Motional Feedback Dipole Subwoofer with a current amplifier?

A new thread is created:

MFB and CFB using Piratelogics Electronic​


After have been experimenting a lot with MFB in the 90ties i went to dipole subwoofer and QUAD electrostatics.
Now i want to add accelerometer motional feedback to the double Dipole woofers.

And after reading some papers
http://rmsacoustics.nl/papers/whitepaperMFBtheory.pdf
http://rmsacoustics.nl/papers/whitepaperMFBdesign.pdf
Read BOTH!

I realize that i also need current feedback power amplifiers.
There is other threads about the MFB itself. Want to know if anyone tried or want to try the combination of:
1. Accelerometer feedback
2. Current feedback amplifier
3. Dipole long throw subwoofer
(4. Chip Amp) not neccesarily.

Signal conditioning:
15 OP amps/channel is needed since im stupid... :-( ?
3 OPs 6order low cut 16 Hz filter
2 OPs for Linkwitz transform Q value compensation
1 OPs for Dipole correction
2 OPs for LP filter at 200Hz
1 OP amp for 200-300Hz notch filter (extending the phase margin) reed the papers linked in the first post.
1 OP for current feedback loop
1 Op for impedance conversion of the accelerometer signal (buffer) on the speaker frame.
3 OPs for acceleration feedback signal conditioning.
1 OP for signal blender between signal Input and acceleration feedback.
2oct 2020. I ended up with keeping my MTAudio active crossover, includes dipole compensation, and for MFB section i use one double 5532 and some FET transistors/channel!

I will use 3*Dayton IB 15" drivers per side. They work in free air.
And i will use ACM-01 as accelerometer, and only one accelerometer/ channel wich means that one out of three woofers will have the accelerometer.
The other two elements will have added mass matching the accelerometer setup.

And each element will have it´s own current feedback chip amp. This will guarantee same excursion on each element!!
Or i could build a more powerful amp and put each element i series. Also this will guarantee same current (from the current feedback amp) into each element.

But thats not my problem now. I want to know how to reduce the nr of OPamps.
20 dec 2019 Update:
I added a magnet to the polepiece and gained 1.1 dB in sensitivity that means that the 200w chip amplifier get an extra "boost" of 60 watt to 260watt (for the same sound pressure), to do the calculation you can use this calculator: SPL Calculator i know you can´t hear 1.1dB increae in sound pressure but my brain knows that i´m better off then original! :)

2 oct 2020, Almost there... see distortion figures #27
Specially thanks to: rscamp, chriscam, bolserst and joe rasmussen
 
Last edited:
Current drive is not necessary, but from a mechatronic view it is better.

Using only one sensor will not work if the drivers share the same, closed box. If drivers share the same compartment they will exhibit as one driver with the same parameters. This is not the case in a open baffle!
 
Using only one sensor will not work if the drivers share the same, closed box.

If drivers share the same compartment they will exhibit as one driver with the same parameters.

So if the drivers share the same box (compartment) and they behave as one driver in there , why using one sensor will not work? I've tried to conclude but I failed. :)
 
I believe that with OB one sensor will work ok, because all drivers share same signal/amplifier channel and have similar air load (=striction only). But only a comparison test will confirm that, without that this is just speculation.

The benefit will come mostly with high output/excursion. How high spl is needed?
 
I have double Dayton IB15" drivers now but will add one up, to three, mostly to keep the excursion down.
15 OP amps/channel is needed since im stupid... :-( ?
3 OPs 6order low cut 16 Hz filter
2 OPs for Linkwitz transform Q value compensation
1 OPs for Dipole correction
2 OPs for LP filter at 200Hz
1 OP amp for 200-300Hz notch filter (extending the phase margin) reed the papers linked in the first post.
1 OP for current feedback loop
1 Op for impedance conversion of the accelerometer signal (buffer) on the speaker frame.
3 OPs for acceleration feedback signal conditioning.
1 OP for signal blender between signal Input and acceleration feedback.
If I was clever enough I would know how to reduce the nr of amps to 6 by combining them.
This nightmere is maybe a thread in the amplifier section.
 
Last edited:
Whether the "hard work" is done in the analogue or digital domain, I recommend seeking out information on David Birt's ingenious self-balancing bridge. It circumvents the need for an accelerometer and comes with a number of other advantages, such as the means to monitor/limit genuine rms power, for example.
 
"David Birt" - can anybody provide some links. Sounds like the Wheatstone Bridge MFB that I used for decades based on Richard Werner's 1957 JASA article.

Yes, BL may not be perfect, but I think MFBs greatest benefit isn't reducing HD (which is terrible very low in freq but maybe never too hearable). Using VC MFB does provide correction for transients and driver resonance. And that's more striking to the ear. And a whole lot simpler than making accelerometers work right (incl. their noises).

I was skeptical of using MFB on one driver to control all the clones. But the evidence I;ve read seems to be that it's OK.

Sure, leave the EQ (and much else, such as high-pass to protect your OB) to DSP. No way to guess what you'll need once in your room with walls behind. I'm one of the millions of deliriously happy users of the Behringer DCX2496, all-singing all-dancing. Pay no attention to the critics.

B.
 
Last edited:
Nice but it assumes Bl to be linear where it certainly isn't. Bl nonlinearity is a primary distortion source in loudspeakers in a bass range.

Pushing a heavy cone suspended using a large rubber surround and starched cotton spider, Bl non-linearity doesn't seem the big issue to me since it can't be hard to make the mechanical aspects of the coil and gap pretty uniform.

