| tade |
I wasn't able to turn up anything in searching, however if this info exists already please point me to it.
could I mount a microphone to the front of my subwoofer and use that for feedback? The voltage input should be proportional to the acoustic output. Also, Is it true that most microphones are fairly accurate in the low range but it takes a decent one to measure high frequencies? So perhaps I could use an inexpensive eletret mic, amplify it, then use an opamp to compare the input and acoustical output, and then adjust accordingly. Maybe also put a low pass filter on this effect to place the feedback below any room modes.
Thanks |
|
|
| mrshow4u |
Meyer Sound used to have a studio monitor that used a mic for feedback. I believe it (mic) was mounted in front of the woofer. I don't see it advertised on their website now(????). This aproach might not work so well on shorter wavelengths (tweeter). I think your largest distortion contribution will be from the woofers
I found the link: http://www.meyersound.com/products/studioseries/x-10/ |
|
|
| Ron E |
I've never done this, but I have thought about it in the past as a way to avoid purchasing an accelerometer. Accelerometers are starting to get cheaper, but they are usually SMD, so more difficult to work with.
If you try this, it would probably be best to use a sealed box and a microphone inside the enclosure. It would then work at low frequencies where the enclosure is non-resonant. Make sure it is an omni mic. |
|
|
| mrshow4u |
| I like the accelerometer idea too!! Velodyne does that on their upper end subs. I think a thing to watch out for is cone modes. If the accelerometer is placed in the wrong spot, the FB info will be out of phase and the thing may bust into oscillation. You could probably test drive the accelerometer in a couple of location and make a phase plot. |
|
|
| poobah |
Any distance between the mic and the cone will create a delay. This delay should not be confused with a pole... 2 different beasts entirely. Is it a showstopper? No, but it sure complicates things. This is why accelerometers mounted close to the voice coil are used.
:) |
|
|
| bobo1on1 |
I am planning to try this out, I already glued a microphone on my subwoofer and the measurements look very good, it's about the same as winisd calculated, including the phase.
If you want to look at it on the scope use something like a 5 volts power supply and a 4k7 resistor in series with the mic.
Be sure not to wire the mic reversed or you won't get any measurements :clown:
Also, you will need a lowpass, use a first order after the differential stage, not on the microphone signal or you might get lots of noise and a rising top end, something you don't want on a subwoofer. |
|
|
| forr |
In his book, in french, available free on the net at
http://www.brouchier.com/livre/index.html
or
http://www.brouchier.com/Le_son/LE_LIVRE.pdf
author Francis Brouchier has proposed to use a tweeter or a closed back (or closed front...) little loudspeaker as a pressure sensing low cost device and to place it inside the enclosure. An experimental easy setup.
Such a technique has been criticised by John Watkinson in one of his Electronics World articles however I think his arguments are valid only at high sound levels. |
|
|
| Svante |
| quote: | Originally posted by Ron E
I've never done this, but I have thought about it in the past as a way to avoid purchasing an accelerometer. Accelerometers are starting to get cheaper, but they are usually SMD, so more difficult to work with.
If you try this, it would probably be best to use a sealed box and a microphone inside the enclosure. It would then work at low frequencies where the enclosure is non-resonant. Make sure it is an omni mic. |
Ah, but if the mic is placed inside the box, one would have to differentiate the signal twice to use it in a feedback loop. There is a difference of 12 dB/oct between the pressures inside and outside of a box at low frequencies.
The accelerometer is IMHO better, since it needs no differentiation, but still is as immune to "other" sounds on the outside as the in-the-box microphone placement.
...but the mic-in-the box version would work for BR enclosures too. |
|
|
| bobo1on1 |
Feedback for BR enclosures is absolutely useless, first the sound from the bass reflex pipe is 180 degrees out of phase which makes it very very hard to get good feedback, second a bass reflex is a resonator, which means the sound builds up slow when signal is applied and decays slow when the signal is turned off.
This means the feedback circuit will applie huge amounts of power at the beginning to get the resonance started and it will apply huge amounts of power at the end to stop the resonance.
You might as well just plug up the pipe and use it as a sealed enclosure. |
|
|
| Calvin |
Hi,
Mics could be used for bass applications up to app. 500Hz
Commercially the German company Backes&Müller (quite renown for their active and fed back speakers) used this methode.
