An accelerometer is still just reading how the dust cap moves around, not the whole cone output. The same is true of almost any sensor except for a mic. But it is wholly inconceivable to use a mic for a whole bunch of reasons.
ATTENTION ALL THOSE LURKING ON THE SIDELINES: make a bridge or even just try to read the signal across one of the speaker leads (you'll need a mic pre-amp like on that old cassette recorder stored in your cellar). Will this take you more than 20 minutes to set up? Shout some obscenity at your woofer and see what it looks like on your scope. Humm a note like the speaker resonance.
ATTENTION ALL THOSE LURKING ON THE SIDELINES: make a bridge or even just try to read the signal across one of the speaker leads (you'll need a mic pre-amp like on that old cassette recorder stored in your cellar). Will this take you more than 20 minutes to set up? Shout some obscenity at your woofer and see what it looks like on your scope. Humm a note like the speaker resonance.
The headphones I am listening to right now have microphones. I will acknowledge that their primary purpose is noise cancellation, but they must also perform distortion reduction (MFB) in order to perform their primary role.
The microphones are mounted in the earcups so that they are close to the canals that lead from the outer ear to the eardrum. The signal they generate is compared with the signal input to the headphones, and a correction signal generated and fed to the drivers. The result is a reduction in the SPL of external bass and mid-bass noise at the entrance to the ear canal. Note that distortion generated by the driver is also considered an error signal and a correction for it is generated.
Although this would also work for larger drivers, I suspect it would not work up to such high frequencies due to the path length differences of the larger driver. It might be worth experimenting with placing the microphone inside the (sealed) enclosure, provided that it was only required to work at frequencies below the lowest internal box resonance.
I haven't used any modern accelerometer. I haven't fooled with MF for maybe 20 years. As a method, using an economically-feasible sensor doesn't sound quite right but nobody, not even me, is so stupid as to say there never will be an accelerometer that will work great.
I'm in a far warmer place than Toronto just now. Moreover, it is my long-standing web custom never to explain things in a cookbook manner for a couple of reasons. If you can take-off-and-fly-right from loose hints, only then are you wise to be in the cockpit.*
So, let point out that "bridge" can be used in two senses.
1. For testing, all you want to do is to derive a signal which reflects the impedance of the speaker. ANY system you'd use to measure impedance can be tried.
2. The theory of deriving MF from the VC is to make the VC one leg of a Wheatstone Bridge. The other side of the bridge is a model of the driver static electric parameters. Therefore, the bridge is balanced (zero output) when the driver does not depart from its static profile; and when it does depart, that is an error signal. Yippee. "Depart" means stuff like "back-EMF" to use a term from ancient times, that arise when the VC makes motions that differ from the input signal.
*Blew up a Klipschorn woofer driver once. Dumb or crazy wild or just a painful part of normal experimenting? Just showing my MF scars to everybody.
Insightful of Don Hills to relate noise-canceling phones to MF. But the issues are so vastly different between sensing outside noise versus speaker error, hardly possible to compare the processes except by the loosest of verbal analogies.
Earlier I said no way on earth to use mics for MF. With noise-cancelation, the mic/sensor is right at the entry to the ear canal and "all" it has to do is listen for non-music signals and then get the ear-speakers to play them back out of phase. Even there, the effect is limited to wave-lengths and phases that can be kept coherent and even knocking off 10 dB at lowish frequencies is a wonderful benefit.
A world of difference when you try to pick and keep in proper phase (keep in mind that different frequencies have different wavelengths and therefor the distance of the mic matters a whole lot) the acoustic output from the driver and the tiny error component of those signals.
Earlier I said no way on earth to use mics for MF. With noise-cancelation, the mic/sensor is right at the entry to the ear canal and "all" it has to do is listen for non-music signals and then get the ear-speakers to play them back out of phase. Even there, the effect is limited to wave-lengths and phases that can be kept coherent and even knocking off 10 dB at lowish frequencies is a wonderful benefit.
A world of difference when you try to pick and keep in proper phase (keep in mind that different frequencies have different wavelengths and therefor the distance of the mic matters a whole lot) the acoustic output from the driver and the tiny error component of those signals.
Last edited:
Thanks for the prompt response. Have you read the JEAS paper by Peerless entitled "Jump Resonance in Audio Transducers" ?
