Hi all. Another year, another EE project. I always seem to levitate towards projects that involve audio since audio it is a hobby of mine. =)
So, I've been thinking of creating a servo controlled subwoofer. Implementation of the feedback circuit is presumably going to be done via a 16bit integer based motorola DSP (don't remember which off the top of my head). I was wondering if any of you digital guys and gals had papers, articles, or anything related to the theory behind servo controlled stuff (doesn't have to be about subs for the related theory to become apparent) done in the digital domain, meaning no differential amps, analog circuits, or the like. Google is usually my friend but all I can seem to come up with is servo controlled sub advertisements. I will continue looking as well as updating the post with progress and eventually code and circuit layouts for the final project (if it so turns out well).
Last year I made an automated equalization unit like the behringer DEQ2496's autoeq feature on the same DSP. Suffice it to say it was a major PITA b/c of the very limited precision the 16bit int platform but by george it worked and earned me a nice grade. It did get a little noisy after heavy equalization b/c of quantization errors but it was the best that could be done with limited equipment. Oh yeah, I might also be able to land my hands on a more powerful 32bit fp DSP is I dig around a bit and ask nicely.
So, I've been thinking of creating a servo controlled subwoofer. Implementation of the feedback circuit is presumably going to be done via a 16bit integer based motorola DSP (don't remember which off the top of my head). I was wondering if any of you digital guys and gals had papers, articles, or anything related to the theory behind servo controlled stuff (doesn't have to be about subs for the related theory to become apparent) done in the digital domain, meaning no differential amps, analog circuits, or the like. Google is usually my friend but all I can seem to come up with is servo controlled sub advertisements. I will continue looking as well as updating the post with progress and eventually code and circuit layouts for the final project (if it so turns out well).
Last year I made an automated equalization unit like the behringer DEQ2496's autoeq feature on the same DSP. Suffice it to say it was a major PITA b/c of the very limited precision the 16bit int platform but by george it worked and earned me a nice grade. It did get a little noisy after heavy equalization b/c of quantization errors but it was the best that could be done with limited equipment. Oh yeah, I might also be able to land my hands on a more powerful 32bit fp DSP is I dig around a bit and ask nicely.
No nibbles yet huh?
Does no one do this stuff besides the big boys (meaning no DIY guys versus actual companies).
Does no one do this stuff besides the big boys (meaning no DIY guys versus actual companies).
I'm wondering how you plan to get a decent feedback from the speaker.
Somme do 't with piezo, but thats far from lineair.
If you could use some optical/sonical/ultrasound distance measurement you could leave the specs of the driver unchanged.
Regards and good fun with your project
Simon
Somme do 't with piezo, but thats far from lineair.
If you could use some optical/sonical/ultrasound distance measurement you could leave the specs of the driver unchanged.
Regards and good fun with your project
Simon
Feedback control of a speaker is a tricky issue. You probably want to ask on the speaker groups anyway. There have been quite a few attempts, but they are often flawed by what I tend to feel is a misapprehension about what it is you can do.
Typically the feedback sensor is an accelerometer, and so the first thing people do is to integrate the input to get velocity, and then sometimes even try again in the hope they get position. Perhaps in the unfortunate hope that the input voltage to the speaker's amplifier is supposed to translate to cone position. Then they try to create a feedback loop. The whole thing is pretty much doomed, ending in a marginally stable mess that often yields very little improvement. Draconian bandwidth limiting of the input helps stability and may yield something useful.
An alternative approach I have been hoping to try is a feed-forward approach. Use the sensor, not as a feedback input device, but as the input to characterise speaker, building a parameterised model of the speaker. Then using this model, dynamically perturb the signal to correct for the modelled errors.
You could certainly cope with Bl transfer non-linearities - both static limits from the magnetic gap geometry and dynamic errors due to flux modulation from the voice coil., thermal non-linearities (you could build a paramaterised dynamic model of the voice coil temperature and thus correct for dynamic resistance changes.)
You would of course also need to know the ideal Thiel Small derived behaviour of the speaker in the particular enclosure. You could also look to adding artificial alignment changes (usually bass boost - but they needn't be) and also - since you have a quite complete model of the driver - you could perform dynamic level limitation - and never drive the speaker outside of its real physical (i.e. excursion and thermal) limits.
Such correction could be useful across a much wider bandwidth than a simple feedback system, and I would bet generate vastly better results.
