Spectrum analyse

[Dest]

I start building metal detector but I dont know what spectrum of frequneces need to pulse in ground for good hunting and discriminating metals like Gold, Fe, aluminiom, silver, Cu...

[Dave J.]

Re: Spectrum analyse

In the case of nonferrous metals, it has more to do with the size of the objects, than with their metallic composition. The response of iron, regardless of size, is not so dependent on frequency.

These issues have been discussed at considerable length on Eric Foster's PI Technology Forum; however, you'll have to read through a lot of posts to find the ones which are relevant.

--Dave J.

[Randy Seden]

Re: Spectrum analyse

Dave,

I posed the question to Eric if he had connected a spectrum analyzer up to the receiver input of his PI detectors and he said he didn't even own a spectrum analyzer. No one on the Forum came forth so my question still begs, what do you see on a S.A. on the return pulse of various metals? The very fact that they have different time constants shows that there will be a distinct differance in the spectral pattern. That's why Anthony Barringer's INPUT system has the response of each type of mineral deposit in a memory bank and simply does a comparison for closest match (check out his patent 3,852,659).

So Dave, why can't we do this with our hobby PI's? I'm sure generating the data base would be labour intensive, but worth it for accuracy.

Randy Seden

[Dave J.]

Re: Spectrum analyse

In the case of PI, doing the analysis in the time domain, rather than the frequency domain, is a lot simpler, both from a mathematical-conceptual point of view, and from the standpoint of circuitry and software. Doing it the hard way won't become mainstream unless someone discovers a distinct advantage in doing so.


--Dave J.

[Randy Seden]

Re: Spectrum analyse

I agree with the fact that the time domain is easier but you still haven't answered my question(don't feel bad, nobody else has either).

Question:Have you actually looked at the return pulse on a spectrum analyzer for different metals?

Randy Seden

[Robert Hoolko]

Re: Spectrum analyse

Randy

The spectrum of an exponential decay is not very interesting. There are no humps or anything else exciting in it.

The plots above show two decay signals and their spectrums. The signals are on the left.

The bottom signal decays 5 times faster than the top one. The most noticeable difference in the spectrum is that the faster decay has a smaller DC component. The high frequency components are a little larger.

Even if the signal is a sum of exponential decays it does not get much more interesting.

Robert

[Randy Seden]

Thank You!!!

Robert,

I should of known you would come thru as you've been providing some very detailed information on the forums with your graphs. A picture is worth a thousand words and now I finally know why everybody always goes with the Time Domain Analysis.

Much obliged!

Randy Seden

P.S. How about the idea of storing the waveforms for differant metals in a lookup table like the patent I mentioned above by Anthony Barringer? That would sure solve alot of problems if you could do all the tests for different depths of gold ore/nuggets from different areas as the silver content varies from nugget to nugget. Now THAT would be the ultimate detector in my humble opinion.

[Robert Hoolko]

Re: Thank You!!!

Randy

The fact that someone got a patent does not mean that their invention will work, even for the use he describes let alone for what we would use it.

One problem with this idea is that metals do not have waveforms. An object will have a repeatable waveform if you hold it at one orientation and at one location under the coil. But as the orientation of the object changes or its position under the coil changes the waveform may change. So a silver dime lying flat on the ground may have a specific waveform, but silver does not.

Another problem is that to record the waveform you have to sample the signal at multiple points while the signal is decaying. For very weak signals you are only going to be able to see the peak of the signal. The rest will be hidden in noise. So you get no waveform information for the weakest signals. For somewhat stronger signals you may get two clean samples, which would be enough to calculate the target time constant. For even stronger signals you might get 3 samples, which would be enough to see if the time constant stays constant or if it changes with time. So you need a very strong signal before you get enough information to start looking for a match in a database.

Robert

[Carl]

Re: Thank You!!!

If you run an oversampling converter, synchronous with the pulse rate, and digitally filter, you can get a significant SNR improvement and literally pull signals out of the noise floor. I think this is one area of general metal detector design that can offer some improvements in sensitivity, as I don't know of anyone using this technique.

- Carl

[Robert Hoolko]

Re: Thank You!!!

I agree about the benefits of oversampling. That really worked out well in my VLF project. Doing it in a PI would require a much faster A/D with a lot more bits and it would require CPU horsepower to crunch the numbers. But I do think it is the right way to go.

Robert

[Dest]

Re: Spectrum analyse

10x but im look for graphs and other usefull information - metal reaction in diferent frequneces - the spectrum of gold fe aluminium silver etc. Up to now i dont know what frequence (or frequences) is better to use for gold (etc).
10x in advance

[Dave J.]

Re: Spectrum analyse

Spectral response for most metals, has more to do with target size than with metallic composition. The smaller the target, the higher the frequency response.

Higher frequencies are usually better for gold, because most of the gold to be found is in small pieces, not big chunks.

--Dave J.