There are lots of observations you can make with a mic or laser tools. The question is how to quantify (or even rank-order) them.
Yes, looking at the feedback signal is helpful. It should reflect the speaker errors it is in place to correct. Again, how to quantify.
B.
Any motion purchased with a higher Q is false (that is, not the signal). It may be apparent on a sim because the sim assumes a flat input signal to the speaker. But all that means is that a flat input signal isn't what that speaker needs to play flat. In the sim, the tail is waving the dog.
MFB corrects any movement of the cone not related to the signal - whether music is playing or just stationary while you are punching the cone. So it resists the punch with all the power it has.
B.
Klippel controlled sound is pretty cool new stuff in this arena. It is based on self-learning software, no sensors required. Klippel Controlled Sound (KCS) - Controlled Sound Technology for Nonlinear Compensation of Loudspeakers | audioXpress
Hopefully minidsp or some other "big" manufacturer licenses it and makes available to all
Hopefully minidsp or some other "big" manufacturer licenses it and makes available to all
What I meant to say is that T/S driver parameters don't matter the same once the driver is mounted in an enclosure. But if the mounting is an OB, then maybe T/S parameters are relevant to setting up the status-quo-ante balance in a Wheatstone Bridge current feedback.
In traditional thinking, MFB based on cone motion (current, accelerometer, etc) is legitimate only if the cone is the only sound generator (sealed, OB, true horn, labyrinth, etc.). It isn't meaningful when you have a tuned box (BR, TL, etc.) because the accelerometer has no means to know the total speaker output. (Think how with a BR, the motion of the cone and the total output of the speaker have a changing relationship above and below tuning point.)
B.
You have to know what the limitations of your measurements are. This goes for regular speaker building as well.
Most of us don't have an anechoic chamber at home, thus most people use gated measurements. This doesn't mean these measurements are invalid! Just the low frequency part is irrelevant. With a close in measurement you don't have the problem of the room.
A close in measurement SHOULD give the same result as the accelerometer. There might be a resonance in the sensor measurement higher up, because the sensor mounting isn't infinitely stiff.
Yes a high Q system introduces ringing that isn't in the original system. This will afaik not be suppressed by loopgain as it can be introduced BY the controller. Just by placing a pole very close to the Im-axis.
Ideally a MFB sub is a brick wall yes.
Yes, MFB only works well in sealed enclosures. No helmholtz resonators, no passive radiators, no transmissionlines, just a sealed box. This is because those other things are not controllable NOR are they observable.
Besides that, those resonators all have a very high Q and ruin the beautiful transient response of our MFB system.
I'm not quite sure we can throw all the T/S parameters out of the window. Lower BL and higher moving mass is still going to lower our sensitivity. Cone stiffness is still going to impact the upper resonance.
I'm still not quite sure what you mean with a wheatstone bridge configuration.
True and a clipping amplifier is even worse. It is a matter of how much resources you are willing to spend. Taking them into account can extract every drop of performance, ignoring them still results in satisfactory performance.
The RMS Acoustics paper "Acceleration Feedback Design" only uses the Thiele and Small parameters to determine that a high BL is beneficial for maximum efficiency and mms is hardly relevant. The feedback controller is designed based on measured transfer functions.
As long as port losses are low (low air velocities), there is a relation between cone motion and system output. The fact that this is frequency dependent, does not matter.
Assuming minimum-phase behaviour and a linear system, resonances are not a problem and can be corrected by minimum-phase equalization.
Last edited:
Klippel controlled sound is pretty cool new stuff in this arena. It is based on self-learning software, no sensors required. Klippel Controlled Sound (KCS) - Controlled Sound Technology for Nonlinear Compensation of Loudspeakers | audioXpress
Hopefully minidsp or some other "big" manufacturer licenses it and makes available to all
This is very interesting. Looks like they are trying to do full-state feedback based on voltage and current and used LQR-control based of a model with some filters for system protection.
I'm very interested how they made that non-linear model. Though Klippel is king here (though I must add that Purifi also showed some interesting things in modelling). Unfortunately that is probably intellectual property and will not become publicly available.