If you have, what response should I use to model the impedance of the bridge arm ?
"long-standing web custom never to explain things in a cookbook manner" ?
This doesn't make sense to me, but I guess it's your prerogative!
Regards
Peter
If you have, what response should I use to model the impedance of the bridge arm ?
"long-standing web custom never to explain things in a cookbook manner" ?
This doesn't make sense to me, but I guess it's your prerogative!
Regards
Peter
Last edited:
Can't recall if I read that article or not.
As I said, there is (1) getting a signal to experiment with and (2) there is final design of MF systems. When you do (2), or so the theory goes, you should cancel out the static model of the driver.
Setting up a bridge just for experimenting, you are just fine with a Wheatstone Bridge that balances just the DC resistance of the VC. Simple as that. (Anyway, even with final systems, you are tinkering only at lowish frequencies and the inductance of the voice coil is not important, generally.)
On the driver side of the bridge, you have the VC and a small resistor (like .2 ohms) because they are in series with the amp output (and you don't want too big a resistor hampering the damping factor). The other side is - obviously - in proportion of 8 to .2 or 40:1 or like that.
Let me explain that I am a nice and friendly old guy: MF is really dangerous and you can blow very expensive drivers. If a person can't think-through the issues based on posted rough-notes, he/she should not be experimenting with MF, fuel-injection maps on motorcycles, or other things that I post about. But if they can conceptualize the issues and think them through better than I ever can, then full speed ahead.
As I said, there is (1) getting a signal to experiment with and (2) there is final design of MF systems. When you do (2), or so the theory goes, you should cancel out the static model of the driver.
Setting up a bridge just for experimenting, you are just fine with a Wheatstone Bridge that balances just the DC resistance of the VC. Simple as that. (Anyway, even with final systems, you are tinkering only at lowish frequencies and the inductance of the voice coil is not important, generally.)
On the driver side of the bridge, you have the VC and a small resistor (like .2 ohms) because they are in series with the amp output (and you don't want too big a resistor hampering the damping factor). The other side is - obviously - in proportion of 8 to .2 or 40:1 or like that.
Let me explain that I am a nice and friendly old guy: MF is really dangerous and you can blow very expensive drivers. If a person can't think-through the issues based on posted rough-notes, he/she should not be experimenting with MF, fuel-injection maps on motorcycles, or other things that I post about. But if they can conceptualize the issues and think them through better than I ever can, then full speed ahead.
Last edited:
If you can achieve 10dB (your example) of external noise reduction over a certain frequency range, you automatically achieve 10 dB of driver generated noise (distortion) reduction over the same range.
I agree that scaling the headphone sized example up to a 12 inch driver is a non-trivial exercise. But if it can be done, it will reduce the relatively high levels of harmonic distortion produced by typical drivers at low frequencies. In a flat frequency response situation these would be relatively innocuous, but because of the ear's response curve at bass frequencies, even a low level of harmonic distortion may be as subjectively audible as the fundamental frequency.
Although I don't know much about the circuits (and how they might be choosing to suppress anything remotely like the driver output in order to avoid positive feedback), I suppose you are right about reducing distortion by that same (modest) amount. We are talking about distances which are only small parts of a wave-length within the headphone box.
Applying that way of using a mic to a driver in the same way is kind of like mounting the mic half an inch from the dust cap. It might work on that basis, but it certainly isn't (and better not) be responding the whole cone... which is why you'd want to use a mic picking up the whole cone sound. When the mic is even as nearby as one foot, some frequencies are way different in phase than others.
Applying that way of using a mic to a driver in the same way is kind of like mounting the mic half an inch from the dust cap. It might work on that basis, but it certainly isn't (and better not) be responding the whole cone... which is why you'd want to use a mic picking up the whole cone sound. When the mic is even as nearby as one foot, some frequencies are way different in phase than others.
links
A few links on the subject to check out:
mfbfreaks : Official Site
Check the Series IV F9638 with 0.1% TDH 30-700Hz at 50W
http://www.diyaudio.com/forums/subwoofers/84614-pressure-based-motional-feedback.html
http://www.diyaudio.com/forums/subwoofers/94796-motional-feedback-subwoofer.html
http://www.xs4all.nl/~loosen/elektronica/mfb/
High g
A few links on the subject to check out:
mfbfreaks : Official Site
Check the Series IV F9638 with 0.1% TDH 30-700Hz at 50W
http://www.diyaudio.com/forums/subwoofers/84614-pressure-based-motional-feedback.html
http://www.diyaudio.com/forums/subwoofers/94796-motional-feedback-subwoofer.html
http://www.xs4all.nl/~loosen/elektronica/mfb/
High g
Hi,
I don´t think that a bridge circuit as supposed by Ben would work in this apllication. If one of the legs of the bridge is the VC of the driver and the other legs are made up from resistors, the output of the bridge can only be 0 under DC-conditions. Under AC-conditions the output will differ from 0, because of the reactive parts of the speakers impedance. There will be a bridge signal even if the driver works perfectly linear (distortion-free).
It should help to also AC-balance the bridge by modelling the driver´s impedance response in one of the bridge´s legs instead of using just a simple resistor. But even if we corrected the bridge this way I doubt that we could really isolate the driver´s distortion artefacts from its ´normal´ behaviour, because the driver´s impedance changes not only over frequency but also it changes with apllied power.
MFB is just usable in a frequency range where the membrane movement can be assumed pistonic. It can´t correct for resonances and vibrational modes. This restricts the usable bandwidth considerably.
Additionally to keep the system stable the upper bandwidth limit needs to be restricted.
Delays introduced by the sensor (think microphones) and feedback circuits (think of signal conditioning) also reduce the usable upper bandwidth limit.
Noise issues have to be regarded. I´ve been reported that MEMS-acceleration sensors, which would be quite attractive otherwise (ADI´s to name one), are noisy (it seems though that Velodyne uses kind of MEMS-acceleration sensor).
Linearity and dynamic issues rule out many cheap sensors. MFB is of low or no use at all if the sensor is not considerably more linear than the driver itself. The dynamic range needs to be greater than the driver´s. If You calculate e.g. the acceleration values, that depend on the applied frequency of the signal and the excursion of the driver, values of a couple of hundreds of g may result. Sensors that work linear over such a huge g-range are costly.
If one of the components of the MFB system can´t fulfill the stringent demands You might think about using a simple EQed system instead.
jauu
Calvin
I don´t think that a bridge circuit as supposed by Ben would work in this apllication. If one of the legs of the bridge is the VC of the driver and the other legs are made up from resistors, the output of the bridge can only be 0 under DC-conditions. Under AC-conditions the output will differ from 0, because of the reactive parts of the speakers impedance. There will be a bridge signal even if the driver works perfectly linear (distortion-free).
It should help to also AC-balance the bridge by modelling the driver´s impedance response in one of the bridge´s legs instead of using just a simple resistor. But even if we corrected the bridge this way I doubt that we could really isolate the driver´s distortion artefacts from its ´normal´ behaviour, because the driver´s impedance changes not only over frequency but also it changes with apllied power.
MFB is just usable in a frequency range where the membrane movement can be assumed pistonic. It can´t correct for resonances and vibrational modes. This restricts the usable bandwidth considerably.
Additionally to keep the system stable the upper bandwidth limit needs to be restricted.
Delays introduced by the sensor (think microphones) and feedback circuits (think of signal conditioning) also reduce the usable upper bandwidth limit.
Noise issues have to be regarded. I´ve been reported that MEMS-acceleration sensors, which would be quite attractive otherwise (ADI´s to name one), are noisy (it seems though that Velodyne uses kind of MEMS-acceleration sensor).
Linearity and dynamic issues rule out many cheap sensors. MFB is of low or no use at all if the sensor is not considerably more linear than the driver itself. The dynamic range needs to be greater than the driver´s. If You calculate e.g. the acceleration values, that depend on the applied frequency of the signal and the excursion of the driver, values of a couple of hundreds of g may result. Sensors that work linear over such a huge g-range are costly.
If one of the components of the MFB system can´t fulfill the stringent demands You might think about using a simple EQed system instead.
jauu
Calvin
People have been publishing impressive results using the bridge concept for more than 50 years. The publication that got me started was from R. E. Werner, RCA Labs (not sure if he worked for Olson), JAES (or JASA, not sure) around 1957.
