Commercial motional feedback woofer available sort of

Even if it's of no use to add this..
As spectator, it seems me that's bad faith from Berento.
You already know it, but...
Ear does not read graph, moreover graph formated to look good for eyes, formating that does not mean anything for ear and brain.
If it were true, classic tube amp would be the worst amp possible, and classic guitar speaker a non sense.
 
Last edited:
Ear does not read graph..

If you are educated in the history of audio engineering, you know about the historic back-and-forth between those seeking electronic test methods and those sure the methods fail to capture subtle sound qualities.

Not sure where the battle stands today - it takes historic perspective to know.

But I do know that the improvements brought by motional feedback in transient response aren't easy to measure electronically. That's an issue I face right now in re-starting my MF R&D.

it is clear as day the improvement you get with MF on impulse inputs. Goes from "clunk" to "click". Anyone can hear that and it is very clear to the eye on tone burst oscilloscope images.

But some of us really want to be able to quantify it too.

Ben
 
it is clear as day the improvement you get with MF on impulse inputs. Goes from "clunk" to "click". Anyone can hear that and it is very clear to the eye on tone burst oscilloscope images.
I'd say it's not clear at all once both concepts have the identical transfer function (near field dustcap measurement giving 100% identical response). If they don't, you are hearing/seeing the result of different transfer functions.
Motional Feeback only does influence the "fine print" (distortion) and the overload behaviour, and (partly) the effects of the system being excited from the "outside world" (response to external sound field), the latter depends heavily on what type of sensor being used.
 
Motional Feeback only does influence the "fine print" (distortion) and the overload behaviour, and (partly) the effects of the system being excited from the "outside world" (response to external sound field)...

You list some stuff MF corrects. Is your point that it does not correct some other stuff? What other stuff?

I can not disagree with you less that all devices with identical transfer functions and which are in all other respects identical as well, should sound the same most of the time.

Ben
 
it is clear as day the improvement you get with MF on impulse inputs. Goes from "clunk" to "click".
I designed and built a prototype motional subwoofer system many years ago. It had dramatically improved transient response, as you said. Also dramatically flattened frequency response, close-miked.

Interestingly, the improved transient response resulted in very underdamped (too "tight") drum sounds on some recordings. I think the original recordings were made with too many blankets stuffed into the kick-drum: the recording engineers were not using monitor speakers with motional feedback, so whatever monitors they were using were adding "boom". They compensated by damping the heck out of the drumset.

But some of us really want to be able to quantify it too.
With motional feedback operating, I can tell you I measured (using a microphone) an 18 dB reduction in 3rd harmonic distortion of the emitted sound at some frequencies, compared to the same speaker operating open-loop at the same SPL and frequency. That was pretty spectacular!

I never tried measuring transient response, but close-micing the woofer with a low-frequency square wave input signal would probably prove very revealing.

IMO, attempts to get the feedback from the voice coil itself are doomed to failure; the voice-coil motor is quite nonlinear at high power and low frequencies, just when you need the benefits of MF most.

So, under these conditions, the signal from the voice coil (which is halfway out of the magnetic gap, and moving through a very non-uniform magnetic field on every peak) no longer represents the velocity of the cone. If the feedback signal itself is corrupted, all the intended benefits of MF are lost.

For the same reason, I am skeptical of the dual voice coil approach as well, because the close proximity of the two coils creates a transformer; the output signal from the motion-sensing coil is corrupted by the current flow in the main voice coil (which does not represent actual cone movement, only drive current from the amplifier.)

I did have success with a home-brewed piezo accelerometer mounted right on the top edge of the voice coil former.

That was a while ago. Today there is probably a suitable MEMS accelerometer in tiny SMD format available off the shelf.

-Gnobuddy
 
Gnobuddy - couldn't be more wisdom packed into fewer words. Thanks.

I'm about ready to compare all those systems including a modern accelerometer I bought (and even just plain old current sensing resistor in series with driver, the simplest method of all).

Which is why I am stalled puzzling over the best quantitative measures to permit comparisons of the methods.

BTW, one other special-device method is a driver with a custom-wound sensor coil and personal magnet system separate from the driver coil.

Ben
 
Thanks for the kind words!

...even just plain old current sensing resistor in series with driver
IMO this is yet another method that does not work when you need it most. Let me explain why:

I think the idea behind the method is that the current through the voice coil is proportional to the force acting on the speaker cone. In principle, then, since f=ma, you know "f", therefore you know "a", and the acceleration of the cone is proportional to the sound pressure level.

