I am also very interested in the approach used in that paper (Werner), plus another by Steiger in "IRE Transactions on Audio" in 1960... but I wouldn't call those approaches "motional feedback". These just implement negative amplifier output impedance via feedback.
-Charlie
Ignoring my own inability to follow all of Owen's text, it is a pleasure to see the brainpower Owen brings to this topic. Let me mention a few points relevant to his comments.
1. Yes, speakers with MF are wonderfully tight but, unless EQ is added, they have less bass. That is partly because almost every bass speaker traffics in resonances and MF counter-acts these resonances. See end of point 2.
2. Positional sensing per se isn't crucial because absolute position isn't relevant for AC systems, provided you can detect and control motion. Likewise, the difference between acceleration functions and velocity functions is simply a matter of EQ, if I understand this. Voice coil MF, for example, is sensitive to one or the other (I don't know which) and requires bass boost to pump the air correctly.
3. ALWAYS one hears advocacy of using accelerometers and other externally mounted sensors. Cone drivers have a long history of development and care in production. Can you really correct a $300 woofer using a 50-cent accelerometer? For that reason, I tend to think the voice-coil-in-a-bridge (which uses the motor mechanism itself to provide the feedback) is best. I'd like to see evidence one way or the other. Some of those Finnish tests of woofers included home-brew MF systems. And, if I recall, they did very poorly. I wish somebody could review those tests for this forum and report their meaning to us.
1. Yes, speakers with MF are wonderfully tight but, unless EQ is added, they have less bass. That is partly because almost every bass speaker traffics in resonances and MF counter-acts these resonances. See end of point 2.
2. Positional sensing per se isn't crucial because absolute position isn't relevant for AC systems, provided you can detect and control motion. Likewise, the difference between acceleration functions and velocity functions is simply a matter of EQ, if I understand this. Voice coil MF, for example, is sensitive to one or the other (I don't know which) and requires bass boost to pump the air correctly.
3. ALWAYS one hears advocacy of using accelerometers and other externally mounted sensors. Cone drivers have a long history of development and care in production. Can you really correct a $300 woofer using a 50-cent accelerometer? For that reason, I tend to think the voice-coil-in-a-bridge (which uses the motor mechanism itself to provide the feedback) is best. I'd like to see evidence one way or the other. Some of those Finnish tests of woofers included home-brew MF systems. And, if I recall, they did very poorly. I wish somebody could review those tests for this forum and report their meaning to us.
Yes, there might be various ways, perhaps expensive, cumbersome, or otherwise challenging to assess cone motion. I welcome hearing feasible plans.
Today's fancy drivers are, well, fancy. So bridge-feedback is a pretty good concept, even if there are problems with the concept. And it involves zero special gizmos outside the ordinary electronics box. Maybe there is an accelerometer (or laser reflection) which works better and doesn't cost much either. Try it out: to start experimenting, all you need to do is see how the feedback signal compares to the input signal. Later, you can connect that feedback signal into the amp loop.
My impression is that the main "mountain to climb" is dealing with a heavy cone flapping in thin air... not the fine points of motor linearity that are little detectable on music anyway.
Today's fancy drivers are, well, fancy. So bridge-feedback is a pretty good concept, even if there are problems with the concept. And it involves zero special gizmos outside the ordinary electronics box. Maybe there is an accelerometer (or laser reflection) which works better and doesn't cost much either. Try it out: to start experimenting, all you need to do is see how the feedback signal compares to the input signal. Later, you can connect that feedback signal into the amp loop.
My impression is that the main "mountain to climb" is dealing with a heavy cone flapping in thin air... not the fine points of motor linearity that are little detectable on music anyway.
Here some nice MFB details in this link.
http://www.diyaudio.com/forums/multi-way/174276-wmtmw-baekgaard-2.html#post2315042
http://www.diyaudio.com/forums/multi-way/174276-wmtmw-baekgaard-2.html#post2315042
Last edited:
I’ve found that the Measurement Specialties model ACH-01 accelerometer is quite a good device for use in MFB systems. Prices have increased significantly over the past few years from about USD13 (early 2000’s) to a current USD32 from Mouser and Digikey. IIRC, bwaslo used this device in an article on MFB.
One benefit that I would like to explore using MFB is the degree by which it can be used to reduce power compression, especially in subwoofers. Another area I’d like to investigate is its effects on standing waves (in the enclosure) passing through the cone of a bass/mid-range drive. If anybody can point me to any info on these issues I’d be grateful. Hopefully, I may be able to find the time to research this myself in the near future.
Regards
Peter
One benefit that I would like to explore using MFB is the degree by which it can be used to reduce power compression, especially in subwoofers. Another area I’d like to investigate is its effects on standing waves (in the enclosure) passing through the cone of a bass/mid-range drive. If anybody can point me to any info on these issues I’d be grateful. Hopefully, I may be able to find the time to research this myself in the near future.
Regards
Peter
Very interesting about the interior spurious sounds. I've had a theory for a long time that MF speakers absorb room sounds - more or less the same idea as yours. For sure, speakers act as microphones and using my old bridge systems, I could pick up those room sounds and, presumably, "subtract" them from the amp output. This is one of the most fascinating aspects of MF and just might rank almost as significantly as cone motion in its value to reproduction systems.
Is there any way to present the specs on that accelerometer in nomenclature the same as for drivers (or vice versa, of course)?
Is there any way to present the specs on that accelerometer in nomenclature the same as for drivers (or vice versa, of course)?
Please pardon me for quoting myself, but I found the accelerometer has a spec for "linearity". For the ACH-01 it is .1 to 1.0%, kind of a big range. How would that relate to harmonic distortion in a driver?
