Most servo subs seem to use a piezo sensor of some sort. There is apparently a problem with phase shift in the sensors, though.
Is this the main cause for instability in servo subs or is the loop time delay the main cause?
I am also curious if it is possible to use ultra sound transmitters and receivers to generate the position sensing. Are they fast enough? Is it too complicated? It would have to involve some dopler theory, at least...
Give me your thoughts!
/Marcus
Is this the main cause for instability in servo subs or is the loop time delay the main cause?
I am also curious if it is possible to use ultra sound transmitters and receivers to generate the position sensing. Are they fast enough? Is it too complicated? It would have to involve some dopler theory, at least...
Give me your thoughts!
/Marcus
Both can be a problem. Piezos don't tend to be too flat or too sensitive below their resonance, so they need to be pretty heavily compensated. If we don't try to run the subwoof too high in frequency, the loop time delay isn't too much of a problem. One does, however, need to keep in mind the phase lag due to the LF rolloff of the sub when adjusting servo parameters. For my own system, I used a test signal a couple octaves up from fundamental resonance to tune the system, and it seems to be quite stable. Judging from some of the responses here, it looks like I'll have to write this one up and put it on my web page. So many projects, so little time...
Ultrasonic should work, but it can be awfully complicated. There IS a better way, if I can get some uP whiz to work with me on it.
Ultrasonic should work, but it can be awfully complicated. There IS a better way, if I can get some uP whiz to work with me on it.
Are you talking about sampling the sensor response and making the feedback in the digital domain?
BTW, did you use a piezo transducer from an old tweeter or have you tried the accelerometer from Analog Devices?
Their accelerometer is quite expensive and I've heard about problems with the piezo chrystal. That is why I came up with the ultra sound thing. I know that ultra sound is used in highly advanced medical systems to keep track of things, so it should be possible.
/Marcus
BTW, did you use a piezo transducer from an old tweeter or have you tried the accelerometer from Analog Devices?
Their accelerometer is quite expensive and I've heard about problems with the piezo chrystal. That is why I came up with the ultra sound thing. I know that ultra sound is used in highly advanced medical systems to keep track of things, so it should be possible.
/Marcus
No, it's a completely different sensing method, based on some work I did 20 years ago at Nicolet on rapid scan Fourier Transform infrared spectrometers. We had to control a moving mirror of some considerable mass at moderate frequencies (i.e., woofer range) to within a fraction of a wavelength of red light. The technology available then was pretty complex and expensive; it would be pretty easy and inexpensive to do nowadays.
I'm using the AD accelerometers. The price was right (free). I don't know what these actually cost in the real world, but though they're much better than piezo units, they're still not the most optimal technology.
I'm using the AD accelerometers. The price was right (free). I don't know what these actually cost in the real world, but though they're much better than piezo units, they're still not the most optimal technology.
It is potentially patentable, though given the expense versus payoff, I doubt we'd go that route. I'm not in the speaker business, and the chances of finding a remunerative licensee are small. This is purely for fun and interest.
If I can find a suitable collaborator, we'll absolutely publish the design.
If I can find a suitable collaborator, we'll absolutely publish the design.
Well what you are thinking is probably doable, but maybe not the way you are thinking of. Similar things have been done in the past, and certainly it has been done with a laser before.
You can not simply bounce off ultrasound to measure the distance based on the return time. The best accuracy you will ever get will be a significant fraction of a wavelength, and the update rate will be low.
You could generate an ultrasonic tone, and then look at the frequency spectra of the reflected tone. The doppler shift gives you the motion of the cone. That would likely be the best way to do it. However, any way you look at it, you have the travel time of the sound, and hence you have phase shift so you are back to square one.
What you are talking about is regularly done in a lab environment with speakers during their design and measurement, i.e. to see cone breakups, etc. Could you do it cheaply, probably, but there could be much cheaper ways of doing it.
Alvaius
You can not simply bounce off ultrasound to measure the distance based on the return time. The best accuracy you will ever get will be a significant fraction of a wavelength, and the update rate will be low.
You could generate an ultrasonic tone, and then look at the frequency spectra of the reflected tone. The doppler shift gives you the motion of the cone. That would likely be the best way to do it. However, any way you look at it, you have the travel time of the sound, and hence you have phase shift so you are back to square one.
What you are talking about is regularly done in a lab environment with speakers during their design and measurement, i.e. to see cone breakups, etc. Could you do it cheaply, probably, but there could be much cheaper ways of doing it.
Alvaius
There are accelerometers suitable for this application for under $50, and if you had the time you could conceivably do it with analog components.
Marcus brought up signal processing in the digital domain. I think this is something that could be done with an inexpensive computer-on-a-chip kit.
The computer would take the analog inputs from the preamp and the accelerometer and convert it to digital domain, where it it would be very quick and easy to adjust PID values and monitor them in real-time.
Just A Thought,
-Joe
Marcus brought up signal processing in the digital domain. I think this is something that could be done with an inexpensive computer-on-a-chip kit.
