Computer Interfacing
Discussions about interfacing and electronics

Decay curve interpretation

Will the PI metal detector be fast enough to differentiate between ferrous and non-ferrous metals?


       Computer Interfacing Forum Index -> Pulse Induction metal detector
Author Message
New User

Joined: 19 Jan 2009
Posts: 1

Jan 19, 2009 4:06 pm

It looks interesting so far. My question is will it be fast enough to differentiate between ferrous and non-ferrous, or will it look at the end of the expected decay curve and see the level above null? Will it see the difference and say there is metal, but not a type?
Site Admin

Joined: 13 Mar 2007
Posts: 145
Location: Netherlands and Kazakhstan

Feb 06, 2009 9:06 am

Yes, it will be fast enough to make decisions about the curve shape in the first part of the decay curve. The integrated ADC of the DSP in the project has a sampling frequency of 1 MHz.

Mar 16, 2009 7:28 am

I was thinking of using something like an ad876 20MSPS(max) with signal processing performed by an ARM-7. Not sure what max conversion speed would be possible.
Between pulses there is time to do a fair bit of work with an ARM.

My first thought was to do simple area under the curve measurements. No metal = background. Keep a long term running average value of n data points. Also keep a second running average using less data points.
By comparing LTA and STA (long term average/short term average) you can pick differences and possible signals.
This is how I used to detect seismic activity when a drifting baseline was present. Which is often the case when a detector moves over mineralized ground
By suitable selection of LTA, STA (and threshold) you can get pretty reliable detection of a slope change.
In this case, not slope change but area under the curve change.
From your graph this would probably allow discrimination. Indeed by splitting the area under curve into an upper and lower segment, it looks like a certainty.

I used this scheme in a pic18f PI design a couple of years ago. It worked well but I eventually got bogged down in details and it never left the lab!

Mar 16, 2009 7:32 am

The pic18f used LTA/STA of the time to decay to a fixed voltage. It didn't try any signal processing for the obvious reasons


Mar 17, 2009 6:27 am

Though have now moved to your thinking that a dsPic might be the way to go.
dspic30f2020 is capable of 2MSPS and is relatively cheap. Also makes the hardware easier not needing any support chips.
I still like the idea of an ARM-7 (lpc2000)for processing. So much memory available makes for ease of programming in C.

Mar 18, 2009 11:27 pm

Target Identification:

I have posted some oscilloscope pictures of different target responses on the Geotech, link?
I don't know how to post pictures here.
I would like to use a PIC18F4550 to convert these responses into an audible operator interface.
If this works good then you could move on and improve the signal processing with DSP.


Apr 28, 2010 3:35 am

I added note on that diagram (Pi decay) to make more clear about metal detectors which can and can't differentiate signal from the coil.

Since I can't find and contact author of that diagram, I think that copyright is not abused.

And, I am not registered user, so I can't post that change in diagram here. Sorry.


New User

Joined: 28 Apr 2010
Posts: 1
Location: Vinkovci, Croatia

Apr 28, 2010 3:40 am

Heh, nice forum. Now registered, but still can't upload picture.
New User

Joined: 30 Oct 2010
Posts: 2

Oct 30, 2010 7:45 pm

@Lammert: Although the DSP has a sampling frequency of 1 MHz, what will actual sampling rate be inclusive of saving the A/D value to a variable?

       Computer Interfacing Forum Index -> Pulse Induction metal detector
Page 1 of 1

Running on php BB © 2001, 2009 php BB Group
   Lammert Bies     Interfacing     Sitemap     Forum