Motional Feedback Speaker Project - Circa 1981

That's a nice example!

A loudspeaker is spring-mass combination with a certain resonance. If you take away the main ( voltage drive) damper it will go resonate like hell if it has to rely only on his suspension and compressed air (CB) in the cabinet. If you drive a speaker with pure current it will selfdestruct if used in a BR or open baffle.
 
F=BIL
That means:
F=Force on the cone
B=Magnetic fild strength in the gap
I= CURRENT (Not voltage....! )
L=Length of the wire in the magnetic field (B)

In this formula you do not find Voltage. Since the cone is moved by CURRENT!

Ok... but Voltage gives me current yeah?
Yes as long as the impedance of the speaker element is constant.
If you look on an impedance chart you will see that there is no such thing as flat impedance. Add to this that the impedance is changed with excursion and power input. It is changed when the coil is getting hot after some minutes, even eddy currents in the polepiece creates impedance changes. Then I can tell you that the element will generate back EMK that in turn gives a voltage back to your nice amplifier which in turn is picking up it´s feedback signal from the output where your Back EMK is landing...
This voltage is feed to the input stage and messing up with your input signal...
Not a good ide at all, if you ask me....The Ace Bass technology was invented in Sweden back in the early seventies and a company called Audio Pro was founded. Their early subwoofer with current feedback or negative output impedance sonded really powerfull. It had 2x6,5" elements. Sounded like 2x12".
Can't resist... I can't follow them. The musical signal is voltage modulated. The feedback signal is voltage modulated. Of course I can't transfer power without current, but I don't understand that "current drive".
 
@chriscam

You sure know a lot more about MFB in practice than me, so maybe you can shed some light on my ideas.
I have build small and large loudspeakers, I really like your modern interpretation of the old Philips MDF theme. Sure a great sounding speaker. I tried to copy the original, around 1979, when the MFB speaker was available for little money. I see the WAF advantage of MDF, no question, but there is also a place (maybe not a market!) for large speaker systems. Please don´t see MDF limited to space saving, you would throw away it´s main advantage, the improved reproduction quality. Small woofers are nice, but large ones better, you will never change this. Large woofers have lower acceleration while moving identical volumes of air, so correction should be easier.
Today we have digital possibilities like never before, to deal with room resonance problems. If these are consequently solved, the “normally” not audible distortions with low frequencies, get more relevance.


The voice coil is not ideal to get a signal, because it heats up and get´s disturbed when the magnetic field is left with high excursion. Not the best place to get information.
So we look for other ways to get the theoretical best signal from the moving cone. You did not comment on the electret microphone Backes&Müller used for the woofers of their products. They where reverence speakers for decades. They used large speakers and huge volumes, while the electronics in the bass section where really simple. Maybe not the worst idea. Patent´s have run out.
Modern acceleration sensors are pure overkill, as you (me) do not want to go digitally with the correction and fast ones get expensive. Does not make sense as long as an analog correction is faster anyway.
Even piezo sensors are not perfect. Optical has been done, too, but I see no advantage.
So, if we start at the beginning, after ruling out rocket science, I see a very simple way of getting a precise, independent voltage signal that represents the cone movement.
Simply take a small, long high impedance coil, fix it at the dust cap and let a magnetized metal rod that rests on the pole piece, give a stable magnetic field. Or, other way round, as most decent woofer have a center hole in the magnet anyway, fix a piece of (carbon fiber) rod on the dust cab, put a tiny neodymium magnet in the other end and let it move in a small coil, as described.
All parts, even carbon fiber dust caps, are available for the hobbyist at low cost. Have you ever build small structures with carbon fiber parts and epoxy glue? It is amazing how solid with virtually no weight such parts are.
So independent coil and field. Magnetic pick ups for analog records (remember) where build that way. So in the end you get the very best signal you can have, or is there something better than an analog voltage representing the actual movement of the cone, for feeding the correction signal without heavy processing?

Sure, this is so simple, you can´t patent it, because anyone with a technical background will find this solution with common knowledge. But as I see it, there are no really successful patents with loudspeakers anyway, as the big company´s don´t see any pressure to improve their products, as junk sells best. I´m in the market for more than 45 years now and some of the best patented ideas have just been ignored for decades, because old stuff in new clothes still sold.

