Ground noise

[wyndham]

Since I recieved no responces from my earlier post I'll try to restate the question as it wont be so vague.
In the design of a PI detector, what I term the "ground noise/signal" that has to be cleaned up some how. How is this best done.
How much of the unwanted signal is the result of the TX coil itself inducing the noise and how much is it the natural noise or signal comming in from the earth.
Can the problem signal, where ever it comes from be mapped, stored and subtracted or are filters the better way of cleaning up this or any kind of noise.
Please bear with me since I am at the bottom of this learning curve and the questions aresimple compared to where most of you are.

Carl, did you get my PM about the HH?

Thanks for yor time Wyndham

[Carl-NC]

Sorry for not replying, I'm out-of-town with very limited Internet access, and what access I have reminds me of the 2400-baud days.

Your boom box simile is pretty close... more like trying to hear a bird with laryngitis chirp after each boom. There are 2 types of noise: uncorrelated, and correlated. Uncorrelated means random noise, which is averaged out by the integrating filter.

This leaves the correlated signal (desirable) and the correlated noise, which is more accurately described as interference. Some types of interference can be mostly canceled with the differential sampling integrator. Some types of interference are best dealt with using a DD coil and differential receiver.

I'll check your PM & answer this weekend... this connection is just too painful to try to do it now.

- Carl

[wyndham]

Thanks Carl

Didn't realize you were out of town. No rush. Thanks for the info, Wyndham

[Reg]

Hi Wyndham,

To be honest, the answers to your questions could fill a book, and what is worse, some of your questions do not have consistent answers.

First, the ground signal can be thought of as merely a target or object that can be detected. It is not a noise per se, but a signal much like a very deep huge object. As for getting rid of this type of signal, the conventional way is to take a second sample, amplify it, and subtract a sufficient amount of this second sample from the main sample that the ground signal is eliminated.

So, lets say your main sample is at 15 usec, then you would take a second sample, maybe at 30 usec or 40 usec and ultimately subtract the second from the first in the proportion necessary that the ground signal is gone.

Now, there are other techniques being used to accomplish the same thing. On some of the more sophisticated designs, the ground signal is sampled, held in memory and then subtracted as necessary.

Others may use another coil to sample the ground signal of sorts.

As for noise, Carl stated the main differences. As to the down and dirty of it, noise can be generated either internally, or is picked up by the coil. The internal noise can be caused by multiple things including internal clock pulses maybe from a uprocessor, power supply, etc. They can also be caused by simple shot noise from within an IC itself. This is why it is important to use a very low noise amp for the preamp stage.

As for filtering, there are many techniques used. Simply sampling at a much higher rate helps because the noise will average out to be less. Additional capacitance across amps helps as does simple low pass filters.

Unfortunately, if the noise is more than a random blip, then even filtering doesn't help that much since additional filtering will simply change a sharp pop to a smoother warble.

Maybe something like a noise blanker might help.

The bottom line is there is no simple answer to your question.

Reg

[wyndham]

Reg, thanks for the reply. I would like to carry this one step further
In my picture of the working PI in my mind, I see the tx as, by need, putting out a lot of signal and that signal also causes(my term) ground chatter that has to be dealt with thru filtering.
Is it a feasible direction to find a better way to eliminate this chatter, so the tx can better induce to the target or maybe a better question is "where is the barrier to better PI signals"
It would also seem that the earth is absorbing/dissipating or hidding some of the signal from the target as well.
I can't help but feel that there might be something similar to the spectral division of visible light in this area of the band and that each element should respond to that frequency. Some thing of a harmonic induction. Well not to get too far off. thanks Reg,
Wyndham

[Ferric Toes]

Hi Wyndham,

One of the beauties of PI is that the transmitter is fully switched off when the receiver is sampling the return signal. In fact , the short delay before sampling ensures that there can be no transmitter noise superimposed on the return signal. That is one of the reasons that PI excels in detection range over other methods. Any ground eddy currents have time to die away too, and the only thing that is left, is the magnetic lag decay of maghemite ground minerals. The search coil of course, does pick up stray electromagnetic noise from radio transmissions and power lines, which does occur during the sampling time. This is the main source of noise in a PI. In fact for a nickel at 14in, the signal you want is buried in maybe 10 times as much noise, which has to be filtered out by integration, or averaging.

