Negative Feedback - destination reason?

[Little Ricky]

I notice that that on most amps the NFB goes from the speaker to the phase inverter, but on the 5D5 (wide panel tweed Pro) it goes to the power tube grid.

What is the reason for sending the NFB to different endpoints? What sound results from these differences?

Also, could you have two separate wires sending NFB to two different locations, maybe with different resistor values, or a variable resistor 'blend' pot, to send some to the power and some to the PI?

The 5D5 has a paraphase splitter, just like the 5D8, but the 5D8 NFB goes to PI.

 

[jazzguitar]

Basically the designers intended to make an amp that amplifies on a linear curve throughout the frequency range. None does that. One way to improve that is to feed some of the output back to some position in the circuit which now is a new input. The polarity is chosen so that the new input signal is in reverse (hence "negative") to the amplified signal, so the amplifier amplifies a little less, but more true to the signal at the input jack as the "errors" are canceled out, to a degree.

With decades of tube amp development and design, the electronics engineers tried many versions of installing negative feedback. The usual going to the phase inverter is just the most common. Each has its own influence on sound, there is no general statement valid for all circuits.

 

[ThermionicScott]

I love the one-off experiments that sometimes showed up in these early amps. There are lots of ways to skin the proverbial cat, and Leo tried a bunch before settling on the design elements we're more familiar with.

At first glance, it's tempting to see the feedback loop connect to one output tube's grid and go, "What the hell was he thinking? It's only affecting one tube!" But since this is a paraphase inverter, as you noticed, by wiggling this grid a little bit in counterreaction to the output signal, that correction is carried into the other PI stage, and passed along to the other output tube as well. A somewhat roundabout way to do it, and Leo obviously wasn't satisfied in the end.

If I had to speculate, I think Leo was aiming for a cleaner sound, but was trying to keep the negative feedback from impacting the feel of the amp as much as possible, and a 1 Meg feedback resistor from the 8Ω output load would do that. (That resistor still has a significant effect because R2 of this particular divider is 270k, after all.)

- Scott

 

[CoolBlueGlow]

ThermionicScott makes a good point, and of course everything has context. NFB reduces distortion at the expense of gain (and impedance). In the early 50's gain was expensive and therefore still pretty sacrosanct.

Hard to imagine as well, but when the 5D5 was designed, those higher gain 9 pin miniatures like the 12ay7/ax7 were still new kids on the tube block in those days.

The cost of gain is why we see 6SJ7 pentode preamps in the earliest Fenders, like the very early versions of the Dual Professional, etc. Gain cost money and those 6SJ7 pentodes were the cheap gain miracle one tube solution. They were plentifully available in the post WWII surplus market. Lots of other cheap amps used them up into the late fifties, like the 1334 Silvertone.

Anyway, Leo moved away from small signal pentodes for reasons he apparently kept to himself, but his use of those then-newfangled miniature 9 pin double triodes in his early 50's designs were a bit of gamble too. (Octal dual triodes or the octal pentodes were still the dominant players in many designs - and would be into the sixties for several well known amp mfg.)

So, with regard to the feedback path, it is possible he was purely focused on inexpensively reducing output stage distortion to acceptable levels in a design that had only two gain stages including the PI. Maybe he figured the distortion problem was dominant in the output section due to those crummy beam power tubes? He was probably just listening and guessing, just like so many of us still do.

Again, context. Even in the late 40's early 50's, those newfangled beam tetrodes like the 6L/V6 were still viewed with some suspicion by more than a few audio amp design folks. They were not triodes, after all. Leo probably read all the available dope on the hot new Hi-Fi circuits. Maybe he secretly pined for the ultra clean results provided by the Williamson circuit?

Remember too that when Leo was getting into this thing, the Williamson circuit that had made such beam power tubes as the 6L6/V6 practical for hi-fi was itself only about 10-15 years old at that point. It was considered fairly exotic, and it was the UL transformer that made it all possible. Quite a few of those early descriptions of the Williamson circuit extol how "almost triode like" the UL/Williamson it makes those (crummy beam power tubes perform, especially reducing third harmonic distortion.

(Of course, in those days, when an amp distorted, everyone frowned at the player and said "turn down, your amp is crackling!" )

I reckon Leo hated distortion but also would not/could not spend the extra bucks on ultralinear tapped output transformers. By themselves, a UL OPT cost about as much as the rest of the parts on the 5D5 chassis combined!

Who knows, maybe he was trying to find a cheap shortcut. He was definitely known for that kind of value engineering and tinkering. That's why I do admire him. No pretense.

All speculation, of course...except the part about those crummy sounding beam power tubes.

 

[wmprivett]

You can add a pot and vary the nfb. It makes a nice difference. I just added it to a 5f1. I like it.