Negative Feedback and Inductors

[PeanutNore]

Inspired by the Ampeg midrange circuit and the "boost" circuit in early Orange / Matamps, I've been experimenting with adding an inductor to a Marshall-style presence circuit, tuning the inductor and capacitor value to form essentially a bandpass filter that dumps negative feedback signal to ground, increasing the overall gain for the frequencies that it passes.

I couldn't find any recent examples of anyone doing this in their own builds, and I have seen some speculation that this would be impractical or would not work due to the limitations of real life inductors, so I wanted to share my experiences and hopefully clear things up.

I wired up a 25k pot as a variable resistor and placed it in series with a 220nF cap (100v film) and a 150mH inductor (10mm diameter radial with 5mm leads) to provide a center frequency of 876hz. This is wired between the point where the NFB connects to the PI / driver stage and ground, and when turned up the 25k series resistance is eliminated and frequencies within an octave or so of 876hz are shorted to ground.

I have installed this circuit in two very different amps, a 22 watt Orange Matamp type circuit with 6V6s and a 5 watt 5150 lead channel clone with an ECC99 output tube in parallel SE. In both it's very effective at boosting mid frequencies and does not cause oscillation or add any noticeable noise at any setting of the knob. The amount that it is able to boost the mids is limited by the amount of negative feedback in the circuit, so it's more pronounced in the OrMat clone which I have set up with a lot more NFB than the mini 5150.

I have a variety of similar inductors of various values between 1mH and 100mH on their way to experiment with other center frequencies - I know from building effects pedals with gyrator EQs that boosting 1800hz can sound really nice.

I am also planning to do some experiments with using an inductor and cap in series to bypass a cathode resistor. I've seen a thread on here from 2008 (link attached) where it's argued it wouldn't work, but I'm unconvinced and have the means and opportunity to test it experimentally. I'm currently working on a 5 watt mini SLO type thing with a push-pull ECC99 power amp and there's a few spots I plan to test it out in there. I'll report back with an update as things progress.

But yeah I definitely encourage anyone who is interested to try adding an inductor to a presence control. You can use an online calculator to figure out what values you need, but caps between 100n and 1uF and inductors between 22mH and 150mH are the range that seems most ideal.

[sparkies]

I am also planning to do some experiments with using an inductor and cap in series to bypass a cathode resistor. I've seen a thread on here from 2008 (link attached) where it's argued it wouldn't work, but I'm unconvinced and have the means and opportunity to test it experimentally. I'm currently working on a 5 watt mini SLO type thing with a push-pull ECC99 power amp and there's a few spots I plan to test it out in there. I'll report back with an update as things progress.
.

The series notch filter will work in the cathode circuit. It just be inserted into the opposite circuit function. So the effect will be inverted.

Midband response in the tube is govern by the dynamic inductance of the cathode circuit as this inductance reactance slows the current flow at effected frequencies. This negative effect is enhanced by stray inductance in the cathode circuit. So to cancel this effect (that I do in compressors, mic pres, and amps) is install a non inductive resistor that aggressively nulls the stray inductance in that part of the circuit. Get a Mills MRA5 resistor the size of the cathode resistor and use that, a/b with a bypass cap. Because in a lot of cases, the cap becomes a useless item as there is no insertion loss in AC with the resistor.

[PeanutNore]

The series notch filter will work in the cathode circuit. It just be inserted into the opposite circuit function. So the effect will be inverted.

The filter I'm using is technically a band stop filter, it attenuates the targeted frequencies by sinking them to ground. Because it's attenuating a negative feedback signal the overall effect is like a bandpass instead.

Here's my hypothesis/ understanding of why this will work in a common cathode circuit:

When you have a common cathode amplifier that's cathode biased using a resistor, any change in current through the tube changes the voltage drop across that cathode resistor.
Because the input voltage controlling the current in the tube isn't the absolute grid voltage alone but rather the grid to cathode voltage, when the cathode voltage changes this also changes the grid to cathode voltage and as the grid voltage drops, the cathode voltage also drops, reducing the magnitude of the change in grid to cathode voltage.
So the negative feedback signal is created by the cathode resistor and feeds into the cathode as an additional input.
With the LC filter connected where the cathode resistor and cathode meet, AC current flowing from the cathode at the target frequency has a low impedance path to ground through the LC circuit.
Because this current is not flowing through the cathode resistor, it does not cause a change in cathode voltage so if the grid voltage is changing at our target frequency, the cathode voltage stays relatively fixed at its bias point instead of following it. This results in a greater grid to cathode voltage differential and therefore higher gain at the target frequency.