If you find a problem with correcting bridged amplifiers, because the signal is inconvenient, just take an simple audio transformer to get it ground free. At low frequency even cheap ones are nearly perfect and you have no DC problem. No need to go Lundahl! The 6dB drop can work as advantage, saves a filter. Tried it?
 
Last edited:
@turbowatch,

The quality of a feedback loop depends heavily on the sensor ability to mimic physical motion into an electrical signal, please read the original Philips technical research document which explains in detail why voicecoil mounted piezo based accelerometers outperform any other solution.

Dustcaps tend to sing their own song & suffer from deformation due to non linear / non axial forces. Compared to accelerometers microphones have a very limited bandwidth and are sensitive to picking up third party noise.

The main reason for using small drivers is with cone breakup and it's impact on D3 reduction, large drivers have a tendency to breakup earlier then small ones effectively limiting the servo loop's ability to reduce THD. Doing D3 reductions upto 400hz - a common low note low pass - requires the driver to be clean upto 1200hz.... hence my predisposition to daisy chaining small ones :)

Sorry, I really appreciate your input and don't wanna sound negative but getting a sensor to do what it needs to do isn't easy, been trying to do that since 2012 and only recently have been able to build accelerometers that seem to work fine.
 
...The main reason for using small drivers is with cone breakup and it's impact on D3 reduction, large drivers have a tendency to breakup earlier then small ones effectively limiting the servo loop's ability to reduce THD. Doing D3 reductions upto 400hz - a common low note low pass - requires the driver to be clean upto 1200hz...

A lot of interesting discussion has been added to this thread.

Again, I think the use of small or large drivers and the required bandwidth all depend on the application and the corresponding performance goals. There is no single right answer.

It would be very desirable for the feedback for the woofer in a two-way bookshelf speaker to be functional to about 4 X the crossover frequency. This could be 6 kHz or higher. In contrast, a sound system with a large floor-standing stereo pair may employ a subwoofer with a 50 Hz crossover. The subwoofer would only require feedback to perhaps 300 Hz or so to do a very good job of addressing in-band distortion whilst providing much higher SPL at lower cost than multiple smaller systems.

Or we could build an MFB dream system using an MFB bookshelf speaker pair with multiple MFB subs.... :)
 
Back in Post #12 we modeled the transfer function for the AD80671/MFB in a 25 litre enclosure. I used an online filter tool to make a circuit that provides a similar response. I'll put this in the feedback path to model the speaker response. I think this would be a valid way to model it with the obvious exception that higher frequency nastiness will be present in the real transfer function for the speaker.

741956d1552238743t-motional-feedback-speaker-project-circa-1981-a-ad8067_mfb-pass-filter-model-25-litre-enclosure-jpg
I use a very fast and efficient way to model driver transfer functions. You just need to remember that the impedance of a 159.2 µF capacitor or 159.2 µH inductor is 1 Ohm at 1 kHz (160 µ will introduce an error of 0.3 % only) and to use a hand calculator to do the operations shown on my drawing. The resonance frequency is set by the same numeric values of C and F and the damping by the numeric value of the resistor. Values are unrealistic if looked as filters but the aim is to get fast and easily modifiable simulations. I invented this method fifteen years ago when studying transforms and servoed drivers.


attachment.php
 

Attachments

  • forr_Spkr f and Q transfer function.png
    forr_Spkr f and Q transfer function.png
    10 KB · Views: 493
Last edited:
@chriscam

