Sounds good. Thanks. Looking forward to further development and testing. Your approach is certainly fertile for the future.
But not many people are aware of the daunting challenges to testing DSP programs. Not easy to test for DSP software performance let alone input-output performance. Big challenges. That's why I am keen to see the test done acoustically because hearing better sound is our chief purpose here.
Guess I was just a bit ticked-off when you told me to sign off the thread. Dunno why.
Ben
Footnote: for a long time, I've been asking to see performance specs for DSP gizmos. How much garbage of various sorts do they add to output signals? Funny, in all other areas of sound reproduction we are intent on reading performance specs. Where are the DSP specs?
But not many people are aware of the daunting challenges to testing DSP programs. Not easy to test for DSP software performance let alone input-output performance. Big challenges. That's why I am keen to see the test done acoustically because hearing better sound is our chief purpose here.
Guess I was just a bit ticked-off when you told me to sign off the thread. Dunno why.
Ben
Footnote: for a long time, I've been asking to see performance specs for DSP gizmos. How much garbage of various sorts do they add to output signals? Funny, in all other areas of sound reproduction we are intent on reading performance specs. Where are the DSP specs?
Last edited:
be patient my friend. I could have turned up with the finished product 6 months later and you could have said "why didn't you put this feature in it or that feature etc". In this way I spread some of my cards on the table and get some feedback and criticisms at the same time.
Regarding DSP the specs that are given don't necessarily translate into sound quality because the DSP is just a computation engine and nothing more and nothing less. It's like saying does your Intel i7 sound better than an Intel i5 ?
Some DSP's like the one I am using do have extra audio peripherals so they may have an effect on the sound qualkity if they are used. At the end of the day, given that a DSP has enough number crunching capability the effect on the audio performance has a lot to do with the word depth or how many bits per word, single or double precision, sample rate, the actual algorithms used IIR vs FIR filters , floating point vs fixed point etc. And then you have to interface it to the outside world using ADC's and DAC's and there is a lot of controversy about which part to use. Using digital interface there are a number of different standards and each has their pros and cons such as I2S vs SPDIF etc..... And then there are some DSP vendors who can actually sell you a software plugin to give it a certain tube sound. 😉
At the end of the day you need to work out what your objectives and then focus on that rather than what is going on under the bonnet. Measurements are a good way of sorting out the wheat from the chaff and like you said at the end of the day you want to measure what you are hearing. I have no argument with that whatsoever 😉
regards
david
Regarding DSP the specs that are given don't necessarily translate into sound quality because the DSP is just a computation engine and nothing more and nothing less. It's like saying does your Intel i7 sound better than an Intel i5 ?
Some DSP's like the one I am using do have extra audio peripherals so they may have an effect on the sound qualkity if they are used. At the end of the day, given that a DSP has enough number crunching capability the effect on the audio performance has a lot to do with the word depth or how many bits per word, single or double precision, sample rate, the actual algorithms used IIR vs FIR filters , floating point vs fixed point etc. And then you have to interface it to the outside world using ADC's and DAC's and there is a lot of controversy about which part to use. Using digital interface there are a number of different standards and each has their pros and cons such as I2S vs SPDIF etc..... And then there are some DSP vendors who can actually sell you a software plugin to give it a certain tube sound. 😉
At the end of the day you need to work out what your objectives and then focus on that rather than what is going on under the bonnet. Measurements are a good way of sorting out the wheat from the chaff and like you said at the end of the day you want to measure what you are hearing. I have no argument with that whatsoever 😉
regards
david
Hello Tranquility Bass,
Is there something unique in your implementation of accelerometer based feedback that precludes issues with the servo responding to cabinet motion? Or, are you addressing this behavior via DSP processing.
Tranquility Bass' sensor is moving many millimeters, a well constructed cabinet's walls should move an order of magnitude less.
He hasn't loaded the speaker in a cabinet yet, but if a cabinet is floppy enough for it's motion to be a problem for servo control, that is an issue in itself.
The function of the sensor is not to just respond to large motions of the cone but also to respond to vastly smaller errors of the driver, smaller, I'd guess, than the amplitude of foot-steps on a typical wood floor.
Whether the foot-steps or the counter-action of the cone are audible and/or deemed noxious is another matter. But the system has to be a stable feedback network in any case.
Ben
Whether the foot-steps or the counter-action of the cone are audible and/or deemed noxious is another matter. But the system has to be a stable feedback network in any case.
Ben
motional feedback applied to a driver won't fix up a poorly constructed cabinet. The same rules still apply to cabinet construction 😉
regards
david
Well, that clears things up(haha).
Care to elaborate on what about your physical/electronic implementation method addresses this concern?
DSP fixes are certainly possible, but don't interest me as much.
weltersys & bentoronto already hit on both major points.
1) Feedback loop responding to physical motion of the cabinet unrelated to the input signal.
In particular, a sharp bump from a foot can impart frequency content into the feedback loop well outside the frequency range that the input signal is filtered for. Instabilities can be provoked from bumping the cabinet even if the system appears to be perfectly stable when playing music. This is not generally a problem as most of us aren’t in the habit of kicking our subwoofers, but it can happen. In fact I was recently reminded of this fact during a visit to a local hifi shop. Rapping knuckles on the cabinet of a big name MF subwoofer to try and determine if any internal bracing was used, and I was surprised to see the woofer start oscillating at very low frequency(<20Hz)…amplitude varied reaching max of about 15mmPP and eventually died away after about 10 seconds.
2) Feedback loop responding to physical motion of cone relative to an inertial frame, where as the sound pressure that we want linearized is proportional to the motion of the cone relative to the box. In the case of a very light cone and a very heavy box, they are essentially the same. But, anybody who has experimented with large, heavy, long-throw woofers knows that there is a whole lot of box shaking going on. Once you get a woofer in a box and measure with a microphone, you will find that the odd harmonics will not be reduced as much as the accelerometer measurement suggests. I guess if a down firing woofer is used there would probably be more of an effect on even harmonics, but I did not experiment with that. In any case, the heavier the woofer and the lighter the box, the greater the disparity. Of course distortion will still be much lower than without feedback.
One simple solution for this problem is two use two woofers on opposing sides of the box such that the voice coil forces on the woofer chassis, which are bolted to the box, cancel out. It worked quite well when I tested it some years ago.
Last edited:
Very clear. Thanks. I suspect all boxes shake, except maybe a Klipschorn*.
But remember, we aren't talking about big, obvious, shakes you can feel. We are talking about reducing woofer distortion from say, 2% to .5% and reducing overhang. So the dimensions of interest are quite tiny. They are .5% of the normal cone motion, which itself is usually pretty tiny.
When an accelerometer sensor budges falsely by those tiny amounts, it sends a "correction" feedback to the cone... which produces distortion. Unlike bridge and VC feedback, an accelerometer is not keyed to the motion of the cone relative to the driver basket, but to the outside cosmos. Although not a feasible sensor method anyway, box movement might also be a deficiency of mic sensing.
....unless, heaven forbid, your music room has a concrete floor (and your speaker box is screwed firmly to it), in which case you might be OK.
I believe that feedback based on bridge or second voice coil sensing is pretty much immune to reacting to these external forces and indeed might just correct a wayward cone from budging at all (at its resonance) after rapping real hard on the back of the box.
Ben
*I keep my wine rack on top ot the Klipschorn anyway - very convenient. At other times, with my wife's approval, I have put cinderblocks on speakers. Some people use spike-like feet. Does Monster sell special HiFi cinderblocks?
But remember, we aren't talking about big, obvious, shakes you can feel. We are talking about reducing woofer distortion from say, 2% to .5% and reducing overhang. So the dimensions of interest are quite tiny. They are .5% of the normal cone motion, which itself is usually pretty tiny.
When an accelerometer sensor budges falsely by those tiny amounts, it sends a "correction" feedback to the cone... which produces distortion. Unlike bridge and VC feedback, an accelerometer is not keyed to the motion of the cone relative to the driver basket, but to the outside cosmos. Although not a feasible sensor method anyway, box movement might also be a deficiency of mic sensing.
....unless, heaven forbid, your music room has a concrete floor (and your speaker box is screwed firmly to it), in which case you might be OK.
I believe that feedback based on bridge or second voice coil sensing is pretty much immune to reacting to these external forces and indeed might just correct a wayward cone from budging at all (at its resonance) after rapping real hard on the back of the box.