My two cents anyway 🙂
Typically the feedback sensor is an accelerometer, and so the first thing people do is to integrate the input to get velocity, and then sometimes even try again in the hope they get position. Perhaps in the unfortunate hope that the input voltage to the speaker's amplifier is supposed to translate to cone position. Then they try to create a feedback loop. The whole thing is pretty much doomed, ending in a marginally stable mess that often yields very little improvement. Draconian bandwidth limiting of the input helps stability and may yield something useful.
An alternative approach I have been hoping to try is a feed-forward approach. Use the sensor, not as a feedback input device, but as the input to characterise speaker, building a parameterised model of the speaker. Then using this model, dynamically perturb the signal to correct for the modelled errors.
You could certainly cope with Bl transfer non-linearities - both static limits from the magnetic gap geometry and dynamic errors due to flux modulation from the voice coil., thermal non-linearities (you could build a paramaterised dynamic model of the voice coil temperature and thus correct for dynamic resistance changes.)
You would of course also need to know the ideal Thiel Small derived behaviour of the speaker in the particular enclosure. You could also look to adding artificial alignment changes (usually bass boost - but they needn't be) and also - since you have a quite complete model of the driver - you could perform dynamic level limitation - and never drive the speaker outside of its real physical (i.e. excursion and thermal) limits.
Such correction could be useful across a much wider bandwidth than a simple feedback system, and I would bet generate vastly better results.
My two cents anyway 🙂
The feedforward thing is quite nice and shown to be reliable since it is done often within active systems although not usually done in an adaptive fashion like suggested by Francis.
I don't know how much THD reduction could be achieved by such an approach. But response correction can definitely be done. The advantage of the adaptive system would be the automatic incorporation of manufacturing tolerances and ageing.
There is even a system around that takes the room interaction into account as well (although the manufacturer is not my favourite one 😉 ):
http://www.bang-olufsen.com/web2/systems/product.asp?page=ts§ion=systems&sub=ls&ProdID=544
Regarding servo control:
Another alternative to position- and motion- sensors is the measuring of the actual audio output.
http://www.meyersound.com/products/studioseries/x-10/index.htm
Regards
Charles
I don't know how much THD reduction could be achieved by such an approach. But response correction can definitely be done. The advantage of the adaptive system would be the automatic incorporation of manufacturing tolerances and ageing.
There is even a system around that takes the room interaction into account as well (although the manufacturer is not my favourite one 😉 ):
http://www.bang-olufsen.com/web2/systems/product.asp?page=ts§ion=systems&sub=ls&ProdID=544
Regarding servo control:
Another alternative to position- and motion- sensors is the measuring of the actual audio output.
http://www.meyersound.com/products/studioseries/x-10/index.htm
Regards
Charles
A few years ago I wanted to build a servo feedback sub and therefore gathered all the infos about I could get. It took quite some time to dig everything up but in the end there was enough info to start with. Unfortunately the only limiting factor for me is always time and so I put the project on ice since I wanted to realise other more important projects first. I still have all these infos and documents but don't know what I can make available here as some of the articles are from electronics magazines and there are sure some copyright issus to take care of. Maybe in the afternoon I have some time to look into it what I can spare.
Anyone interested in?
Anyone interested in?
I once owned the original of the following interesting article (in German only unfortunately) about the microphone based servo control of a German BM speaker:
http://www.johannes-krings.com/funkschau2
The subject has already been discussed on this forum about two years ago.
Regards
Charles
http://www.johannes-krings.com/funkschau2
The subject has already been discussed on this forum about two years ago.
Regards
Charles
Hi Using_e
Yes, I am very interested in every thing related to servo-loudspeakers. I've collected quite a lot of documentation (many patents) but still looking for more. Feedback stability is never fully treated.
~~~~~ Forr
§§§
Yes, I am very interested in every thing related to servo-loudspeakers. I've collected quite a lot of documentation (many patents) but still looking for more. Feedback stability is never fully treated.
~~~~~ Forr
§§§
I as well am interested using_e and will be willing to share my progress as the project (hopefully) continues. If you don't want to post, send me an email instead (dswiston AT uiuc.edu)
Tomorrow I will be discussing the idea with a professor of mine. The feedforward idea also sounds interesting so I will bring that up. BTW, this particular professor is the creater of the continuum fingerboard so I'm hoping to get some good info out of him.
Link if anyone is interested:
http://www.hakenaudio.com/Continuum/
Thanks for the ideas/suggestions/info guys. If anyone has more to share, please do.
Tomorrow I will be discussing the idea with a professor of mine. The feedforward idea also sounds interesting so I will bring that up. BTW, this particular professor is the creater of the continuum fingerboard so I'm hoping to get some good info out of him.