But when the coil is displaced well to one side or the other, and moving through a non-uniform magnetic field, the force generated per amp of voice coil current is reduced. So once again, this feedback method fails at large voice coil excursions.

There is a second, equally troubling problem. At low frequencies, near and below the speakers resonant frequency, there is also the restoring force of the spider and surround to deal with. These forces are highly nonlinear, and they use up some of the "push" from the voice coil, so the speaker cone never sees the full force. So f=ma is modified to
(fcoil - fspider - fsurround) = ma; unfortunately for us, we don't know the exact nature of fspider or fsurround, so once again, our feedback signal is corrupted, and unable to provide the true benefits of motional feedback.

Which is why I am stalled puzzling over the best quantitative measures to permit comparisons of the methods.
If you place a microphone within a few millimetres of the speaker cone, you get an accurate indication of what the speaker itself is doing at low frequencies. (i.e., acoustic wavelength much longer than speaker diameter.)

Using this method, I was able to measure the acoustic frequency response with and without motional feedback.

I also measured the electrical frequency response through the electronics, including the sensor on the speaker. This is a necessary preliminary step before applying motional feedback - you have to find out how much gain and phase margin you have, and ensure you have negative (not positive!) feedback before you close the loop. Otherwise, the whole system bursts into full-power oscillations; likely to be highly destructive to your ears, the speaker, and maybe the power amplifier too!

BTW, one other special-device method is a driver with a custom-wound sensor coil and personal magnet system separate from the driver coil.
Ben
I've never encountered that system outside of an old book, but if the coils are within an inch of each other (which is likely), there will still be significant transformer coupling between them. Remember those old phone pickup-coils? You wound a few thousand turns of thin copper wire into a doughnut, wired the ends to a low-power audio amplifier, and you could listen to or record your telephone conversations by simply placing the coil near the handset.

-Gnobuddy
 
but if the coils are within an inch of each other (which is likely), there will still be significant transformer coupling between them

Sony SA-EX100 has a woofer designed to meet your specification. The sensor "voice coil" hangs off the back end.

A smallish woofer in a small box. As is with Philips, a big manufacturer trying to get a quart of sound from a pint-sized driver using MF. Pity.

B.
 
Sony SA-EX100 has a woofer designed to meet your specification. The sensor "voice coil" hangs off the back end.

A smallish woofer in a small box. As is with Philips, a big manufacturer trying to get a quart of sound from a pint-sized driver using MF. Pity.

B.
Ben, perhaps Sony used a Voice-coil as a sensor (don't know about that), but Philips used a ceramic "accelerometer-like set-up" back then... And as far as my listening experience goes, they succeeded in reaching their objectives, quite well.

Besides, Philips not only used the system in small systems. They also produced a more professional version intended for studio use, the 545. The 545 used a 12" MFB woofer and managed 20Hz flat. Sounded quite well, if I may say so?
 
IMO this is yet another method that does not work when you need it most. Let me explain why...

Without disputing the good if cautionary notions of Gnobuddy (a rare modest name on the anonymous web), my personal agenda is to encourage everybody to research MF, the last frontier in audio quality.

Not much to add about feedback sensors external to the driver with modern accelerometers looking like the best bet these days.

But in noting the many shortcomings of voice-coil sensing, ideally using the bridge method, Gnobuddy is emphasizing the limitations rather than the monumental benefits.

Even using just the crude series resistor sensing, you are capturing a large amount of bad stuff that doesn't correspond to the audio input signal: freq deviations around resonances, overshoot and undershoot, distortion, etc.

True, the feedback signal is generated by the same voice-coil and magnet structure that is imperfect* (and true that beyond extremes of motion you are in trouble as always), but over all, beneficial.

Ben
*I'd like to see a comparison of the precision of a modern accelerometer glued to a cardboard dust cap (or coil former, as Gnobuddy did it) compared to the voice-coil irregularities of a quality woofer
 
Last edited:
Even using just the crude series resistor sensing, you are capturing a large amount of bad stuff that doesn't correspond to the audio input signal: freq deviations around resonances, overshoot and undershoot, distortion, etc.
Ben, using a resistor as sensor also captures the inherent L and C characteristics of the speaker you are trying to control. These "qualities" are also depending on the enclosure and the individual production characteristics of the unit. No two units are equal!
I regard the use of a resistor as a sensor as a basic design flaw.

