Feedback Resistor
Moderators: pompeiisneaks, Colossal
Feedback Resistor
Why is the NFB resistor only connected to the 8 ohm tap?
Re: Feedback Resistor
> Why is the NFB resistor only connected to the 8 ohm tap?
Why is it "always" connected to "8"? It isn't. Fender 5F6A only had a 2 ohm tap; when Marshall adapted the plan he used the same NFB resistor but tapped at 16.
Why is it always connected to just one tap, not all of them? Because they "should" be the exact same signal except different voltage/current ratios. In practice they are exact-enough. And if they weren't, which tap should the NFB loop listen to?
When you build an amp for just ONE load impedance, you may take NFB from that tap, and adjust NFB ratio accordingly. You occasionally see this on custom-made Hi-Fi amps, which live by numbers instead of by sound, and are working at the ragged edge of too much NFB.
OTOH, you sometimes see a PA amp with 4-8-16 ohm taps, 25V and 70V taps, and NFB not taken from any of these, but from a dedicated NFB winding. As long as it senses what is happening in the transformer, it can correct 90+% of errors.
In guitar amps, we rarely run a huge amount of NFB, don't want to totally correct all "error" (we call it "tone"). You pick one handy tap, then fiddle the NFB resistor to taste. As a starter, you can steal tap and resistor values from a very similar existing plan and duplicate as close as you conveniently can, but don't sweat being a tap or a few-K off. When it's all "working", you should play with the resistor value (or the tap-point, which apparently was more convenient for Jim).
Why is it "always" connected to "8"? It isn't. Fender 5F6A only had a 2 ohm tap; when Marshall adapted the plan he used the same NFB resistor but tapped at 16.
Why is it always connected to just one tap, not all of them? Because they "should" be the exact same signal except different voltage/current ratios. In practice they are exact-enough. And if they weren't, which tap should the NFB loop listen to?
When you build an amp for just ONE load impedance, you may take NFB from that tap, and adjust NFB ratio accordingly. You occasionally see this on custom-made Hi-Fi amps, which live by numbers instead of by sound, and are working at the ragged edge of too much NFB.
OTOH, you sometimes see a PA amp with 4-8-16 ohm taps, 25V and 70V taps, and NFB not taken from any of these, but from a dedicated NFB winding. As long as it senses what is happening in the transformer, it can correct 90+% of errors.
In guitar amps, we rarely run a huge amount of NFB, don't want to totally correct all "error" (we call it "tone"). You pick one handy tap, then fiddle the NFB resistor to taste. As a starter, you can steal tap and resistor values from a very similar existing plan and duplicate as close as you conveniently can, but don't sweat being a tap or a few-K off. When it's all "working", you should play with the resistor value (or the tap-point, which apparently was more convenient for Jim).
Re: Feedback Resistor
A little complement to PRR's answer.
Personally I would say that for example the Marshall JTM45 has rather high amount of feedback. At least compared to other guitar amps.
As PRR already has said, when Marshall copied the Fender Bassman 5F6-A to make the JTM45 they used the same values on the feedback resistors (yes, there’s two) as in the Bassman, 27k and 5k (the 5k is the Presence potentiometer). As the output transformer in the JTM45 is made for a 16 ohm speaker load, the voltage on the output is about 2.5 times higher than in the Bassman, which is made for a 2 ohm speaker load. Therefore about 2.5 times more voltage is also feed back in JTM45 than in the Bassman, despite that the resistors have the same value.
Don’t forget that the amount of feedback is not only depending on the series feedback resistor from the output transformer, but also on the resistor to ground that it’s connected to (which is the Presence pot in many amps). For example: Doubling the value of the series feedback resistor is the same as choosing half the value of the resistor to ground.
For more information, please read Randall Aikens tremendous explanation in this matter, here: http://www.aikenamps.com/GlobalNegativeFeedback.htm
I really recommend you to read the information there!
I would say that the amount of feedback changes the sound of the amp rather much.
Many amps have no feedback like this in the output stage at all. VOX AC30, many Matchless amps, Fender Vibro-King and more.
Take care,
Sven-Johan
Personally I would say that for example the Marshall JTM45 has rather high amount of feedback. At least compared to other guitar amps.
As PRR already has said, when Marshall copied the Fender Bassman 5F6-A to make the JTM45 they used the same values on the feedback resistors (yes, there’s two) as in the Bassman, 27k and 5k (the 5k is the Presence potentiometer). As the output transformer in the JTM45 is made for a 16 ohm speaker load, the voltage on the output is about 2.5 times higher than in the Bassman, which is made for a 2 ohm speaker load. Therefore about 2.5 times more voltage is also feed back in JTM45 than in the Bassman, despite that the resistors have the same value.
Don’t forget that the amount of feedback is not only depending on the series feedback resistor from the output transformer, but also on the resistor to ground that it’s connected to (which is the Presence pot in many amps). For example: Doubling the value of the series feedback resistor is the same as choosing half the value of the resistor to ground.
For more information, please read Randall Aikens tremendous explanation in this matter, here: http://www.aikenamps.com/GlobalNegativeFeedback.htm
I really recommend you to read the information there!
I would say that the amount of feedback changes the sound of the amp rather much.
Many amps have no feedback like this in the output stage at all. VOX AC30, many Matchless amps, Fender Vibro-King and more.
Take care,
Sven-Johan
Re: Feedback Resistor
Thanks, great info and link!
I meant to put this in the TW tech discussion forum (TW has NFB resistor on 8 ohm tap).
I meant to put this in the TW tech discussion forum (TW has NFB resistor on 8 ohm tap).
Re: Feedback Resistor
> the Marshall JTM45 has rather high amount of feedback
Yes.
The 5F6A is pretty darn perfect. Most of its ideas were old, but they got put together well.
One thing that was still evolving was NFB. In 1950 it was rare. By 1960 it was routine in Hi-Fi amps but guitar amp fashion lagged. The 5F6A runs "small" NFB. Enough to reduce but not suppress speaker bass resonance (need to leave some boom to overcome the bass-loss of open-back cabinets). Enough to take some of the low-level kink out of cool-biased tubes. Incidentally Leo hung a cap on the network and got an interesting high boost, different from the old TREBle network. (In particular: it lessens the NFB-induced IM, up top where it tends to be un-tonal.)
Some of the older 1950s amps are downright muddy to modern ears. A few years later, Jim had been exposed to a lot of hi-NFB Hi-Fi amps. They are generally cleaner than lo-NFB or no-NFB amps. And anything "new" can be "good".
Also I believe Jim was mainly selling closed-back or very large open-back cabinets, with drivers tuned different than what Leo was using. Also by Jim's time, most bands had a strong bass player, and the guitar was not expected to fill the bottom octaves.
So Jim gave us amps that can be a bit shy in bass, pushing the power up to the tenor-alto range and biting-through crowd noise. Clean up to fairly high level with a narrow zone before gross clipping (softened by the soft EL34 tubes he got cheaper than 6L6).
So just those few changes from 5F6A (bass-trim, more NFB, speakers, EL34) gave a New Sound. It also was cheaper for some poor brits to buy from Jim than from Leo, poor brits who were Changing The Music. And Jim (unlike Leo who retired in here) needed and wanted the business, and worked with promising customers to build bigger amps and speakers and slave-amp systems. Wall Of Marshall Stacks. Very different sound than a Twin or an AC30, and quite necessary in larger venues in days when the house PA system was usually smaller than the guitar-rig.
Yet NBF comes and goes again and again. Ampeg made some amps with ample NFB. Yet they sold the VT-40 with no NFB at all. And it works very well, and the later versions are not radically different from a Marshall. (I was startled at my measurements of "my" modded VT-40; I'd been taught to use heavy NFB to get "good spec numbers" and done it a lot to make bad amps less-bad; the spec-numbers of the raw VT-40 are quite good.) Both low-NFB and no-NFB amps are popular today. (Few guitar players are real interested in high-NFB amps like the early Sunn; some bassists are still fond of high-NFB tubes.)
I'm thinking the NFB has a lot to do with the speaker you use. Closed-back will need some NFB or a lot of bass-response tweaking. Open-back is wonderfully synergistic with raw (no-NFB) pentodes. But choice of speaker resonance, box or baffle size, pickup, style, make a big difference also.
I just came off a session where we talked of Baroque pipe organ in modern rooms. The point was made that even if you could crate-up Bach's own organ and reinstall it in our 1873 chapel or a 1972 concert hall, it would NOT be the same sound. People do play this game: write-up a "specification" for a pipe organ in a certain style, and then install it in a room which is very different from any of the old rooms. Our 1873 room won't hold bass the way Bach's stone churches did. (Most new rooms suck; even JSB wudda switched to Guitar.)
Organ builders "bench tune" to get an approximate balance, then install and spend months tuning the balance to the room. A bass rank which seems powerful in a small shop may be lame in a big but wimpy structure. A pipe that seems tame in the shop may screech-out in the room. Like you may fiddle the pickup poles to balance string loudness, pipe-dudes have tricks to change the power and flavor of pipes.
And a guitar amp also needs a rough bench-tune (does it seem to play OK?) and then a "room tune".... not a specific room because guitarists rarely have the luxury of staying in one place, but some trials of the cabinet and speakers and guitar and player in several typical rooms and usually a series of changes to balance the sound. NFB is one powerful tweak in the toolbox.
> the feedback resistors (yes, there’s two)
Quite true. And it is the ratio not the absolute values. But it is often convenient to say "the NFB resistor". The lower one is usually "fixed" by some other factor. It may be a cathode bias resistor and has to be right for DC. Or it may be made lower than cathode impedance. In the 5F6A/JTM45 case, it is a compromise between "low", an affordable Presence cap, and available pots. Once the lower resistor is fixed by other factors, the only free variable (the number you can change) is the value of the "top" resistor (and the tap it is connected to).
Yes.
The 5F6A is pretty darn perfect. Most of its ideas were old, but they got put together well.
One thing that was still evolving was NFB. In 1950 it was rare. By 1960 it was routine in Hi-Fi amps but guitar amp fashion lagged. The 5F6A runs "small" NFB. Enough to reduce but not suppress speaker bass resonance (need to leave some boom to overcome the bass-loss of open-back cabinets). Enough to take some of the low-level kink out of cool-biased tubes. Incidentally Leo hung a cap on the network and got an interesting high boost, different from the old TREBle network. (In particular: it lessens the NFB-induced IM, up top where it tends to be un-tonal.)
Some of the older 1950s amps are downright muddy to modern ears. A few years later, Jim had been exposed to a lot of hi-NFB Hi-Fi amps. They are generally cleaner than lo-NFB or no-NFB amps. And anything "new" can be "good".
Also I believe Jim was mainly selling closed-back or very large open-back cabinets, with drivers tuned different than what Leo was using. Also by Jim's time, most bands had a strong bass player, and the guitar was not expected to fill the bottom octaves.
So Jim gave us amps that can be a bit shy in bass, pushing the power up to the tenor-alto range and biting-through crowd noise. Clean up to fairly high level with a narrow zone before gross clipping (softened by the soft EL34 tubes he got cheaper than 6L6).
So just those few changes from 5F6A (bass-trim, more NFB, speakers, EL34) gave a New Sound. It also was cheaper for some poor brits to buy from Jim than from Leo, poor brits who were Changing The Music. And Jim (unlike Leo who retired in here) needed and wanted the business, and worked with promising customers to build bigger amps and speakers and slave-amp systems. Wall Of Marshall Stacks. Very different sound than a Twin or an AC30, and quite necessary in larger venues in days when the house PA system was usually smaller than the guitar-rig.
Yet NBF comes and goes again and again. Ampeg made some amps with ample NFB. Yet they sold the VT-40 with no NFB at all. And it works very well, and the later versions are not radically different from a Marshall. (I was startled at my measurements of "my" modded VT-40; I'd been taught to use heavy NFB to get "good spec numbers" and done it a lot to make bad amps less-bad; the spec-numbers of the raw VT-40 are quite good.) Both low-NFB and no-NFB amps are popular today. (Few guitar players are real interested in high-NFB amps like the early Sunn; some bassists are still fond of high-NFB tubes.)
I'm thinking the NFB has a lot to do with the speaker you use. Closed-back will need some NFB or a lot of bass-response tweaking. Open-back is wonderfully synergistic with raw (no-NFB) pentodes. But choice of speaker resonance, box or baffle size, pickup, style, make a big difference also.
I just came off a session where we talked of Baroque pipe organ in modern rooms. The point was made that even if you could crate-up Bach's own organ and reinstall it in our 1873 chapel or a 1972 concert hall, it would NOT be the same sound. People do play this game: write-up a "specification" for a pipe organ in a certain style, and then install it in a room which is very different from any of the old rooms. Our 1873 room won't hold bass the way Bach's stone churches did. (Most new rooms suck; even JSB wudda switched to Guitar.)
Organ builders "bench tune" to get an approximate balance, then install and spend months tuning the balance to the room. A bass rank which seems powerful in a small shop may be lame in a big but wimpy structure. A pipe that seems tame in the shop may screech-out in the room. Like you may fiddle the pickup poles to balance string loudness, pipe-dudes have tricks to change the power and flavor of pipes.