Link if anyone is interested:
http://www.hakenaudio.com/Continuum/
Thanks for the ideas/suggestions/info guys. If anyone has more to share, please do.
Ok, I've now scanned, enhanced, converted and optimized all the articles I found up to now. They are now available as PDF files. Just note that they have been converted in the best possible quality so if you cannot read a word or a part of an article I cannot do anything about it because I don't have this article in a better quality either. So please don't mail and ask for hires scans or so because they are already up in the best quality possible.
They have been converted to Acrobat 7 format to get the smallest file size possible so please don't mail and inform me that the files are corrupted. They aren't just use Acrobat Reader 7 to open the files and you will be just fine. I've tested them all and they work.
Files will be up on the following link for a few hours only. I think it's clear why. If someone misses the first phase just drop me a message and we will find a way.
Download Link
Enjoy
using_e
They have been converted to Acrobat 7 format to get the smallest file size possible so please don't mail and inform me that the files are corrupted. They aren't just use Acrobat Reader 7 to open the files and you will be just fine. I've tested them all and they work.
Files will be up on the following link for a few hours only. I think it's clear why. If someone misses the first phase just drop me a message and we will find a way.
Download Link
Enjoy
using_e
hopefully your university library will have the Audio engineering society journal JAES - the place for articles on speaker distortion/control/equalization
http://www.klippel.de/background/introduction.asp should be a good source to start a literature search
depending on your skills/ambitions the topic could reach grad thesis level complexity
http://www.klippel.de/background/introduction.asp should be a good source to start a literature search
depending on your skills/ambitions the topic could reach grad thesis level complexity
using_e:
Thank you so much for the articles.
jcx:
We do have AES journals avaliable, thanks for the suggestion, it didn't register with me to look.
So far I've only spoken with the professor and had a chance to speak about securing a DSP to use.
Thank you so much for the articles.
jcx:
We do have AES journals avaliable, thanks for the suggestion, it didn't register with me to look.
So far I've only spoken with the professor and had a chance to speak about securing a DSP to use.
Hi Using_E,
Thanks for the articles.
Another source for servo-subs is patents (Free patents on Line).
My own collection of articles on the subject has now about forty authors and is 70 Mb of hard disk space (most articles in PDF).
Sorry, I am unable to establish the full list for now.
~~~~~~ Forr
§§§
Thanks for the articles.
Another source for servo-subs is patents (Free patents on Line).
My own collection of articles on the subject has now about forty authors and is 70 Mb of hard disk space (most articles in PDF).
Sorry, I am unable to establish the full list for now.
~~~~~~ Forr
§§§
Have you seen:
http://www.rythmikaudio.com/servo_product.htm
It isn't exactly what you are looking for, if you are looking for a digital solution, but has anyone tried this out? It has always seemed pretty cool to me.
http://www.rythmikaudio.com/servo_product.htm
It isn't exactly what you are looking for, if you are looking for a digital solution, but has anyone tried this out? It has always seemed pretty cool to me.
Nice link. I'll have to take a wonder over to the patent website and lookup the exact implementation they use (cicuitry). If anything else, it serves as good read. They make a point of explaining what some causes of the woofer distortion are.
They also seem to poo-poo on the accelerometer based approach. The "mechanical" ground issue seems to be a valid one and would dictate the sub inclosure be heavy and dead to vibrations.
If anyone wants to add or disagree with something I said feel free. Discussion is good!
An accelerometer is on order and my lab partner is working on securing a DSP that we can work with. Hopefully solid progress can be reported soon.
They also seem to poo-poo on the accelerometer based approach. The "mechanical" ground issue seems to be a valid one and would dictate the sub inclosure be heavy and dead to vibrations.
This statement doesn't convince me of anything other than to implant a seed of doubt, after all, they are selling something.
This claim seems to have more validity to its claim. After all, in some strange instance say the sub were to tip over or fall off something while on, the corrective system could damage it or the amplifier. A DSP based approach seems to lend a hand at incorporating protective circuitry without degrading signal quality though.
If anyone wants to add or disagree with something I said feel free. Discussion is good!
An accelerometer is on order and my lab partner is working on securing a DSP that we can work with. Hopefully solid progress can be reported soon.
Unfortunately the "mechanical ground issue" is valid for ALL other approaches as well ! So claiming that their's isn't susceptible to it is pure marketing BS IMHO.
I have never heard a Rythmik SUB admittedly and they may be fine sounding devices.
But the figures they present don't look that promising at all: A reduction form 15% k2 to 10% is only an improvement form very bad to bad and doesn't make one think that the principle is very effective. I know that distortion audibility is not very high at 20 Hz. But the driver excursion was only 1/4" p-p, which is still in the linear range of many mid-woofers.
Regards
Charles
Indeed, I thought the table comparing DSP with analog was either simply naive, or just plain bogus. There is nothing the DSP can't do the op-amp based design can, and the DSP has some intrinsic advantages, in particular it can include elements of time.
The other approaches rather run the range of very simplistic, to almost but not quite there.
Those designs that ignore the Theil Small behaviour of the cone are doomed to failure. But any design that seeks to correct the behaviour of the cone in a simple feedback loop should be doing some serious stability analysis, and may find some unfortunate problems. The very modest improvements seen in almost all of the works points to this.
The DSP based approach with active change to the filter coefficients is getting there, but also misses the point. The distortion mechanisms in a speaker are largely not minimum phase or model-able as something invertible with a FIR.
There does seem to be an element of hammer and nails (when all you have is a hammer, everything looks like nail.) What is needed is a model of speaker motion that includes the distortion mechanisms in their basic physical form. The issue of correcting the distortion then becomes one of parameter estimation, something there is a good body of knowledge of. But the first task is to decide on a minimum set of parameters, and finding orthogonal parameters is crucial to success. If the parameters are coupled the task of estimation, and then correction will be vastly harder. Orthogonal parameters are very likely to fall out of an analysis of the physical nature of the speaker. Applying the wrong model is likely to yield a slew of coupled parameters and an infeasable problem.
Got right such a model will find that the sensor system will spend most of its time doing nothing more than very slight fine tuning of the parameters, if anything at all. This essentially removes any issues of system stability from view.
The other approaches rather run the range of very simplistic, to almost but not quite there.
Those designs that ignore the Theil Small behaviour of the cone are doomed to failure. But any design that seeks to correct the behaviour of the cone in a simple feedback loop should be doing some serious stability analysis, and may find some unfortunate problems. The very modest improvements seen in almost all of the works points to this.
The DSP based approach with active change to the filter coefficients is getting there, but also misses the point. The distortion mechanisms in a speaker are largely not minimum phase or model-able as something invertible with a FIR.
There does seem to be an element of hammer and nails (when all you have is a hammer, everything looks like nail.) What is needed is a model of speaker motion that includes the distortion mechanisms in their basic physical form. The issue of correcting the distortion then becomes one of parameter estimation, something there is a good body of knowledge of. But the first task is to decide on a minimum set of parameters, and finding orthogonal parameters is crucial to success. If the parameters are coupled the task of estimation, and then correction will be vastly harder. Orthogonal parameters are very likely to fall out of an analysis of the physical nature of the speaker. Applying the wrong model is likely to yield a slew of coupled parameters and an infeasable problem.
Got right such a model will find that the sensor system will spend most of its time doing nothing more than very slight fine tuning of the parameters, if anything at all. This essentially removes any issues of system stability from view.
There was once a JAES article dealing with exactly that but I can't even remember what year but I think it's more than 10 years back.
And be aware that even a model that would be called basic would not intrinsically mean simple.
Regards
Charles
you mean like:
http://www.klippel.de/pubs/The Power of Loudspeaker Models.pdf
(sorry to make it so tough on you guys, thought you could figure out to click "online pubs" in the klipple intoduction link i gave before: http://www.klippel.de/pubs/default.asp )
http://www.klippel.de/pubs/The Power of Loudspeaker Models.pdf
(sorry to make it so tough on you guys, thought you could figure out to click "online pubs" in the klipple intoduction link i gave before: http://www.klippel.de/pubs/default.asp )
Well that presentation pretty well gets my vote 🙂. (And I did miss checking out the publications.)
I'm not sure why their modelling is either Generic (which in general I would agree is a bad idea - although arguablely the Voltera kernel is all encompasing - and thus potentially useful) or based upon memoryless models. It strikes me that creep problems can't be modeled without memory. But I'm no mech eng. 🙂
As a guide to a very wide ranging EE project, this would have to be close to spot on.
I'm not sure why their modelling is either Generic (which in general I would agree is a bad idea - although arguablely the Voltera kernel is all encompasing - and thus potentially useful) or based upon memoryless models. It strikes me that creep problems can't be modeled without memory. But I'm no mech eng. 🙂
As a guide to a very wide ranging EE project, this would have to be close to spot on.
- Status
- Not open for further replies.