In the meantime I also had the opportunity to have a glance at the IPAL system. Basically it comes down to a microphone used as the feedback sensor. It will probably work fine, provided the distance between unit and sensor are small enough, for that distance determines the cut-off frequency and phase-shift between unit and sensor, also some time-delay has to be provided for. I guess the upper frequency of IPAL could be lower than with an accelerometer type MF system.
 
Ben, using a resistor as sensor also captures the inherent L and C characteristics of the speaker you are trying to control.


In the meantime I also had the opportunity to have a glance at the IPAL system. Basically it comes down to a microphone used as the feedback sensor.

Using the Wheatstone Bridge method, you can balance-out L and C. But even with a resistor, their dependency on freq doesn't matter. My first experiments were with a bridge but later ran MF for a few decades just using a resistor. Apropos Gnobuddy, major distortions at resonance and overshooting can be addressed even with low feedback ratios and otherwise less-than-ideal circuits that only make small progress on distortion. And that's because today's Rice-Kellogg subs make such gross distortions.

The patentable aspect of IPAL is differential pressure... they were too cute to just call it a mic! Seems to me, they must be using their microprocessor to correct for the weird and wonderful changes of pressure inside the box ("differential") or maybe just ignoring the pressure inside the box. Dunno.

About how close a mic needs to be to the driver, you are right, the distances are small and get smaller with increasing freq. In fact, if you do the math, you are in big trouble with phase at quite small distances.

Ben
 
Last edited:
Hi esgigt,

Post #91: "...the IPAL system. Basically it comes down to a microphone used as the feedback sensor..."

As you well know IPAL's "DPC-Differential Pressure Control" is a bit more complicated than that, and none of this would work without the "specifically designed DSP core". My guess is that the designers picked the differential pressure transducer because of cost, availability and reliability. The claimed system results are simply not achievable without DSP, so that to me is the heart of the system, with the feedback mechanism(s) being secondary.

Regards,
 
Hi esgigt,

Post #91: "...the IPAL system. Basically it comes down to a microphone used as the feedback sensor..."

As you well know IPAL's "DPC-Differential Pressure Control" is a bit more complicated than that, and none of this would work without the "specifically designed DSP core". My guess is that the designers picked the differential pressure transducer because of cost, availability and reliability. The claimed system results are simply not achievable without DSP, so that to me is the heart of the system, with the feedback mechanism(s) being secondary.

Regards,
Come on... don't let them fool you with that mumbo-jumbo... If you analyse the provided documentation the system only comes down to a pressure sensor (microphone) and some processing electronics to avoid problems with phase differences and time-delay..

And I do NOT guess... I read the paper ;)
 
BTW, one other special-device method is a driver with a custom-wound sensor coil and personal magnet system separate from the driver coil.
Something like US patent number 5,493,620 by Robert E. Pulfrey (1996)
I think there are many others similar.
Stanislaw Drozdowski used an Hall effect sensor, patent number 4,821,328 (1989).

Considering the acceleration servos, I think that the success can only be obtained using dedicated drivers with the sensor already mounted at the manufacturing process (some Philips have been sold for the DIY market).
As far as I see, gluing an accelerator on the cone as done my some diyers does not give the promised results.
 
Last edited:
About how close a mic needs to be to the driver, you are right, the distances are small and get smaller with increasing freq. In fact, if you do the math, you are in big trouble with phase at quite small distances.
Ben,

If you (or I) do the math, you find that one wavelength at 100 Hz (the typical upper range of an IPAL, or most any "real" sub woofer) is 11.3 feet (3.47 meters) long, a bit over 11 ms (milliseconds) time duration. The IPAL processors have only 10us (microseconds) latency, less than one millionth of the duration of that upper frequency's wavelength. Obviously, 10us is far less than one millionth of the duration of frequencies lower than 100 Hz, so the IPAL system has no trouble whatsoever dealing with the phase duration encountered in it's typical pass-band bandwidth.

As far as distance of the mic to the driver, since the speed of sound is constant regardless of frequency, getting it closer simply increases the signal to noise ratio at the inverse distance ratio, each halving of distance resulting in a 6 dB increase in level. That said, in the very near field around the perimeter of a driver, the inverse distance rule does not apply, so there is plenty of latitude afforded in the placement of the pick up transducer to accommodate the location where it best serves the enclosure design it is tuned for.