It's also interesting how it still works WITH a additional resonator (port, PR, TL).
I hope it will be implemented commercially because the implementation is relatively cheap since no sensors are required. Just a 2 channel ADC and DSP would be enough. Though licensing is probably not going to be cheap.
I think the question for today is how to measure improvement in transient ("fast") performance, esp if there's a quantitative score possible so as to document performance.
Ben, maybe I misunderstand your question here, but to my best of knowledge Bolserst has already explained to you before that (woofer-)bandwith and its associated phase behaviour(the Bode plot) determine transient response of the system. No matter if it is by MFB, analogue Linkwitz Transform or digital implementation: in case e.g. 3 woofer systems have exactly the same SPL response, say 25 Hz to 200 Hz with 12 dB slopes on either side of the passband , the first with MFB, the second by L-T., the third by digital equalization, their respective transient responses will be the same, whether one likes that or not. In both the Commercial MFB thread and the Servo Controlled Sub and now in this thread you come back bringing this up.
We all know audiophiles are obsessed with so called “fast bass”” (not to mention DDR) which basically is nonsense for well or critically damped 12 dB roll off systems. Differences in transients have to do with bandwith of and roll off, not with “fastness” of bass.
Ben, maybe I misunderstand your question here, but to my best of knowledge Bolserst has already explained to you before that (woofer-)bandwith and its associated phase behaviour(the Bode plot) determine transient response of the system. No matter if it is by MFB, analogue Linkwitz Transform or digital implementation: in case e.g. 3 woofer systems have exactly the same SPL response, say 25 Hz to 200 Hz with 12 dB slopes on either side of the passband , the first with MFB, the second by L-T., the third by digital equalization, their respective transient responses will be the same, whether one likes that or not. In both the Commercial MFB thread and the Servo Controlled Sub and now in this thread you come back bringing this up.
We all know audiophiles are obsessed with so called “fast bass”” (not to mention DDR) which basically is nonsense for well or critically damped 12 dB roll off systems. Differences in transients have to do with bandwith of and roll off, not with “fastness” of bass.
The model was made based on relevant T/S parameters. In chapter 3 of the "Sensorless Veleocity Feedback Subwoofer" paper the author explains how modelling can be improved. He mentioned that ignoring the semi inductance was a bad choice and it should be modelled.
Besides, what harm does it do to have a more accurate model?
The Klippel concept is not a feedback error-correction. It is a kind of wannabee-engineer fantasy that if you knew everything about how your speaker converts electricity into sound, you could finagle the signal in reverse. A feedback system is like having a policeman watch the cone and at the first sign of error, react to correct it.
The Wheatstone Bridge feedback pattern is like using a little current-sensing resistor to send back-EMF from the voice coil. But instead of a little resistor, you set up a bridge configuration and when the back-EMF differs from the input signal, the bridge goes unbalanced and the feedback signal is generated.
For a BR, above box tuning, the port output adds (that is, it is in-phase) with the cone. Below tuning it is opposite. In other words, the total speaker output does not have a well-behaved relationship to cone motion. And the region around box (of driver) resonance is the key band for MFB implementation.
Whether there is "fast" bass independently of band-width of a driver, a question for another day. But without exactly repeating Lord Kelvin's famous quote*, if you can't measure something you can't talk about it. Nice to have pretty pictures but better to have numbers representing audio parameters. "Quantitative" is the word.
B.
* actually, he had a bunch of famous quotes, one of which I passed every day when a university student in Chicago. Here's one: "I often say that when you can measure what you are speaking about, and express it in numbers, you know something about it; but when you cannot measure it, when you cannot express it in numbers, your knowledge is of a meagre and unsatisfactory kind."
The Wheatstone Bridge feedback pattern is like using a little current-sensing resistor to send back-EMF from the voice coil. But instead of a little resistor, you set up a bridge configuration and when the back-EMF differs from the input signal, the bridge goes unbalanced and the feedback signal is generated.
For a BR, above box tuning, the port output adds (that is, it is in-phase) with the cone. Below tuning it is opposite. In other words, the total speaker output does not have a well-behaved relationship to cone motion. And the region around box (of driver) resonance is the key band for MFB implementation.