The DIY-magazine Elektor published this circuit
jauu
Calvin |
|
|
| Circlotron |
| quote: | Originally posted by bobo1on1
...a bass reflex is a resonator, which means the sound builds up slow when signal is applied and decays slow when the signal is turned off. This means the feedback circuit will applie huge amounts of power at the beginning to get the resonance started and it will apply huge amounts of power at the end to stop the resonance. |
I imagine that if you had a driver with low Qes and suddenly applied a LF signal at the resonant frequency of the box+port, while the port output is gradually building up it doesn't place much load on the driver and so the cone excursion is large, and gradually reduces as it's load increases by the port coming "up to speed". What the port momentarily lacks, the larger cone movement makes up for. Is this true in practice???
A mfb of some sort will work for a sealed box where cone movement equates to spl but with a bass reflex at the tuning point for example the cone movement gets very small and most of the sound comes out the port. MFB is going to make a mess of this because if it keeps the cone movement directly proportional to signal level you will end up with a =huge= peak at box+port resonance. |
|
|
| bobo1on1 |
| quote: | Originally posted by Circlotron
I imagine that if you had a driver with low Qes and suddenly applied a LF signal at the resonant frequency of the box+port, while the port output is gradually building up it doesn't place much load on the driver and so the cone excursion is large, and gradually reduces as it's load increases by the port coming "up to speed". What the port momentarily lacks, the larger cone movement makes up for. Is this true in practice???
|
Not really, the output on my woofer is 10 db lower than the output from the port at resonance, that's why it needed a bassreflex enclosure in the first place ;)
About two years ago I build an mfb sub using a piezo loudspeakerthingy and a schematic from elektor, it worked pretty good but I had some problems with clipping and I didn't have a scope yet so I didn't know what the problem was.
It was pretty cool though, if you used the volume pot from the amplifier you could turn the woofer from tight to boomy without changing the volume. |
|
|
| phase_accurate |
The principle brought up by Calvin has been discussed several times here. NP for instance used it for active absorbers and he gave some comments on it as well.
The owners of the original patent - B & M - used an electret mic capsule that was glued to the driver itself in order to achieve the lowest delay possible. They placed it in the grove between cone and dome.
And most importantly: It was monted sideways !!!!!!!!!!! Otherwise it would work partially as an accelerometer instead of a pure pressure detector.
There is still a commercial sub available that uses the principle: The Manger subsonice.
Here you will find an extract from the German mag Funkschau describing the (then new) B&M:
http://www.johannes-krings.com/funkschau2
Regards
Charles |
|
|
| tade |
this is cool. I think ill buy a mic and borrow an oscilliscope and see what my sub is doing.
Thanks |
|
|
| bobo1on1 |
I just did a small test with a microphone, I could set the feedback to about 70% microphone signal and 30% line signal, that is the maximum feedback I could use without oscillation.
The frequency response with and without feedback is included, remember this is in room with the mic close to the sub. |
|
|
| bobo1on1 |
And the schematic.
I need a better lowpass to prevent oscillation. |
|
|
| bobo1on1 |
| One other interesting feature I noticed is that it linearizes the phase as well as the amplitude, this should make the woofer faster. |
|
|
| bobo1on1 |
I just tried a second order lowpass but it doesn't do any good.
I did put a shelving lowpass at the input, you can see the feedback effect even better now.
Now all I need is a highpass :clown: |
|
|
| phase_accurate |
What frequency was it oscillating at ?
In order to increase stability you'd have to put your mic as close to the driver as possible. High filter orders in the loop don't help to increase stability either. Even your first order LPF might probably be working better if changed to a lag filter. I.e. connect a small resistor in series with C1 like 1 k for instance.
Regards
Charles |
|
|
| tade |
| this makes sense about the phase. Because the woofer lags behind its signal, the feedback loop sends a much high voltage (should be infinite) to get the sluggish piece of mass to move more quickly. I am liking your results. How is cone excursion? Any chance of mounting the mic to the cone itself? |
|
|
| bobo1on1 |
I did mount the microphone directly to the woofer.
When I turn up the feedback it starts oscillating at 800 hertz, when I used a second order lowpass it started oscillating at 16 hertz, that's where the phase is 180 degrees out of phase because the woofer acts like a second order highpass.
I didn't really come up with a solution for that.
Phase response with feedback is flat from 40 hertz up though. |
|
|
| phase_accurate |
If it was oscillating at around 800 Hz then I'd really try the LAG filter instead of a simple lowpass. Furthermore I'd make the cap a little larger. You need most of your NFB at low frequencies.
