PI metal detector for really small nuggets

[green]

I have preamp & preamp_inverted that pass through SPDT, they are rotated each cycle and feeded to single ended integrator onto 2 channels (target + ground).
Very interesting chip Max7409 you're using for filtering I may check it out.

Rewired using amplifier normal out and amplifier invert out into a 1C integrator. Had thought the added delay time do to invert out might be a problem but the 1C integrator seems to work better. Have removed mAX7409. Worked good with Tx coil disconnected. Seemed to have some offset drift with Tx connected, maybe board layout.

[green]

6666,
The key to optimizing one variable such as coil size is to understand how other variables may interact. In the advancer search box enter Optimizing Target Responses to read more about this.

The first thing you need to know is the Time Constant range for the nuggets you are seeking. Here is an example. If you are seeking a nugget with a 2uS (microsecond) TC, the fully stimulated target will decay to fully decayed energy in 5 Target TC, time constants. A 2uS target will fully decay in 10uS. If you are sampling at a 10uS delay, guess what...no signal left to detect with any size coil. When comparing coil sizes, make sure your delay is fast enough to detect your target. If your nuggets have a target TC of 1uS then you need to have a delay faster than 5uS just to potentially detect that nugget. Generally, smaller coils are more effective in detecting smaller targets. Noisy or mineralized soil is another variable that may need a different coil style to minimize the effect of mineralized soil.

Generally, faster sampling produces a stronger RX signal. By integrating many RX signals you can improve the sensitivity to smaller targets but only as long as the target is within the coil area. Look up the theory behind the concept of a lock-in amplifier to see how integrating many signals improves the signal to noise ratio.

Good research requires knowing how related variables interact, and trying to isolate variables when trying to optimize any one of them by creating good mental models of:
1. Fully stimulating a particular target
2. Sampling soon enough to capture target energy before it fully decays
3. Coil style proven to be effective in particular environments, ground types and targets sought
4. Coil size for targets sought
5. Unique coil design features to overcome limitations such as putting the first stage preamp module in or near the coil housing.

This forum contains a lot of good information to help you more clearly see how these things interact or even see a hole that you can creatively fill.

Joseph J. Rogowski

Searched 1 thru 5. Some good threads. Not to repeat at lot of the information, I'm wondering if we can make the goal to detect a piece of #9 lead shot with a PI detector. First in air and then in ground, maybe 1 or 2 inches. More if possible.
1. Fully stimulating a .64us TC target shouldn't be a problem.
2. Including a decay chart for target and ground with 2 different ground decay slopes(Tx time and shape are a couple things that effect slope). Decreasing delay time .45us doubles target signal. How best to decrease delay time .45us without decreasing signal?
3, 4 and 5 need answers.

Some of the variables. Coil style, coil size, number of turns, mono or both Tx and Rx, Tx current and shape, Rx coil to amplifier connection and amplifier. Any others?

[bbsailor]

Searched 1 thru 5. Some good threads. Not to repeat at lot of the information, I'm wondering if we can make the goal to detect a piece of #9 lead shot with a PI detector. First in air and then in ground, maybe 1 or 2 inches. More if possible.
1. Fully stimulating a .64us TC target shouldn't be a problem.
2. Including a decay chart for target and ground with 2 different ground decay slopes(Tx time and shape are a couple things that effect slope). Decreasing delay time .45us doubles target signal. How best to decrease delay time .45us without decreasing signal?
3, 4 and 5 need answers.

Some of the variables. Coil style, coil size, number of turns, mono or both Tx and Rx, Tx current and shape, Rx coil to amplifier connection and amplifier. Any others?

