AI can't invent. AI isn't creative.

[ivconic]

I'm not sure I understand how you're going to do that practically?
In other words; can you immediately make a usable metal detector out of it?
...
The real problem is that the transmit signal leaks straight into the receive coil and the ground response swamps the tiny target signal, so you must cancel both in the analog front end before the ADC, and no amount of digital processing fixes a saturated or buried front end.

[ivconic]

BTW
(This is how the AI analyzes the situation)
New titles, latest news!
XP Detectors — Emergency Relocation to China!
XP — MADE IN FRANCE. RELOCATING FOR SURVIVAL!!!
That I was asked; XP should have moved to China a long time ago!

[Aziz]

...
Aziz do I need an amplifier when my noise floor is around -150 dbv ?

No, you don't need.
Except, you want to go below 0.4 nV/ rt Hz.
At some point, the EMI noise, Coils resistor noise, input impedance noise will dominate anyway.

But I need the damn amplifier to overcome the ADC noise of appr. 6 nV/rt Hz (Sound card noise).
Cheers

[ivconic]

Post #46 above, exactly my point too.

[ivconic]

The problem with AI is similar to the problem with some "nerd" who just graduated from MIT.
In theory, everything seems ideal.
In practice, you are surprised by a very cold shower every day.

[Aziz]

Post #46 above, exactly my point too.

Yeah, you pointed out: The ground noise!

We should focus to the issue with the help of AI to reduce it. There is big potential there.

[ivconic]

I'm still a big proponent of correlation filtering and fft analysis.
For months, when I have free time, I work on it a little at a time.
But the problem is too serious to easily achieve some tangible results.
That's why I don't want to claim and announce anything prematurely, because I haven't achieved anything worth bragging about yet!
Ha, ha, ha!

[moodz]

I'm not sure I understand how you're going to do that practically?
In other words; can you immediately make a usable metal detector out of it?
...
The real problem is that the transmit signal leaks straight into the receive coil and the ground response swamps the tiny target signal, so you must cancel both in the analog front end before the ADC, and no amount of digital processing fixes a saturated or buried front end.

LOL .... its only an ADC ...not a metal detector. The only point is that the noise floor is below what any reasonable amplifer can achieve with around 10 to 20 nanovolts of sensitivity but a dynamic input range of +/- 3.2 volts.
Of course the transmit signal leakage is going to be there ... but its not going to be near 3.2 volts ... so as long as the ADC does not overload its just math.

Below is coil leakage ( on the detector using this ADC ) at 50 mV with a 0.5 mV target at 30 degrees .... on the actual hardware = no problemo.
The envelope is the audio output response .. red lines are thresholds etc