Regards
Charles |
|
|
| bobo1on1 |
| There isn't much to gain from that as it will still oscillate at 16 hertz. |
|
|
| phase_accurate |
| quote: | | There isn't much to gain from that as it will still oscillate at 16 hertz. |
How do you come to this conclusion ? Did you actually try it ?
Regards
Charles |
|
|
| bobo1on1 |
That's what it did with a second order lowpass, made from 2 10k resistors and 2 100nF caps.
Pherhaps I can take the output from IC1D, pass it through a lowpass and feed it to the inverting input. |
|
|
| capslock |
Maybe it would make sense to put the LP filter after the microphone amp 1B and hence in front of the subtractor 1D?
The least it will do is keep higher frequency acoustic signals picked up by the mike from entering the subtractor and ending up in the loop that simply cannot deal with them. |
|
|
| bwaslo |
BoBo1on1,
The 800Hz oscillation is because of the cone breakup resonance. I did a similar MFB sub (but using an AHC-01 accelerometer) and there was a huge peak in the woofer/accelerometer response at around the same frequency. I measured loop gain, and without EQing, that was the highest gain point and the phase moved fast through there, just about guaranteeing oscillation.
What I did to fix it (and what may work for you) is to put a passive Twin-T notch network, tuned to 800Hz, in the feedback path. With an integrator in the loop there was a region between about 300Hz and 700Hz where gain was reasonably far down, so the phase shift from the Twin-T sitting at a little higher in frequency didn't cause a stability problem. I also loaded the Twin-T with a smallish resistor (compared to the resistors in the T itself), which gave a bit of phase lead, so the notch ended up not hurting the phase much at all.
The MFB mod was done on a Dayton powered sub (with the ports blocked). Worked really well. There should be an article on this in AudioXpress in a few months... |
|
|
| bwaslo |
| Here's a plot of the result response measured nearfield. |
|
|
| bwaslo |
Here is a plot of the loop responses before and after filtering.
The trace marked by circles is the "after" (gains offset for comparison).
The "before" is what comes out of the accelerometer when driving the amplifier directly. The "after" is what comes out of the accelerometer when driving into the filter circuit where the accelerometer will connect when the loop is closed. |
|
|
| moamps |
Hi bwaslo,
I have a few questions.
Where and how did you mount the accelerometer?
The peak in FR at 1kHz looks to me like the resonance of the mechanical system consisting of the dustcap/accelerometer's mass .
Did you measure the THD?
Regards,
Milan |
|
|
| bwaslo |
>The peak in FR at 1kHz looks to me like the resonance of the mechanical system consisting of the dustcap/accelerometer's mass .
Could be. The sensor is glued to the dustcap (which is inverted). I figured the cone was flexing where the cap bonded to it, pretty far up from the voice coil.
I only made a cursory look at THD, it's about 1% (at 90dBSPL two feet) away down to about 30Hz, then climbs from 30Hz down till it is about 15% at 20Hz. Not amazing, but pretty good for under $200. |
|
|
| phase_accurate |
The original B&M used the same mic position as this Manger sub does:
http://www.manger-msw.de/images/pro...sonice_2_kl.jpg
It wouldn't be bad if bobo101 did a sweep of the driver-mic response as well. One doesn't have to be scared of phase-shift introduced by EQ networks if they really represent the complementary of an error with minimum-phase properties.
Regards
Charles |
|
|
| capslock |
@all: Any opionion as to where the filter should be placed, in front of or after the subtractor?
@bwaslo: could you sketch your filter topology, please?
Thanks! |
|
|
| phase_accurate |
| quote: | | Any opionion as to where the filter should be placed, in front of or after the subtractor? |
After the subtractor IMO.
Regards
Charles |
|
|
| capslock |
| What is your reasoning? |
|
|
| bobo1on1 |
Just now I wondered why the bass was so loud when listening to the radio, it turned out I had my soundcard configured for bass redirection while using stereo spdif output.
So I don't need a shelving lowpass at the input.
The circuit is picking up interference from my wireless mouse though, I need some small caps at the input.
The microphone pics up sounds from my mains as well, but it is alot lower than the signal from the woofer.
The reason any filters should be after the subtractor is to get a sort of flat frequency response at the amplifier's output.
If a lowpass was placed before the subtractor the feedback circuit will make a shelving highpass at the amplifier's output, something you don't want. |
|
|
| phase_accurate |
| quote: | | What is your reasoning? |
Since the error is in the forward path of the loop it is best corrected there.
