Electromagnetic interferences

[samu]

Hi, im new in the forum and firstly i want to say thank you for everybody for your work.

I want to build a metal detector to be placed under a conveyor belt in a mine. I don´t need much accuracy i only need to detect big metals. The longer distance is about 30 cm.
The main problem is that there is a lot of machines working very near. This is a lot of electromagnetic interferences.

the question is: ¿PI or BFO or VLF or other?.

Thank you very much.

[Max]

Hi, im new in the forum and firstly i want to say thank you for everybody for your work.

I want to build a metal detector to be placed under a conveyor belt in a mine. I don´t need much accuracy i only need to detect big metals. The longer distance is about 30 cm.
The main problem is that there is a lot of machines working very near. This is a lot of electromagnetic interferences.

the question is: ¿PI or BFO or VLF or other?.

Thank you very much.

Hi,
pulse induction could be the right solution there. Avoid vlf and bfo cause too weak performances. With pi you haven't discrimination but have good detection also on small objects not only big ones.
Also coil is much less critical than in vlf/tr/bfo to make. Think that a 300Hz separation between each md could avoid problems due to interferences.

If you plan to detect iron objects or iron-like (cobalt etc) you can think also on small magnetometers with prebuilt sensors e.g. fluxgate to do the job.
Take a look at Carl's fluxgate mag project in mag section.

Best regards,
Max

[samu]

Thank you Max,

I also need to detect non-ferrus metals so i have to use a metal detector.

Now, im looking for a design. I want to use a microcontroller based design because i think is more flexible. Any recommendations will be welcomed.

Thank you.

[Max]

Thank you Max,

I also need to detect non-ferrus metals so i have to use a metal detector.

Now, im looking for a design. I want to use a microcontroller based design because i think is more flexible. Any recommendations will be welcomed.

Thank you.

Hi samu,
a good starting point is the Mark Stuart's PI that uses a pic micro (16x84).
You can find schematic in the project section of geotech site. I've built it some years ago and works well. You can find also programs .hex ready to use/modify about it. Take a look at project section of the forum regarding Stuart's circuit and its mods.

With this PI you have much freedom on coil shaping and size...and good detection performances on all metals.

Best regards,
Max

[Geo]

Ηι. Τhe Mark Stuart's PI is not a good choise because there is not second sample to reduce the noise.
"The main problem is that there is a lot of machines working very near . This is a lot of electromagnetic interferences ."
Bye

[Max]

Ηι. Τhe Mark Stuart's PI is not a good choise because there is not second sample to reduce the noise.
"The main problem is that there is a lot of machines working very near . This is a lot of electromagnetic interferences ."
Bye

Hi Geo,
yes you are right on second sample but unit was sensitive and a good starting point I think. He need to program different frequencies for different devices to avoid much problems. I've done with 2 stuart's some time ago and this solution worked ok on beach hunting with detectors not much dinstant (1meter or more) and small coils of 8''.
Yes...it's not the best and better if he could realize double sampling and then using e.g. some deltapulse PI removing the older osc/timing (556/4538s) section and then wiring a pic or something to get timings, but that requires much expertize on pulse detectors to be done...and a bit of programming too.

FOR SAMU: if you have electronic and programming skills you can use deltapulse instead of stuart's with less interferences problems, but you need to modify schematic and put some microcontroller and programming it for right timing.


Best regards,
Max

[samu]

I like to think that im an adventurer so i´ll study the delta pulse. I was looking for information about in the foro but i only found some eschematics and a lot of people with doubts. To do any modification, it is neccesary to understand how it works, so i need more than the eschematic. If anybody knows where is the information please tell me.
I´ve some programming and electronic skills but i think this is a good way to learn more.

Thank you one more time .

[samu]

Many questions

I read about delta pulse but i don´t know what is the diference between it and normal PI. I saw a signal diagram in wich are 4 signals. I suppose that after the pulse, the "signal" pulse is used to discriminate the normal decay and helps to avoid ground detection. Then, the threshold (sens) pulse is used to setup the time to check the "echo" amplitude. ¿the shorter is the sens time, the more sensible is the detector?. I think that those signals doesn´t provide a mechanism to avoid interferences so it may be on the width of the noise pulse signal.

I also read about problems with AC power source. I want to use the metal detector without batteries.

About the coil, may works better with a figure-8 coil?

Thank you in advance.

[Max]

I read about delta pulse but i don´t know what is the diference between it and normal PI. I saw a signal diagram in wich are 4 signals. I suppose that after the pulse, the "signal" pulse is used to discriminate the normal decay and helps to avoid ground detection. Then, the threshold (sens) pulse is used to setup the time to check the "echo" amplitude. ¿the shorter is the sens time, the more sensible is the detector?. I think that those signals doesn´t provide a mechanism to avoid interferences so it may be on the width of the noise pulse signal.

I also read about problems with AC power source. I want to use the metal detector without batteries.

About the coil, may works better with a figure-8 coil?

Thank you in advance.

