Kinergetic Subwoofer Circuit Analysis

[MLVW]

I have a pair of old Kinergetic Subwoofers (SW-100) for my stereo, which are long out of production. The speakers came with a passive control box that includes a low pass filter and circuits to compensate for the mechanical structure of the speakers. This was done in an attempt to eliminate "one-note" bass.

The control box is in short supply on the used market and the guy that use to repair them has long since retired. With DSP now more commonly available, I’d like to try and develop a DSP replacement for the control box as a community service.

My background is physics, so I have some familiarity with circuit design, but the “Mechanical Structure Compensation Circuit”, which contains a complicated OpAmp circuit in the feedback loop is beyond my skill level.

I am wondering if anyone is able/willing to help me analyze that circuit? Alternatively, is there an advanced circuits textbook that could push me along?

What I know from reading the patent filings is that this is a feed-forward designed to compensate for cabinet resonances. So, I think this is somewhat analogous to an RIAA correction for an LP.

I am including a copy of the circuits and signal path from the patent filings. I will need to fiddle with the resisters to tune the circuit, but I need to start with an understanding of the circuit.

Thanks for any help anyone can provide, Mike

 

[danadak]

Post the values of components in the schematic.

 

[Tony Stewart]

Patents obfuscate the real operation in spades and make it sound full of intel. The real intel is in the transfer function and the ratio of every component gives a lot of strings to fiddle.

Without any values or real theory, it's just an oscillator in the center section. In practice it should be a variable f and Q notch filter.

Also, the resistor "B" is useless between 2 voltage sources, one being gnd or 0V.

A better idea is to put a reflective tap on the cone and reflect and IR beam to servo control the cone to the input signal.

 

[crutschow]

A better idea is to put a reflective tap on the cone and reflect and IR beam to servo control the cone to the input signal.

That's an interesting idea, but how does the IR beam convert cone movement into a signal linearly proportional to the cone movement?

I have a sub that uses an extra sense winding on the cone to give a feedback signal proportional to cone movement.

 

[MLVW]

I don't know the values of any of the components. Once I figure out what it is doing, I will build a model and hook up a mic and signal generator to collect data at a variety of frequencies. Hopefully I can use the data to determine the values.

 

[crutschow]

I will build a model and hook up a mic and signal generator to collect data at a variety of frequencies. Hopefully I can use the data to determine the values.

Given the complexity of the circuit, I don't see how you can to do that from those simple measurements.

Look up subwoofer graphic equalizers.
They will likely do all you need to correct the bass response to smooth any resonant peaks.

 

[Tony Stewart]

A parametric equalizer, can be modeled in a DSP.

 

[crutschow]

A parametric equalizer, can be modeled in a DSP.

Of course.
Just wanted to avoid re-inventing the wheel.

 

[Tony Stewart]

That's an interesting idea, but how does the IR beam convert cone movement into a signal linearly proportional to the cone movement?

A wide angle light source will have 2nd order inverse Frisian path loss. A narrow beam will have 1st order inverse path loss. I was thinking this could be made linear for position feedback.
servo optical concept with TIA amplifier

If one can implement an inverse Laplacian on a mic samples of repetitive impulses applied to the Sub, could error correction be used to compensate for wall reflections and room dynamics.

 

[audioguru]

All modern fairly high quality woofer and subwoofer speakers have detailed specs for you to design the enclosure type, size and overall frequency response.

 

[MLVW]

I have a DSP that I use for room mode corrections. This is something different. It is a cabinet compensation that is more along the lines of RIAA correction used on LP's. I am not going do directly convert the analog topology into a digital process, but I am trying to understand in general terms what the circuit is doing. I have a software program (REW) that sends out a SIN sweep and records the speaker output, so I can get data on what the box is doing, but it will be easier to interpret if I know where I am headed based on the circuit.

