VLF MD with digital signal processing : Bee-Buzz 1

[Carl-NC]

OK, I didn't see that there were more pages.

[Aziz]

And how this old information for low noise amplifiers with input power transistor and ultra low source impedance is connected to real metal detector's schematic?

Hi,

I do connect the RX-coil directly to the input of the preamp (note, the preamp is AC coupled). RX coil resistor will be in range of 1 - 2 Ohm to keep the noise very very low.
But the coil will pick up a lot more EMI noise.

You can't use the preamp on a PI. Only VLF/LF detector type.
Aziz

[Detectorist#1]

Hi Aziz,
Thank you for the answer.

[Aziz]

Hi,

I really enjoyed reading all the 6 pages. It is well done and a good source regarding amplifier noise. Go back to his home page and you will find more interesting things.

http://www.dicks-website.eu/index.html

http://www.dicks-website.eu/low_nois...rt1/part1.html


You can bet, that some transistors will be placed on my purchase list.

0.2 nV/sqrt(Hz) input voltage noise is easily possible.
Aziz

[moodz]

Does not matter how low your input noise is ..the insertion noise of the amplifier raises the noise floor above the input noise floor of any competent ADC ....and ruins the noise response of the system. The target signals are not buried in low levels ... They are actually the difference in a signal composite of transmitter and wanted / unwanted signals ..the task is to separate the wanted signals from unwanted/ transmit signals nothing to do with low noise inputs. We are not trying to receive a signal reflected off the moon ...that requires a low noise amp ...our targets are less than a meter away it's just sampling and mathematics.

[Aziz]

Hi Moodz,

I find it still interesting to push the limits.
With a simple design. This is soo cool.

Such a front-end hasn't been used yet.

Aziz

[Aziz]

Hi all,

I am still coding the reference C source code of the demodulators (FFT, Goertzel and Lock-in) to share with you. After making 1001 changes, I have found and fixed an evil bug in my experimental detector software on which I code the new demodulators at the moment. I am not finished yet and it will take some time. They will be tested on real conditions too.

BTW I have attached a 1050µH RX-coil to the sound card input and the noise floor didn't rise. So all the signals and EMI noise is still below the noise floor of the ADC system (-120 dB or so). An RX pre-amp would be necessary if we want to detect much much deeper. Maybe a super-duper-ultra-low-noise pre-amp?

On the TX side, the Sound BlasterX G6 has really huge power to the TX-coil on the headphone output. No output amplifier would be necessary. On line-out output, we would need an output amplifier to detect much deeper. Power really matters!

Cheers
Aziz

[Aziz]

Hi all,

I am really curious to how we can process all the bunch of demodulator outputs.


In my special case with dual resonant frequency F1 and F2, I will have at least 12 demods with magnitude and phase info. 6 demods for TX-reference signal and 6 for RX-signal. On each resonant frequency there are 3 demods around the resonant frequency Fr:
Fl, Fr, Fh,
dF = Fr - Fl = Fh - Fr,
dF is the bin frequency resolution,
Fl=next lower,
Fh=next higher frequency

If I take N = 256 (block size), than Goertzel is efficient compared to FFT if the following condition is fullfilled:
P < log₂(N),
N = block size
P = number of demods
so, P < 8, it is efficient to use Goertzel.

Note that I would require without Goertzel 2 FFT's: one for TX-reference and one for RX-signal. So 6 is < log₂(N) and therefore it is convenient to use Goertzel.

BTW, you don't really need Lock-in Amplifier demods. I am also getting rid of the Lock-in Amplifier. Goertzel is beautiful and delivers same results. You will see the comparison, when I publish the c-reference code soon. Be patient please.
We will be able to have a super-duper-dual-frequency-VLF/LF-killer-detector soon.

Cheers,
Aziz

[Aziz]

Hi all,

I have finished the c-reference code for FFT, Goertzel and Lock-in Amplifier for your convenience. The license model is the "The MIT License". You can do what ever you want to do with it. No warranty will be given. Just adapt & arrange to fit your demands!

