Correcting speaker resonance without removing output stage distortions

[anson]

Hi folks,


I really like the way that negative feedback corrects for speaker box resonance, removing the exagerrated "boominess" out of amps and making for a tighter sound. But not only does NFB reduce the severity of resonance, it also reduces disitortions introduced to the waveform as it passes through the PI, the output tubes, and the OT. What if I wanted to only correct the speaker resonance, but not the other other distortions? That way the distortions from the PI and output stage would be audible, and, possibly, good sounding. Does anyone have ideas whether this is possible and what type of circuit it could be?

I'm thinking that taking the signal from the speaker tap and sending it to the grids of the power valves might work, which will still allow the PI to distort at least. I'm hesitant to just go and try it though with out knowing more. And I'm wondering what approaches would allow the output valves to go uncorrected in their clipping, too.

I admit, this whole thing might me a crazy idea, but I'm trying to think pragmatically about wat the functions of an amp do and what is desireable and what is not. And it seems that correcting for output stage distortions is generally inconsistant with the rest of hi-gain guitar amp design, which involes distorting waveforms left and right almost as much as possible. So I'm wondering if correcting output stage distortions is not much more than a lingering legacy of the days when guitar amps were adapted from hi-fi design, or maby just a necessity if one wants to reduce resonant frequencies in the speaker.

Thanks to everybody who contributes. I'm ready to try some experiments and report back!
A

[Steve Conner]

Generally speaking, everything that's bad for hi-fi is good for guitar. The speaker resonances improve the sound, at least from a classic rock perspective. The boominess is perceived as "chunk" and the treble boost from a high output impedance interacting with voice coil inductance helps you cut through the mix.

If you want to hear what it's like without them, you can run a tube amp (with no NFB) into a resistive dummy load and then amplify it again with a solid-state hi-fi amp. This damps the speaker and blocks any interaction between output stage and speaker. I've tried it and it sounds a bit flatter and duller than the same tube amp driving the same speaker directly.

I also tried a different solid-state amp with a high output impedance. Still no interaction between tube and speaker, but the resonances were undamped and it livened things up nicely.

In a push-pull tube output stage, you can apply feedback to the cathodes of the power tubes from the OT secondary. You ground the 4 ohm tap and connect the cathodes to 0 and 16.

[jazbo8]

If you can show the schematic of the amp, we can take a closer look at what can be done with the NFB. Another way to deal with the speaker resonance without messing with the NFB too much, is to use a Zobel network across the speaker.

[Chuck H]

There is also a lot of good info on "de tuning" enclosures in the cabinet section of this forum. Basically it's calculated porting to balance the speaker and cabinet resonances to flatten the LF response.

[cbarrow7625]

It sounds like what you are asking about is similar to what a typical resonance control does. The higher frequencies are shunted to ground while the effect of the feedback is unaltered for lower frequencies.

Your application would require a lower knee frequency than the typical presence control but the effect would be the same. Experiment (and measure) with larger presence caps.

To get the knee frequency closet to the driver resonance you could use a steeper slope than the typical 6dB per octave presence control slope by building multiple R-C constants into the "presence" circuit. Higher orders of slopes in the negative feedback circuit could cause instability, but it would be somewhat reduced at the lower frequency cutoff frequencies that you are considering.

[anson]

I use a mid-scoop in the nfb circuit. Which of course makes for an audible mid-boost. It sounds very sweet!

The frequency response curve represents the signal taken from the junction of the 500k and 100k ohm resistors, which are there to bump the level of the nfb signal down before it gets to the PI.

Cbarrow, yes, I tried using a steeper than -6bd slope a while ago. I used two low pass circuits in series, and had the two resistors going to ground as a double ganged control. It sounded really good--got lots of mids. I can't remember why I moved on from that... I think I wanted a way to get rid of the hissy hashy highs that I could hear when the presence was urned up, which is why I went for the mid-control.

The 25,000 and 1 ohm resistors in the schematic act together as a potentiometer. I change them manually for a pot effect, since I haven't figured out yet how to put a pot in LT Spice. They control the amount of mid-scoop. When 1ohm is inserted into the upper resistor, and 25k in the lower resistor, the signal is very flat and it sounds like an amp with the presence all the way down.

The 2k resistor can be varied and is effectively a "sweep" control, and varies the frequency of the center of the mid-depression useful values are from 1k to 25k ohm.

Steve, I think I should give your cathode nfb method a try. If the ground is at 4ohms, then would 0ohms and 16ohms tap make it unbalanced? Should it be 0 and 8ohms instead?

[jazbo8]

[ATTACH=CONFIG]25416[/ATTACH]
Steve, I think I should give your cathode nfb method a try. If the ground is at 4ohms, then would 0ohms and 16ohms tap make it unbalanced? Should it be 0 and 8ohms instead?