Hi samu,
you could also think at hammerhead project that is well documented here (thanks Carl) in a pdf file containing everything you need to know how it works.
Now on deltapulse and PIs witth double pulses:
after tx switchoff you have always a delay I call "the first delay" that cold be from some few uS (5-10) to some hundred of uS (usually never more than 200us).
First: original deltapulse was designed for deeper search on big items using large 1mtx1mt coil...so maybe you need tune it to your needs and you do so modifing coil and timings.
For delta diagrams (original):
tr2 is signal for gate of mosfet that is tx period 9.2mS nothing, then a pulse of 80uS.
tr4 "signal" show you first delay (90-170uS) and sample window (45uS).
tr5 show you second sample (equals to first = 45uS) and inter-delay means delay between the two samples (160-900uS).

WHY THIS TWO SAMPLES: simplifing... a differential integrator is used to cancel out low noise from signal path and also to minimize drift related to components warming...that is very slow...but could be a problem in mono-sample designs. Double sampling/diff integrator cancel noise cause noise is everytime present with similar level and then subtraction cancel it due to double sampling-diff integration and useful signal is still present there.
Another important thing is that after say 100us or more useful signal vanished and remains only noise ! So if diff-integrator subtract these two samples (first and second) noise is virtually cancelled out.

You can find more deep explaination on hammerhead article and on many other sources also in this forum (there is a good article of Reg Sniff but I can't remember where I saw last time).

Good timings for best detecting small objects could be:
TX 100-150uS (coil of 300-350uH)
DELAY 15-20uS (good also for gold...and other precious metals also 35+ for iron detection)
RX1width like it is 45uS in original (or also 15-20uS better noise insensitivity)
RX2width idem above RX1
INTER-DELAY (samples separation) of 150-200uS

With these data you could detect very well gold and aluminium and almost every good conductor too (silver, copper etc).

For AC...you can use any good filtered linear dc supply without any problem. Just need good filtering on output you can also do with say 10000uF cap or more to smoth out ripple noise 50 or 60Hz.

For coil you could use figure-8 reducing much ac external noise coupled with coil...but this way you lose some depth instead using a normal monocoil (round or rectangular or square etc). Figure-8 are needed only is external noise is a big problem e.g. you have an hi-voltage ac line just above your detector(s).

Best regards,
Max

[samu]

Thank you very much

Now i need some days to study all the information you gave me . Maybe it´s easier to improve the stuarts microcontroller program (to add double sampling) than to add a microcontroller to other design.
I´ll tell you all the advances.

In other way, im really gratfull with your help. If i can help you in anyway don´t doubt about ask me for.

Thank you very much.

[Max]

Now i need some days to study all the information you gave me . Maybe it´s easier to improve the stuarts microcontroller program (to add double sampling) than to add a microcontroller to other design.
I´ll tell you all the advances.

In other way, im really gratfull with your help. If i can help you in anyway don´t doubt about ask me for.

Thank you very much.

Hi samu,
you are welcome. For stuart's you can add a second sample by program but then you need also a diff. integrator that there isn't ; so you need to modify circuit to get it work as a double sampling PI.
If you study the hammerhead you could find anything about the diff. integrator you need or you could think also removing some osc/timing components (osc. and some ttl/HC monostables) and then wire directly a pic to get right timing by programming it.
Very easy to do cause you only need to put in the microcontroller but not to do any change in diff. integrator section or other parts.
Hammerhead is a very good PI and a really flexible design to experiment with,
and of course is also well documented.


Best regards,
Max

[samu]

Diferential integrator

I think that a diferential integrator do the next:

int(S1-S2), that is equal to int(s1)-int(s2).

I can implement a double digital integrator by software in the microcontroller.

Also i can implement a digital correlation filter if de micro is powerfull enough.

An important question: in one side of the metal detector there will be some metal parts of the conveyor belt, ¿is necessary to shield the coil?.

Thank you.

[Max]

I think that a diferential integrator do the next:

int(S1-S2), that is equal to int(s1)-int(s2).

I can implement a double digital integrator by software in the microcontroller.

Also i can implement a digital correlation filter if de micro is powerfull enough.

An important question: in one side of the metal detector there will be some metal parts of the conveyor belt, ¿is necessary to shield the coil?.

Thank you.

Hi,
"
An important question: in one side of the metal detector there will be some metal parts of the conveyor belt, ¿is necessary to shield the coil?.
"
depends on how much metal is...expecially the area of it that the coil could "see". No shielding can avoid saturation of input preamplifier by too metal near the search coil... at the end you need to find a metal clear area or reduce coil dimensions to avoid too much eddy-currents in parts of the apparatus be detected as targets.

On integration yes one could also do it in software if micro (and programmer) is smart enough...but this requires that you feeds data inside the MCU and only way to do that is converting some analog into digital by e.g. an ADC circuit (with good resolution). But this way complicates schematic...I think is easyer to add some op. amps...it's your choice.

Also correlation could be a really good idea...but requires much programming.
Think you could use some C++ compiler or something similar for some smarter MCU...not pic16x that have too few memory and speed (and a very essential instruction set).

Best regards,
Max