I've been looking for websites and forum topics which deal with techniques to measure the exursion of a driver in real time, but I haven't found any. Probably I don't know the keywords.
In case it's not entirely clear what I want to achieve, let me summarize the idea: I would like to attach to a woofer a device which measures this woofer's excursion. The signal is fed into the computer responsible for generating the audio signal. A software running on this computer compares the reading to the audio signal and shapes the audio signal accordingly.
This raises a couple of questions:
I've set up a small system with a digital crossover (BruteFIR) and applied digital room correction. Measuring the excursion of the woofers seems like a fun thing to go on with.
Grateful for any advice
Pagnol
In case it's not entirely clear what I want to achieve, let me summarize the idea: I would like to attach to a woofer a device which measures this woofer's excursion. The signal is fed into the computer responsible for generating the audio signal. A software running on this computer compares the reading to the audio signal and shapes the audio signal accordingly.
This raises a couple of questions:
- What to correct?
- Is it at all possible to perform a useful correction, given the latency with which the correction arrives at the speaker?
- How to accurately measure the excursion?
I've set up a small system with a digital crossover (BruteFIR) and applied digital room correction. Measuring the excursion of the woofers seems like a fun thing to go on with.
Grateful for any advice
Pagnol
You are basically limited to low frequencies because of the phase shift betwwen the actual excursion and the driving signal at higher frequencies making the system unstable.
Also at higher freqs the cone will no longer be a single piston and have partial vibrations.
Google 'klippel' - Wolfgang Klippel has done a lot of (professional) work on that and has a functioning system.
jan didden
His office is located in the city in which I live. Maybe I'll find an opportunity to learn more (I notice he's offering internships).
I think that is how Klippel does it. He 'measures' the driver parameters in a calibration cycle and then adapts the drive to compensate for the non-linearities.
It's pretty sophisticated - not only can he get much more undistorted output from a driver, but it also has a protection and soft-clip subsystem.
It's all software, of course
jan
Do you think something could be achieved using the Kinect? I suppose its frame rate is too low to make it apt for a real time application in audio. However if its resolution, accuracy and frame rate were remotely sufficient, one might attempt to measure such things as excursion and deformation of the cone. Maybe useful statistical data about a driver could be gained by running the setup (Kinect immediately in front of the woofer) many times.
I hesitated sending this off because the idea strikes me as heavily impractical. I do nevertheless because it might still provoke thought.
Thank you for the hint. That looks pretty much like what I'm looking for. The method he employs is rather elegant.
I hesitated sending this off because the idea strikes me as heavily impractical. I do nevertheless because it might still provoke thought.
Thank you for the hint. That looks pretty much like what I'm looking for. The method he employs is rather elegant.
Last edited:
Doesn't a mic provide what you are looking for (using a bit of calculus)?
Tongue-in-cheek answers aside (which is what you deserve for posting a naive-like question as if you intend to invent some break-through system nobody ever thought of before), there's 60 yrs of motional feedback research to consider.
Not clear to me the difference between before-correction and after-feedback, except to say that only the latter makes any sense in banishing the many and instantaneously varying faults of cone drivers.
Ben
Tongue-in-cheek answers aside (which is what you deserve for posting a naive-like question as if you intend to invent some break-through system nobody ever thought of before), there's 60 yrs of motional feedback research to consider.
Not clear to me the difference between before-correction and after-feedback, except to say that only the latter makes any sense in banishing the many and instantaneously varying faults of cone drivers.
Ben
Last edited:
Sure, but with feedback you need a lot of gain to make any meaningful correction.
Which is inherently unstable.
With feed-forward you only need to apply as much correction as the distortion itself. (sort of)
If you correct non-linear and linear distortion, or at least minimize them by doing a little bit of correction. You should be left with only "the many and instantaneously varying faults of cone drivers." Which I'm not sure you can correct without introducing other distortion like feedback always does.
