Sound imaging confusion

1. Bad idea.
2. Try running some numeric estimates.
3. Drop it.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Hi. It's a little late to drop it. It's mostly built (prototype) and I'm seeing results. The range is over 100 ft. Here's what I was thinking for a try at imaging. Imagine a memory arranged in a 3d array driving 3d glasses. Pick some arbitrary point and compute the triangulation from all sensors to it and "and" their results. If the result is not 0, place a number (intensity) there, subtract that number from the sensor memory, and go to the next one. Since I only have 4 sensors, there will be a lot of false points. By moving the sensor array a fixed distance and doing another pulse, I can add 4 more points for the calculations, reducing the number of false points. What do you think? Larry

^^^^^^^^^^^^^^^^^^^^^^^^^

100ft - I don't buy it. There is inverse cube law of the magnetic field, the iverse square law of the acoustic wave plus the exponential attenuation of the both fields in the ground. Unless you are looking for something huge, this is not going to work.

Here you are assuming that the media is absolutely uniform, magnetically, acoustically and electrically inert, there is no noise and there is only one solid big target of a simple shape in the range. This is what they call the spherical horse in vacuum.

For the real horse, you have to build a model and estimate the parameters of the model as the multidimensional optimization task. This requires heavy computations and may lead to the peculiar results if the model is not adequate for the situation.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Hi Vladimir. Thanks for your response. The over 100 ft thing is actually

Larry,

Solving the nontrivial problems is my profession. If you are interested in our services, the contact email is at the web site.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Hi Vladimir. I really don't have the money to afford paying for help. I just want a really cool toy. This thing is probably patentable, but it's probably more trouble than it's worth. If you want, I'll send you the schematics and you can go do the patent thing. I'll probably take what I have and run with it. I have enough stuff to pump a megawatt into the pulse. I'll have to see if it interferes with my atv. Thanks for your input. Larry

Larry Snyder snipped-for-privacy@pacific.net posted to sci.electronics.design:

I suggest you study how sonar works. Especially in search modes.

Larry... can you explain in just a little more detail.... the EM pulse propagates at almost c down into the ground and physically shakes the metal thing in the ground. It 'resonates' and this sound propagates radially outward and upward in a spherical wavefront I assume, at what ever the speed of sound is in dirt. If it hits all 4 mics simultaneously, you can tell you are directly above it. Is this about right?

Hi Bob. Right. Sound propogates at about 5000 ft per second through dirt. My a/d converters run at 1.5 us which translates to about 1/10 inch resolution. This is stored in a memory and dumped to my laptop after it is through. What I end up with is 4 strings of data. The actual waveform looks like a step folowed by a ringing per object. The ringing frequency is determined by the mass/length, the step determined by conductivity. Non magnetic materials are repulsed by the pulse and magnetic materials are attracted. With a single object, there is no problem determining 3d position through triangulation. With 2 objects, there can be 4 positions and so forth. With an infinite number of sensors, a perfect picture is possible. What I'm going to try is a sample(4points), move the array a foot or so, take another sample(8 points), move again(12 points) and so on. I haven't figured how to take a running array yet, but worst case is to do maybe 20 and call it good. I saw a thing years ago about some guys from a university that were getting

3d images by dragging a thing around and shooting shotgun shells into the dirt. The took a lot of samples. They got reasonably good pictures of dinosaur bones. I'm going to try to chase that down. Any ideas will be appreciated. Thanks for your interest Larry

Look up crosscorrelation. That should yield the optimal match between sensors, with the time difference indicating relative position. The HP3562a and maybe newer models has this function built-in. [Not my thing, so I never used the feature. They also have an ap note for the box with similar physical measurements.]

Since you have the the ADCs, this should be doable with mathcad or octave. What you might want to do is upload to one of the binary newsgroups the signals you have digitized. Geeks could then process them. [I did this with some codar signals.]

Not quite. Time to each mike depends on the distance from the source.

OK. You probably stand the best chance of getting something vaguely usable from this if you have the mikes arranged in a linear array and go for a sequential 1-D search EW followed by NS. The most effective microphone spacing is one which maximises the system imaging capability by avoiding redundant baseline lengths.

One possible solution are Golumb rulers. The first few are perfect:

1 : 01 1,2,3 : 01 3 1,2,3,4,5,6: 01 4 6

After that there are some baseline lengths that are unmeasured. The relative phases of signals arriving at the mikes will give you a more nearly unique triangulation on the target. And some of the ghosts can be ruled out. Also look up phased array radar and aperture synthesis literature for more details of how to interpret the signals (or seismology - but remember you are looking at a source rather than reflections).

I hope you never find anything like a WWII UXB that may take violent excepton to being jolted by a magnetic pulse.

Regards, Martin Brown

Thanks Martin. I never considered getting blown up. I'm going to try signal processing. The step is all I really need to determind position. I can differentiate the signal before digitizing it. I'll know more later. Thanks for your interest. Larry P.S. The distance this is going is only 12 ft. My software programming sucks.