Universal PI Micro

[Tinkerer]

Originally Posted by matabele

Yep - live not too far from there, where the 'goway' birds cry.

PI metal detectors generally record 3 channels - each corresponding to an integral of the response over a certain period (channel) after different delays. One is generally a ground balance - the other two are on either side of the 'knee' in the response signal.

The problem is that ground and target responses are highly variable - the trick is to 'tune' the instrument to separate the target response from the ground signal in the presence of noise for the particular conditions encountered. This can not be done in the general case, as conditions vary tremendously (hence the 'salt', 'deep' etc. settings on many instruments.)

An instrument which records the entire decay curve could be customised by the user to match the specific conditions encountered - the whole response being compared to a ground response for the particular location and coil used.

With the TINKERERS_V1, I have been using 4 different samples, placed at strategic points in time.

http://www.geotech1.com/forums/forumdisplay.php?f=53

This is working well and gives good discrimination.

I would like to improve on that method and extract 10 or even 20 samples in quick succession to capture a whole time section.
The reason for this can be better explained when you look at the pictures at:

http://www.geotech1.com/forums/showthread.php?t=15441

Where I try to show that there is a lot of information available, if one knows where to look for it.
When locking at the oscilloscope screen, I can see movement at certain spots in time that indicate the TC of the target and if it is iron or non magnetic metal. Sampling at these spots let's me extract this information.

This information has thought me how to produce a TX pulse of a different shape, that gives more information about the target, because it excites core eddy currents not only skin effect eddy currents.

I believe that if this is taken a step further, like you indicate, recording a significant section of the decay curve, much more information can be extracted and much better compensation for mineralized ground and other PI problems could be designed.

Unfortunately, I am not capable of doing the programming. Are you?

Tinkerer

[matabele]

Until recently, it was not possible to build a time domain instrument at reasonable cost, as this requires a dynamic range exceeding 100dB. Most multichannel instruments, therefore, resorted to multi discreet channel or swept frequency techniques in the frequency domain.

I believe that it is now possible to build a quality time domain instrument with off the shelf components - however, many difficulties must be overcome. For example your current IB instrument does not appear to receive a classic time domain response (which should resemble an exponential decay curve.)

There are several problems to be addressed in order to increase the bandwidth (dynamic range.) Early stage sampling corresponds to higher frequencies, late stage sampling to lower frequency:
Using a digitiser for TD measurements

For small shallow targets we require fast switch off for early stage sampling, and for larger deeper targets we need better resolution in the late stage. This requires work on a broad band receiver loop, with low noise preamp, something the ELF community have been doing for some time.

To 'tune' the response it is also possible to transmit a wave rich in the frequencies of interest rather than a square wave. There are two approaches to this:
- either use a FPGA to create an arbitrary waveform generator - something like this
- or (my preference) - to use a PWM to create the desired wave somewhat in the fashion of magic sinewaves. A couple more links about how to do this yourself.

http://en.wikipedia.org/wiki/User:Kieff/Sine_from_square_waves
http://massmind.org/techref/io/pwm/harmonic.htm

Finally, we need to work on the uC. My preference is to piggyback on the robotics community, and use either an Arduino, or my preference the Pinguino. This will cut down development time, and enable additional tinkerers to get involved!

Hope the links stimulate some ideas.

[matabele]

Looking through some reference material, it would appear that the response from your receiver coil resembles the voltage response shown for the loop on this link:http://www.vlf.it/minimal2/minimal2.html

The article explains why a preamp functioning as a current to voltage amplifier is better in these circumstances than a voltage amplifier. Some additional theory can be found here

A realisation of this circuit using an OP27 is here and one based upon an LT1028 is here.

It may also be possible to build a satisfactory (and cheaper) current to voltage input preamp utilising an OPA134 or perhaps even a NE5532.

[matabele]

The current mirror (LM394) at the front end of the SD2000 perhaps does the same thing?

[Tinkerer]

Originally Posted by matabele

The current mirror (LM394) at the front end of the SD2000 perhaps does the same thing?

Thanks for the many interesting links and information. It will take me some time to look at all that.

The "Pinguino" sounds interesting.

If we can combine intrinsic metal detector knowledge, with digital programming, we will end up with something very useful.

Team work can get us there.

Tinkerer