VLF MD with digital signal processing : Bee-Buzz 1

[moodz]

It looks pretty cool but I've never worked with Matlab but I can try it. The biggest problem is that Matlab will never replace real tests because a metal detector behaves completely differently on the ground than in the air and it's best to do all the tests on the ground with real equipment.

In order to do the real tests you have to devise the real circuit ... my experience is that most people apply lots of tweaks to circuits that they think will work.
If there is an understanding of how the different variables change then the correct modifications can be made to the circuit so that it will do what you want in real testing. Potentially you could simulate a full ground / coil / detector in matlab however that is not the intent that I meant. Its more of a case of quickly generating the code harness that can then be tweaked in the real testing and also a way of communicating the model so people can contribute. Its a top down modular approach to solving the problem rather than starting at the bottom ... building a circuit then making it work.
My experience with top range detectors is that the air test and ground test dont show a huge difference in target ranges ... this is in contrast to "home built" detectors which usually dont get the ground balance / cancellation right.

[Marchel]

It looks pretty cool but I've never worked with Matlab but I can try it. The biggest problem is that Matlab will never replace real tests because a metal detector behaves completely differently on the ground than in the air and it's best to do all the tests on the ground with real equipment.

[moodz]

If the prototype design were up to me, I would build it on this STM32H7 board, which also has a 16bit ADC and a small LCD that is enough for common information. On the top side of this board, I would connect a board with a few buttons and on the bottom side, a board with TX, RX and an audio amplifier.

https://vi.aliexpress.com/item/10050...Cquery_from%3A

Hey that board looks OK and could do the job .

But seriously Carl is right there needs to be less talk and more action. Like at least a diagram of the signal flow. Following this thread is like listening in to a a crowd at the pub .. everyones got an opinion but you cant find them when its their turn to buy a round of beers.
I know Carl does not like it but you really do need to choose a better CPU than the blue pill ( which name sounds like it needs to be inserted somewhere LOL ).
Also I think Carl also wont like it but you need to model some of this stuff as well ... as it will help other people understand what is going on ... rather than trying to interpret a huge block of obscure code in a undersized platform like the blue pill.

So my 2 cents ( and there are many ways to skin a cat - so this is not the only way ).
Get on a Matlab trial ( you have 30 days for free ) and you can get 20 hours a month on line access for free also.

1. Draw your model and run it in simulink- heres the IQ demodulator


and here is what is in the IQ block


2. You can edit the params in each block like frequency / phase etc as much as you want. Add or delete blocks as required.

3. Now right click on the IQ block and select generate code. Done the C code for the IQ demod is autmagically generated.

4. You can even get Matlab to squirt the code into your chosen platform. ( you WILL have to put more bits into it though )

If you have your model all this takes less than a minute.
I can post the model too ... but maybe you should learn by doing.

PS I have not used MATLAB for years ... its come a long way.

moodz

[Marchel]

If the prototype design were up to me, I would build it on this STM32H7 board, which also has a 16bit ADC and a small LCD that is enough for common information. On the top side of this board, I would connect a board with a few buttons and on the bottom side, a board with TX, RX and an audio amplifier.

https://vi.aliexpress.com/item/1005006632336183.html?spm=a2g0o.productlist.main.1.57bfXwVAXwVAKX&algo_pvid=dfc303d8-e7aa-4584-8c13-4e9785f37cba&algo_exp_id=dfc303d8-e7aa-4584-8c13-4e9785f37cba-0&pdp_npi=4%40dis%21EUR%2118.41%2118.41%21%21%21136.75%21136.75%21%402103834817363700044394355e1e28%2112000037878649641%21sea%21SK%21178235430%21X&curPageLogUid=mogAGVjvFHLk&utparam-url=scene%3Asearch%7Cquery_from%3A

[Marchel]

Yes, I work for FTP (Bounty Hunter, Teknetics, Fisher).



It is both. You can't get high performance results with low performance hardware, no amount of software can fix a bad ADC. However, it is often a useful step to start out with lower ambitions and simpler hardware and work your way up. For a 'why' see my next comment.



You can talk about the process for months on end, but the most valuable step is to build a circuit and get it working. For that, all you need is a TX driver, preamp, micro, and coil. Build it, make it work! You'll learn a whole lot more from that than continuing to talk about it. This thread is now 11 pages long and your very first post was a reasonable plan of what you should do.

I also think that until hardware is made and real tests and measurements are done, the theory is worthless.

[Carl-NC]

Are there any professionals employed by commercial metal detector manufacturers on this forum?

Yes, I work for FTP (Bounty Hunter, Teknetics, Fisher).

its not the adc/ics, its the process : signal processing, user interaction, adaptability and the over all functionality of the device.

It is both. You can't get high performance results with low performance hardware, no amount of software can fix a bad ADC. However, it is often a useful step to start out with lower ambitions and simpler hardware and work your way up. For a 'why' see my next comment.

my objetive is to first get the processes in shape, and then implement.

You can talk about the process for months on end, but the most valuable step is to build a circuit and get it working. For that, all you need is a TX driver, preamp, micro, and coil. Build it, make it work! You'll learn a whole lot more from that than continuing to talk about it. This thread is now 11 pages long and your very first post was a reasonable plan of what you should do.

[Marchel]

My detector design works with the XP Deus X35 coil so the TX signal is rectangular, and what TX signal do you want to work with ?.

[Atul Asthana]

I have been into metal detectors for a long time, but I am self-taught in programming and I am still learning. It would be great if a VLF digital metal detector with direct sampling was designed that would be fully functional and available to everyone. There are people on this forum who have built their own fully functional metal detectors, but they keep their secrets to themselves, it would be great if people who have already built a functional metal detector joined.

we will try and build a fully functional direct sampling vlf md, after we've done some research into most of the aspects, inclufing complexity of the software and of writing the code.

those, who have made their own, are not bound to share their designs (the sofware and hardware), but could share design philosophy / processes, giving us insights into the working principles or techniques and other aspects.

so lets work on with whatever knowledge we have or we can gather, and create some thing simple and useful.

[Atul Asthana]

Based on my discussions with several embedded system designers and leaders from commercial embedded companies regarding the use of the ESP32 and a 24-bit codec, I would appreciate it if experts could analyze and share their opinions on this topic:

Upgrading to an ESP32 and a 24-bit codec offers potential benefits like increased processing power and higher resolution, but these may not always translate to significant real-world improvements in a metal detector. The ESP32 allows for more complex algorithms (e.g., filtering, frequency domain analysis), while a 24-bit codec reduces quantization noise and increases dynamic range.

However, the limiting factor in most metal detector/similar designs is often not the microcontroller or ADC resolution, but rather analog noise (thermal, EMI) and ground effects. Optimizing the analog front-end (low-noise op-amps, shielding, PCB layout) and implementing effective signal processing techniques are far more crucial.

A well-designed STM32-based system with its internal 12-bit ADC, using the optimized techniques, can achieve exceptional performance, far surpassing traditional analog VLF metal detectors. This is because digital signal processing allows for precise control, noise reduction, and advanced target identification that are simply not possible with purely analog circuitry. Therefore, focusing on software optimization and careful analog design is the most effective approach for achieving high performance.

[Atul Asthana]

I think that even a metal detector with an internal ADC can work well, the Mole metal detector is proof of that.

http://www.radiodom.org/index/krot_stm/0-32

its not the adc/ics, its the process : signal processing, user interaction, adaptability and the over all functionality of the device.

if the process reasonably corelates to the requitements, any other sufficiently powerful processor, even with an external audio codec will do the job very well.

my objetive is to first get the processes in shape, and then implement.
later, we could look at esp32 (which seems much more capable, sans its adc, which can be substituted by an external codec) would make a better / more capable md, where even kalman filters etc can be implemented for noise management.

[Atul Asthana]

I have been into metal detectors for a long time, but I am self-taught in programming and I am still learning. It would be great if a VLF digital metal detector with direct sampling was designed that would be fully functional and available to everyone. There are people on this forum who have built their own fully functional metal detectors, but they keep their secrets to themselves, it would be great if people who have already built a functional metal detector joined.

Are there any professionals employed by commercial metal detector manufacturers on this forum?

It would be incredibly valuable to hear their insights. My knowledge of metal detectors is quite limited and primarily drawn from older sonar and navigation radar technologies, dating back 30-40 years.

Any expertise or information shared would be greatly appreciated.


and, does this design violate any patents in any country, though it should not be, being 30-40 years old.

or can I learn from any expired patents?

[Marchel]

great,


thank you. I will talk to my favorite programmers to experiment, though they seem to be more comfortable with stm32mx cube.

I use MCUViewer for debugging because it has better graphics and better features, but that's probably just my opinion.

[Marchel]

I think that even a metal detector with an internal ADC can work well, the Mole metal detector is proof of that.

http://www.radiodom.org/index/krot_stm/0-32

[Marchel]

I have been into metal detectors for a long time, but I am self-taught in programming and I am still learning. It would be great if a VLF digital metal detector with direct sampling was designed that would be fully functional and available to everyone. There are people on this forum who have built their own fully functional metal detectors, but they keep their secrets to themselves, it would be great if people who have already built a functional metal detector joined.

[Atul Asthana]

For this detector, I have set the lpf corner frequency of 1 Hz for ambient response and 100 Hz for target response.

What are the typical values used by commercial detectors?

Should I implement a selectable range, or allow a continuous selection from, for example, 1 Hz to 0.1 Hz, enabling users to choose a value that best suits their needs?