DIY SubWoofer with Velocity Feedback
This article has been updated since it was first published in 2021. I made a second pass at the filter, simplifying it considerably. Now it is just a 2nd order low-pass filter with adjustable Q, so the dip in the response can be removed and the phase shifted. Also, I am driving the sub with a 100W amplifier, rather than a bridged lm-3886 stereo amp. About the same power, but less trickery.
Expectations
One side effect of velocity feedback is it can lower the cutoff frequency below the driver and cabinet resonance. With feedback this 10" subwoofer has response down to 10Hz. This sub is just for music, and not for special effects. I would want a 15" sub for home theater. Also, keep in mind that this feedback will not help on a ported speaker box. The port will change the ratio between cone motion and loudness at different frequencies. The amplifier will cause a huge peak at the port/cabinet resonance, in an effort to keep the velocity up.
Box and SD270A-88 Driver Response
Plot of the driver in the enclosure, without velocity feedback. The woofer is in an enclosure with an internal volume of 73 liters, giving QTC = 0.669, f3 = 43Hz. The free-air resonance of the Dayton Audio SD270A-88 is 26.2Hz.
A dual voice coil driver has two nearly identical windings on the same coil form and moving in the same magnetic field. One coil can be driven and the other used to provide velocity feedback, if the feedback signal is compensated [1]. My plots do not look exactly like theirs, because their velocity is coil voltage w.r.t. laser-measured velocity (an error signal), where mine is absolute coil voltage with a zero reference.
With any dual voice coil low frequency driver, the signal is not really a perfect match to the velocity of the cone. There is mutual inductance between the two coils that causes a 180° phase shift and a dip in the amplitude response. The frequency and amplitude of this phenomenon is different for each driver.
The disturbance in response can be counteracted by a 2nd order filter, with f0 = 146Hz. That puts the feedback more in line with the actual cone velocity. It will work without the filter, but it won't be anything near flat response. In fact my first pass at velocity feedback, in 1995 or so, was a 15" Eminence DVC woofer with a 25W amp I cobbled together using 2N6055 and 2N6053 Darlington transistors, and straight feedback from the voice coil. I dropped the whole mess in the cardboard box the driver came in and man did it sound good. Not the best, but it taught me a little about velocity feedback. I thought had I invented it. Doh! People have been working on it since the 1930's.
The carnival trick of any motional feedback system is pushing gently on the cone. It is like pushing on the wall, if the system is tight enough. If the cone does move in, it will stay for a while, then slowly move back out. All the while using amplifier power to hold the cone while the suspension is trying to put it back in place. The cone re-centers itself at a rate determined by the low-frequency cutoff of the amplifier. Don't try it with a speaker protector on the output, though.
Velocity Feedback Circuit
The circuit consists of several parts - a special feedback filter, a summing amp, an input low-pass filter, and a differential amplifier. The input summing amp sums the left and right input signals into a mono signal. The input filter then filters out frequencies above 200Hz. A pot controls the level of the input signal. The feedback filter takes as input the output of the non-driven voice coil, which can be several volts. It must not be greater than the signal, so there is a pot to adjust it down to the correct level. These two pots are to control the ratio of signal to feedback. There is also a pot to adjust the Q of the filter. That control determines the amount of peaking at the cutoff frequency. The peaking is opposite the dip in the unfiltered feedback. From there the signal goes to the differential amplifier, where it is subtracted from the filtered sum of the two preamp channels. The output from the differential amplifier is the error voltage, the difference between what we want the driver to do and what it did, which drives the power amplifier.
The circuits are powered by an on-board ±15V supply, driven by 2 15VAC windings on the power transformer.
Finished Board
I built the VFB circuit on a 100mm x 100mm PCB. This is a picture of the final board. It goes in the subwoofer amplifier case, between the inputs and the amplifier, and connects to one voice coil of the driver. See the build page for more on the other parts of the power amp and filter.
After compensation, the response was flat within ±2.6dB from 10Hz up to 180Hz. YMMV, but I think it turned out good.
Adjusting the Feedback Circuit
The adjustment procedure needs to be done after the system has been built and otherwise tested. You can set the input level and "SIGNAL" to some small level, and the "FEEDBACK" pot all the way down, and then test the amplifier without the feedback. If you have a dual-channel scope, you can look at the signal and the feedback, and verify that the two are in reasonably correct phase, before cracking open the feedback pot. Correct phase means the feedback peak is lined up with the zero-crossing on the rising edge of the input signal. That is the place on the sine wave of maximum velocity. If it is lined up with the falling edge, you have to switch the 2 wires of the feedback cable.
There are 3 adjustments that control the behavior of the subwoofer, and a front-panel-mounted pot to adjust the input level. If you don't use the level pot, you need to put a pair of trimmers on the board, or somewhere in the inputs, to adjust the level to match your main speakers. After that, the sub will track the volume of the main speakers without adjustment. Don't try to control the output voltage - it won't work. Don't try to get the maximum feedback, because the error amp will be dangerously close to oscillation, and the last step would probably throw it over.
The adjustment process starts with all pots, including the external input level pot, at zero. Keep the sub away from you. A small mistake makes a big noise. The amplifier will be amplifying the oscillation of the feedback circuit at full power. Power everything up and apply a 20Hz signal to the inputs. You shouldn't hear anything from the sub.
Raise the level on the level pot a little, and the "SIGNAL" pot until you hear the signal. The output might be loud, because there is minimal feedback. Adjust the external level pot to get the level down to a reasonable loudness.
Raise the level on the "FEEDBACK" pot a little bit, and very slowly. The 20Hz should get quieter. If it gets louder, or motor-boats, you have the feedback signal backwards. If you aren't familiar with motor-boating, it is the sound an audio amp makes when it is oscillating. Its loudness does not match the level setting. The power amp is running full tilt boogie all the time. The feedback circuit has a gain of 6x. It gets loud.
If the 20Hz signal turns into "motor-boating", turn the feedback down. Repeat the two steps until you feel comfortable with the output, or until you can't make any more adjustments without causing motor-boating. That is the point of maximum feedback. You probably want to back off on the feedback a little more.
Now it is down to the "Q" adjustment. To adjust it, you need a sweep signal and a microphone to measure the frequency response of the sub. On the plot, you will probably see a peak in the response at around 146Hz, and a corresponding freak out of the phase. Adjust the "Q" pot a little and run another sweep. The peak should be either larger or smaller. The goal is to adjust the peak out of the frequency response plot by adjusting the Q of the filter. It may take a while, due to the time it takes to run a sweep. There is a limit to the Q adjustment to prevent oscillation of that circuit.
Results
REW was used to measure the feedback response again, this time at the output of the feedback compensator, resulting in the plot to the left. This plot primarily serves to appease curiosity, as the filter's tuning was carried out based on the speaker output curve, with the amplifier being driven by the error signal.
After tuning the filter, there is a noticeable difference in the frequency and phase response compared to the previous response.
Fine Print
Nothing is free. The Dayton SD270A-88 is a nice woofer, but it has a 10" cone. A 10" driver can only move so much air, so it can only get so loud at any given frequency. When you exceed that loudness, the driver gets nonlinear, causing a flat spot on the feedback. The amp will push at full power to try to get the driver to move farther. It makes a very loud click or pop noise as it applies full power to the driver. There is no way around the physics of it. If you need to go lower and/or louder, you need a bigger cone or more of them. That is always true, but with velocity feedback it is more obvious when your woofer gets overextended, since it won't go gracefully nonlinear. The circuit is designed for the SD270A-88, and would need some research to modify for a different driver. The PCB would probably remain the same, but some resistor values will change. Dayton Audio has an SD315A-88 12" DVC subwoofer driver, but no 15" version.
Although harmonic distortion is generally improved by a factor of 2 to 3 above resonance, it is still considerable below resonance. Not as much as trying to get 10Hz out of a 43Hz system without feedback, but bad enough. And it takes power to keep the distortion down when operating below resonance. The feedback has to overcome the cone suspension which favors the resonance of 43Hz, not a 10Hz signal.
1. Radcliff, Clark J., Gogate, Sachin D. (1996) "Velocity Feedback Compensation of Electromechanical Speakers for Acoustic Applications"