For reference: "Audio Distortion and Feedback" by Nelson Pass
FIRST WATT ARTICLES
I have no reason to believe that speakers with feedback would show any less increase in harmful IM distortion than a transistor.
Not to mention that noise would probably be amplified as well.
It is probably some reason to why feedback isn't used in speakers on a large scale, other than cost and complexity.
Feed-forward vs feedback is an interesting concept and there is a lot of info and discussions out there about it.
In a completely closed box, a mic(or other pressure sensor) should be able to measure excursion as long as no standing waves are allowed to modulate the pressure measured.
"If you know acceleration, you know position. That's why people normally use accelerometers for this application."
Accelerometers are used as acceleration is related to sound pressure so there is no need to convert it. Especially if you use current to drive the VC.
Absolute position is not relevant as pressure (and thus SPL) is not affected by it. (unless you are in a completely sealed room).
Velocity isn't useful ether as the only thing you get is a DC offset of pressure.
Which is inherently unstable.
With feed-forward you only need to apply as much correction as the distortion itself. (sort of)
If you correct non-linear and linear distortion, or at least minimize them by doing a little bit of correction. You should be left with only "the many and instantaneously varying faults of cone drivers." Which I'm not sure you can correct without introducing other distortion like feedback always does.
For reference: "Audio Distortion and Feedback" by Nelson Pass
FIRST WATT ARTICLES
I have no reason to believe that speakers with feedback would show any less increase in harmful IM distortion than a transistor.
Not to mention that noise would probably be amplified as well.
It is probably some reason to why feedback isn't used in speakers on a large scale, other than cost and complexity.
Feed-forward vs feedback is an interesting concept and there is a lot of info and discussions out there about it.
In a completely closed box, a mic(or other pressure sensor) should be able to measure excursion as long as no standing waves are allowed to modulate the pressure measured.
"If you know acceleration, you know position. That's why people normally use accelerometers for this application."
Accelerometers are used as acceleration is related to sound pressure so there is no need to convert it. Especially if you use current to drive the VC.
Absolute position is not relevant as pressure (and thus SPL) is not affected by it. (unless you are in a completely sealed room).
Velocity isn't useful ether as the only thing you get is a DC offset of pressure.
If tweaking the input to drivers (or the equivalent, tweaking the poles pieces, windings, spider, etc.) and maybe a bit of room equalization too was all that was needed to clean up speakers, we'd have perfect sound long ago.
The historic direction of development is in the enlargement of the feedback loop.
Ben
The historic direction of development is in the enlargement of the feedback loop.
Ben
I like most of Pass' publications – but the feedback article is the exception
is wrong - Nelson admits exaggerating the examples for effect - no real feedback amplifier internal stage-to-stage distortion follows his numbers
Because the loop gain is high the signal level in real feedback amps decreases by large factors as you move from output to input
input diff pair contribution to the distortion budget can be ~ -120 dB - the "distoriton of the distortion" in the early stages isn't an issue with competent high feedback design
many of your/Nelson's objections to high negative feedback are ill informed when applied to audio amplifiers
but
applying feedback to distributed mechanical systems is difficult for many reasons, the total amount of loop gain that you can employ is strictly limited by the phase shift of the propagating waves in the mechanical structure, added stability margins needed to cope with mechanical resonances in the physical structures between your actuator and sensor
and sound waves are AC velocity/pressure waves in air - a dynamic cone woofer driver motion is mass controlled above its low frequency resonance by the mass of the moving cone, coil surround... - the radiating sound in the room air load hardly signifies in analyzing/controlling the cone motion
and it is the cone velocity you want to control to most directly control radiated sound - acceleration sensing is technically convenient, integrating the acceleration signal to get velocity reduces measurement noise - position sensing requires differentiating the position signal to get velocity - amplifying high frequency noise
is wrong - Nelson admits exaggerating the examples for effect - no real feedback amplifier internal stage-to-stage distortion follows his numbers
