Concept Design

[Carl-NC]

Although I don't consider the product definition complete, I will go ahead and open up discussion on the Concept Design because I think people are itching to tackle this. It is always dangerous to start designing before you fully understand the end requirements but in this case I think we know enough to come up with a reasonably safe bet.

This is the time to discuss the basic technical approach. It is not the place to develop the architecture or pick out the micro, but the Concept Design will begin to point to towards what these will look like.

The next 23 posts are from the original "Challenge" thread and are relevant to this discussion.

[detectormods]

A design challenge.

Currently there is increased interest in gold all over the world.
There are gold rushes happening at several places. The gold rushes I am most familiar with are in Guyana and Venezuela.
This region presents a special challenge for metal detectors.

• Heavy rains and mud during the wet season.
• Very hot and dusty during the dry season.
• Several hundred thousand gold diggers gold who can barely read and write and work in the most primitive and dangerous environment just to survive.

These gold diggers have to produce about 2 grams of gold per day just to survive. The gold is alluvial gold, scattered over thousands of square miles. Gold nuggets of the size of many pounds have been found, but the gold comes in all sizes from gold flower size to large nuggets unevenly distributed.
The total gold reserves of the entire region are immense, but have remained largely untapped until recently.
After drilling some 1700 test holes in just one claim, by a Canadian gold mining company, the claim has been declared as the second largest gold reserve in the world.

A metal detector suited for the special environment in this region would sell in many thousands per year if it was available.

Hybrid Detector sounds like the go.

[Tinkerer]

So let's get started with the design.

I propose
Others might have different ideas. Please criticize and propose alternatives.

Start with:
One 8" DD coil
One 12" DD coil
Square wave constant current bipolar TX
Pulse repetition rate from 40us to 200us, that is 25k to 5K CHIRP

RX
Differential input preamp
DSP, Digital Signal Processing

[Carl-NC]

Pulse repetition rate from 40us to 200us, that is 25k to 5K CHIRP

RX
Differential input preamp
DSP, Digital Signal Processing

Are you suggesting a linear chirp or a few discrete pulse steps? If linear, would you dump everything into a single demod channel? I would assume 3 or 4 discrete steps, each with its own demod channel(s). And would probably weight the short pulse higher, something like this:

200 100 40 40 40 40 100 200 100 40 40 40 40...

Either run analog demods or preferably use a 24-bit SAR ADC and direct sample.

[Tinkerer]

The first try (last year) gave the results seen in the picture above. (Real circuit)
The Flyback shows that the peak current is not the same with different cycle length. The cause was the slope in the TX current.
Now we have achieved the correction of the slope, so that every TX pulse is of equal peak current no matter the length of the cycle.
Trying to keep the PRR (Pulse Repetition Rate) as high as possible, I think that maybe 2 switchable modes accentuating the desired average response for larger or smaller nuggets might be a good idea.
A coil size of 8", would be used more for smaller nuggets. More shorter cycles.
The 12" coil for larger nuggets. More longer cycles.
With the same peak current for each TX pulse, the demodulation is essentially the same.
The difference is in the decay time for the targets with a TC that is too long to decay completely before the next TX pulse.

[Carl-NC]

A coil size of 8", would be used more for smaller nuggets. More shorter cycles.
The 12" coil for larger nuggets. More longer cycles.

That's an excellent way to simplify the detector, have the coil size determine the mode. It's easy to convey that a small coil will find small shallow stuff while a large coil will find large deep stuff. And you don't have to worry about accidentally using the wrong mode.

[Tinkerer]

Here is a post from 2016. Has any of this been tried and implemented since then?

01-04-2016, 10:18 AM
Originally posted by Monolith View Post
When we look at the dates on the posts above, we see that the square wave TX was postulated by Dave Johnson in December 2001, Further discussed in 2002, implemented experimentally in 2010-12 by Deemon and finally sold as commercial metal detector by MINELAB in 2015.
Others have played with the idea in the meantime, but we might say that the basic idea took 15 years to be converted into a commercial product.
By the way , guys , the game isn't over yet What Dave proposed and Minelab implemented ( current square wave TX ) is only one part of the thing ... but the entire design of mine does contain something more . I mean another 2 ideas of my approach - current signal pickup instead of voltage and usage of balanced ( or auto-balanced ) mono coil . And another important thing is that I use quite a different method of current reverse - quasi-resonant power chain with L-C energy exchange , the first time appeared in that unipolar recuperative PI machine ( published here ) , and then transferred to my square-wave approach ( from the other topic ).


