Hi, simplest way to remedy driver motor effect on current through the voice coil, which makes the force moving the cone and making acoustic sound, by increasing impedance in the circuit between speaker terminals. Basically add series impedance, or even better use high output impedance amplifier, aka current drive. Now any changes in driver parameters due to excursion, or increasing temperature, hysteresis, what ever, microphoning, makes only minimal variation to circuit impedance because load for the driver is high and current in circuit, the same current that also flows through voice coil, is pretty much exactly what the amplifier feeds in, thus the cone is accelerating exactly as told to. Use DSP to straighten out the response.
They can be both, exceeding a limit in an instant or over time like thermal compression where accumulated heat affects the performance of the driver.
Describing things in your own words instead of using established terms usually leads to confusion.
That does bey no means help against BL(x) and BL(i) nonlinearities. But it helps against problems with Le(x). And it cancels the lowpass pole caused by Le, which can be a big plus when using motional feedback (MFB).
There are still many problems which have to be solved with MFB: Drivers can only be used in their strictly pistonic range, otherwise there is no exact correlation betwen accelaration and sound pressure. The sensors have to be highly linear - also over time because their nonlinearity is the dominant factor in the whole feedback system. I always wondered what kind of microphone they used at Meyer in their X-10 monitor given that this thing must be linear up to SPLs of > 120 dB (and at the same time have very low noise).
Regards
Charles
Hi,
Ok. Thanks to Fluid comments and your answer to it i think it's clearer.
I think there is a misunterdanding in our differents target: mine is to use whatever tool possible which can minimise some artefacts generated at loudspeaker stage ( this is the most harmful place in the reproduction chain just before the room). It include use of FIR for some ten years now in my case. I use complementary filters ( xover) as they helped mitigated some issues i had but wasn't in any way a magic bullet and introduced some drawback as i have a use of live monitoring of input source.
I tryed some FIR tools for other duty ( the approach Mark mentioned and initially started the thread) and not been convinced by end results. Same thing with DRC.
However, there is enough example of users which managed impressive results, documented it and are satisfied with them:
Wesayso, Cask05, Mitchba,...
So worth being aware of what they do/did, try it and then makes your own pov.
From some previous discusion and this one i see your approach as a try to solve every issue a loudspeaker have or could have. I understand but don't see the point for a system in home use.
By specifying your design target and wise choice of part and approach you can mitigate 90% of the issue you list. Just by design choice and knowing what is gonna be your typical use and not deviate from it.
In fact i can see 2 use of this kind of approach: live and the tiniest boombox.
In both you'll push the use past or close to the boundary of SOA of system and in this case then a complete monitoring of the loudspeaker/amp can be needed. In the live case to protect the system in the boombox to squeeze the last drop of juice tiny system could output.
In between this two extreme i don't see the point to include something like this if the loudspeaker have been planed according to a set of design criteria.
Whatever you do you'll hit limits at one point or another, so... better design well into SOA and then use tools to help what can be in my view.
Motion feedback is a powerful tool. I would not use it to counteract any issue in the world but it's powerful!
The Philips loudspeakers John K talked about were astonishing. Last pair i've heard was ten years ago and they kept comparison to entry level studio monitors quite well.
There is a whole diy project in France call 'the Zippy' including motion feedback amplifier boards developed by the designer ( JCB, which is also an acoustician, maintenance engineer, etc,etc,... well known and respected in Paris/ France studio's world ( he was ten years ago)).
Sub ( dayton)+ beyma coax ( 12ax30 iirc). I think the threeways are motion feedback controled independently ( and iirc he used the motion feedback to implement a Linkwitz Transform into the loop for the sealed dayton 15"...).
It's all there but in french.
https://www.homecinema-fr.com/forum/diy-general/zippy-enceinte-integralement-asservie-t29923222.html
Apologize for the OT.
Hi Charles,
I wonder if they not ended with some optical cells instead of a mic?
And i think the anecdote Mark reported was true for inwall instalation ( iirc the way loudspeaker interact with the room is changed when inwall making them much sensible to perturbation comming from the room... but i may be wrong).
I wonder if they not ended with some optical cells instead of a mic?
And i think the anecdote Mark reported was true for inwall instalation ( iirc the way loudspeaker interact with the room is changed when inwall making them much sensible to perturbation comming from the room... but i may be wrong).
Last edited:
However, if they're compensated instantaneously they will be compensated over time, but the converse doesn't seem to be true.
What many people (including JBL) show for power compression variation is the SPL/phase at high power. The problem I see with this approach is that it only introduces a different set of conditions (thermal, power, ageing etc.) but the measurement technique remains the same as before.
It therefore, counteracts parameter variations only but not non-linearity, as frequency response / impulse response based measurement is a method for linear systems.