Because the loop gain is high the signal level in real feedback amps decreases by large factors as you move from output to input
input diff pair contribution to the distortion budget can be ~ -120 dB - the "distoriton of the distortion" in the early stages isn't an issue with competent high feedback design
many of your/Nelson's objections to high negative feedback are ill informed when applied to audio amplifiers
but
applying feedback to distributed mechanical systems is difficult for many reasons, the total amount of loop gain that you can employ is strictly limited by the phase shift of the propagating waves in the mechanical structure, added stability margins needed to cope with mechanical resonances in the physical structures between your actuator and sensor
and sound waves are AC velocity/pressure waves in air - a dynamic cone woofer driver motion is mass controlled above its low frequency resonance by the mass of the moving cone, coil surround... - the radiating sound in the room air load hardly signifies in analyzing/controlling the cone motion
and it is the cone velocity you want to control to most directly control radiated sound - acceleration sensing is technically convenient, integrating the acceleration signal to get velocity reduces measurement noise - position sensing requires differentiating the position signal to get velocity - amplifying high frequency noise
Last edited:
I am almost speechless... never before have I heard an argument that more clearly parallels "the bee can never fly" kind of arm-chair theorizing.
Let's take as one example, "limited by the phase shift...." OK, tell me, is it any insuperable challenge to engineering design to work within the freq compass where phase shift isn't a big problem?
Does it need anything more than adding a trivial RC circuit to work with a "mass controlled" driven cone element? Or is it better not to have feedback for such an uncooperative element (like every cone driver ever made)?
Numerous studies have shown the benefits of motional feedback for low frequencies - not to mention the experience of lots of users, like me. It is however, a tricky strategy to commercialize. But so is landing payloads on Mars.
Ben
Let's take as one example, "limited by the phase shift...." OK, tell me, is it any insuperable challenge to engineering design to work within the freq compass where phase shift isn't a big problem?
Does it need anything more than adding a trivial RC circuit to work with a "mass controlled" driven cone element? Or is it better not to have feedback for such an uncooperative element (like every cone driver ever made)?
Numerous studies have shown the benefits of motional feedback for low frequencies - not to mention the experience of lots of users, like me. It is however, a tricky strategy to commercialize. But so is landing payloads on Mars.
Ben
Last edited:
BJ Lurie's feedback work covers mechanical systems well – he was trusted enough for JPL to let him design controls on $billion space projects
search diyAudio for my posts mentioning Lurie to get a link to his paper on Bode's phase integral relation as a fundamental “conservation law” in feedback theory
his website, book chapter are still avaiable from the WayBack Machine
I have designed, built, characterized, calibrated electromechanical motion controls, analog and digital, writing the DSP code myself
watched real hardware under motion control as I turned up the gain, shut down machines before they shook themselves apart
BJ Lurie's works have been the best for me expanding on undergrad level control theory I learned from the Mech E dept
my interest in electronics was sparked by audio, I have read amplifier and speaker design, including servo feedback articles in JASA current issues from mid '70s onward, have spent time in the stacks too
so my armchair opinion may have some weight (beyond just my a--)
[edit: here you go, my "there is nothing so practical as a good theory" post]
search diyAudio for my posts mentioning Lurie to get a link to his paper on Bode's phase integral relation as a fundamental “conservation law” in feedback theory
his website, book chapter are still avaiable from the WayBack Machine
I have designed, built, characterized, calibrated electromechanical motion controls, analog and digital, writing the DSP code myself
watched real hardware under motion control as I turned up the gain, shut down machines before they shook themselves apart
BJ Lurie's works have been the best for me expanding on undergrad level control theory I learned from the Mech E dept
my interest in electronics was sparked by audio, I have read amplifier and speaker design, including servo feedback articles in JASA current issues from mid '70s onward, have spent time in the stacks too
so my armchair opinion may have some weight (beyond just my a--)
[edit: here you go, my "there is nothing so practical as a good theory" post]
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.