I should see if I have my test results from Bell Labs using pulses. Frankly, the scope Polaroids of the pulses with and without MF aren't a tiny bit as impressive as the sound: one sound is CLICK and the other sound is cccllllooook. Clear to your ear as could be. And that is what would convince any doubter that MF is fabulous.
I should see if I have my test results from Bell Labs using pulses. Frankly, the scope Polaroids of the pulses with and without MF aren't a tiny bit as impressive as the sound: one sound is CLICK and the other sound is cccllllooook. Clear to your ear as could be. And that is what would convince any doubter that MF is fabulous.
For those who have examined Calvin's post, are there any points he makes which seem to you: (A) true, (B) significant or important in degree, AND ALSO (C) cast doubt on the value of MF or bridge MF?*
If so, let me know and I'll offer my thoughts on those items either in support of his reasoned point of view or otherwise.
*There's no value in hearing my 2-cents amateur opinion on anything else.
If so, let me know and I'll offer my thoughts on those items either in support of his reasoned point of view or otherwise.
*There's no value in hearing my 2-cents amateur opinion on anything else.
Last edited:
so...you're saying something that has been shown to work in real life won't work on paper?
The ACE BASS literature is adamant that it has no connection. Seems all but identical to me.
There are various ways to look at MF and I am too rusty in the concepts of positive feedback, output impedance, etc. to be trusted here. But.... the amp acts as it is the mirror image of the driver's impedance. In other words, if the driver impedance goes up like to 80 ohms (like at resonance) the amp also behaves as if its internal resistance was 80 ohms.... so the loudness of the driver is cut and the sound output is "correct." The amp therefore behaves as if it has negative output impedance: when the driver seems to be asking for more oats, the amp gives it less.
C'mon... if that comment doesn't stir your blood and make you want to fool with MF tonight, well, nothing in audio could.
Did you know that when you push on a driver cone with MF, it feels very very stiff, almost as if it is pushing back (because it is).
Hi,
Your reply Ben finally prooves, that we agree to disagree.
But instead of just throwing mud, it´d be much more helpful, if You had decently argumented against my reply instead of spilling poison.
So, please explain what the simple bridge circuit corrects for? How do You separate normal driver behaviour from its distortion artefacts? Or in other words, what are the signal components the bridge generates and how do You extract the correction signal from the bridge signal?
Regarding MFB. No, pedroskova, I´m not saying MFB doesn´t work. I´m just listing a some limits and restrictions of MFB.
As with any technique MFB functions only within limits and if precisely executed. If one wants to build a well functioning MFB he/she should know about the restrictions, otherwise the end results will be non optimal. One of the more demanding constraints regards the linearity of the sensor. Unfortunately there seem to exist more bad examples than good ones, where the sensor´s capabilities are not sufficient.
The working of the MFB shows in a linearized amplitude response and lowered distortion values. To achieve considerably lower distortion values the sensor must perform alot more linear over a wide dynamic range than the raw driver. If not, the reduction in distortion values will remain small to negligible and the MFB functions rather only like an equalizing network. In that case a simple EQed box might give even better results with less circuit complexity and hassle.
jauu
Calvin
Your reply Ben finally prooves, that we agree to disagree.
But instead of just throwing mud, it´d be much more helpful, if You had decently argumented against my reply instead of spilling poison.
So, please explain what the simple bridge circuit corrects for? How do You separate normal driver behaviour from its distortion artefacts? Or in other words, what are the signal components the bridge generates and how do You extract the correction signal from the bridge signal?
Regarding MFB. No, pedroskova, I´m not saying MFB doesn´t work. I´m just listing a some limits and restrictions of MFB.
As with any technique MFB functions only within limits and if precisely executed. If one wants to build a well functioning MFB he/she should know about the restrictions, otherwise the end results will be non optimal. One of the more demanding constraints regards the linearity of the sensor. Unfortunately there seem to exist more bad examples than good ones, where the sensor´s capabilities are not sufficient.
The working of the MFB shows in a linearized amplitude response and lowered distortion values. To achieve considerably lower distortion values the sensor must perform alot more linear over a wide dynamic range than the raw driver. If not, the reduction in distortion values will remain small to negligible and the MFB functions rather only like an equalizing network. In that case a simple EQed box might give even better results with less circuit complexity and hassle.
jauu
Calvin
Hi,
Your reply Ben finally prooves, that we agree to disagree.