The computer would take the analog inputs from the preamp and the accelerometer and convert it to digital domain, where it it would be very quick and easy to adjust PID values and monitor them in real-time.
Just A Thought,
-Joe
I believe the "problem" with piezo devices is that they provide a voltage output that is proportional to acceleration when what you want is an output that is proportional to position. The piezo signal needs to be integrated to get the velocity, and integrated again to get the position info.
If you attach a piece of foil to the speaker cone and place a fixed metal plate (screen) in front of it you form an air dielectric capacitor. If you put a fixed amount of charge on the cap, the voltage across it will be proportional to the position of the speaker cone. The problem with this is bass driver excursion is pretty large and so the capacitor you can make is very small because the stator plate must be placed far enough from the cone so the cone won't bang into it. Capacitance decreases rapidly with increased spacing between the plates. So to get a useable signal, you'll have to put a lot of charge on the small cap. That requires a high voltage power supply- not too difficult by itself, but it all starts getting pretty complicated. You'll need a
VERY high Zin amplifier to monitor the voltage on the cap.
MR
If you attach a piece of foil to the speaker cone and place a fixed metal plate (screen) in front of it you form an air dielectric capacitor. If you put a fixed amount of charge on the cap, the voltage across it will be proportional to the position of the speaker cone. The problem with this is bass driver excursion is pretty large and so the capacitor you can make is very small because the stator plate must be placed far enough from the cone so the cone won't bang into it. Capacitance decreases rapidly with increased spacing between the plates. So to get a useable signal, you'll have to put a lot of charge on the small cap. That requires a high voltage power supply- not too difficult by itself, but it all starts getting pretty complicated. You'll need a
VERY high Zin amplifier to monitor the voltage on the cap.
MR
I worked on a bipole feedback sub last year for a school project. Was all in the analog domain using an AD accelorometer. Noise was a huge problem and really limited the amount of feedback we could add. Also mounting the accelorometer took some engineering to damp high freq vibrations in the cheap plastic cone. Low pass filters in the feedback loop mess with the phase. Still, after much tweaking, we had a sub that was +- 1 db from 16-200 hz using 2 10" drivers in a 1.5 cubic foot box, at decent SPLs. Wasn't too reliable though; the accelorometer fell off during our demo for the prof! Hearing the sub try to correct for the loose banging of the chip on the cone was not pretty.
- John
- John
Yes, but double integration of a sine leads to an inverted sine, hence no integrators are needed in the feedback for that reason. Maybe for other reasons, though, as cogs mentioned.MRehorst said:
I've though about that to. I think that the speakers position is proportional to the current. The reason is that the magnetic field in the speaker coil is proportional to current and in the end it's the two magnetic fields that are determining the position.cogs said:
However, since feedback is applied it doesn't matter if we monitor voltage or current. Since we know that the voltage from the amp is a correct copy of the music signal that is the only thing we can rely on. The current is created when our correct voltage "hits" the non-linear speaker load.
/Marcus
I like the capacitor idea!
Here is a way that may make it simpler to sense:
Mount a metal tube to the magnet, on the same axis as the rest of the speaker. Mount a wire or smaller diameter tube to the dust cap, so that it can travel in and out of the tube. Leave enough space so that the wire does not touch the tube wall. As the diaphragm moves in and out, the wire moves in and out of the tube and changes capacitance.
This is potential more linear than puttinga plate on the cone and an exteranl plate. In that case, the capacitance will not change linearly with the spacing, unless the plate area is much larger than the plate spacing. The non-linearity is due to changing of effective plate area becuase of the effect of fringing fields.
There will be fringing fields with the wire and the tube also, but they will act more like a constant capacitance in parallel with the variable capacitance, if a reasoanble section of the wire is always out of the tube.
John
Here is a way that may make it simpler to sense:
Mount a metal tube to the magnet, on the same axis as the rest of the speaker. Mount a wire or smaller diameter tube to the dust cap, so that it can travel in and out of the tube. Leave enough space so that the wire does not touch the tube wall. As the diaphragm moves in and out, the wire moves in and out of the tube and changes capacitance.
This is potential more linear than puttinga plate on the cone and an exteranl plate. In that case, the capacitance will not change linearly with the spacing, unless the plate area is much larger than the plate spacing. The non-linearity is due to changing of effective plate area becuase of the effect of fringing fields.
There will be fringing fields with the wire and the tube also, but they will act more like a constant capacitance in parallel with the variable capacitance, if a reasoanble section of the wire is always out of the tube.
John
I'll use "w" as omega. i.e., 2 pi f. For a constant amplitude signal A*sin(wt), the acceleration is -A(w^2)sin(wt), because of the nature of the double differentiation- chain rule, twice applied. So an accelerometer's output will scale to displacement but for the square of frequency.
It would be a lot easier with an equation editor!
It would be a lot easier with an equation editor!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.