The 3.1415 Metal Detector

I've  done quite a bit of research into hand held metal detector  technologies. Beat Frequency Oscillator (BFO) is quite old.  Very Low  Frequency (VLF) is 25+ years old technology.  Pulse Induction (PI)   technology is the latest technology out there for metal detecting, and  good quality PI metal detector units can retail well into the thousands  of dollars. 

An excellent discussion of metal detection technologies can be found here:

For  those of us with ADHD who don't have the time/will/want to digest  humbling technology write-ups,  there are some excellent animations  available here how the various metal detector technologies work:

PI  technology pulses a search coil with current.  While the current is  flowing in the search coil, it creates a strong magnetic field around  the search coil, and this magnetic field imparts energy to any ferrous  of non-ferrous object hidden in the soil within range.   When the  transmit current is rapidly switched off to the search coil,  the object  found in the soil will rapidly shed its newly excited energy.  This  shed energy is detected by the search coil in a listening mode.  The  best lay man's analogy is like sonar where a signal is sent out and it  bounces off of an object and then received back at the point of  transmission, but instead of using sound, such as sonar, PI metal  detectors use magnetic coupling to transmit a pulse and then "listen"  for a magnetic received signature.

With  the fast sampling ADC contained on the evaluation unit,  it should be  fast enough to directly sample the received signal from the target  object located in soil  The received data is then crunched by the  onboard DSP.  Should this unit be used for gold nugget detecting, a  sample must be collected within 10us of turning off the transmission  pulse.  Searching for coins and other metal objects makes for more  forgiving (longer)  timing margins.  TX/RX receive circuit design is  crucial to support fast unit step response margins.  The most critical  element of a metal detector is the search coil, and with proper  construction,  getting a valid RX signal within 10us after ending the TX  pulse is possible controlling the inductance and coupling capacitance  along with "tuning" the coil after construction.  In geological  conditions where BFO and VLF metal detectors cannot operate,  the PI  metal detectors excel in highly mineralized and magnetically "hot" soils  because a relatively novel ground balance hardware trick is used and  the magnetically "hot" soil's signature decays too rapidly for the metal  detector to detect. The downside is that PI metal detectors cannot  discriminate between ferrous and non-ferrous objects easily.  If you are  using a PI detector to hunt for gold, such as here in Arizona,  discrimination is not wanted because there is very little trash in the  soil for false alarms (such as beer caps),  and gold has magnetic  signatures of either ferrous or nonferrous materials  depending on the  mass of gold detected.  Yes,  here in Arizona, there is literally gold laying on the ground, or just beneath it.

Fortunately,   for this author,  the internet abounds in home built P.I. detector  technology, however almost none of it is micro processor controlled.   Those existing internet PI metal detector designs available usually use a  low horsepower MCU, such as a PIC to control only the hardware timings  of the TX/RX signals.   Renesas' new high horsepower MCU with DSP  capabilities offers new possibilities to light battery powered metal  detectors.

Some existing PI metal detector designs:

The HammerHead - Quite a good reputation for a home build unit.  Not MCU controlled.

The GoldPic PI - a PIC controlled unit that attempts to directly sample data using an external ADC.

The  QED - A MCU/DSP "vaporware" development effort with a lot of  unsubstantiated pre-hype trash talk that has been ongoing for several  years now, but has never been released into the "wild".

My progress to date is that of research of metal detection technologies,   selecting  a suitable technology given the constraints of the evaluation hardware,  adapting the schematics, and reading the instruction book for the eval  unit, including the RTOS part.  Have also done quite a bit of research on search coils and how to build them fast enough to support gold hunting. 
  • Still curious as to your progress.  This is the same information posted back in Jan. 8. Although the "Laymans" description is a bit off, but sufficient for those that don't work with EM fields and Metals.

  • Yes this was a copy/paste from my original checkin post.   I am starting a construction blog and this was the first posting.   I have some photos to post, but they are on my camera at home. More to be added later.


  • I have constructed a fast coil. Attached are construction photos. 

    Teflon coated wire and polyethelene wrap are used because of their low dielectric constants which is good for fast coil recovery times.

    The wire insulation has thickness which is good for reducing interwinding capacitive coupling and speading up coil speed, as opposed to enameled coated wire.

    Scotch No. 24 wire mesh is used as shielding because it provides EMI shielding while the mesh provides reduced capacitive coupling compared to a solid shielding.