I got this Philips papers 40 years ago, together with a lot of schematics and some parts, when a friend of my parents, who worked at Valvo/ Hamburg/Germany, cleaned out his lab. One day he dropped a huge box and suggested, I could build some MDF speakers from it. I Think Phillips had lost interest in further development of MDF at that time. A lot in speaker chassis tech has changed till then, and so some passages are much less important than in 196x. I have found documents as far back as 1952 with this stuff, if I remember right, even from some German tech´s.
Before the internet changed our life, I was a frequent customer at public library's and read anything available about loudspeakers, even had frequent fliers buy me international magazines. When you are young, you always think you can build something better than there is. If you are clever too, you try to learn what is already there, before you start experimenting.
As I have been building loudspeakers ever since, I know about cones breaking up and other nasty stuff, quite well.
See, I don´t want to rate any system in better and worse, I´m quite sure more that more than one will have comparable good results. Also, something that works for one given speaker might not be best for another. I (we?) will never have build speakers especially for MFB. We will use something that fit´s the basic requirements and modify it for correction. You ruled out the microphone type, I think the idea is still interesting. Maybe better for larger cones than for your tiny WAF woofers? By the way, I have no commercial interest, like most here, but build and experiment for my self. I´m the type of guy who first does things theoretical, wasting time and money on trying funny ideas is not my desire. That´s why I ask here, not because I want to copy any PirateLogic stuff. They are safe, because our far east copy cat´s don´t know anything about what sounds good, you are only in danger of getting know how robbed with cheap ****.

Do you have covered the basic´s of B&M microphone corrected woofers? Maybe have a look here.
http://thomas-gehrke-online.de/attachments/File/AFB_Mikrofon_Gegenkopplung_Funkschau_25_1984.pdf It´s German but the schematics don´t need explanation for you.
The principle had much more success for a longer time than the Philips MDF and their speakers are still considered very high end.

See, there is not much difference in the electronics, so if you can do Piezo MFB, Microphone or even a pick up coil, separated from the voice coil, will not be something new.
With Piezo you run into just the same problems as with other sensors, look at Piratelogic what they did to prevent flexing of the sensor. (-:
You can do the same to tame a coil or any other sensor.
Maybe just think for a second why for turntables piezo are always worse than magnetic pick ups and moving coil rules the high end? It might be just a question of implicating it in the right way?
 
Back when the Phillips small MFB speakers were introduced, I was a dealer. Had a request to provide PA for a high-school gymnastics event, and thought that these would work.



I took 4 pairs of the MFB speakers, a pair at each corner of the gym, and FILLED the place with quite decent sound, even for the after-performance clean-up, when some of the kids brought music of their own to play... A very successful event from the sound standpoint. No, we didn't fry any of them.
 
...The only “feed back” comes from the voice coil/ series resistor, the other measures are called feed forward, as far as I know, as they do not react to anything the cone/ voice coil will do.

Maybe we should first see what the basic current drive really does in a construction, then add the rest and look at it again. ...
In the end the resistor is not the best way at all. Any real signal sensor should do better, I think about glueing a microphone to the cone for correction, not more expensive in the DIYS world.
Big debate about using the VC to produce its own feedback using Back-EMF versus the various issues using sensors.

A more sophisticated way of using the VC is by locating it in a Wheatstone bridge and feeding the cross-connection into the two inputs of an op-amp.

With a bridge, the theory is that you balance the bridge for the stationary (blocked) impedance of the VC which cancels the resistance and the inductance of the VC. So the out-of-balance feedback voltage which is generated is mostly Back-EMF. In practice, not too much different than Turbowatch2's current feedback with a single series resistor in the minus lead.

How could anybody possibly imagine they could get quality sound from a system as inferior in design concept as a subwoofer without feedback? I really love rscamp's scope traces. Folks should look closely at the over-shoot at the end of the tone-burst with and without MFB.

B.
 
Last edited:
The "Motional Feedback Speaker - Circa 1981" which hasn't been operated since 2000 is working!

I didn't want to rip a power amplifier out of my sound system so I used an old receiver as the amplifier. Being an integrated amplifier, the volume (amplifier gain) and tone controls are active so first I made sure the transfer function was okay by defeating the tone controls and such and checking with REW. This confirmed there were no amplitude or phase anomalies and left the volume control on the receiver as essentially the feedback loop gain adjustment.

I replaced the 1-1/2A fuses with 1A fuses for a little extra safety margin, plugged in the accelerometer signal to the MFB controller and slowly turned up the amplifier gain. Eventually when exciting the diaphragm with the fingers I began to see the familiar oscillation around 10 Hz. MFB is in da house.