It is possible to make a noise cancelling coil system, which consists of a receive winding in a figure of 8 shape. The noise picked up on one half is equal to the noise on the other half, which automatically subtracts out. the trade off is less depth, and objects such as coins and rings give a null signal where the windings cross. If you had a noise free coil, you are then left with the noise generated in the front end amplifier, as the limiting factor.

Another plus for PI, is that because the transmit pulse and the receiver sampling are separated in time, you can theoretically increase the transmitter power by any desired amount. As always, there are tradeoffs, not least the size of battery you would have to carry around.

As Reg says, you could write a book - which one day I will do.

Ferric.

[bbsailor]

Ground Noise...one more thing

Ferric,

Great answer.

There is one more thing that needs to be consideed when looking at ground noise...coil sweep speed. This is related to the speed of the main pulse clock and the amount of samples collected during the integration period. If you have a longer integration time to allow more samples during the integration period, you must sweep the coil slower. That is why higher frequency PI machines in approximately the 10 Kpps to 15Kpps range are used with low power pulses so you can have more samples during a shorter integration period and still sweep the coil at a good speed for fast ground coverage.

The integration circuit and the sampling RX window timing circuit actually work like a lock-in amplifier. Go to the following web site to see a good explaination of what lock-in amplifiers are and how they work.
http://www.cpm.uncc.edu/lock_in_1.htm

This will give you and other forum readers a good understanding about how to extract desired signals from noise using sampling windows. Simply, by adjusting the period of the sample window to synchronize with the desired signal, you can extract the desired signal even though burried in high noise.

You can pick up a lock-in amplifier, very inexpensively, on e-bay. Try to replace the differential input to the PI integrator IC chip of the Hammerhead with the lock-in amplifier to see how much more sensitivily you can squeeze out of adjusting integration time with coil sweep speeds. This is only for bench-top experimentaion unless you can find some lock-in amplifier IC chips.

The above web site will provide a very good tutorial on various lock-in amplifier applications and techniques, all of which are very related to PI circuits, timing and signal extraction.

bbsailor

[wyndham]

Thank to both on this topic. now what if there were a way to sample the "magnetic lag decay of maghemite ground minerals". and subtract this out.
Say you had a 3 coil system TX,RX and a third a coil or device (daughter board) that only reads the mag response an subtracts this out(where I don't know), but would this be an ideal setup?
I think there may be a way(fools rush in, you know), your thoughts
Wyndham

[Sean Goddard]

Has anyone tried using DSP to CORRELATE the Rx signal? Using this technique it is possible to extract even the minutest of signals from noise.

I've seen military radio Rx's pull signals that were only a few dB above noise floor (-123dBm). this equtes to being able to hear ANY radio transmission on the planet, from anywhere on the planet.

I think the main problem with beeprers at the moment is that the designers are mainly stuck in the dark agaes of analogue and are not 100% aware of what is possible using modern DSP techniques.

OK, we might not be able to break the laws of physics, but we can compensate for them in most cases.

[bbsailor]

Ground Noise

Sean

When you talk about extracting a small signal from large background of noise, you are talking about what lock-in amplifiers do well. See the link below for a good set of articles. http://www.cpm.uncc.edu/lock_in_1.htm

These are very informative. "Boxcar integrators" also could be useful so do web search on these words.

Actually, PI metal detectors synchronize the RX circuit to operate between the TX pulses. All lock-in amps have an input for a synchronizing signal. Some lockin amps have a two inputs in the form of a differential amplifier that could replace integrator circuit in the Hammerhead circuit.

The only tradeoff for extracting a very weak signal from noise is how much time you are willing to wait for the desired signal to be filtered through the noise. With a detector head moving across the ground at a few feet per second, the lock-in response needs to be pretty fast. If you could scan the search head very slow you could extract some pretty weak signals.

This would be a very interesting tradeoff analysis.

bbsailor