Open-loop systems do not have correcting mechnisms but closed-loop systems do have (at least) one mechanism built around them to guarantee linearity. How do you follow that ?
I don't know if that would help but the direction of thinking seems to be correct. That is, linearise the system before applying any methodologies meant for linear systems.
Last edited:
SOA, the treshold at which a system is non linear... same thing.
So basicaly i think we are in agreement. As i'm sure ( no i know) are other members participating! 🙂
Current drive have interesting properties for sure (i've seen it used to drive Atc dome mid in some place. Doesn't seemed to bring nasty and the mids were good sounding) . As does motion feedback or other means... it depend what you want to achieve.
So what are FIR tools good at and what are the drawbacks?
So basicaly i think we are in agreement. As i'm sure ( no i know) are other members participating! 🙂
Current drive have interesting properties for sure (i've seen it used to drive Atc dome mid in some place. Doesn't seemed to bring nasty and the mids were good sounding) . As does motion feedback or other means... it depend what you want to achieve.
So what are FIR tools good at and what are the drawbacks?
Last edited:
People often assume home systems to be linear as they generally use low power. However, there could be significant cone excursion, especially in the bass units of even home systems. I say this because people buy woofers to use them to the fullest, but a compensation scheme that requires high linearity would result in under-utilisation of its excursion capabilities.
That is very true and double differentiation of the output of an interferometer is a bad idea.
Using a microphone to feed back SPL instead of force/acceleration would inject sounds from other channels, the vicinity and room effects into the feedback path, which I think may not be of much help.
Yes, definitely. It makes sense to measure the acoustic output when the parameter that is to be controlled is also the acoustic output. No easy task with a vented box anyway.
There are also systems using differential pressure sensors like Powersoft's Ipal.
Regards
Charles
There are also systems using differential pressure sensors like Powersoft's Ipal.
Regards
Charles
Krivium, thanks for the words of confidence earlier.
Been meaning to say so, but have been tied up on another DIY project, putting in a couple of ductless mini-split heat pumps in my home.
Never done any HVAC before, it's kind fun taking on new challenges....and saving a lot of money for more DIY audio 😀
So you've heard the X-10 ! Impressions?
Found this old brochure: https://www.audioheritage.org/vbulletin/attachment.php?attachmentid=5800&d=1109853988,
and link on ASR about them. https://www.audiosciencereview.com/forum/index.php?threads/meyersound-x10-speakers.12437/
Any chance you've heard the Bluehorns?
I'm a big Meyer fan, and student of their philosophy/design goals. It was listening comparisons around year 2000, with some of their boxes vs my electrostats, that transformed my view on audio.
My stats had a bit better clarity and detail, but the Meyers were nearly too dang close to believe.... And lordy, how the they could rock!
So bye-bye stats being my main speakers anymore....
I got UPA-1p's and a 650-p sub for smaller rooms/setup, and a pair of MTS4a's for shoot-for-the moon big.. Used them all in a very large surround setup.
The X-10 came out right about the time I bought all those, so I had keen interest in them, even if the price was way out of reach.
Anyway, I'm bringing up the Meyer stuff because it kinda comes full circle to the way the thread turned to discussing large signal issues.
One of Meyer's oft stated goal is to produce the highest linear SPL possible.
Personally, I think having fully linear SPL, including linearity needed for peak headroom/transients, to rank right below frequency response, in terms of factors affecting sound quality. I even put it ahead of DI.
(I don't mean having some enormous max SPL capability; I mean maintaining linearity for whatever system SPL, and bass extension, one desires/designs to. )
Lack of SPL linearity, particularly for the low freq end of response, is the major SQ killer in home audio, imho.
Heck, in many cases, it's not even about lack of displacement or motor non-linearities. Pushing a few numbers around shows home audio amps often have no way of delivering the peak voltages and current needed for transients.
Anyway again, studying Meyer's speakers and Bob McCarthy's "System Design and Optimization" book opened my eyes, that good audio is more about good solid engineering, tha all the esoteric minutiae I had been following.
It's also been the starting source of my fascination with the importance of phase.
Been meaning to say so, but have been tied up on another DIY project, putting in a couple of ductless mini-split heat pumps in my home.
Never done any HVAC before, it's kind fun taking on new challenges....and saving a lot of money for more DIY audio 😀
So you've heard the X-10 ! Impressions?
Found this old brochure: https://www.audioheritage.org/vbulletin/attachment.php?attachmentid=5800&d=1109853988,
and link on ASR about them. https://www.audiosciencereview.com/forum/index.php?threads/meyersound-x10-speakers.12437/
Any chance you've heard the Bluehorns?
I'm a big Meyer fan, and student of their philosophy/design goals. It was listening comparisons around year 2000, with some of their boxes vs my electrostats, that transformed my view on audio.
My stats had a bit better clarity and detail, but the Meyers were nearly too dang close to believe.... And lordy, how the they could rock!
So bye-bye stats being my main speakers anymore....
I got UPA-1p's and a 650-p sub for smaller rooms/setup, and a pair of MTS4a's for shoot-for-the moon big.. Used them all in a very large surround setup.
The X-10 came out right about the time I bought all those, so I had keen interest in them, even if the price was way out of reach.
Anyway, I'm bringing up the Meyer stuff because it kinda comes full circle to the way the thread turned to discussing large signal issues.
One of Meyer's oft stated goal is to produce the highest linear SPL possible.
Personally, I think having fully linear SPL, including linearity needed for peak headroom/transients, to rank right below frequency response, in terms of factors affecting sound quality. I even put it ahead of DI.
(I don't mean having some enormous max SPL capability; I mean maintaining linearity for whatever system SPL, and bass extension, one desires/designs to. )
Lack of SPL linearity, particularly for the low freq end of response, is the major SQ killer in home audio, imho.
Heck, in many cases, it's not even about lack of displacement or motor non-linearities. Pushing a few numbers around shows home audio amps often have no way of delivering the peak voltages and current needed for transients.
Anyway again, studying Meyer's speakers and Bob McCarthy's "System Design and Optimization" book opened my eyes, that good audio is more about good solid engineering, tha all the esoteric minutiae I had been following.
It's also been the starting source of my fascination with the importance of phase.
Sorry for asking such DSP noob questions - despite having implemented a crossover software - I just derived filter coefficients from existing work (translating pylab code to C) or plainly copied them from the interwebs. 😉
So I guess the frequency of the filters reside in the "s² + 1.28299239215439s + 2.47283345436118" formulae? Could you point me to a resource that describes a) what this form is called and b) how to calculate biquad coeffictions from these forms? I'd like to try and add 8th order bessel lowpass to pulseaudio crossover rack, or maybe even build a complete LADSPA plugin for the hp/lp cases of the described IIR phase linear filters here in this thread...
Last edited:
I'm not an electrical engineer, nor an expert in control theory or communication theory... but this line of reasoning strikes me as slightly misleading
Would it be a true statement that "Linear is as Linear does" ? In other words, a system is linear if it behaves linearly, regardless of whether it uses local or global feedback. A resistor behaves linearly without any feedback, to use a simplistic example.
As @fluid points out, most speakers perform well on dynamic compression tests. It is not hard to find drivers in which THD is -40 dB over their intended frequency range. That sounds pretty linear to me, without the benefit of feedback.
j.
Would it be a true statement that "Linear is as Linear does" ? In other words, a system is linear if it behaves linearly, regardless of whether it uses local or global feedback. A resistor behaves linearly without any feedback, to use a simplistic example.
As @fluid points out, most speakers perform well on dynamic compression tests. It is not hard to find drivers in which THD is -40 dB over their intended frequency range. That sounds pretty linear to me, without the benefit of feedback.
j.
The parameter variations due to power compression are usually not severe in (adequately-sized) home system drivers, but the excursion-induced non-linearity is still present, as it is an inherent property. A driver may appear to be subjectively linear if X<<Xmax, and you may have it that way if you like it. However, what I was talking about was the mathematical correctness of the technique.
I only wanted to mention how inherently non-linear systems with feedback could easily be more linear than ones without it and mathematically speaking, an amplifier without feedback must be considered non-linear, much like a typical loudspeaker. A resistor is not a complicated electro-mechanical system with moving parts, so I think it's a poor comparison here.
I only wanted to mention how inherently non-linear systems with feedback could easily be more linear than ones without it and mathematically speaking, an amplifier without feedback must be considered non-linear, much like a typical loudspeaker. A resistor is not a complicated electro-mechanical system with moving parts, so I think it's a poor comparison here.
I have a python impl which you can see at https://github.com/3ll3d00d/beqdesi.../src/main/python/model/jriver/filter.py#L2128 (am adding it to my jriver DSP tool). I guess it is a bit shrouded in jriver specific stuff but probably you can read through that to see the dsp parts (but otherwise happy to give pointers via pm)
Last edited:
Motional voice coil feedback has been around even before the 60s. As a young Electronic Countermeasures EE and audio freak, I saw Philips, Grundig and the Japanese doing it then.
Do any of you remember the Integrand, a servo'd audio system that even compensated for drapes and people in the listening room?