I hope, I haven't overlooked bugs in the code. The complete C-project can be opened via Visual Studio (I have actually used the Visual C++ 6.0). It is a Win32 console app, which writes the calculated results into a text file (DemodTest.txt).
In the main program, there are different tests and you can see, how to use the API.
The setup functions will be called once per demodulator of specific frequency (k) of course. But I sweep the k from 0.. N-1 to get the whole complex spectrum just like FFT would do. You can also set k negative to demodulate negative bin frequencies.
The complex output of FFT, Goertzel and Lock-in Amplifier is the same (for N=32). Forget FFT and Lock-in Amplifier and use the Goertzel. Use even block size N Goertzel. Otherwise you will have spectral leakage effect.
The API and the functions are commented. It is kept as simple as possible. You can also use the modules in C++ projects.

The ASSERT_()-macros are for debugging mode and will be empty if compiled in release mode (no debugging). The scaling of the result is correct compared to many other publications. If you set iFlags DEMOD_POWER for decoding, the total spectral energy will be calculated for positive bin frequencies. If set to DEMOD_COMPLEX, to total spectral energy is halved into negative and positive bin frequency.
These are the binary-or flags for the variable iFlags

  // Binary or Flags
  //
  DEMOD_COMPLEX        = 0, // I/Q or Re/Im won't be scaled for full complex spectrum output
  DEMOD_POWER          = 1, // I/Q or Re/Im will be scaled for positive spectrum output only (scaling by 2 for k=1 .. k = N/2 - 1, for k > N/2 set to 0)
 
  DEMOD_POLARCALC      = 2, // Calculate additionaly polar coordinates (Radius/Magnitude & Phase)
  DEMOD_EPSILONLIMIT   = 4, // Set output to 0 below EPSILON for stable phase calculation
 
  DEMOD_STANDARD       = DEMOD_POWER|DEMOD_POLARCALC|DEMOD_EPSILONLIMIT, // switch everything on

​

You can also calculate polar result (magnitude, phase).

The interesting modules are:

preproc.h : Select REAL type definition (either float or double)
RealType.h : Type definition for REAL (either float or double)
Demod.h : Demodulator result output structure
fft.h, fft.c : FFT-Modules
Goertzel.h, Goertzel.c : Goertzel-Modules
Lockinamp.h, Lockinamp.c: Lock-in Amplifier Modules
​main.c : Main test programm with many test cases.
BinCalculator.xlsx : Excel table to play with.
DemodTest.txt : Output text file with results.

Cheers,
Aziz

[Aziz]

Hi all,

I have been for a week in Berlin and had no time for the project.

Regarding the choke L1:
In a resonant TX-tank circuit, we can not afford any losses in the core material. So the choke L1 will be a tight winded air core coil choke.

I have to purchase a 1 mm diameter enamelled magnet wire to solve the choke issue.

There a many ready to use inductances in the hi-fi area, but they are totally overpriced.

Aziz

[Marchel]

For those who want to make a metal detector based on STM32, I have completed the PCB for the STM32H7 development board. The materials and information are in this post.

https://www.geotech1.com/forums/foru...elopment-board

[Atul Asthana]

Apologies for being away for so long and not following the discussions.
I had to spend over a week in Heart ICU, due to my mrdical condition and I am still recovering.

Will take me some time to latch on to the discussions here.

[Aziz]

Hi Atul,

get well soon. No worries, I'll keep the thread a live.

Aziz

[dbanner]

Apologies for being away for so long and not following the discussions.
I had to spend over a week in Heart ICU, due to my mrdical condition and I am still recovering.

Will take me some time to latch on to the discussions here.

Good that you're recovering.

[Atul Asthana]

Hi Atul,

get well soon. No worries, I'll keep the thread a live.

Aziz

thank you so much Aziz and dbanner