Steve's talking about a push-pull output stage, which requires symmetrical FB to both cathodes, as shown in the example below.

[loudthud]

The 4 ohm tap is the center tap voltage wise of a 16 ohm winding. Do the math! Calculate the voltage of 100W at 4 ohms. Next calculate the voltage for 100W at 8 ohms. Notice anything? Finally calculate the voltage for 100W at 16 ohms. Light bulb should be on at this point.

You sort of want two opposing things. A low output impedance (high damping factor) is what damps the resonance. The typical pentode or tetrode output stage has a high output impedance and feedback is what lowers it. But you don't want feedback. If you want a low output impedance without feedback, you have to move away from the typical output stage. A triode or ultralinear output stage has a lower output impedance. Perhaps not as low as you would like. A big resistor across the output or resistive attenuator would lower the impedance seen by the speaker. Do I have to tell you how bad that sounds?

Solid State amps had very low output impedances. That sounded so bad that now they use mixed mode feedback to raise the output impedance. If you want to lessen the effect of resonance, use a notch filter or a gentle bass rolloff. One last thing to try, use several speakers in parallel with staggered resonant frequencies.

[loudthud]

To illustrate the staggered resonance, see the attached impedance plot of my bass rig. One cabinet is a 16 ohm JBL E-140 15 inch woofer. The second cabinet is a pair of EVM-10s connected in series for 16 ohms. The parallel connection of the two cabinets has a much flatter impedance curve.

[anson]

The 4 ohm tap is the center tap voltage wise of a 16 ohm winding. Do the math! Calculate the voltage of 100W at 4 ohms. Next calculate the voltage for 100W at 8 ohms. Notice anything? Finally calculate the voltage for 100W at 16 ohms. Light bulb should be on at this point.

Thanks, I get it now.

In a push-pull tube output stage, you can apply feedback to the cathodes of the power tubes from the OT secondary. You ground the 4 ohm tap and connect the cathodes to 0 and 16.

Right now, each of the output tubes on my amp is connected to ground through a 1ohm resistor. To try what you mention, would I get it to work by putting the 0 and 16ohms taps at the junction between the cathode and a much larger cathode resistor? I must admit I can't interpret all of the symbols on teh image from Jazbo's post #7--the circles with crosses in them and the symbols coming off the cathodes that look like input/output jacks don't trigger anything in my brain.

[jazbo8]

The 1 Ohm resistors are for checking the bias and should be left where they are, so the connections go from the cathodes -> 1 Ohm resistors -> OPT taps (0 Ohm or 16 Ohm, with 4 Ohm grounded). Unless you are doing self-bias, you do not need the large cathode resistors. The example shown has jacks for monitoring the cathode currents, so you can ignore them. To make sure that you got the connections right, upload a schematic here for comments before you actually try it on the bench.

[anson]

The 1 Ohm resistors are for checking the bias and should be left where they are, so the connections go from the cathodes -> 1 Ohm resistors -> OPT taps (0 Ohm or 16 Ohm, with 4 Ohm grounded). Unless you are doing self-bias, you do not need the large cathode resistors.

But if I connect the 4ohm OTtap to ground, and both the 0ohm and 16ohm OT taps to the 1ohm cathode resistors as described, then that leaves only 1ohm resistance between those taps and ground all the time, which seems to me to be a situation waiting for the OT to be toasted, since the taps are are all so near to ground. How can such a tap drive a 16ohm speaker when it already spilling so much current to ground through a 1 ohm resistor? It seems like a couple hundred ohms would be minnimum. I admit I might be missing somehing here.



I did an experiment I've done dozens of times already--I disconnected the NFB circuit and I dig the sound of mids and highs, but the lows are too much. So I put a 4-way selector which allows me to choose the series capacitor in the preamp that determines the low frequency roll off. I can control the bass alot better now, both the level and the roll-off frequency and the result is good. I'll just have to see how good it sounds cranked at rehearsal now.

As a side experiment, I also put such a 4-way capacitor selector switch that allows me to choose the high frequency roll off of the treble control. I like that too!

If you want to hear what it's like without them, you can run a tube amp (with no NFB) into a resistive dummy load and then amplify it again with a solid-state hi-fi amp. This damps the speaker and blocks any interaction between output stage and speaker. I've tried it and it sounds a bit flatter and duller than the same tube amp driving the same speaker directly.

That's pretty much the definative experiment that I was looking for. Flattening the response of the speaker, but still allowing the PI and output tubes to carry forth their harmonic distortions. So, you say it sounds crap?

Thanks for your input,
A

[Steve Conner]

Easy. You lift the bottom ends of the 1 ohm resistors off ground and connect them to the OPT secondary taps.

I didn't say it sounded crap. I said "a bit flatter and duller", not quite as good as the tube amp driving the same speaker directly, but not hopelessly bad either. Easily fixed by adding a bass and presence boost elsewhere in the circuit.