Subs show pretty tolerable distortion north of maybe 35 Hz but HD gets quite bad below. While cone excursion is increasing at low frees, the shape of that THD curve does not suggest that Bl non-linearity is what is causing the great increase below 35 Hz since the same Bl non-linearity would be doing its evil influence at all frequencies to the same extent.

The engineers at Rythmik, linked by Juhazi in the post above, make a pretty coherent case in defence of their second VC approach which depends on Bl linearity. Sony once made a driver with a feedback winding. While cannibalizing a DVC driver to get MFB doesn't compute too well, using a high-turns special winding does make a lot of sense (as does the bridge approach, at least to me).

B.
 
Last edited:
"David Birt" - can anybody provide some links. Sounds like the Wheatstone Bridge MFB that I used for decades based on Richard Werner's 1957 JASA article.

Yes, BL may not be perfect, but I think MFBs greatest benefit isn't reducing HD (which is terrible very low in freq but maybe never too hearable). Using VC MFB does provide correction for transients and driver resonance. And that's more striking to the ear. And a whole lot simpler than making accelerometers work right (incl. their noises).

I was skeptical of using MFB on one driver to control all the clones. But the evidence I;ve read seems to be that it's OK.

Sure, leave the EQ (and much else, such as high-pass to protect your OB) to DSP. No way to guess what you'll need once in your room with walls behind. I'm one of the millions of deliriously happy users of the Behringer DCX2496, all-singing all-dancing. Pay no attention to the critics.

B.

David Birt was a BBC engineer and I believe his invention was granted two patents. There is other discussion on this forum that mentions a commercial implementation too.

His self-balancing bridge is indeed a Wheatstone bridge, but with an additional dc servo that allows monitoring and compensating for the changing coil resistance that otherwise prevents simpler bridges from being used effectively.

And yes, displacement-dependent Bl distortions are still evident, but they can be compensated in the digital domain if required. By chance, David Birt also penned a JAES paper in which he describes a somewhat "DIY-orientated" approach to this compensation that is different from other published methods employing "less fun" non-linear mathematics.

One word of caution, however, is that the distortions due to eddy currents in the motor structure are reflected in the coil impedance too: They can become more evident if the operating bandwidth is increased too much. If the system implements current drive, thereby obviating these coil impedance effects, then the bridge-derived feedback can serve to add them back into the transfer function.

But I whole-heartedly agree that the efforts made to implement MFB can be worth it. Although I would not recommend using one feedback signal to control several drivers, even if they share the same acoustic space: Manufacturing tolerances are such that the chance of resonances being sufficiently well-matched is small.
 
Although I would not recommend using one feedback signal to control several drivers, even if they share the same acoustic space: Manufacturing tolerances are such that the chance of resonances being sufficiently well-matched is small.

Why not? The Fusion mfb controler Rob ( Munnig Schmidt) and I designed can be configered for two woofers in a impuls compensated sub. Rob has several ic subs running with one sensor....
 
Why not? The Fusion mfb controler Rob ( Munnig Schmidt) and I designed can be configered for two woofers in a impuls compensated sub. Rob has several ic subs running with one sensor....

Unless the two (or more) drivers are tightly matched, their power input will not divide equally. The issue will be most obvious around the nominal resonant frequency, where the actual resonances will exhibit a spread in their centre values. The notion of two or more drivers moving in unison in a single cabinet, for example, is just an approximation, the usefulness of which is exceeded in an example such as this.
 
Unless the two (or more) drivers are tightly matched, their power input will not divide equally. The issue will be most obvious around the nominal resonant frequency, where the actual resonances will exhibit a spread in their centre values. The notion of two or more drivers moving in unison in a single cabinet, for example, is just an approximation, the usefulness of which is exceeded in an example such as this.

I don't think you know who RMS is:

Prof.ir. Robert Munnig Schmidt - High Tech Institute

LS1s-dmf | GrimmAudio
 
Nice discussion, thanks. Some thoughts:

1. I believe Infinity made a multi-driver single-accelerometer sub; that and other people seem to find it is OK even if not ideal. Sounds inelegant to me. Even with drivers shifty in exact resonance to one another, still (as with other imperfections of any of the MFB systems.. such as non-pistonic cones) it doesn't matter as much as the benefits help.

2. When I was young, nobody fussed about hot VCs and rising resistance. As far as MFB goes, the speed of the heat/cool cycling is too slow and the feedback loop band is too fast for DC offset to be a bother, I'd guess. I have trouble conceptualizing why I need to conceptualize any hot copper problem with or without MFB.*

3. Poor David Birt, lived before the age of Google searches (likewise for most of my own non-audio work).

B.
*aside from speaker fires
 
Nice discussion, thanks. Some thoughts:

2. When I was young, nobody fussed about hot VCs and rising resistance. As far as MFB goes, the speed of the heat/cool cycling is too slow and the feedback loop band is too fast for DC offset to be a bother, I'd guess. I have trouble conceptualizing why I need to conceptualize any hot copper problem with or without MFB.*

3. Poor David Birt, lived before the age of Google searches (likewise for most of my own non-audio work).

B.
*aside from speaker fires

The issue of coil resistance is that any changes mean the bridge will no longer be balanced and therefore the feedback will be in error. The thermal time constant is irrelevant.

The issue of the dealing with thermal modulation is a different matter and often misappropriated. Thermal modulation in a single driver is inaudible because of the long time constant involved. But in a multi-way system it can act to change the frequency balance as the sensitivity of a band (usually the bass) changes relative to the others.

I will endeavour to find some references for the ingenious Mr. Birt...