Whether there is "fast" bass independently of band-width of a driver, a question for another day. But without exactly repeating Lord Kelvin's famous quote*, if you can't measure something you can't talk about it. Nice to have pretty pictures but better to have numbers representing audio parameters. "Quantitative" is the word.
B.
* actually, he had a bunch of famous quotes, one of which I passed every day when a university student in Chicago. Here's one: "I often say that when you can measure what you are speaking about, and express it in numbers, you know something about it; but when you cannot measure it, when you cannot express it in numbers, your knowledge is of a meagre and unsatisfactory kind."
Last edited:
Folks,
referring back to the thread title... A subwoofer servo/feedback system which was not yet mentioned in the discussion, but is relatively simple to do, is an Acoustic Feedback system which uses an electret microphone capsule glued on the piston or fastened on a bar across the front of the piston. The actual sound pressure level is measured and "linearises" the sound output in real-time. A closed Box design builds the basis.
So, basically, there's no need for "complex" inclusion of TSPs in the design per se, just a realistic interpretation of fsc and Qc and hence building a cabinet size not limiting efficiency and low cut off too much is required. On the electronics side, a modified PI control defines the lower cut-off. The power of the amp, the power handling capability, piston area and excursion limit of the woofer limit the max sound pressure level and min cut off frequency.
The method is known since the 1980s and here's a small company that adopted the AFB priciple successfully for their sub-woofers in all possible sizes up to18" several years ago.[ Sorry, everything is in German...]
Greetigs,
Wifried
referring back to the thread title... A subwoofer servo/feedback system which was not yet mentioned in the discussion, but is relatively simple to do, is an Acoustic Feedback system which uses an electret microphone capsule glued on the piston or fastened on a bar across the front of the piston. The actual sound pressure level is measured and "linearises" the sound output in real-time. A closed Box design builds the basis.
So, basically, there's no need for "complex" inclusion of TSPs in the design per se, just a realistic interpretation of fsc and Qc and hence building a cabinet size not limiting efficiency and low cut off too much is required. On the electronics side, a modified PI control defines the lower cut-off. The power of the amp, the power handling capability, piston area and excursion limit of the woofer limit the max sound pressure level and min cut off frequency.
The method is known since the 1980s and here's a small company that adopted the AFB priciple successfully for their sub-woofers in all possible sizes up to18" several years ago.[ Sorry, everything is in German...
]Greetigs,
Wifried
Acoustic Feedback using a mic was also succesfully implemented by Mayer Sound in their X10
Meyer Sound X-10 Powered Loudspeakers
And this for a vented box.
The mic is mounted on the bar across the woofer. I would guess the distance from cone to mic, limits the upper frequency for which this will work.
An added benefit of this arrangement will be improved room integration as for lower frequencies the sound "recorded" by the mic will include room modes, which will then be attenuated. So build in bass trap
Think I like the notion of mounting the mic on a bar better than attaching it to the cone, and thereby submitting it to quite high G force.
Meyer Sound X-10 Powered Loudspeakers
And this for a vented box.
The mic is mounted on the bar across the woofer. I would guess the distance from cone to mic, limits the upper frequency for which this will work.
An added benefit of this arrangement will be improved room integration as for lower frequencies the sound "recorded" by the mic will include room modes, which will then be attenuated. So build in bass trap
Think I like the notion of mounting the mic on a bar better than attaching it to the cone, and thereby submitting it to quite high G force.
Whenever MFB is discussed, always the first and most obvious idea is to use a mic* but it never seems a good idea in the end, for a whole bunch or reasons such as the phase, distortion of the mic, question about what sound reaches the mic mounting spot, etc.
Cousin to the mic notion is a pressure sensor in the box, used by somebody I can't recall.
Times change. Maybe DSP is fast enough today. Maybe the Meyer system has a very narrow bandwidth which gives the designer more scope and less trouble.
B.
* that little blather from Meyer about supersonic jet engines and university inspiration seems about as desembling as any marketing I've ever seen for speakers. And other discussions of their miraculous gear has a disruptive effect on my digestion; much praise and little examination.
Last edited:
And this for a vented box.
