Moodz' Awesome Gold Pulse Induction Version 3 - MAGPI V3 Project

[pito]

Attached is pdf file with explanation of my construction. = and you have 1600V ?

[moodz]

Just about to commence laying out a PCB from the breadboard. Maxed out the TX power and speed that will reasonably work with off the shelf mono coils. The final figures for the MAGPI V3x will be something like 10 Khz TX frequency with 1.2 Amps peak current and 1.2 KV flyback. Damped in 650 nanoseconds and two channels ( one for discrim and one for "all targets" ). The waveforms are near perfect damping.
Preamp is 100X - red trace. Flyback and coil current shown.

[Detectorist#1]

Hi Teleno,

I have not use spiral construction of the coils. The mechanical construction is relatively simple.
Attached is pdf file with explanation of my construction.

[Teleno]

Hi Teleno,

In my project of PI MD I use Two TX coils with bifilar winding and two RX coils in gradiometric connection. The detector works in two modes - "ALL metal" and "Fe only". In "Fe only mode TX coils are not used (no pulses).
The input signal is only from electromagnetic induction in two RX coils. In this mode is very important to use Op Amp with lowest possible noise. Nevertheless, the sensitivity for Fe objects is reduced in comparison with ALL metal mode. The profit is that recognition of Fe objects is 100% reliable.

I have trouble encapsulating a spiral mono coil, I can't even imagine how to do it with a gradiometric one, kind of a cilinder I guess.

I'm working on a simpler scheme in which I discharge a cap on the coil for a fixed time (mcu timer) and the residual voltage is proportional to the inductance, so any changes caused by ferrous could in principle be detectable. Will have to verify what the sensitivity is. The residual voltage is also used as the analog ground to simplify the power supply.

[Detectorist#1]

Hi Teleno,

In my project of PI MD I use Two TX coils with bifilar winding and two RX coils in gradiometric connection. The detector works in two modes - "ALL metal" and "Fe only". In "Fe only mode TX coils are not used (no pulses).
The input signal is only from electromagnetic induction in two RX coils. In this mode is very important to use Op Amp with lowest possible noise. Nevertheless, the sensitivity for Fe objects is reduced in comparison with ALL metal mode. The profit is that recognition of Fe objects is 100% reliable.

[Teleno]

Hi Moods!
It is very strange for me how you have the claim for sampling noise reduction with using of LM6171 as front end. This IC have 130nV and 8pA noise at 2Hz.
Where is "hidden the dog"?

The target decay is exponential so the additional op-amp noise is compensateď for by a larger signal of earler sampling.

[Detectorist#1]

Hi Moodz,
Really very impressive - 500ns! In my practice (not very big), every attempt to use delays under 2.5us haves bad results - the screen of the coil stops to works as screen and the connecting cable starts to be more sensitive
to near hands than the coil. It seems for me, these 2.5us are "magic" border! Do you have some explanation for this effect? Or I not understand what really happens?
Thank you in advance!

[moodz]

Glad you like it ...
Fast coils are fast because the capacitance is low .. however if you utilise the energy in the coil ( 1/2 LI squared ) the capacitance term is not there and the time constant of a shorted cap approaches zero... so capacitance is less consequential.

So regardless of how "fast" the coil is my latest circuit can damp a typical PI coil in around 500 nanoseconds ... much faster than anything else I am aware of

[Detectorist#1]

Hi Moodz,

Thank you for these additional explanations! I know, the real things are more complex than seems at first look.
The idea for use current load for fast damping in not bad, but using of fast coils allow to make first sample after 2.5us TX switch-off and this is more fast than the time in your time diagrams.
The path for improving PI MD become more and more complex, but this will not stops our common efforts!
Success!

[moodz]

THIS IS HOW YOU DO DISCRIM ON THE MAGPI >>>>

I have got to do some other stuff so I will describe the MAGPI discrimination method.

If you have been reading the thread the MAGPI is basically a single channel machine and uses an X sample ( reactive ) to control the damping via a feedback loop and an R sample ( real ) for the actual target detection.

In normal MAGPI operation you dont really see metal discrim because the damping feedback loop clamps the X sample to a relatively fixed DC level that controls the damping mosfet.
HOWEVER if you first "open" the control loop temporarily and thus keeping the damping mosfet at a fixed bias for a short time ( maintained by gate capacitor ) the X sample will now provide target discrim information. So if the R sample timing is now moved to X sample timing you get discrim info on the target. ( ie Ferrous / non Ferrous ) QED as they say.
You could arrange for this to happen by a push button so that you can ID a target or it could happen all the time.

See post #293 for image of result.

Pic below ...

[moodz]

Hi Moods!
It is very strange for me how you have the claim for sampling noise reduction with using of LM6171 as front end. This IC have 130nV and 8pA noise at 2Hz.
Where is "hidden the dog"?

To answer your question ... the Magpi 3 is a "demonstrator" platform not a "uber detector". I spent all of 5 minutes selecting that opamp. The reason it was chosen ( say rather than a NE5534 and I am not saying the NE5534 is choice either) is that it can take rail to rail overload voltages at the input whereas the NE5534 has the equivalent of diodes across the inputs so any overload voltages above 0.6 volts cause current to flow in the opamp inputs. This is not good as it a causes errors in the current damping that the V3 uses.
Feel free to choose your own opamp ... V3 is for experimenters.

For instance the noise cancelling loop still does a great job ( even with the 6171 ).

And importantly PI ferrous / non ferrous discrimination is there in plain site ( bad luck .. patent guys ) ... it just was not implemented in V3 ... but can be with simple software change.

[moodz]

Sometimes the focus on a single tree hides the benefits of the forest!!

Could not have put it more succinctly than that

[Orbit]

Sometimes the focus on a single tree hides the benefits of the forest!!

[KingJL]

Hi Moods!
It is very strange for me how you have the claim for sampling noise reduction with using of LM6171 as front end. This IC have 130nV and 8pA noise at 2Hz.
Where is "hidden the dog"?

Sometimes the focus on a single tree hides the benefits of the forest!!

[Detectorist#1]

Hi Moods!
It is very strange for me how you have the claim for sampling noise reduction with using of LM6171 as front end. This IC have 130nV and 8pA noise at 2Hz.
Where is "hidden the dog"?