With cone drivers (and many other areas in hifi), the distortion specs relate to extreme outputs and there is little clear association between those extremes and the amount of problems at ordinary or even loud domestic levels*. Therefore a cone driver with, say .1% distortion at ordinary listening levels can't be helped any too much by an accelerometer with 1% distortion.
How would phase response in the accelerometer screw-up any chance of having good results when in the loop?
The Philips system miklos refers to is a good example of wrong-headed engineering. Yes, using a 50-cent accelerometer can really help a lot a badly awful junk loudspeaker system. That might be good system engineering: take something cheap and awful and with little added expense make it somewhat less awful. Reports on that speaker I've read say it was pretty impressive for something made from a hunk of cheap commercial parts and thin wood. Other systems using MF and high-end design appear commercially from time to time; some impress reviewers mightily.
*and in the case of S/N in amps, the every-day levels may actually be far worse and doubly-so for high power amps when running at low power. But for loudspeakers, very hard to know how much distortion you have at the usual half-a-watt power level. On the other hand, speaker resonances are always "playing" at all power levels.
Last edited:
Great potential but the next step is conceiving a testing strategy. For example, playing a pure tone into the driver and seeing what is the spectral composition coming out of the accelerometer or the frequency plot or stuff like that. Without that kind of evidence, you never know if the accelerometer or piezo element is up to the job (or if you can define a narrow band, say 5-300 Hz, where you can use it).
That is easy to test with soundcard and arta used. Send test signal in speaker and receive the output of the element and with a THD measurement it will show very quick.
You have to compare it to the thd of the driver it shout show the same thd figure.
I just found this option for you UK folks.
200G acceleration Max
360 Hz max frequency.
http://www.farnell.com/datasheets/37677.pdf
10mv/G Output
I cant say it works but fun to glue on an old woofer in an old box, and check the oscilloscope and preamp direct into the computer.
you could then at least measure the power compression for a driver?
I would guess these sort of SMD devices and an oscilloscope might show some output, could be fun.
If some one made a custom preamp into a ADC or computer might give results interesting to us all, hopeful we attract the eye of a clever mechanical engineers who can apply control theory, and tell us what is impossible, stuff like that I cant bluff my way through.
I saw another at around 7 UKPounds (about 10 Euro). Unfortunately Farnell offloaded their consumer sales in Germany to another company with a tragic web interface
What is the Acceleration of a speaker cone?
Which devices do you think people should play with, as I will usually go with the cheapest option I can find, when I get around to it which wont be for a long time.
Also does any one know how to build a preamp for one of these, I am guessing a FET opamp high input impedance?
Regards
Owen
200G acceleration Max
360 Hz max frequency.
http://www.farnell.com/datasheets/37677.pdf
10mv/G Output
I cant say it works but fun to glue on an old woofer in an old box, and check the oscilloscope and preamp direct into the computer.
you could then at least measure the power compression for a driver?
I would guess these sort of SMD devices and an oscilloscope might show some output, could be fun.
If some one made a custom preamp into a ADC or computer might give results interesting to us all, hopeful we attract the eye of a clever mechanical engineers who can apply control theory, and tell us what is impossible, stuff like that I cant bluff my way through.
I saw another at around 7 UKPounds (about 10 Euro). Unfortunately Farnell offloaded their consumer sales in Germany to another company with a tragic web interface
What is the Acceleration of a speaker cone?
Which devices do you think people should play with, as I will usually go with the cheapest option I can find, when I get around to it which wont be for a long time.
Also does any one know how to build a preamp for one of these, I am guessing a FET opamp high input impedance?
Regards
Owen
Those Freescale accelerometers can be made to work, but they are noisy. IOW, when the speaker isn't playing anything you'll hear a low-level roar from the woofer. A more sensitive model from the same series will cause less noise, but will not be able to handle large signals -- dynamic range limitations. Acceleration handling needed is
xmax*(2pi*fmax)^2/(9.807 m/sec^2)
For a sub to handle excursion of 5mm, operating up to 100Hz, that comes to 201g.
You can handle more excursion at lower frequencies, but won't be able to work with more than that at 100Hz or higher.
xmax*(2pi*fmax)^2/(9.807 m/sec^2)
For a sub to handle excursion of 5mm, operating up to 100Hz, that comes to 201g.
You can handle more excursion at lower frequencies, but won't be able to work with more than that at 100Hz or higher.
This is off-topic.... but relevant.
Why is there so much enthusiasm for using a paste-on accelerometer instead of a bridge circuit? Why is there such strong feeling that gluing on a tool (like an accelerometer) to the dust cap produces results which are somehow more precise or scientific? Why does a complex solution seem smarter that a direct solution like a bridge?
An accelerometer seems like an inherently weak-linked strategy, even it it can be made to work.
Why is there so much enthusiasm for using a paste-on accelerometer instead of a bridge circuit? Why is there such strong feeling that gluing on a tool (like an accelerometer) to the dust cap produces results which are somehow more precise or scientific? Why does a complex solution seem smarter that a direct solution like a bridge?
An accelerometer seems like an inherently weak-linked strategy, even it it can be made to work.
Personally, I should like to see the feedback loop include the cones movement.
I should be most interested in feeding back the cone position rather than acceleration, as I think that will provide the greatest bandwidth for the feedback loop that I should like to extend all the way up to 2-300Hz even if the musical signals will only be less than 100Hz.
The reason I like accelerometers is they have the potential to be easier than constructing your own sensor, although a photo diode and a infra red LED might be enough, and I do agree the quality of the sensor will have a big effect on the quantity of feedback that you can apply.
I would also be interested in just analyzing the speakers movement and comparing it to the source so if a motional feedback project fails I would still have interesting data to read.
If we could maybe reach more than 10dB feedback around the cone movement to amplifier input their is in my opinion potential for better performance than with a bridge, this belief which may or may not be true.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.