Let's made a PC-base metal detector with usb interface !!!

[Mrand]

Maybe I will give the new impulse in the discussion reminding the TX circuit Troy X5.
Mrand

[Aziz]

Hi,

you all are welcome to simulate the feeder circuit. Attached file is the source code for LTspice. Remove the ".txt" extension and load into LTspice.

Enjoy,

Aziz

[simonbaker]

I am adding this new comment: my above comments probably may not apply to your circuit because you chose capacitor values that make it behave more like a "series resonant" circuit than a "parallel resonant" circuit.

So my new question is: suppose you simply substituted a parallel configuration for the series configuration without the extra feed capacitors in your very first diagram, and also remove the current limiting resistor. You would need to assume some Q for the LC circuit. In other words, what is the difference between a basic series LC configuration and a parallel LC configuration for use in a metal detector?

[simonbaker]

Enjoy is correct! This is very interesting.

I have been thinking about coil resonance, maximizing current, and saving energy for a while.

Here is a question/comment: Correct me if I'm wrong: I understand why the current limiting resistor in the series resonant circuit, because at resonance the impedance of the LC goes to zero. Theoretically we can use a very high limiting resistance and still have a large current in the coil because it exchanges current and energy with the capacitor (I think) without going through the resistor. If we make the current limiting resistor low, we just waste energy in the resistor, there is probably a good trade-off value we want to find.

But I don't think we need any limiting resistor with the parallel resonant circuit because the impedance of the LC goes to infinity - the current in the coil is only limited by the Q of the LC. I guess if the Q is very high, the current limiting resistor doesn't play much part anyway, but better to take it away completely, no?

Just some thoughts without calculations. I'll try to do some calculations and correct my thoughts if I understand it differently.

[Aziz]

Hi guys,

here are some interesting spice simulation results and the example feed circuits. It compares the normal series LC feed with the proposed one (I don't know how it is called). Both circuits take the same signal source (earphone output, 20 ohms impedance, ±2V = 4 Vpp) and the coils are exactly same with same parasitric characteristics. C1 of the first circuit is matched to have the same resonant frequency of 10.63 kHz. See the coil parameters on the attached picture.

The proposed circuit will load the signal source much lower than the first one despite of more voltage and current on the coil.
The current limiting resistor R1, R2 is necessary due to zero impedance of the resonant circuit.

Other combinations of the capacitor values could have better frequency response. If you like, then try to find a better one.

Enjoy,

Aziz

[Aziz]

Are you saying you found a way to increase the Q of your coil resonance with a clever feed? How is it different from a normal coil / feed? Are not all coils designed to resonate with as little power as possible?

Here is an experiment -- put coil on mineralized ground -- how does it affect the voltage (Q)?

I think your research is very interesting to help us design MDs. Thank you as always.

Yes, that's true. I have increased the Q of the resonant circuit much. This is a typical narrow band filter. And this is predestined to lock-in amplifier to react to phase and frequency changes.
It makes sense to use HF litz wire for the search coil to minimize the coil losses (higher Rpar).
I will make more simulations on multi-pole resonant filters.
At the end of the week, I will see the difference to my previous experiments.
It's getting very interesting.

Aziz

[simonbaker]

More good news:

3.
Taking a very clever feed circuit, the coil current can be increased despite of the low current feed from the sound-card. This can be achieved by getting the LC circuit totaly in resonant so the sound-card delivers power only for the coil/capacitor/cable losses. My sound-card delivers only 4 Vpp and with a clever feed, the coil voltage can go upto 40-50 Vpp!

This is the most good news I can tell. To design the feed circuit proper, I have to measure some coil parasitic parameters (Rpar, Cpar).


This exploration omits the need for power amplifier for the TX coil.

Aziz

Are you saying you found a way to increase the Q of your coil resonance with a clever feed? How is it different from a normal coil / feed? Are not all coils designed to resonate with as little power as possible?

Here is an experiment -- put coil on mineralized ground -- how does it affect the voltage (Q)?

I think your research is very interesting to help us design MDs. Thank you as always.

[Aziz]

Hi Sean,

thank you for your offer. I will have soon a brand new one sponsered by my parents with high fidelity sound-card built-in with 4-6 hours operating time.
1 GHz will be enough for the laptop MD. Even lower to save battery time.

Enclosed the AC signal analysis (frequency response on the coil). There is an excellent frequency response to minimize the 50/60 Hz, 100/120 Hz line injected ripple noise. Also the 1/f noise will be removed excellent.

Now it makes sense, to develop the laptop MD further.