Those details, and the knowledge base required for properly tuning the IPAL as a complete "bulletproof" system are another reason why it is only offered to OEM, so consumers can purchase complete systems that already have all the "bugs" worked out. As an example, a large dance club installed an early run of IPAL subs, were very pleased with the performance, but the drivers started burning out after a short period of time, within a month, IIRC. The manufacturer took back the burnt units, re-coned the drivers, and re-tuned the system to withstand the extreme abuse some forms of dance music can present, for instance signals with under 3 dB crest factor, more average power than in a sine wave are common to some genres! The repair service was done at no charge to the customer, and the manufacturer also provided loaner cabinets so the club had no lack of bottom end while the adjustments were made.

Pushing a system to the "bleeding edge" can be an expensive proposition, obviously it is in the best interest of a manufacturer to "leave some meat on the table" rather than give their customers "enough rope to hang themselves" ;^) .

Art
 
Considering the acceleration servos, I think that the success can only be obtained using dedicated drivers with the sensor already mounted at the manufacturing process (some Philips have been sold for the DIY market).
Indeed the best option is to mount the sensors during manufacturing. But have you ever seen such a Philips unit on the inside? It's a quite simple laid in unit in the voicecoil former, which also can easily be added afterwards, provided one can easily remove the dustcap and some other requirements are met:
- no compression due to the sensor
- no physical contact possible between the sensor and the magnetic core.

@weltersys: the speed of sound is influenced by airpressure and moisture... ok, the system will hardly be affected by those factors. As far as time delay is concerned, the system must take that time in account to be able to function properly. Remember it is a feedback system and those systems can only perform optimal when timing is perfect, otherwise all kinds of unwanted side-effects are created.
 
Last edited:
Hi esgigt,

Post #94: "...I do NOT guess... I read the paper..."

No problem here.

I'm pretty sure that calling the "differential pressure transducer" a microphone can lead to confusion; e.g.: in Post #92 bentoronto says: "...The patentable aspect of IPAL is differential pressure... they were too cute to just call it a mic! Seems to me, they must be using their microprocessor to correct for the weird and wonderful changes of pressure inside the box ("differential") or maybe just ignoring the pressure inside the box. Dunno."

As an aside, I found a nice masters thesis on "Optical Position Sensors with Applications in Servo Feedback Subwoofer Control", by F. Antonio Medrano.

https://www.mat.ucsb.edu/Masters/medrano_MASTERS2009.pdf

That's another way to get a feedback signal.

Regards,
 
Hi esgigt,

Post #94: "...I do NOT guess... I read the paper..."

No problem here.
....
As an aside, I found a nice masters thesis on "Optical Position Sensors with Applications in Servo Feedback Subwoofer Control", by F. Antonio Medrano.

https://www.mat.ucsb.edu/Masters/medrano_MASTERS2009.pdf

That's another way to get a feedback signal.

Regards,
Thanks...
... also for providing the paper. I've heard several times of optical feedback, but this paper gives a nice insight of how such a system could be implemented. It appears less complicated than I imagined.

I'm pretty sure that calling the "differential pressure transducer" a microphone can lead to confusion; e.g.: in Post #92 bentoronto says: "...The patentable aspect of IPAL is differential pressure... they were too cute to just call it a mic! Seems to me, they must be using their microprocessor to correct for the weird and wonderful changes of pressure inside the box ("differential") or maybe just ignoring the pressure inside the box. Dunno."
Well, how one wants to describe the sensor... to me any type of air-pressure sensor is applicable as a mic... I don't care for fancy names ;)
 
@weltersys: the speed of sound is influenced by airpressure and moisture... ok, the system will hardly be affected by those factors. As far as time delay is concerned, the system must take that time in account to be able to function properly. Remember it is a feedback system and those systems can only perform optimal when timing is perfect, otherwise all kinds of unwanted side-effects are created.

You are quite right, weltersys seems to know more about the speed of sound than he/she knows much about feedback.

In the purest case, we are comparing the signal INTO the driver with something a ways downstream. And we are not talking about 100-foot waves at 10 Hz but about small fractions of a cycle at 300+ Hz.

The reason I mention 300+ Hz is because the feedback circuit has to work for fundamental notes (which go way beyond the nominal crossover freq) and obviously the harmonic distortion products (which go way, way, way beyond the nominal crossover freq). Can't fix harmonic distortion if those frequencies are outside the reach of the feedback circuit.

If at any point in upper (or lower low) frequencies, the phase (and/or feedback circuit gain) goes sour you've got howling destruction!!!

Phase and gain are issues for all types of feedback, but much less of a problem for accelerometers and even less problem for voice-coil methods.

Ben
 
Last edited: