PICKINI - an easy to build PI detector

[F117]

PICKINI = PI detector, PIC based, KINE = motion, PICKINI= typical garment in the natural habitat of this MD, the beach.
This is a very simple, straightforward, easy to build PI detector. I am in no way an analog electronics specialist, so I shamelessly copied the front end of the blind squirrel detector and hooked it up to a simple Microchip controller. Since software is more my speciality, the shortcomings of the front-end electronics are somewhat compensated by the signal processing software + audio feedback.
There is only 1 control: a 4K7 potmeter for sensitivity combined with the power switch.

So far, we built 6 units, and all perform equally well. Air tests show about the sensitivity one can expect from a 20 cm diameter coil: about 25 cm for my gold wedding ring. 20 - 30 cm for euro coins. All depends on the quality/size of the coil.
That is why I spent some more time on field tests instead of simple air tests. I normally search with an XP GoldMaxx Power which is like most IBdetectors an excellent detector on fields, but is close to worthless on the beach. With the Pickini detector I searched fields that I had gone over with with the XP GMp and still recovered small items at greater depth.
On the beach, it never gets boring, although a lot of the finds are trash.
You can find all the data + PCB + software here:

http://users.telenet.be/willaert/MD/..._en/intro.html






Have fun,
- Bernard

[sinclairuser]

Thanks Bernard.

[johnandles]

Thanks Bernard for sharing this detector.
Have you done any tests on very small gold?.
Cheers
John

[F117]

Have you done any tests on very small gold?.

No, for 2 simple reasons:
1. The only place in Belgium where there is gold to be found is the beach (wedding rings, earrings, ...).
2. I don't have any natural small gold nuggets. As mentioned, it will detect my wedding ring, but this is probably not pure gold

- Bernard

[johnandles]

No, for 2 simple reasons:
1. The only place in Belgium where there is gold to be found is the beach (wedding rings, earrings, ...).
2. I don't have any natural small gold nuggets. As mentioned, it will detect my wedding ring, but this is probably not pure gold

- Bernard

Thanks for your reply.
I live in Australia,( natural gold deposits..but where? )... and I would like to build this detector.
Is the software, timings etc optimized for a particular mineral?
Is it possible for you to write new software so timings to be externally adjustable?
So many questions...
Regards
John

[brucester]

Congrats on your detector.

You have just finished what i am starting

Well done

[F117]

Is the software, timings etc optimized for a particular mineral?
John

Hi John,
the software is pretty straightforward:
The flyback decay pulse is fed into the internal comparator of the PIC.
It slices the pulse at 3.75 V.
The output of the comparator enables an internal 16 bit counter, running at 20 MHz.
It takes a running average of this counter value as reference, which is very stable.
Any changes in counter value mean detection.
The sensitivity potmeter results in the number of pulses that are accumulated in the counter. The more accumulated values, the more sensitive to changes.

So, there is no 'delay' adjustment, unless you want to set the opamp DC differently with the multiturn potmeter. This will influence the point where the pulse is sliced inside the comparator.

Regards,
- Bernard

[F117]

Congrats on your detector.

You have just finished what i am starting

Well done

Thanks !
Feel free to try it out. Once you have the PCB, it takes less than 1 hour to stuff it.
- Bernard

[johnandles]

Hi John,
the software is pretty straightforward:
The flyback decay pulse is fed into the internal comparator of the PIC.
It slices the pulse at 3.75 V.
The output of the comparator enables an internal 16 bit counter, running at 20 MHz.
It takes a running average of this counter value as reference, which is very stable.
Any changes in counter value mean detection.
The sensitivity potmeter results in the number of pulses that are accumulated in the counter. The more accumulated values, the more sensitive to changes.

So, there is no 'delay' adjustment, unless you want to set the opamp DC differently with the multiturn potmeter. This will influence the point where the pulse is sliced inside the comparator.

Regards,
- Bernard

Thanks Bernard for the explanation, I guess I was comparing your design to a 'standard' approach to P.I.
I have just ordered the micro, and printed out the Eagle board file.
I have a programmer that can program the hex...so here we go..soon!
I am looking for sub gram gold.
Similar to what my gold bug 2 can find.
Thank you again
Regards
John

[jap]

Goooooood,thanks Bernard.
best time
j.

[ODM]

Good results for the effort and nice looking hardware

I've got two things I'd like to hear in more detail; What kind of running average algorithm do you use? And how does the sensitivity control math work?

[F117]

What kind of running average algorithm do you use? And how does the sensitivity control math work?

Hi ODM,
The "running average algorithm" is limited by the amount of RAM inside the PIC controller and is not as sophisticated as I would have liked:
The potmeter value is read as an ADC input. This gives a 10 bit value. This is reduced to a 6 bit value (0..63). This is the number of pulses that are accumulated in the internal counter (pulse width of the flyback * 50 ns [@20 MHz]). Even when there is an overflow in the internal counter, this snapshot value is compared to the reference value. The reference value is stored whenever "n" consecutive snapshot values (accumulated pulse values) are identical. With identical, I mean binary identical to the lowest bit. Now, "n" is fixed to 4. This results in relatively fast motion response when sweeping at 1m/s.
Suppose, the sensitivity is set to worst case = maximum = 63. This means that 64 flyback samples are accumulated before obtaining a snapshot value. A value of 63 is in the high sensitivity range, with 500 Hz pulses. Total accumulation time = 2 ms * 64 = 128ms --> 8 complete snapshot values / second.
With a sweep speed of 1 m / s, a snapshot value spans 1m /8 = 12.5 cm. So, worst case, when sweeping that fast at maximum sensitivity, you get a signal 12.5 cm [5 "]next to the target. Intuitively, you will decrease sweep speed and pinpoint the target.
So far, I could only crank up the sensitivity potmeter to max when there are no houses/civilisation in the vicinity, like on a large field or beach.
To summarize: the "running average" does not take into account 'anomaly' values, only the real stable values are taken as reference. "Sensitivity" is limited only by the internal counter frequency ( stable 20 MHz xtal reference ) and the "normal" sweep speed.

Regards,
- Bernard

[ODM]

Thank you for the description, I like the averaging method.

I might try and fool around with the pulse width value detection to make for continuous detection when I get around to building a similar decay pulse width PI, using similar fixed point algorithms as I described in my BFO thread. There's plenty of mcu clocks between PI pulses for some rather obscure algorithms even with an 8bit micro.

Will probably be building that the next time I order parts ... and have the time to sit down with them! Fortunately there is still some time for protos before the ground melts over here.

[6666]

Excellent project thanks
and comes with "Technical " back up.

20 - 30 cm for euro coins. All depends on the quality/size of the coil.

Which size coil gave the 30cm result ?
thanks

[F117]

Which size coil gave the 30cm result ?

The coil type that gave overall best results is a flat spiral wound coil - ID 14cm OD 24cm - 30 turns:



But without the shielding tape !
This is the coil that can be seen on one of the detectors on the picture at the top of this topic, embedded in a flat plate.
So far, I only tried this coil and a 20 cm classic round coil - the other one on the picture.
It's also surprising (or not) that the small euro coins, that are made of iron, can be as easily detected as the larger ones.