TX Cascade for Multi-Frequency

[JoyJo]

Only the digitized signal. No analog demodulators. According to the scheme, the signal is amplified, and them goes to the ADC (AD7685).

[waltr]

I did play with this idea of generating a multi-frequency TX then amplify and direct sample the RX waveform.

In code (PIC32MZ EF), the frequencies were separated and I-Q demodulated.
Never completed the code to get this to work properly. Mostly got lost in figuring out the math needed.

If you can understand the math required then this can be done totally in software.

The important hardware will be the RX front end and the ADC.

[Carl-NC]

Since you are direct sampling, only one ADC is needed. Walt is correct, do software filtering & demodulation to separate the channel frequencies.

[JoyJo]

Thank you for your advice!
And what is the potential benefit of multi-frequency? Is it more effective to determine the influence of the soil or a more informative vdi? In general, the most tangible problem is the influence of soil, because of which colored targets sound like black metal. By the way, can you recommend literature to read about vector cutting of soil? The vector method is better than the phase cutting of the soil. In fact, we have a ground+target (S) signal on the receiving circuit. How can the vector clipping determine the soil vector (G). If you then subtract S - G, then the required vdi of the target is plucked.

[Carl-NC]

MF helps with both soil and VDI. Once you separate the frequencies you can reject the ground exactly like in a single frequency detector. You can also do a subtraction directly on the reactive & resistive components across the frequencies (X_f1 - k_xX_f2, R_f1 - k_rR_f2). This is best figured out by doing actual ground/target testing to see how things behave over frequency. The best literature to read are patents, they are difficult but very informative.

[JoyJo]

How, in your opinion, is it best to process the ground signal in a single-frequency variant?

[Carl-NC]

I've never compared the two methods so I can't say if one works better than the other. I suggest starting off with the SF method because it's easier to understand.

[Altra]

You could do a fft/dft. I'm no expert but have dabbled and believe its a valid approach.

FFT of your samples will give you the (R)eal and (I)maginary numbers. Which you can then derive the magnitude and phase for each IR pair in each frequency bin. Requires a fixed sample rate that obeys Nyquest. The more samples and twiddle factors the higher the frequency resolution.

In metal detectors the "Real" is the resistive and the Imaginary the reactive components. Phase = atan(I/R) and the magnitude = sqrt(IxI+RxR). Once you have these for each frequency of interest, you can do as Carl suggested, the mixing across frequencies to get your GBs, vdi's.

Using Microchips dsp functions. I did a audio spectrum analyser a long time ago. I also did an analog dual freq detector where I learned how to mix the various signals to disc, gb or even null salt water response.

Here's a app note for 16bit dsp functions, they have similar functions for pic32

https://ww1.microchip.com/downloads/.../90003141a.pdf

All the other processor brands have dsp libraries.

Then again maybe its as simple running your samples through a set of firmware BP filters?

Add: pic32 example
http://hades.mech.northwestern.edu/i...f_Analog_Input

[JoyJo]

I've never compared the two methods so I can't say if one works better than the other. I suggest starting off with the SF method because it's easier to understand.

The off-frequency method is already working. In deel checked. Now there is just an optimization of work with the soil. Video, however, in Russian.
https://www.youtube.com/watch?v=z7_JB6_nPFE
https://www.youtube.com/watch?v=oA_TS8IAh4U
https://www.youtube.com/watch?v=DC0fO4tO77k
https://www.youtube.com/watch?v=TVZkgTvEmU8
https://www.youtube.com/watch?v=eFgT06boolk
Not for advertising, but for the sake of demonstrating the achieved results.

[JoyJo]

You could do a fft/dft. I'm no expert but have dabbled and believe its a valid approach.

FFT of your samples will give you the (R)eal and (I)maginary numbers. Which you can then derive the magnitude and phase for each IR pair in each frequency bin. Requires a fixed sample rate that obeys Nyquest. The more samples and twiddle factors the higher the frequency resolution.

In metal detectors the "Real" is the resistive and the Imaginary the reactive components. Phase = atan(I/R) and the magnitude = sqrt(IxI+RxR). Once you have these for each frequency of interest, you can do as Carl suggested, the mixing across frequencies to get your GBs, vdi's.

Using Microchips dsp functions. I did a audio spectrum analyser a long time ago. I also did an analog dual freq detector where I learned how to mix the various signals to disc, gb or even null salt water response.

Here's a app note for 16bit dsp functions, they have similar functions for pic32

https://ww1.microchip.com/downloads/.../90003141a.pdf

All the other processor brands have dsp libraries.

Then again maybe its as simple running your samples through a set of firmware BP filters?

Add: pic32 example
http://hades.mech.northwestern.edu/i...f_Analog_Input

We, in fact, also use FFT. The metal detector of us is built on a microcontroller from STM: STM32F407
So far, for the single-frequency version, the question is how to optimize the fight against the influence of soil. The microcontroller is now loaded with the main algorithm by about 20%, so there is enough power for mathematical calculations. It remains to understand how and what to calculate. How can you only get a signal from the ground? In real time, we have a mixed target+ground signal. And how to get or calculate only the ground signal?

[ArchibaldSTM]

Dear JoyJo ... You forgot to say that the project Stalker IB is commercial... $$$

[Carl-NC]

So far, for the single-frequency version, the question is how to optimize the fight against the influence of soil.

Are you familiar with how ground balance is achieved in analog detectors?

[JoyJo]

TheYou are both right and also wrong. The project has a share of commerce. You need to buy a counter-controller with a pre-installed loader only from the developer. All updates are free. Our region (Russia, Belarus, Kazakhstan and a number of other countries) knows the well-known metal detector Quasar. The developer of the quasar is a genius. But a lot of dishonest sellers began to earn on its development. Therefore, buying a microcontroller is copyright protection. In addition, the diagrams of the device are freely available. The cost of the microcontroller is not large. 30% of this cost is the purchase of a microcontroller, which the developer carries out for his own money. Therefore, this is just a protection against theft of the developer's intellectual property.

[ArchibaldSTM]

I know these arguments about intellectual property --- they are made by all commercial vehicle developers. I just said that you forgot to say that the project is paid. Then many who want to repeat it will not be disappointed when they learn that you have to pay money. That's all

[JoyJo]

I don't share your point of view. Having the ability to assemble a metal detector that is not inferior to branded analogues for the cost of components only is not the same as building a car and making money on it. The example you have given is similar to the argument of a small offended child.