Ben
*I keep my wine rack on top ot the Klipschorn anyway - very convenient. At other times, with my wife's approval, I have put cinderblocks on speakers. Some people use spike-like feet. Does Monster sell special HiFi cinderblocks?
Last edited:
This is true, voice coil feedback signal is proportional to relative motion between woofer and the box; just what we want. However feedback from a voice coil has its own linearity issues from BL variation with position, which is one of the main sources of woofer distortion in the first place. The voice coil feedback signal is in phase with the velocity of the cone, which is NOT in phase with the position. So correction of the BL-vs-position variation isn't a particularly simple task, although I'm sure DSP could handle it.
I could be wrong, but I think the voice coil MF systems available today tend to be less prone to reacting to external forces mainly because they use much lower loop gain.
I respectfully ask if you have good reason to say motor irregularities are the "main source" of problems... say with modern high-quality drivers? Perhaps some of those irregularities are negated by listening to the voice coil or second winding (or possibly doubled!!!)?
Also, why should absolute position of the cone be much of a concern (assuming restoration speeds are very low frequency)?
MF is very tricky. And you may well be right that contemporary commercial systems use a trivial amount of feedback in order to be stay stable. Nobody has ever sold a system that could be used except with a dedicated driver.... till now?
Ben
Hi,
I'm pretty sure that a second voicecoil wound onto the same former will give more probs simply due to crosstalk and BL- and L-modulation than a decent piezo-sensor.
A coil may be used advantageously though if it is wound separately and features its own magnetic circuit. That's btw how B&M does it since long, one of the few protagonists of MFB for decades.
jauu
Calvin
ps. using MFB and DSP is no excuse to choose crappy ingredients in first place. Anybody expecting good results from bad ingredients will be disappointed. Wonders are something related to religious beliefs, not with technical hardware.
I'm pretty sure that a second voicecoil wound onto the same former will give more probs simply due to crosstalk and BL- and L-modulation than a decent piezo-sensor.
A coil may be used advantageously though if it is wound separately and features its own magnetic circuit. That's btw how B&M does it since long, one of the few protagonists of MFB for decades.
jauu
Calvin
ps. using MFB and DSP is no excuse to choose crappy ingredients in first place. Anybody expecting good results from bad ingredients will be disappointed. Wonders are something related to religious beliefs, not with technical hardware.
Last edited:
I didn't say it was "the" main source, I said it was "one of the". Look at any Klippel measurement set for a modern high-quality woofers and you will see that BL linearity and suspension linearity at extremes of stroke are of similar concern. My experience some 10 years ago was that voice coil MF distortion reduction ability tapered off(on a % reduction basis) as excursion increased, right where distortion reduction would be most beneficial.
The feed back signal generated by the voice coil is proportional to velocity and BL. Since BL changes with absolute position of the cone(and voice coil) it is a major concern. In fact the patent you linked to in the other MF thread is all about correcting the BL linearity problem.
http://www.diyaudio.com/forums/plan...iver-cone-position-fedback-4.html#post3554268
When discussing previous attempts at voice coil MF the inventor states:
"...Although this bridge scheme was economical, it failed to reduce significantly the nonlinear distortion of the sound radiated by the speaker...primarily due to the nonuniformity of the magnetic flux lines cut by the moving voice-coil and the resulting nonlinearity of the BL factor."
Calvin makes a good point that VC feedback can be used to good advantage with a separate VC and magnet structure optimized for linearity with little regard for efficiency. Whether that approach is more cost effective than an accelerometer, I don't know.
There are a few issues that deserve airing.
First and speaking generally, lack of a good motor is - at least conceptually - perfectly fixable and has been the object of many quality driver manufacturers. Likewise for the issue of the cone momentarily sitting slightly off-center (absolute error) in the magnet being not a big problem with good drivers. My guess is that nobody makes a surround as good as a good magnetic motor. But the errors from both benefit from negative feedback to help them better "balance" with the rest of the audio system.
Second, it is true to say that all kinds of distortion hell breaks loose at the extremes of motion. But at that kind of loudness (and near driver self-destruction levels), who cares? My hope is to make woofers work better at ordinary levels. At those levels, the magnetic circuitry is so-so in linearity and all the other "one of the main" problems may become more prominent.