What I claim - we need to use all these ideas simultaneously to obtain the best possible result - as I had shown in my topic . As I can see now - nobody implemented all this stuff in the commercial device yet , so here is still a great field for experiments , research and development ..... as for me , I have completely solved all the "theoretic" and "electronic" problems , but stopped now with some "mechanical" troubles . I mean the proper coil winding , isolation , shielding , mounting and so on . Then I need to make a good box for all this electronic stuff .... so I really don't know how much time I need to finish it in the working device - but I'll finish it anyhow, at any rate

[Tinkerer]

I looked a bit around the PIC32MK0512MCJ064.
There is a development board called
"DT100113 Development Board, Curiosity Pro, PIC32MK0512MCJ064, 32bit PIC32 MCU"
and costs around 80-90e. But it is nowhere to be found in the local stores, here in Serbia.

It can be found here though:
https://in.element14.com/microchip/d...bit/dp/3579863

Something I haven't researched yet; even if I managed to get somehow (I don't know how) that development board;
what development software environment is most rewarding for it?
How complicated is it to set up that development software environment on a local system with Win10.
What is needed from a software tool to be able to sit down and work on code in one moment?
And at the same time, don't bother yourself with other minor side things.
...
In general, I'm not against switching to another development system and a much stronger processor.
But what always deters me from that decision is the complexity of setting up the system, collecting the necessary software and additional software tools,
setting it all up correctly and aligning it to the needs of further work.
I'm always sick of that, I almost never managed to do it right. Such systems are too cumbersome, complicated, often online dependent.
That's why I love the Arduino IDE. Because everything is so easy and simple with it.
You don't waste time on overly complicated and "profane" nonsense, but you are immediately focused on bare programming.

Maybe we should first define what we really need, to achieve our goal.
A mostly digital, microprocessor controlled PI?
Mostly or fully DSP (Digital signal Processing) ?
Direct sampling?
Analog integration?

With the goal defined, can we find a consensus for the processor?
There are so many different aspects of the metal detector design. As a group, in teamwork, we should divide the tasks, such that each can work on a specific aspect that he is most familiar with.
- Coil development
- Housing, maybe 3D design and prototype 3D printing
- Power supply, batteries, parts research, RX, TX, so much work to do.
- Drawing the schematic, PCB design and manufacturing (I like your idea, ivconic, of assembly in China)
- Coding
- Field testing

As a team we have a chance. It is really too much work for one alone.

[Willy Bayot]

Following the ideas given in the posts of Carl, Tinkerer and IDEFIX(#52), we can define a list of potential project subjects.
First, I suppose that:

• We want to exclusively concentrate on PI-related Technologies.
• We do not want to re-invent a Nth+1 version of traditional PI systems.
• We want to make a maximum profit from the current digital processing technologies.

What is the current state of art in DIGITAL PI systems?
What functions are still subject to improvements?

Let's try to collectively answer to those two questions!!

[ivconic]

1) What is the current state of art in DIGITAL PI systems?
2) What functions are still subject to improvements?

1) Easy one. What ML does.
2) Tough one. I would say a price. Costs. Instead of listing functions and improvements... I would mention the most important factor that needs to be improved, which is the price of such a device for the end user (compared to ML).


I know these are the answers you didn't expect nor ask, but that's how it is.
ML has already done what can be done. But the problem is the price of that technology.
It is inaccessible to most detectorists.
Especially the ones that Tony was talking about and for which Tony proposed the whole story.
Besides, ML detectors don't fit ideally into the format Tony suggested. They are not simple turn on and go type.
It would be ideal to make a detector with ML performance, but with as simple a UI as possible and with
as few options as possible that will not confuse the user and waste his time on additional settings.
The point is to simply and quickly locate the gold nugget.
The GEB is mandatory. But how do you make it simple as possible for the user?
Auto tracking GB is the one of the solutions. Not ideal.
The previous detectors with it showed all the flaws. Too slow and can't catch up with faster movements of the coil or more often soil changes etc.
One button and "pumping" method?
Whenever the detector becomes louder than the treshold; do the "pumping" procedure.
A simple "click-clack" switch that selects between "Auto track GB" and "Manual GB".
Let's just define what all of the front panel commands need to exist to satisfy the conditions Tony listed.

