Please review: "improved" Mark Stuart's PI detector

[Janne]

Thanks for your reply. Yes, indeed I've used twisted wire for connecting the coil. Didn't realize the capacitance issue you mentioned. I'll try normal untwisted flat wire and see how it works.
The coil itself measures 340µH.

[Janne]

Hi,
I tried untwisting the coil wires, and also swapped the cable to a straight one.. not much but mayby slight improvement on the damping. 2 pictures of the amplifier's output waveform;

Before;



After;

[kt315]

it's good clear demonstrative picture what damping resistor does. i hope this will help.

[Carl-NC]

Nice, I like that.

[Janne]

Hi,

I've seen that picture before, but it's not clear for me what it's representing. Is it the amplifiers' output with different values of damping resistors?

[Qiaozhi]

Hi,

I've seen that picture before, but it's not clear for me what it's representing. Is it the amplifiers' output with different values of damping resistors?

It's the output of the preamp for different values of damping resistor. The optimum value is found when the signal moves smoothly down to zero. This is known as "critical damping". If the waveform is ringing, then it is underdamped, and too low a resistance will cause overdamping. The critically damped position theoretically gives the best response to all targets.

[Janne]

Thanks for your reply. It looks like I've put my design into the underdamped category, as that extra "ringing colum" on the waveform wasn't there with the previous 1k damper resistor.

[Janne]

I've finished writing the first version of the software, and were able to do some actual field tests.

Before getting there I also had to do some modifications to the hardware. I had to reduce the gain of the main amplifier to 1000, with high gain I had problems with offset drifting. Guess the LM318 was a bad choise, but at the time I didn't know what parameters to pay attention to.

What the program does, is to sample the coil every 3ms. These samples are then first averaged, and after that fed to a butterworth low pass filter, with adjustable cutoff frequency. Seems to work nice. The conditioned signal is then fed up to the phones, which has a total of 984 steps in the range of 8-12500 tones /s. seems to be enough for the steps to not be noticeable.
Zero point can be calibrated automatically with a push from a button..
What is still missing is battery voltage measurement, and adjusting the gate time based on that.

Practical performance seems to be ok, It is possible to locate 1€ coil from grass about 20-25cm away. And I already stumbled upon an old lawnmover blade someone had buried

[6666]

Great work what chip are you useing ?

[Janne]

Great work what chip are you useing ?

Hi the main micro is ATMEGA168, but the current program would also fit into an ATMEGA88.
Took a couple of pictures of the prototype.. it's made from electrical conduit =) I need to pick a crutch from somewhere to make the handle.

For some reason this refuses to attach images, links to them instead.

http://pics.ww.com/v/Janne/Electroni...90008.jpg.html
http://pics.ww.com/v/Janne/Electroni...90009.jpg.html
http://pics.ww.com/v/Janne/Electroni...90011.jpg.html

I'm also attaching the firmware and final version of the schematic for anyone interested to study.

[Nicolas]

Nice work Janne, I like that.

[6666]

Janne what is the function of Pot 1, 2 and 3 ?
thanks

[Janne]

I intended to use them to control the sample pulse timings, but in the end I hardcoded those values into the program. Only POT1 is currently in use, and that controls the cutoff frequency of the butterworth filter, essentially determining how fast the detector respods to targets.. Longer time --> less chatter, but coil has to be swept slower.

[ODM]

Looking interesting. Apparently an AVR DAC's resolution is enough for such work then.