LIke global NFB controls, the magnitude of the effect would be limited by the amount of negative feedback present in the first place, so this would potentially have a larger impact on stages with large cathode resistors than those with smaller ones.
With this in mind the first thing I plan to test it on is a cold clipper stage with a 39k cathode resistor.
Bypassing these with a capacitor usually doesn't sound good, but this is a much narrower band of frequencies and may have better results.

[Helmholtz]

I don't see a reason why this interesting idea should not work.
The lower the impedance between cathode and ground, the lower the feedback voltage and thus the higher the gain.
A series wiring of L and C forms a series resonant circuit having an impedance minimum at the resonant frequency.
This results is a gain boost around the resonant frequency of up to 6dB (at least in theory).
The parallel cathode resistor damps the resonance and broadens the boosted band.
The larger the cathode resistor, the narrower the band.

[sparkies]

The filter I'm using is technically a band stop filter, it attenuates the targeted frequencies by sinking them to ground. Because it's attenuating a negative feedback signal the overall effect is like a bandpass instead.

Here's my hypothesis/ understanding of why this will work in a common cathode circuit:

When you have a common cathode amplifier that's cathode biased using a resistor, any change in current through the tube changes the voltage drop across that cathode resistor.
Because the input voltage controlling the current in the tube isn't the absolute grid voltage alone but rather the grid to cathode voltage, when the cathode voltage changes this also changes the grid to cathode voltage and as the grid voltage drops, the cathode voltage also drops, reducing the magnitude of the change in grid to cathode voltage.
So the negative feedback signal is created by the cathode resistor and feeds into the cathode as an additional input.
With the LC filter connected where the cathode resistor and cathode meet, AC current flowing from the cathode at the target frequency has a low impedance path to ground through the LC circuit.
Because this current is not flowing through the cathode resistor, it does not cause a change in cathode voltage so if the grid voltage is changing at our target frequency, the cathode voltage stays relatively fixed at its bias point instead of following it. This results in a greater grid to cathode voltage differential and therefore higher gain at the target frequency.

LIke global NFB controls, the magnitude of the effect would be limited by the amount of negative feedback present in the first place, so this would potentially have a larger impact on stages with large cathode resistors than those with smaller ones.
With this in mind the first thing I plan to test it on is a cold clipper stage with a 39k cathode resistor.
Bypassing these with a capacitor usually doesn't sound good, but this is a much narrower band of frequencies and may have better results.

You need to look into what is called AC analysis of small signal amplifiers. Because the AC circuit aspects are the same regardless of amplification device (tube/soliid state).

The stop band frequency in a series notch is higher impedance than the other frequency, So when you applied it in the negative feedback you passed all the other frequencies at a lower impedance which causes them to be at less gain (by more negative feedback) . The exact opposite happens in the cathode circuit, As lower impedance frequencies would ignore the cathode resistor in the gain formula, and the stop band would include the cathode resistor in the gain formula. Which would cause all frequencies but the stop band to be amplified more.

Doing the same function in the cathode circuit that you applied in the negative feedback would be accomplished with a band-pass filter which is the inverted circuit function of a series notch filter.

[sparkies]

The downside of applying these filters in an amplifier is that it will shift the phase of the the signal at different frequencies inside the audio band. So it might sound ok by itself, but turns into a tone nightmare for anyone capturing and mixing a recording or using it with a live PA.

[PeanutNore]

This morning I tested it out using a 100mH inductor, a 220nF cap, and a 25k pot wired as a variable resistor in series from the cathode to ground, paralleled with the 39k cathode resistor on the cold clipper stage of my mini 5150/6505+ clone.

It definitely works - as I decrease the resistance of the 25k pot the amount of midrange increases, as well as the amount of preamp distortion (in a good way). If the series resistance of the pot is decreased below about 3k3, it starts to get too compressed for my taste - I think this is due to blocking distortion in the gain stages after the cold clipper (it's the 3rd triode in the preamp, there's two more gain stages and an anode follower after it).

If I installed this permanently in an amp, it would definitely be on the cold clipper stage and I would probably wire a fixed series resistor in place of the 25k pot to permanently set the boost level and put the whole thing on a switch, labeled "gain boost" or "mid boost" or something more clever. Basically I would use the 25k pot to find the point where blocking distortion starts, then back it off a little bit and measure the resistance and choose the next highest standard resistor from that.

[Chuck H]

Or... you could just put a 3.3k or 4.7k resistor in series with the load leg of the pot so it can't be dialed into ugly distortion territory Best of both worlds?