TX Cascade for Multi-Frequency

[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

[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.

[JoyJo]

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

[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]

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]

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

[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.

[JoyJo]

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

[Carl-NC]

Are you wanting to use analog demodulators or direct sampling?

[JoyJo]

Do you need a multi-channel ADC for this? We now have an input stage on one operational amplifier, then the signal goes to the ADC, and then to the microcontroller.

[Carl-NC]

The easiest way to process is to treat each frequency like a separate metal detector. Separately ground balance each one, then scale their VDIs to the same scale. For example, the White's V3 scales each frequency (2.5k-7.5k-22.5k) to 6.592kHz, which they had been using for many years.

[JoyJo]

And how can a multi-frequency signal be processed? And according to the vdi, it is not clear: each frequency will have its own vdi, but there is no point in displaying three vdi values. And when averaging, the vdi will not be accurate.

[Carl-NC]

No, arbitrary choices. But they are evenly spaced at 1:4:16.

[JoyJo]

And what is the reason for the choice of such frequencies 4, 16, 64? This is 2 to the power of n, where n is 2, 4, 6.

[JoyJo]

But I want to use a lower voltage of 3.5-9V. To be able to be flattered by two 18650 batteries.