Regards
Charles |
|
|
| capslock |
Note that the Elektor example puts filtering right at the microphone. They have also configured the subtractor to work as a shelving integrator.
AT first sight, this looks like a double (2nd order) low pass, but I have the feeling that there is some lead compensation. Will have to plot the transfer functions at some point... |
|
|
| bobo1on1 |
Here you can find an article from dutch elektor, they put the shelving lowpass at the microphone input and a second order lowpass at the output.
They used a piezo tweeter as an accelerometer.
I wonder what good the shelving lowpass does, except make the woofer go lower, except the woofer already goes very low due to positive feedback at low frequencies. |
|
|
| phase_accurate |
| quote: | | I wonder what good the shelving lowpass does, except make the woofer go lower, except the woofer already goes very low due to positive feedback at low frequencies. |
This so-called shelving lowpass is actually the approximation of a PI controller if it is placed in the forward path of the loop.
To see what Elektor did with the piezo acceleration sensor example I would have to undig the article (I remember that I have it somewhere).
Regards
Charles |
|
|
| bobo1on1 |
| That would make sense if I was heating my room, but not for a subwoofer imho. |
|
|
| capslock |
| I was referring to the Elektor microphone based circuit that was posted on the first page of this thread. |
|
|
| bwaslo |
>@bwaslo: could you sketch your filter topology, please?
Sorry to take so long getting this on the board, I was out of town and didn't have any files with me (other than a few plots on my laptop).
Here is a sketch of the topology (sorry for the crummy drawing). I took the parts values off, as I don't know how AudioXpress would feel about me giving out the design before they publish it. But you can see what is happening from the topology, I think. Notice that the ACH-01 accelerometer output is inverted, so its signal gets fed back in-phase (to the inverting input of the summing stage, and then inverted again by the leaky integrator stage). R8A gives an adjustment for the loop gain. R9, 10, 11 and C5,6,7 make up the Twin-T notch (and when loaded by R12, there is a relative high frequency boost from the notch, which gives a phase lead). You can calculate the Twin-T notch frequency to take out any peaks in your loop, using the formula at
www.radio-electronics.com/info/circ...otch_filter.php |
|
|
| capslock |
| Thanks, I wasn't familiar with the topology. |
|
|
| bobo1on1 |
| Bwaslo how much feedback can you achieve without oscillation? |
|
|
| bwaslo |
>Bwaslo how much feedback can you achieve without oscillation?
For the article design, I'm using about 10dB feedback (at ~70Hz, the highest point in the "raw" response), leaving 8dB or more of gain margin over the full range. The feedback is at about 0dB at 20Hz, and phase shift is what kicks the response up below that frequency, from slight peaking. I could use more feedback, but the peaking starts to look ugly and scary if taken too far, and I wanted to keep it moderate for the article.
I've done another in which I had stiffened the cone and dust cap with a coat of epoxy (to move the resonance to above 1200Hz) and was able to use about 15dB of feedback, without visible peaking, with that one. But that approach will only work if you can measure the resonance and custom tune the notch for it, since I doubt the resonance shift would be very repeatable from smearing some uncontrolled thickness of epoxy on the cone.
BTW, I just yesterday found out that the distortion graph I uploaded previously is wrong -- I forgot about low frequency resolution limits and didn't use a long enough measurement for the plot to be correct below about 30HZ. It will probably look better than I showed at lower frequencies (though I guess it could instead be worse than shown... the FFT size wasn't long enough to show anything valid about the very low frequencies). I'll remeasure and post the correct distortion measurement after I get back home again and have some time. |
|
|
| bwaslo |
I finally got a chance to rerun the distortion measurements on the servo-d Sub120. This is the design that will be published in AudioXpress.
I set the sub up in my listening room, about 1 meter from a wall and with the microphone 2 meters away, a meter from the floor and toward the center of the room. That's a pretty typical listening distance for small rooms. Then made measurements at various drive levels.
Here is the series of fundamental sweeps, to show how the output compresses. You can see the room lift along with the usual modes and nulls. The design has a highpass filter at about 16Hz, but the response is still strong below that in this case because of the room. The input drive at lower frequencies is rolling off fast, so you can see that the compression lets up below about 10Hz -- the output there is mostly being peaked by the room. Which would be advantageous if I had any CDs with actual content that low! |
|
|
| bwaslo |
Here is the fundamental (top trace) and summed harmonic distortion components (lower trace) for the highest level sweep at which the woofer still was moving linearly at all frequencies.