The mic is mounted on the bar across the woofer. I would guess the distance from cone to mic, limits the upper frequency for which this will work.
An added benefit of this arrangement will be improved room integration as for lower frequencies the sound "recorded" by the mic will include room modes, which will then be attenuated. So build in bass trap
Think I like the notion of mounting the mic on a bar better than attaching it to the cone, and thereby submitting it to quite high G force.
The distance adds a constant delay and thus reduces the phase margin. 1cm distance is about 30uS of delay (which obviously scales linearly with distance). Sampling also introduces a delay in the order of uS. The max SPL of a measurement mic might actually become relevant once you push the subwoofer to some higher levels.
Yes the microphone will pick up room modes and room problems, but that is only at the point of measurement. This also assumes that the microphone is relatively far from the subwoofer (and thus a large delay). And yes the subwoofer can correct for these problems, BUT it will only correct at the point of measurement. Anywhere else in the room it will not correct. This might lead to "faulty" corrections, because you tend to sit somewhere else as the microphone stands.
Personally I still think acoustic treatment and MFB sub with accelerometer and current drive is the best way to go.
Totally agree..
I did a demo the other day of a current feedback dipole subwoofer and we did comparison with an ordinary driven sub (yes i did eq ing and level adjustment) and the difference in sound is huge.
We both felt that the current feedback sub was playing much lower during sine sweep. But still they measured the same level!
The difference is distortion. Removing distortion gives the illusion of lower level. Very noticable and interesting demo.
Blind testing logic means you must have identical acoustic FR. No one doubts you can equalize any speaker to any curve, at least for blind testing.
Nice FR is not a main benefit of MFB today with DSP. But extending the band without too much distortion is a big benefit.
Once FR is the same, and only then, can blind comparisons proceed. Hard to say the consequences of distortion on the perceived loudness of the non-MFB speaker.
B.
A/B testing indeed does indeed only make sense with equal FR (unless you are testing FR related things of course). His findings don't surprise me though. We perceive the harmonics as bass notes. In fact, lots of clubs in the netherlands use the fake bass principle where they lower the fundamental but increase the 2nd and 4th harmonic. It gives the perception of more bass while not having excessive noise going through the walls. Very effective, not really my thing for hifi though.
Transconductance amps are much more effective when used with MFB though. The high sensitivity near the resonance frequency increases loopgain and thus reduces distortion in the most critical area!
Using transconductance amplifiers still reduce the distortion caused by Le(x), but they create a big peak near Fres of the woofer. You can fix it with sharp filters etc, but I'm not a big fan of that. If you really don't want to use MFB (for whatever reason) you can try and create a dynamic output impedance and try to create voltage amplifier for frequencies near fres and current amplifier for higher frequencies. Tricky to design though, but quite effective for midwoofers. I believe bruno putzeys uses something like this for the midrange in his Kii three. Kii Three is a really a result of awesome engineering where cheap woofers (like 30 euro a piece) result in a phenominal loudspeaker. But that's a different topic entirely.
My new MFB PCB is on the work bench and will be tested soon.
MFB will do the Eq-ing for me since that loop is the outer one, and voltage controlled with dipole correction of course (analogue filter). Then i will use RePhase to shape up the phase and frequency response. ESL63 is used above 200Hz and needs a little eq as well after my mods. Dipole woofers has a lot of excursion and by that benefit most from CFB and MFB.
Measurements will follow on both frequency response and distortion.
Planning to do outside measurements this summer.
MFB will do the Eq-ing for me since that loop is the outer one, and voltage controlled with dipole correction of course (analogue filter). Then i will use RePhase to shape up the phase and frequency response. ESL63 is used above 200Hz and needs a little eq as well after my mods. Dipole woofers has a lot of excursion and by that benefit most from CFB and MFB.
Measurements will follow on both frequency response and distortion.
Planning to do outside measurements this summer.
On one side it's a shame that I can't read about the very interesting project. On the other hand I'm glad it got picked up by a very influencial company. MFB has, unfortunately, been a bit under the radar in commercial designs and it would be great if there was an easy to acces platform for development.
Is there a release goal planned?