Since I like classical cathedral pipe organs, I recently tried to reproduce them in my RAV4 SUV. The RAV4's Class D-driven subwoofer poops at 35 Hz. I needed 16. I'd considered using my Sunfire True Subwoofer's Kevlar-coned driver/2.7 kW amplifier in the RAV4, with a 12VDC to 120VAC psueudo-sinewave inverter, but decided the Sunfire's 14 pound magnet would screw up the RAV4's compass. It really screwed up convergence in my ancient Sony Trinitron TVs. I have some funny stores about designing hardware for 0.35 Tesla superconducting MRI systems.
I recently tried a 12 inch Aluminum cone driver in a 1.5 cu ft sealed enclosure and found the THD went ballistic (>30%) below 24 Hertz. I tried second VC, Current, Voltage, Accelerometer, MEMS pressure sensor FB and and even a 10 GHz Doppler radar modulator FB, with tinfoil on the dust cone. Nothing worked. Never got around to Time-of-Flight LIDAR FB.
While the above or combos thereof, will reduce nonlinearity from the driver's magnetics and suspension, they don't address the real problem, non-pistonic, eigenmode or drumhead-like LF breakup of the cone. That's the ill-defined effect that really sets a subwoofer's "Lowest Usable Frequency".
One fix is a super-stiff driver cone. Another is a bladed-fan rotary driver. But you'll have to mount it in a bus.
If it sounds crappy, use the crap as inverse FB to reduce itself. Phase correction is a bear to avoid creating a howling banshee, but it's obviously been done. According to my crusty, trusty 1955 Edition of Frederick Terman's 1932 Electronics and Radio Engineering, a 30 dB FB audio system needs faithful inverted phase response to well-below 100 milliHertz, to avoid gain peaking and/or oscillation. Those ULF FB systems are susceptible to all sorts of listening room disturbances.
I'd like to use my ADAU1701 DSP, but ADI's SigmaStudio canned s/w seems to poop below 1 Hz. I hate to spend my time DIYing or importing FIR and/or IIR routines.
Do any of you remember the Integrand, a servo'd audio system that even compensated for drapes and people in the listening room?
Since I like classical cathedral pipe organs, I recently tried to reproduce them in my RAV4 SUV. The RAV4's Class D-driven subwoofer poops at 35 Hz. I needed 16. I'd considered using my Sunfire True Subwoofer's Kevlar-coned driver/2.7 kW amplifier in the RAV4, with a 12VDC to 120VAC psueudo-sinewave inverter, but decided the Sunfire's 14 pound magnet would screw up the RAV4's compass. It really screwed up convergence in my ancient Sony Trinitron TVs. I have some funny stores about designing hardware for 0.35 Tesla superconducting MRI systems.
I recently tried a 12 inch Aluminum cone driver in a 1.5 cu ft sealed enclosure and found the THD went ballistic (>30%) below 24 Hertz. I tried second VC, Current, Voltage, Accelerometer, MEMS pressure sensor FB and and even a 10 GHz Doppler radar modulator FB, with tinfoil on the dust cone. Nothing worked. Never got around to Time-of-Flight LIDAR FB.
While the above or combos thereof, will reduce nonlinearity from the driver's magnetics and suspension, they don't address the real problem, non-pistonic, eigenmode or drumhead-like LF breakup of the cone. That's the ill-defined effect that really sets a subwoofer's "Lowest Usable Frequency".
One fix is a super-stiff driver cone. Another is a bladed-fan rotary driver. But you'll have to mount it in a bus.
If it sounds crappy, use the crap as inverse FB to reduce itself. Phase correction is a bear to avoid creating a howling banshee, but it's obviously been done. According to my crusty, trusty 1955 Edition of Frederick Terman's 1932 Electronics and Radio Engineering, a 30 dB FB audio system needs faithful inverted phase response to well-below 100 milliHertz, to avoid gain peaking and/or oscillation. Those ULF FB systems are susceptible to all sorts of listening room disturbances.
I'd like to use my ADAU1701 DSP, but ADI's SigmaStudio canned s/w seems to poop below 1 Hz. I hate to spend my time DIYing or importing FIR and/or IIR routines.
Last edited:
BTW, LF cone breakup isn't very level-sensitive. It breaks up at 1 Watt, almost as well as at 10 Watts or more, because it's generated by resonance, not by nonlinear cone in-and-out motion.
Until just now, I wasn't aware that Meyer or any product actually used pressure FB. My 100 Pascal MEMS analog pressure sensor had too much Johnson noise for audio FB. Once I'd realized my problem was cone breakup, a lower Pascal-rated pressure sensor, Doppler RADAR or Time-of-Flight LIDAR FB would be wastes of time.
Until just now, I wasn't aware that Meyer or any product actually used pressure FB. My 100 Pascal MEMS analog pressure sensor had too much Johnson noise for audio FB. Once I'd realized my problem was cone breakup, a lower Pascal-rated pressure sensor, Doppler RADAR or Time-of-Flight LIDAR FB would be wastes of time.
Last edited: