFB has always confused me. I've always took it for granted that it works, but I need to read up more to full understand exactly what's going.

The power tube is creating a singal that(some poriton) is used to modify the signal that is creating it.

It's kinda seems to me like jumping into a time machine to prevent yourself from jumping into the time machine.

[HotBluePlates]

The big thing to realize is the process is generally described as though the feedback didn't exist, then follow the signal around the loop and see what effect it will have.

In my opinion, Gerald Weber reached a wrong conclusion as a result of this, explaining that cathode-biased output tubes should have "more compression" somehow. He described the effect of the local negative feedback due to no bypass cap, and his description starts with a signal, describes increased current flow through the cathode resistor, then a counteracting effect to reduce tube current.

To help us understand the how/why, we think about it step-by-step. In actual practice, it happens nearly instantaneously (in the absence of some other circuit feature to slow the process). So rather than compression (which in recording practice has an audible effect), it's more like a governor. You slam the pedal down and just can't get more than a certain amount.

Regardless, the important thing to know is "open loop vs closed loop". Open loop is the circuit without feedback. "Closed loop" are the conditions with feedback applied. In order to get negative voltage feedback, there must be more voltage gain in the open loop condition than in the closed loop condition. The difference in gain is the amount of feedback (gain reduction) applied. If you run out of available gain, such as a stage within the loop runs into cutoff or saturation, your loop no longer has gain for the feedback to function, and the circuit will revert to an open-loop clipping condition.

Anyway, these are the things we're thinking about regarding the point of application of the feedback.

If you look up Pete Millet's site, there's a series of books by Norman Crowhurst called "Basic Audio" that describes a lot of aspects of audio electronics in simple terms. There is another book on that site by Crowhurst that includes a discussion of feedback that I can barely follow. Regardless, even the simple version is technically accurate and helps speed understanding.

[PRR]

> you will need a voltage divider to isolate the tube cathode from the speaker

No divider necessary. Cathode (via bias R-C) to OT secondary.

Same or different tap as speaker; use ears.

> it's not enough (16 ohm tap) or really not enough (8 ohm tap).

Taking the "speaker is nominal resistor" assumption, even 1:1 feedback around a unity-gain stage is "some" feedback. Perhaps hardly-any. However many of the big Fenders work this way. 12AX7 long-tail has gain of 25, p-p 6L6 to low-Z tap has gain of 0.4, total gain is 10. NFB divider may be 820+100, loss of 9. Any high-NFB fan will say "none". Any Fender lover will note the amp sounds different than a no-NFB amp.

But the speaker is not a resistor. 8 ohms will be 40 or 80 ohms at bass resonance. The pentode output stage's gain will want to rise sharply. No NFB, the speaker booms, or slaps (or is selected to not be offensive). With "unity" nominal NFB and 5X-10X load and gain, now we have real NFB and real control of speaker bass resonance.

[jeff]

 Well I tried it out, and it works. I like the way it sounds.
 
 I can't really tell you for sure what the difference is between the way it is now and the way it was because I just rewired it without using an A/B switch, but I like it. I'll have to drill a hole to put an A/B switch(ARGGGG how come pots and switches don't mount in the same hole???) Then I can describe better what's actually different. I think I can tell the difference but want to A/B it so I'm not comparing, from the memory. Not - knowing what I did, thinking what I'm "supposed" to hear, and how that differs from what I actually remember, influence my result.

 Thanks for the help. I like how you guys don't just say this is what to do, do it! but give good explainations of what's going on and why/how this works. I'm more interested in that so thank you for your time.
          Jeff

[HotBluePlates]

... ARGGGG how come pots and switches don't mount in the same hole??? ...

Push-pull pot, maybe? Doesn't help if you don't have one on hand, though...

As for explanations, that's why I like it here too! I learn something every time PRR posts (and a lot of other folks who might think they don't have much to add).

Besides, if you actually learn from the explanations, you don't have to ask the same question multiple times for different amps.

[jeff]

Push pull pot. Brilliant!

[HotBluePlates]

Push pull pot. Brilliant!

Time for a Guinness!! 

Spoon! Brilliant!!! 

[kilowhat]

In push-pull form this was once a popular mod for Dynaco power amps. The improvement is small but it is free. No resistors necessary: there is so little forward gain that you don't want _any_ return path loss.

Hi guys,
Know this is a necro but found the thread interesting and since OP is long done with it and named it broadly, for future reference i figured i could hijack (i know, two capital crimes in a single post, the nerve!) and expand the topic to push-pull stages as well.

Im having trouble visualizing a signal injected in a PP cathode as other than common mode. Doesnt seem as elegant, and only half as effective, but I guess with individual bias resistors you potentially could inject in on just the one side, does anyone know if this is what PRR was referring to? Tried googling clues to this particular dynaco mod, but no dice.

I really appreciate the beauty in the speaker damping side of modest NFB, and would like to keep it as local as possible (split load pi making it hard). But i get almost equally erect on the thought of limiting bias shift(when moving into the B side of AB) with a common resistor.

Many thank /kw

[PRR]

Think OT with center-tapped secondary. One end to each P-P cathode, DC cathode current from the CT.

This comes up in some Hammond (Leslie?) and tape-recorder amplifiers.

[pdf64]

Compare the power tube cathodes of the Leslie 147 http://www.captain-foldback.com/Leslie_sub/Leslie_schematics/147.GIF to the 122 http://www.captain-foldback.com/Leslie_sub/Leslie_schematics/122.GIF

[kilowhat]

Wow, thanks guys, how eloquent! And so obvious(!) after youve been led atop the giants shoulder.
I take it the standing DC current in the secondary from one cathode is offset by the out of phase other? (like in a PP primary: counteracting the flux, avoiding core saturation and negating need for an air gap).

If so, am i safe assuming on a PP OT with 16r secondary, a 4r tap (half the number of turns) would suffice as secondary CT?? Or 2r tap of a 8r sec. The only other hiccups i can think of would be that the 8r-16r side might be wound with skinnier wire (spec'd for lower amperage). But probably a very non-issue since cathode current, because of OT ratio, should be at least one order of magnitude less than the current generated on the secondary by the primary, right?
But what about HT isolation? If a tube is welded shut shorting closed (do they ever?) couldnt things get potentially ugly with B+ lingering outside its 'galvanic cage'?

Also, perhaps a pair of SE OTs could work, with primaries as well as secondaries in series, then their point of connection used as CT for prim respectively sec? Btw, just for giggles, anyone know if an SE OT is normally air gapped on the secondary as well? (i know even less about tranny construction than the dark arts of tube wizardry!)

Sorry for the cerebral vomit. /kw

[kilowhat]

Also also, cant straighten out in my head if a virtual ground established by a non polar cap straight between the cathodes, of say the 6550s in that Leslie 147 schematic (great example, thanks pdf!), would mess up any of the NFB magic? Because if not, perhaps we could rid that electrolytic bypasser for a smaller film cap(!) across cathodes. And im probably imagining but about every time i try i could swear i hear a difference tween film and lytic caps in a cathode.
Or wait, cant remember if that virtual ground business only worked properly with individual bias resistors.. Anyone?

As usual, floating from one state of confusion, into the next /kw

[PRR]

You are clearly smart enough to work this through and tell us.

HOW did I forget the McIntosh unity-coupled output?? Cathode feedback up the wazoo.

[tubeswell]

Another possibility

[92Volts]

Another possibility

I did something similar to that, grounding the driver stage directly through the speaker tap of the OPT: https://el34world.com/Forum/index.php?topic=22711.msg246624#msg246624

It's sounded and worked fine for me. Note, it's a triode-mode output with little gain (even from the combination of driver plus the power tube) so this isn't as much NFB as it looks like, about 3dB voltage or 6dB power.

Because the DC resistance of the OPT windings is low, little DC current flows through the speaker. Because it's a single-ended OPT, it would tolerate quite a bit of DC current, but only passes 3-5ma from the driver stage in this case.

FWIW, I haven't yet fried a 12AX7 with the 500v+ B+ voltage. As I hoped, the 250v voltage on the plates has either prevented trouble, or 12AX7s aren't that delicate even at higher voltage-- as long as you aren't also exceeding dissipation or current limits.

[kilowhat]

Hi guys, thanks for your replies, all interesting! Will take time this weekend to answer(!), as well as develop thoughts ive had on the subject since last posting. Just wanted to throw this proof of concept out to get some blitz-opinions from you guys!

Its basically a bare boned marshall major output section (split load pi, PP au7 driver, minus master vol and UL), with a lesliesque cathodes-to-OT-secondary-NFB-thinqie going on, but instead of tying power section to secondary, i go for the au7 driver cathodes (a la 92v's example!). Since the OT i have (hammond 1750ka) only has a 2r tap and no CT on the secondary (and this is where i go of the deep end), i figured i might maybe could perhaps do a fixed humdinger type virtual ground, which would double duty as both a reference to ground for the secondary, as well as bias resistors for the PP au7 driver stage.

If the concept is at all viable (is it??), i guess it could be applied to output cathodes as well, with some caveats i could think of:
* the individual resistors cannot (only the shared can) be bypassed without shunting AC between taps (like the virtual-gnd-cap-tween-cathodes-aka-sec-taps brain fart in my last post).
* as output cathode resistors are normally less resistive than driver ditto, they will also act a load on the output stage, like a parallelled resistive power soak, as well as provide speaker damping, but passively, unresponsively, uglily.
* they will dissipate dearly, and will need to be rated accordingly.
* they will induce current NFB, (im not too well versed in practice but) which in my head looks ugly in inductively loaded stages, but beautiful in resistive ones.

As of now, a 2r speaker/cab can shoot up to atleast ~44r (which it probably wont, because NFB bitches!) before theres any loading to talk about.
Im trying to figure out if the small resistors would work as poor mans fuses if a driver triode would weld shut and start shunting HT onto the secondary. I think the OT and speakers loop potentially could translate into smoke, but the path to ground would clearly blow (hopefully open) and unless touching speaker terminals/wires would probably not risk lives.

BUT!!
All this is based on the assumption that the 0r/common secondary tap - normally grounded - is tied there mainly/only for reference, and do not normally carry any current to write home about. But is this actually true?? I cant see any reason why it should (carry current that it, not not be true), but i dont actually KNOW (obviously tried googoling). Please help!

/kw

[kilowhat]

On second thought, i suspect ive drawn a PFB loop? If so, switching primary or secondary OT phase would be appropriate.

[PRR]

It could be simpler.

It is NOT current feedback.

It has the unlikely but dangerous risk that an OT primary to secondary short will lift both speaker terminals to a high DC voltage, yet not enough to blow the B+ fuse. Actually it will blow the part-Watt cathode resistors, and then will be full B+.

[92Volts]

BUT!!
All this is based on the assumption that the 0r/common secondary tap - normally grounded - is tied there mainly/only for reference, and do not normally carry any current to write home about. But is this actually true?? I cant see any reason why it should (carry current that it, not not be true), but i dont actually KNOW (obviously tried googoling). Please help!

/kw

It does carry current. Current always travels in a loop-- if it goes "out" of the 2r ("live") tap from the OPT it must come back "in" the 0r ("neutral") tap.

However, the 0r tap's connection to ground may not carry much current. It could just as easily float, though as PRR notes this could be a safety hazard.