 

[MLVW]

I should add that the patent filing is silent on the values of all the resistors and capacitors, but as noted, I am not trying to replicate the analog circuit in detail. I will use data to build a parametric model based on the designer's intent of the circuit.

 

[Tony Stewart]

I can make it oscillate in theory by having 180 phase shift with gain > 0 dB and with more than a dozen degrees of freedom make completely different transfer functions, but I can't control each pole or zero independently or make it do what I think is reasonable.

You can play with it here with any part value in real-time such that it affects the response. It is easy to use the wrong value and dominate all the others. ( click output selector) mouse-wheel on RC values , showing 10KHz spectrum only)

1st stage: partial treble cut above RC
2nd stage high-pass+low pass to null pass thru summing junction
3rd stage: treble boost above 1/RC

1st stage is easy: Acoustic load is fundamentally a real R in parallel with L + DCR coil resistance. Negative feedback allows transimpedance for a capacitor to simulate this.


A parametric or graphic equalizer makes more sense to me.

 

[MLVW]

Thanks. In reflecting on your comment, the speaker range is 20-120 hz and I am guessing the circuit must apply a ramp on the amplitude over the specified range. I am not sure of the slope or amplitude of the ramp, but it is clearly more complicated than a lowpass filter. I wondered if there wasn't some sort of software out there. Thanks!

 

[Tony Stewart]

Ramp control can be any+/_ fractional amount of a 1st order 6 dB/octave to a 60 dB notch within a 1/3 octave

 

[crutschow]

I have a software program (REW) that sends out a SIN sweep and records the speaker output, so I can get data on what the box is doing, but it will be easier to interpret if I know where I am headed based on the circuit.

The purpose of the circuit is to correct the speaker frequency response to get as close to a flat output as feasible, so just design the DSP response to do that, based upon your measurements.
I see no reason to try to emulate the analog circuit otherwise, as that's all it's trying to do.

 

[Tony Stewart]

As long as you place the sub in the corner of a room with reflections then the wavelength delay % will be negligible and wall reflections will add.

Remounting the speaker in to a labyrinth tuned duct cabinet can make it more efficient at sub-audible frequencies.

Definitive speakers used a passive rear speaker with added mass to reduce the pressure changes to an internal sealed box with push-pull for front:rear.

Whereas disco's in the 70's in Hull Que. simply used dual 15" subwoofers in each corner of the centre dance stage with 8 Crown DC200's, 1 for sub. pair powering each corner. I found it from the subsonic waves while visiting Ottawa on biz. to be in an old converted Catholic cathedral a few blocks away from the seismic effects. The mid's and treble were broadcast with quadraphonic horizontal wide-beam horns.

The only other next most memorable theatre was "the Us" Festival in San Bernadino, Ca. USA in the 80's that was Steve Wozniak's idea of early Apple fame. The 500 kW of audio power was magnificent with massive arrays of 1m speaker cabinets stacked like two drive-in theatre screens, in stereo, along with the 30 or so rock bands of its time with 3 days of camping on the slope of the mountain with some 1/4 million others.

• DTS 5.1 Surround
• Dolby Digital 5.1 Surround

 

[MLVW]

Tony and Crutschow: Thank you for the input on my project. It has me thinking that what I need to do is put the mic in front of and close to the sub so that I get a near-field transfer function between the input signal and the sound - both with and without the box. If this box does anything at all, I should get a general sense of the correction that is being applied . And, as noted, see what is needed to get a flat transfer function. I suggested that it might be a ramp correction to the frequency response, but I suppose it might also be a band-pass filter. I guess we will see what the data suggest.

As a follow-up for purely educational purpose:

1) what software did you use to model the circuit?

2) 1st stage: you mention this is a treble cut. Is there any distinction with a low-pass filter?

3) "2nd stage high-pass+low pass to null pass thru summing junction." This is the stage that iI am most baffled with. Your description sounds like you think it might be a band-pass filter with yet to be determined coefficients. But, given the complexity of the circuit, wouldn't there be an easier way to design such a circuit?