Aziz

[Sean_Goddard]

I have been SAYING this was the way to go for the last 15 years and no one listened..Why do I bother??

WELL DONE AZIZ, you are truly a genius BTW Would a Pentium 3 1GHz laptop do you? I have one but there are plenty on Ebay cheap if you look.

I was planning to give this machine to my sister, but you can have it if it will do what you require. Needs a new CD rom drive. The one in there is KAPUT.

Let me know here or PM me with your address. BTW you will have to pay the shipping, but that's the ONLY thing you pay for. Laptop is free.

Sound card in it is not good though, so it may not be any use. What about a USB sound card??

[Aziz]

More good news:

3.
Taking a very clever feed circuit, the coil current can be increased despite of the low current feed from the sound-card. This can be achieved by getting the LC circuit totaly in resonant so the sound-card delivers power only for the coil/capacitor/cable losses. My sound-card delivers only 4 Vpp and with a clever feed, the coil voltage can go upto 40-50 Vpp!

This is the most good news I can tell. To design the feed circuit proper, I have to measure some coil parasitic parameters (Rpar, Cpar).


This exploration omits the need for power amplifier for the TX coil.

Aziz

[Aziz]

Very good news:
1.
I experimented with simple line-out impedances (200 ohms on my sound-card measured) till now. I have found the right software adjustment of my sound-card to change this impedance to ear-phone output. The impedance now falls down to ~20 Ohms!!! I have now without an external amplifier more power (current) for the transmit coil.

2.
I will put both left and right input channels together to acquire same signal on the rx coil. The signal-to-noise ratio will be of 1.41 times better, when I take the average of the left and right channel ( signal = 0.5*(left+right) )

So, I will make a new measurement experiment on a very simple laptop MD architecture without an external circuitry (only search coil connected to laptop).

Now, I can break new records.

Aziz

[Aziz]

Hi guys,

I have continued this project again for preparing some new ideas and breaking new detection records..
But lots of software changes to implement..
I will let you know about the results.

Aziz

[simonbaker]

Thanks Aziz.

I think I would need to see your actual equations or algorithm to fully understand, maybe someday .

I will follow your progress. Cheers.

-SB

[Aziz]

I forgot the main reason for two stages and Boxcar integrator:
Saving battery power!

This is becaming an important issue to laptop MD and therefore reducing the computation power and saves lots of miliamperes!

[Aziz]

Hi SB,

usual one channel - in-phase - lock-in amplifiers use following parameters:
- lock-in frequency f (signal frequency / reference frequency)
- phase shift p (of the reference oscillator)
- TC (time constant for the low pass filter for the mixed channel)

The output of the one channel lock-in amp is depended on the phase shift p and at the maximum output value, the phase shift of the signal to reference can be read out directly. So the user has to adjust the phase shift for maximum output value and read then the phase shift and output value. One channel lock-in amps usually used former as analog lock-in amps.

The two channel - or quadrature-phase (I & Q) - lock-in amplifiers do not need the phase shift information from the user. The output of the IQ-lock-in amp represents a unity complex vector (channel I=in-phase, mixed with cos(2*pi*f); channel Q=quadrature-phase, mixed with sin(2*pi*f) ) and the output magnitude R and phase P can be computed directly. Any phase shifts of the reference oscillator will only phase shift the output phase. The magnitude will not be changed on phase shifts. I & Q lock-in amps usually used in digital lock-in amps. My laptop MD is implemented in this way.

The time constant TC is the low pass filter time constant for each channel (I and Q). The -3dB cut-off frequency will be fc = 1 / (2*pi*TC). It defines the filter response and the higher TC is, the higher the signal-to-noise ratio and the lower the response time to changes it will be. So using higher TC will slow down the detection speed. To much TC will slow down the computation time. Taking some reasonable TC's for desired detection speed is necessary.

Now the Boxcar integrator:
The output of the lock-in amplifier (channel magnitude and phase) can be further low-pass filtered using Boxcar integrator. It has some decimation facility to the huge amount of data and will speed up the computation time. Different time constants for the Boxcar can be taken to implement different sweep speeds of the coil. So a higher TC will be used for pin-pointing to get a better sensitivity and stability.

So a two stage of system:
1. Lock-in Amplifier with its own TC
2. Boxcar integrator with its own TC applied to stage 1

If one takes a much higher TC for the lock-in amp, the Boxcar integrator can be neglected. Of course, the laptop has much more computation power to handle this huge of data in real-time! The advantage of implementing more Boxcar integrators is to react to different changes in response time.

Thats all. There is no mystery but a much potential to explore.

Aziz