About that link which I meant only to illustrate bridge circuit design.... who knows why they said that? Many commercially motivated publications (including this thread?) are intent on using cheap drivers to achieve quality sound. That's a reasonable business to be in (for people like Philips, RCA, and others in the past) but hardly of interest to this forum. Also, each person (including this thread?) is intent to show that nobody has solved (patented?) a system worth a darn... until they showed up with their solutions. So that explains their diss'ing of VC feedback.
Ben
First and speaking generally, lack of a good motor is - at least conceptually - perfectly fixable and has been the object of many quality driver manufacturers. Likewise for the issue of the cone momentarily sitting slightly off-center (absolute error) in the magnet being not a big problem with good drivers. My guess is that nobody makes a surround as good as a good magnetic motor. But the errors from both benefit from negative feedback to help them better "balance" with the rest of the audio system.
Second, it is true to say that all kinds of distortion hell breaks loose at the extremes of motion. But at that kind of loudness (and near driver self-destruction levels), who cares? My hope is to make woofers work better at ordinary levels. At those levels, the magnetic circuitry is so-so in linearity and all the other "one of the main" problems may become more prominent.
About that link which I meant only to illustrate bridge circuit design.... who knows why they said that? Many commercially motivated publications (including this thread?) are intent on using cheap drivers to achieve quality sound. That's a reasonable business to be in (for people like Philips, RCA, and others in the past) but hardly of interest to this forum. Also, each person (including this thread?) is intent to show that nobody has solved (patented?) a system worth a darn... until they showed up with their solutions. So that explains their diss'ing of VC feedback.
Ben
Last edited:
I think I know which subwoofer you are talking about 😉 I chose a dsp rather than analog solution because it is easier to change the design without changing the hardware, and secondly there are some things you just can't do in analog hardware that are easy to do on a dsp. The draw back is that there is a large learning curve for dsp.
Not correct. For a point source radiator, far-field sound pressure is proportional to radiator acceleration and not velocity so the signal from the accelerometer is an accurate representation of the sound pressure for wavelengths which are a lot greater than the circumference of the radiator 😉
regards
david
The other issue which is not addressed is how the bridge goes out of balance when the voice coil heats up thus introducing more errors into the feedback signal 🙁
regards
david
The miniscule influence of heating on resistance can be built-into the bridge, if anybody wanted to bother.
But the effect on resistance and bridge balance is trivial, except under test conditions.
The "error" is just a minor issue of proportion, not the introduction of false information.
Ben
With a temperature coefficient of 0.4%/°C for copper it's not hard to see a significant increase in voice coil resistance and total imbalance of the bridge to make it worse than without the bridge !!
regards
david
Last edited:
Correct, by motion relative to the box I was talking about acceleration, not velocity.
I agree that for a radiator mounted in a perfectly rigid, immobile enclosure the far-field pressure will match the output from the accelerometer. However, as soon as you let the box get shoved around by the voice coil forces, they will not match. Take for example an extreme case like a box that is mounted on wheels and weighs the same as the moving mass of the woofer. The far field pressure response will correspond to the acceleration of the cone relative to the box...you can think of it as an air pump and the far field response is related to the pressure pumped in and out of the box. The accelerometer would read the acceleration of the cone relative to an inertial reference frame and would output roughly half the magnitude it would for an immobile enclosure.
In practical cases with heavy enclosures, it should be a small error.
Of more concern than the voice coil heating(which as bentoronto mentioned can be compensated for) is the inductance modulation distortion terms. These terms are change in inductance with coil position, and change in inductance with coil current. According to resent Klippel papers, these are among the most prominent distortion products remaining in modern long throw woofer motors. These inductance modulation terms unbalance the bridge in a dynamic fashion that can't easily be compensated for. A separate VC sensor would not suffer from their effects on the main coil inductance.
Very Cool! thanks for mentioning the B&M MFB, I was not familiar with it.
It has been discussed at least once on the forum here:
http://www.diyaudio.com/forums/mult...feedback-mfb-inductive-sensor.html#post762711
Some pics of their sensor attached.
I noticed their website mentions a capacitive sensor as well....hmmmmm...
Attachments
Last edited:
- Status
- Not open for further replies.