This shows the difference in levels between fundamental (only) and all distortion components below 12th order. The measurement method (log-swept sine) is able to measure fundamental level by itself.
To get to %THD, if you need those values, take the difference between the fundamental and the distortions at any frequency and level, call it "D", (a positive value in dB). From that calculate the fractional equivalent "E", as : 10^(-dB/20). Then the percent distortion is
%distortion = 100*E/(1+E).
|
|
|
| bwaslo |
| Here is with 5dB more drive, the woofer is getting to near its limits between about 15Hz and 25Hz. |
|
|
| bwaslo |
| Here is with 10dB more drive. 90dBSPL is pretty much all there is below 45Hz. But this seems to be quite adequate for music listening (for me anyway...but I don't go for rock concert levels). |
|
|
| bobo1on1 |
I just discovered a bit of a problem with my microphone setup, I noticed the amp was clipping at not so high levels, so I watched with the scope and noticed that it was only clipping on the negative rail.
It turned out the micrphone was clipping on the negative side of the wave, which caused the feedback circuit to boost the signal, causing the amp to clip.
I guess it's not normal for a microphone to deliver a signal of 500 mV.
I'm thinking about placing the microphone about 5 cm away from the woofer, the spl should be alot lower, also room modes are better equalized. |
|
|
| BAM |
| I wish Rod Elliot or BrianGT or someone else known for their electronics kits would come out with a servo feedback project with an accelerometer we could just glue onto the back of any speaker cone, and with controls that we could adjust for our specific driver and design. It would have a mono line-level input and a line-level output which enters into the amplifier. With that level of ease, any project could be servo-controlled. The servo-correction signal would need an adjustable low-pass filter and an adjustable high-pass filter. These functions could be accomplished by small PCB-mounted pots, with set-and-forget simplicity. |
|
|
| bobo1on1 |
Such a circuit would be quite hard to tune without an oscilloscope, also people with a scope are in general able to build such a circuit themselves.
It only works when a certain woofer is supplied, like the kit from rythmik audio.
I would like so see someone like Rod Elliot to post an article on subwoofer feedback though. |
|
|
| BAM |
| The Rhythmik woofer doesn't incorporate an accelerometer, but uses a specially-wound second coil to sense the voice coil's position within the magnetic gap. This would be a kit that includes the accelerometer and the associated circuit. The accelerometer would simply be adhered to the cone of the woofer and connected with flexible multi-stranded leads, not built into the driver. Something would need to be done to compensate for the shorter excursions of smaller drivers, but this could be incorporated into the adjustments that the user would make. And that adjustment might require an oscilloscope, but I don't see where the assumption comes from that everyone who owns an oscilloscope will be an electrical engineer who can develop their own circuit anyway. It would be much simpler to order the kit from Elliot Sound Products and tune it to match the woofer being used, than to design a circuit from the ground up. |
|
|
| Baldin |
I really like the notion of usong a mic for feedback. It will both work for lowering the distortion in the sub, correcting Q and freq response, at the same time working as active room correction.
In fact this is more or less "just" an active bass absorber, that has an extra input for producing bass sound :)
bobo1on1: how does it sound? |
|
|
| bobo1on1 |
Imagine having a woofer made from concrete, not just the case, but the cone as well.
It sounded incredibly tight, I had to turn up the volume around 5 db because it sounded less loud, even though the measuring equipment said it was at the same volume.
I did have a problem with the microphone clipping one way, at first I thought it was the amp but when I checked it with the scope the amp was only clipping to the bottom rail. |
|
|
| forr |
Hi Baldin
---I really like the notion of usong a mic for feedback. It will both work for lowering the distortion in the sub, correcting Q and freq response---
A mike would do about the same job as an accelerometer.
It is very little known what are the subtle effects of the different kinds of motion feedback. As far as I could determine them :
Acceleration feedback alone makes the moving mass apparently higher, so the frequency resonance is lower. The then higher Qt has to be corrected through the associated electronics.
Velocity feedback alone only lowers the Qt making the driver having an ascending frequency response at a rate of 6 dB/o. The associated electronics can differentiate the feedback signal, making it an acceleration information (as used by Dan Ferguson in its AudioXpress articles ) or it can integrate the input signal (as used by Russell Breden in its Electronics World article) to get extension of the low frequency response.
Position feedback alone enhances the stiffness of the suspension, then the frequency resonance and the Qt are enhanced. |
|
|
|