And a guitar amp also needs a rough bench-tune (does it seem to play OK?) and then a "room tune".... not a specific room because guitarists rarely have the luxury of staying in one place, but some trials of the cabinet and speakers and guitar and player in several typical rooms and usually a series of changes to balance the sound. NFB is one powerful tweak in the toolbox.
> the feedback resistors (yes, there’s two)
Quite true. And it is the ratio not the absolute values. But it is often convenient to say "the NFB resistor". The lower one is usually "fixed" by some other factor. It may be a cathode bias resistor and has to be right for DC. Or it may be made lower than cathode impedance. In the 5F6A/JTM45 case, it is a compromise between "low", an affordable Presence cap, and available pots. Once the lower resistor is fixed by other factors, the only free variable (the number you can change) is the value of the "top" resistor (and the tap it is connected to).
Re: Feedback Resistor
Sorry, but I don't agree with you here. The resistor to ground is, as I see it, only (or mainly) for the feedback circuit. In a long tail pair phase inverter the bias for the tube is set by another resistor. Then there's also the "tail resistor". Amps with no global negative feedback (often) don’t have the feedback resistor to ground. The tail resistor and the capacitor from the second triode's grid are there connected directly to ground.PRR wrote:...But it is often convenient to say "the NFB resistor". The lower one is usually "fixed" by some other factor. It may be a cathode bias resistor and has to be right for DC. Or it may be made lower than cathode impedance. In the 5F6A/JTM45 case, it is a compromise between "low", an affordable Presence cap, and available pots. Once the lower resistor is fixed by other factors, the only free variable (the number you can change) is the value of the "top" resistor (and the tap it is connected to).
Randal Aiken also has a good article about this: http://www.aikenamps.com/LongTailPair.htm
Take care,
Sven-Johan
Re: Feedback Resistor
> I don't agree with you here.
Cool!
> The resistor to ground is, as I see it, only (or mainly) for the feedback circuit.
Well, there is always some constraint on resistor values. And in practice, one of the two resistors has a tighter range of "good" values. When coming off a low-Z output, usually the lower resistor (though in low-gain low-noise low-power preamps, it may be the upper resistor).
Quite often NFB is injected at the cathode-bias resistor. Champ 5F1, Vibrolux 5E11. We can fiddle that 1K5 to 1K or 2K, but not much more.
Alternatively it is injected at a resistor "much smaller than cathode impedance". Champ AA764. Cathode impedance is like 1K. "Much smaller" was picked as 20X less or 47R for only 5% loss of potential gain. We could fiddle that up to 100 ohms for 10% loss of gain, OK. We could fiddle down quite far but at the time, small resistors stopped at 22R. Given free choice we could use 1R but then the other resistor scales to 57 ohms which is sucking ~10% of our precious output power. Yeah, a wide range is possible in this scheme, but not arbitrarily wide.
> In a long tail pair phase inverter the bias for the tube is set by another resistor. Then there's also the "tail resistor".
Sure. In fact we should ask why the lower FB resistor is in the common tail at all. Simple logic suggests it should just drive the other-side grid, not the tail. In which case its resistance magnitude is very arbitrary. More than 1R so the network does not load the output, less than 1Meg or 100K so it will push the grid well and doesn't catch squeal.
When you add Fender's Presence control, you have additional constraints. 1-ohm pots are rare. 100 ohm is easy, but then the cap scales to 5uFd. Since there is a DC bias (since it is in the DC tail path), it could be an electrolytic.... Leo didn't do that in 5F6A because he came from 5F6 where the Presence network is not biased-up with DC.
But in guitar amps (and in no other field), that lower FB resistor IS in the common tail. Why? Not sure. However it affects the balance and 2nd harmonic. And the size of that fed-from-output resistor determines how much the effect is. I think it actually reduces even-order distortions, as long as the feedback loop has "ample feedback". However, guitar amps usually live in "small feedback". And many different resistor ratios have been used. Sometimes the 5K pot is shunted with a 1K5 resistor. Tail resistors vary from 6K to 22K, more than plate resistors which also kick-down to the cathode network. Then you find unbalanced plate resistors. Are these to improve balance (as expected), or to UNbalance a too-sterile amp? OR to balance the clip-levels? (It is possible to be balanced for small-signal and unbalanced in clipping, or vice-versa.)