Here's the poor little 8" driver trying to reproduce 16 Hz. The video frame rate makes it look like the amplitude is pulsing but in reality it is steady.
 

Attachments

  • MFB First Test 16 Hz.mp4
    4.1 MB
These charts compare the Output to the amplifier with and without feedback.

In the amplitude response from REW, I didn't match the input amplitudes so these can't be directly compared, but note the relative reduction in Output with feedback where the speaker response would be greatest around 100 Hz and the indication of resonance issues, particularly close to 4 kHz.

In the scope/analyzer view from VA, the loudspeaker was driven with a fairly high amplitude 20 Hz signal. I would guess p-p motion was in the order of 7mm or so. A small change in waveform shape is visible when feedback is applied and this is manifested in the ADDITION of 3rd harmonic energy applied to correct the 3rd harmonic distortion in the driver. I probably could have set these views up better but it is the first time using this software.
 

Attachments

  • MFB Output Amplitude Response With and Without Feedback.jpg
    MFB Output Amplitude Response With and Without Feedback.jpg
    275.3 KB · Views: 317
  • Scope of 20 Hz Output to Amplifier without Feedback.JPG
    Scope of 20 Hz Output to Amplifier without Feedback.JPG
    418.8 KB · Views: 330
  • Scope of 20 Hz Output to Amplifier with Feedback.JPG
    Scope of 20 Hz Output to Amplifier with Feedback.JPG
    419.6 KB · Views: 314
PCF without global NFB

A few years back I was playing around with a simple single end class A amplifier with a positive current feedback (negative output impedance).

My small class A amp could really bring out a truly impressive thunderous bass from my large back loaded horn or my much smaller offset driver quarter wave pipe.

The distortion below Fs vanished with PCF (Positive Current Feedback) enabled.

Lots of fun, and a really nice and cheap way to improve the bass reproduction substantially.
 
PCF without global NFB

A few years back I was playing around with a simple single end class A amplifier with a positive current feedback (negative output impedance).... Lots of fun, and a really nice and cheap way to improve the bass reproduction substantially.
This is a better home for your motional feedback thread than the electronics forum, esp the explanation in post #6 in your link, and the nice explanation in the further ancient link you provided, attached here:

A New Look at Positive Current Feedback, November 1957 Radio & TV News -
RF Cafe


In the early 50's, folks were implementing "variable damping" (such as an amp from Fisher I had). That is really your PCF mixed in with the necessary familiar negative voltage feedback, as you do as well.

But that PCF is really motional feedback based on the back-EMF from the voice coil, not something disembodied inside the amp, as some new readers might think.

Werner's JASA article in 1957 (he was at RCA labs and must have had a sub-text mission to patent stuff) took it a step further by getting the back-EMF using a Wheatstone Bridge which enabled him to cancel-out driver inductance, resistance, etc, by balancing the bridge against those elements.

Yes, very odd to test an amp with negative output impedance. As the test resistor goes higher in value, the output voltage from the amp goes down. QED.

It should be remembered that with your approach (and MFB) the amp's managerial control action takes place both for gross erroneous motions (like killing the speaker's big resonance and enhancing a sealed speaker's reluctance to move at the very low frequencies below the resonance*) but also for counter-acting every tiny erroneous bit of distortion, harmonic or otherwise.

MFB with an accelerometer behaves the same way, viewed metaphorically.

Very glad you've posted in this sub-forum. Looks like MFB is a hot topic, perhaps finally displacing the incidence of adherents of sixth-order bandpass boxes who try to shape their Rice-Kellogg driver output using smart plywood instead of electronic smarts.

B.
* it doesn't make any sense to apply MFB to BR and some other boxes
 
Last edited:
it doesn't make any sense to apply MFB to BR and some other boxes

It never makes sense to use a BR (imho).

Negative output impedance works really nice with back loaded horns, back loaded offset driver quarter wave pipes and high order tapped quarter wave alignments like the ROAR.

You just have to design them to work with a negative output impedance from the beginning.

negut-ROAR.JPG