  • I have configured MTU6 to output the two PWM outputs required to drive the coil and enable the amplifiers during the sampling period.

    Attached is a screenshot of the pulse and the output sampling pulse.

  • Here is the breadboarded MOSFET driver along with the MOSFET used to drive the Coil pictured in photo.  Flyback voltage of 400V was achieved with a 50us pulse of 10V.  The coil is about 5 ohms which means about a 2A pulse on the transmit cycle.

  • Here is the coil and how it fits into the coil housing.
  • Notice how the number on the display represents 2^13 = 8190?   Yes,  a 13 bit virtual ADC has been created from the 12 bit onboard ADC.  Anything for that extra bit.

  • There's a couple of things youcan do for the pulse section.  A resistor across the coil will help reduce ringing in the coil at the end of the kick-back pulse.  Also you can use a .1 Ohm resistor in series to get an accurate measurement for the current.  It will come across as a Triangle wave form, coil resistance is negligable for those, however the inductance tends to become an issue at the lower pulse frequencies.

  • My furnace controller does 16x oversampling on the temperature sensors to get sub-degree "precision".  Despite the specs saying it's only good to 1C, it seems to give believable sub-degree results.

    I think oversampling ADCs is a common thing to do... and the RX has a 4x oversample mode too!  It must be a good thing to do :-)

  • Yes,  that is exactly how I am increasing the ADC from 12 bits to 13 bits is using the 4 times oversampling build into the ADC.  I have very tight timing requirements for sampling a metal detector voltages in the microsecond range, so I could not oversample by 16 to get two extra bits.Ideally,  I would like to use a faster ADC, but we are giving the internal one a go for the contest.

    For those of you new to this,  you can increase your ADC effective number of bits by oversampling by 4^n where n= to the number of extra bits you want.

    1 extra bit needs 4 samples

    2 extra bits needs 16 samples

    3 extra bits needs 64 samples

    4 extra bits needs 256 samples  (BUT YOU SAMPLE 256 TIMES SLOWER)

  • You busted me getting sloppy on my damping resistor.  It was thrown in at something like 560 ohms and it was a little overdamped.  I gotta get some RG/8 Coax tonight from Radio Snack tonight and determine the critically damp resistance of my coil with coax and amplifier in circuit. This coil calculates to just about 300uH with a free resonnence frequency of about 1.11MHz +/- ???  (with my really crappy test measurement equipment), so I am definitely in the ballpark for getting a 10 us recovery time out of this coil necessary for small gold detection.

  • That first link is a gold mine.

    I've been interested in metal detectors for a very long time.

    Back in 1976, I designed what must be the first computerized detector, though the

    hardware was lacking the horse power, so I never built it....

    On top of the fact that Whites wanted me to sign a NDA that would have given it full rights to the device, and

    gaussian noise - gaussian noise = gassian noise, so the simple idea of sampling a smaller returned signal using a second coil,

    and subtracting that signal out of the recieved signal was not going to effectivly cancel the ground effect.... :

    However, I got hard copies of every patent I could find for metal detectors. The first link is a goldmine of patents!

    PI detectors are great devices, but this chip is also well suited for a standard VLF, using the chip to generate the signal, time the

    phase shifts and increase and decrease the power to determine depth, are all within it's processing power...

  • The ringing may be useful.

    By feeding the signal through a voltage divider, and a threshold detector, you can get

    a pulse transition at each zero crossing point.  The width of these crossings is a measurement of the impedence or loading on the

    coil.  It may well be, that this can be used to monitor ground effects......

  • Because you're using an adjacent coil system there are some interesting things that can be done with a single processor.  For our systems, we utilize a CPLD to generate the pulses required to do some of the noise clean up.  We also use a Bipolar pulse to improve sampling rate.  But then again we also make industrial use detectors.  We utilize the damping resistor, or even a buck wound transmit coil to help dampen the ringing along with improving the timing of the kick back pulse.  We also use limiting Zener diodes to limit the kick back, with an 8V pulse, with a 250V kickback.  We can see down  to 1.5% aperture, with that machine 10" inches between TX/RX antennas means we can see a cube of Carbon Steel about .15" in diameter. This is a scope image for our lower sensitivity model, that shows the current spike to induce the Eddy Current in the metal.