But instead of just throwing mud, it´d be much more helpful, if You had decently argumented against my reply instead of spilling poison.
jauu
Calvin
I object strongly to your accusation. Show where I've said anything remotely disrespectful in the recent posts. Or apologize. On the contrary, I said "...his reasoned point of view..." and I meant EXACTLY that and as a fair description (granted, I don't agree with your reasoning).
What I DID say, is that there is no sense in amateur-me commenting on stuff from your post that any reader already feels isn't true, that any reader feels is of insufficient importance to argue about, or that otherwise doesn't matter to anybody. Seems a fair way to move the discussion forward, if any reader had questions about your post (and that they met those criteria in order to be worth discussing further). I did not say nor did I slyly imply that your post was characterized by any of those qualities.
Frankly, pedroskova has deftly replied to your post. What more could I add?
I avoid writing sarcasm and double-meanings and can get my thoughts across jolly well without them. But some people can read sarcasm and double-meanings into anything.
Last edited:
The great progress of audio in the last 100 years has been through the enlargement of corrective feedback. The last link is the room... and you could say the speaker is the next to the last link.
It is hard to get the speaker into the loop and the path is littered with failed commercial attempts.
The laws of feedback require close attention to phase and, I'd say, that is a big challenge since everything upstream of the speaker has well-behaved phases and everything downstream, the opposite. To help, designers limit the bandwidth in which the feedback operates. Fortunately, there is a fairly narrow area where feedback can be readily applied and/or where it can do much good. Specifically, it is in the lower regions where waving a large chunk of cardboard can't be a good way to make precise sound reproduction.
So MF design is bound by the frequency regions of cone break-up, resonances, close relationship between VC and sound produced, and well as by the characteristics of the sensory system.
When a bridge is used, it produces no feedback signal when everything is honky-dory. The signal arises when the bridge is unbalanced due to gyrations of the VC different from the signal that was fed to the VC. If you install a perfect speaker, you will get no feedback, nor will you ever need any. On the other hand, if your speaker is not perfect......
Tantalizing footnote: MF systems don't pay attention to LF roll-off in sealed boxes. All they "know" is that the VC doesn't seem to be moving as much as it should and so it pumps more oats into the driver, even though the "acoustic suspension" is holding back the cone. How does that sound to you?
It is hard to get the speaker into the loop and the path is littered with failed commercial attempts.
The laws of feedback require close attention to phase and, I'd say, that is a big challenge since everything upstream of the speaker has well-behaved phases and everything downstream, the opposite. To help, designers limit the bandwidth in which the feedback operates. Fortunately, there is a fairly narrow area where feedback can be readily applied and/or where it can do much good. Specifically, it is in the lower regions where waving a large chunk of cardboard can't be a good way to make precise sound reproduction.
So MF design is bound by the frequency regions of cone break-up, resonances, close relationship between VC and sound produced, and well as by the characteristics of the sensory system.
When a bridge is used, it produces no feedback signal when everything is honky-dory. The signal arises when the bridge is unbalanced due to gyrations of the VC different from the signal that was fed to the VC. If you install a perfect speaker, you will get no feedback, nor will you ever need any. On the other hand, if your speaker is not perfect......
Tantalizing footnote: MF systems don't pay attention to LF roll-off in sealed boxes. All they "know" is that the VC doesn't seem to be moving as much as it should and so it pumps more oats into the driver, even though the "acoustic suspension" is holding back the cone. How does that sound to you?
Last edited:
Hi,
jauu
Calvin
Is that really so?? In an earlier post You supposed a simple resistor of the value of Rdc of the drivers voice coil as counterpart in the bridge. In this case the bridge is only balanced under DC conditions. But music is nothing but AC. Under this condition the bridge is not balanced, hence it produces an output signal. The signal consists of a ´normal part´ that is solely due to the frequency dependant impedance response of the driver, compared to the Rdc-resistor. Added to this part is a part due to distortions. So how do You separate the distorted parts from the distortion-free part in the bridge signal?
jauu
Calvin
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.