4) In general, do you have any idea why, as circuit designer, you would put an OpAmp in the feedback loop? This is a bit of an elaboration on the previous question.

It will take me a bit to get everything setup, but I circle back and share the data.

Many thanks again, Mike

 

[MLVW]

One more question about the software: is it possible to look at the A, B, & C taps. I tried to move things around, but that didn't seem to work.

And I think the designers put a HP filter in at around 18 hz, to avoid rattling the thing apart and then a LP filter at around 120 hz, to cut out the high Freq signal. The treble boost must be down in the 20-120 hz range, because the speakers can't handle the higher freq's.

Apparently the guys that started the company did acoustics for the submarine fleet, so it seems they took a novel engineering approach and also apparently a remarkably poor marketing program.

Thanks again, Mike

 

[Tony Stewart]

It is common for negative feedback to use buffered Amps, but many values which you can play with using Falstad's javascript web app, will cause oscillation in the time domain, but Bode plots just go erratic.

http://www.falstad.com/afilter/circuitjs.html?cct=$+1+0.000005+5+50+5+50 %+4+10523.112161612546 a+304+256+400+256+0+15+-15+100000000 r+304+240+256+240+0+33000 w+304+240+304+208+0 r+304+208+400+208+0+100000 w+400+208+400+256+0 g+304+272+304+288+0 O+464+368+512+368+0 a+400+128+304+128+1+15+-15+10000000 g+400+144+400+160+0 r+432+256+432+176+0+100000 w+400+256+432+256+0 r+304+208+304+128+0+12000 r+304+64+416+64+0+4700 w+304+64+304+128+0 w+432+144+432+176+0 w+432+112+432+64+0 w+432+64+416+64+0 r+512+96+512+144+0+560 w+400+96+400+112+0 g+512+144+512+160+0 c+560+96+560+144+0+0.0001+0 w+512+96+560+96+0 w+512+144+560+144+0 c+464+96+464+144+0+1e-9+0 w+464+144+432+144+0 w+432+112+432+144+0 w+464+96+512+96+0 w+464+96+400+96+0 r+160+192+208+192+0+12000 170+112+224+80+224+2+20+4000+5+0.1 g+160+256+160+272+0 w+256+192+256+240+0 r+160+160+256+160+0+470000 w+160+224+160+192+0 r+160+224+112+224+0+100000 a+160+240+256+240+0+15+-15+100000000 c+208+192+256+192+0+3.9e-9+0 w+160+160+160+192+0 w+256+160+256+192+0 S+432+368+400+368+0+0+false+0+4 207+256+240+272+272+0+A 207+400+256+384+288+0+B 207+400+336+384+336+0+A 207+400+352+384+352+0+B w+464+368+432+368+0 a+512+272+608+272+0+15+-15+100000000 r+496+208+608+208+0+100000 w+608+208+608+272+0 w+496+208+496+256+0 r+496+256+432+256+0+120000 r+496+288+432+288+0+100000 g+512+288+512+304+0 w+512+256+496+256+0 w+496+256+496+288+0 w+256+240+256+304+0 w+256+304+432+304+0 w+432+304+432+288+0 207+400+368+384+368+0+C 207+608+272+608+288+0+C a+672+288+768+288+0+15+-15+100000000 g+672+304+672+320+0 r+608+208+672+208+0+120000 c+608+272+672+272+0+3.9e-8+0 r+672+208+768+208+0+5600 w+672+208+672+272+0 w+768+208+768+288+0 w+768+288+784+288+0 207+400+384+384+384+0+D 207+768+288+768+304+0+D This is the long link that shows all the wire connection vectors but the tinyurl was in my previous comment.

I don't know how the design was auto adjusted but a graphic EQ would make a lot more sense to me. You can put the output switch and the input signal to examine each stage and adjust R values with your mouse wheel. and undo ^Z