[Tinkerer]

Following the ideas given in the posts of Carl, Tinkerer and IDEFIX(#52), we can define a list of potential project subjects.
First, I suppose that:

• We want to exclusively concentrate on PI-related Technologies.
• We do not want to re-invent a Nth+1 version of traditional PI systems.
• We want to make a maximum profit from the current digital processing technologies.

What is the current state of art in DIGITAL PI systems?
What functions are still subject to improvements?

Let's try to collectively answer to those two questions!!

What functions are still subject to improvements?[/B]
There is a lot of talk about "multifrequency detectors". Why? Because a single frequency gives best performance on a target with a single specific TC.
If we want to search for small and large targets, short and long TC's, we need to be able to switch to different frequencies. Or, we need to run several frequencies simultaneous. with PI, frequency means mostly Pulse Repetition Rate PRR.
With a traditional PI system this is difficult.

The Square Current Wave TX is ideally suited for variable PRR. A BIG STEP FORWARD WOULD BE A CHIRP PRR, where every target is "illuminated" or "excited" with pulse cycles of various times, to induce maximum eddy currents for targets of every size.

Does a CHIRP already exist on the market?
In my archives I found a scope picture of a CHIRP experiment I did in 2017. For some reason or other I did not continue the experiment.
I definitely think that it would be worth spending some time experimenting with a CHIRP PI.

[Willy Bayot]

1) Easy one. What ML does.
Does anybody on this forum know how the expensive ML system works and could explain it?
I assume the response is 'NO', otherwise, that person should tell us because we are all interested.

2) Tough one. I would say a price. Costs. Instead of listing functions and improvements... I would mention the most important factor that needs to be improved, which is the price of such a device for the end user (compared to ML).
This is indeed a good challenge.
How is it possible to make a system at least as good as the ML 'monster' with a much lower price tag (and a much lower weight)?
What exactly distinguishes the 'big' ML systems to the widely published Digital PI systems in terms of useful functions?

ML has already done what can be done
Do we know what they have really done and how?

Just a note:
When I say 'Digital PI systems', I mean 'Digital Signal Processing' and not simply pulse timing generation by a micro-processor.

[Orbit]

1) Easy one. What ML does.
Does anybody on this forum know how the expensive ML system works and could explain it?
I assume the response is 'NO', otherwise, that person should tell us because we are all interested.

2) Tough one. I would say a price. Costs. Instead of listing functions and improvements... I would mention the most important factor that needs to be improved, which is the price of such a device for the end user (compared to ML).
This is indeed a good challenge.
How is it possible to make a system at least as good as the ML 'monster' with a much lower price tag (and a much lower weight)?
What exactly distinguishes the 'big' ML systems to the widely published Digital PI systems in terms of useful functions?

ML has already done what can be done
Do we know what they have really done and how?

Just a note:
When I say 'Digital PI systems', I mean 'Digital Signal Processing' and not simply pulse timing generation by a micro-processor.

Have you ever used that monster ? I am, for me it's nothing special because simply because that particular monster in the game was gpx 5000 the man with the VLF machine finds a bag at a depth of one meter with a coil of 50 cm without any problems and the monster struggles whether he has something or not ! Don't be mistaken there is nothing special!I didn't try only gpz, I tried the others, they are not bad devices, but not as many people who have not owned them think!

[Willy Bayot]

Have you ever used that monster ? I am, for me it's nothing special because simply because that particular monster in the game was gpx 5000 the man with the VLF machine finds a bag at a depth of one meter with a coil of 50 cm without any problems and the monster struggles whether he has something or not ! Don't be mistaken there is nothing special!I didn't try only gpz, I tried the others, they are not bad devices, but not as many people who have not owned them think!

True. We should not be blinded by their products. They also have their own defects.
Let's go on!!!

[Orbit]

True. We should not be blinded by their products. They also have their own defects.
Let's go on!!!