Considering the 5F6A is the most popular big-amp in history, and it is pretty ordinary except some strange stuff like the NFB injected at the tail instead of the grid, I have to wonder how much freedom we have with that lower resistor.
> Randal Aiken also has a good article
Very very good as far as it goes. (Which is a LOT further than most practical tube theory.)
For the simple long-tail, with triodes, perfect balance is not infinite tail impedance but a function of plate impedances. This point is often glossed-over because it is unintuitive, and because in practice we never have enough voltage to spare. If the tail is much greater than cathode impedance, balance is good enough for real-world design.
He does touch upon the "third input". I'm not convinced he tells the whole story. Several subtle things are happening.
IAC, the value of the lower NFB resistor "may be critical", or may be flexible in ways which are not entirely clear (to me). Sure, the amp will "work fine" with VR1=100R, R7=20K, C3=5uFd. But does it work same-as 5F6A values? Leo may have accidentally mis-wired a 5F6 and hit upon something special.
Cool!
> The resistor to ground is, as I see it, only (or mainly) for the feedback circuit.
Well, there is always some constraint on resistor values. And in practice, one of the two resistors has a tighter range of "good" values. When coming off a low-Z output, usually the lower resistor (though in low-gain low-noise low-power preamps, it may be the upper resistor).
Quite often NFB is injected at the cathode-bias resistor. Champ 5F1, Vibrolux 5E11. We can fiddle that 1K5 to 1K or 2K, but not much more.
Alternatively it is injected at a resistor "much smaller than cathode impedance". Champ AA764. Cathode impedance is like 1K. "Much smaller" was picked as 20X less or 47R for only 5% loss of potential gain. We could fiddle that up to 100 ohms for 10% loss of gain, OK. We could fiddle down quite far but at the time, small resistors stopped at 22R. Given free choice we could use 1R but then the other resistor scales to 57 ohms which is sucking ~10% of our precious output power. Yeah, a wide range is possible in this scheme, but not arbitrarily wide.
> In a long tail pair phase inverter the bias for the tube is set by another resistor. Then there's also the "tail resistor".
Sure. In fact we should ask why the lower FB resistor is in the common tail at all. Simple logic suggests it should just drive the other-side grid, not the tail. In which case its resistance magnitude is very arbitrary. More than 1R so the network does not load the output, less than 1Meg or 100K so it will push the grid well and doesn't catch squeal.
When you add Fender's Presence control, you have additional constraints. 1-ohm pots are rare. 100 ohm is easy, but then the cap scales to 5uFd. Since there is a DC bias (since it is in the DC tail path), it could be an electrolytic.... Leo didn't do that in 5F6A because he came from 5F6 where the Presence network is not biased-up with DC.
But in guitar amps (and in no other field), that lower FB resistor IS in the common tail. Why? Not sure. However it affects the balance and 2nd harmonic. And the size of that fed-from-output resistor determines how much the effect is. I think it actually reduces even-order distortions, as long as the feedback loop has "ample feedback". However, guitar amps usually live in "small feedback". And many different resistor ratios have been used. Sometimes the 5K pot is shunted with a 1K5 resistor. Tail resistors vary from 6K to 22K, more than plate resistors which also kick-down to the cathode network. Then you find unbalanced plate resistors. Are these to improve balance (as expected), or to UNbalance a too-sterile amp? OR to balance the clip-levels? (It is possible to be balanced for small-signal and unbalanced in clipping, or vice-versa.)
Considering the 5F6A is the most popular big-amp in history, and it is pretty ordinary except some strange stuff like the NFB injected at the tail instead of the grid, I have to wonder how much freedom we have with that lower resistor.
> Randal Aiken also has a good article
Very very good as far as it goes. (Which is a LOT further than most practical tube theory.)
For the simple long-tail, with triodes, perfect balance is not infinite tail impedance but a function of plate impedances. This point is often glossed-over because it is unintuitive, and because in practice we never have enough voltage to spare. If the tail is much greater than cathode impedance, balance is good enough for real-world design.
He does touch upon the "third input". I'm not convinced he tells the whole story. Several subtle things are happening.
IAC, the value of the lower NFB resistor "may be critical", or may be flexible in ways which are not entirely clear (to me). Sure, the amp will "work fine" with VR1=100R, R7=20K, C3=5uFd. But does it work same-as 5F6A values? Leo may have accidentally mis-wired a 5F6 and hit upon something special.