Green,

Here are a few things to consider.
1. Assuming a 6 inch diameter mono PI coil swept at a distance of 2 feet per second, how many samples would you detect? Fill in your own real numbers.
2. What is the TX frequency that determines the number of potential samples?
3. What is the damping resistor value to determine the TX discharge slope which determines how close to being fully stimulated the target is?
4. What can be done to minimize the total capacitance seen by the coil to use the highest value damping resistor possible?
5. What mechanical design changes need to be made to optimize the length of the plastic coil stem to keep the aluminum shaft far enough away from the coil to not become a target that is detected and become a possible space to put some electronic active circuits close to the coil.

I believe that putting the electronics connection as close to the coil as possible minimizes cable capacitance and allows a higher value damping resistor to be used. Extending the coil wire up to the plastic stem connection point to the metal shaft eliminates a connection inside the coil that may respond like a target. This means breaking the circuit design apart and being creative with which parts you can remotely control to put close enough to the coil to get the benefits of minimal capacitance seen by the coil.

Once you start doing the assumed numbers and the optimization of all of them, you can begin to build prototypes to test each variable independently and see what tradeoff you will need to make to get the optimal final result. The WOT 15 inch DD coil has the preamp inside the coil for a reason. How many samples that are integrated improve the detection signal and detection distance?

Once you start building a good mental model of all the variables that affect the good detection of your .64us target, you can then start to prototype increments of your design to get the best final result.

This is just how I see your design challenges. Others might have a different approach.

Joseph J. Rogowski

[green]

Green,

Here are a few things to consider.
1. Assuming a 6 inch diameter mono PI coil swept at a distance of 2 feet per second, how many samples would you detect? Fill in your own real numbers.
2. What is the TX frequency that determines the number of potential samples?
3. What is the damping resistor value to determine the TX discharge slope which determines how close to being fully stimulated the target is?
4. What can be done to minimize the total capacitance seen by the coil to use the highest value damping resistor possible?
5. What mechanical design changes need to be made to optimize the length of the plastic coil stem to keep the aluminum shaft far enough away from the coil to not become a target that is detected and become a possible space to put some electronic active circuits close to the coil.

I believe that putting the electronics connection as close to the coil as possible minimizes cable capacitance and allows a higher value damping resistor to be used. Extending the coil wire up to the plastic stem connection point to the metal shaft eliminates a connection inside the coil that may respond like a target. This means breaking the circuit design apart and being creative with which parts you can remotely control to put close enough to the coil to get the benefits of minimal capacitance seen by the coil.

Once you start doing the assumed numbers and the optimization of all of them, you can begin to build prototypes to test each variable independently and see what tradeoff you will need to make to get the optimal final result. The WOT 15 inch DD coil has the preamp inside the coil for a reason. How many samples that are integrated improve the detection signal and detection distance?

Once you start building a good mental model of all the variables that affect the good detection of your .64us target, you can then start to prototype increments of your design to get the best final result.

This is just how I see your design challenges. Others might have a different approach.

Joseph J. Rogowski

Thanks for the reply. Lets look at number 1. Sweep speed 24inches/second, target over coil .25seconds. When we sampled data at work we tried to sample desired signals at least 10 times, 20 times was better. Sample 2000/second(target signal 500 samples) sample 8000/second(target signal 2000 samples). Sample rate either case is more than enough to define signal. Using spice to integrate white noise and a target signal, increasing sample rate 4 times doubled noise and increased target signal 4 times(doubled S/N). I think real circuit similar. If noise is ground signal I'm thinking no increase in S/N. Maybe I'm missing something?

[bbsailor]

Thanks for the reply. Lets look at number 1. Sweep speed 24inches/second, target over coil .25seconds. When we sampled data at work we tried to sample desired signals at least 10 times, 20 times was better. Sample 2000/second(target signal 500 samples) sample 8000/second(target signal 2000 samples). Sample rate either case is more than enough to define signal. Using spice to integrate white noise and a target signal, increasing sample rate 4 times doubled noise and increased target signal 4 times(doubled S/N). I think real circuit similar. If noise is ground signal I'm thinking no increase in S/N. Maybe I'm missing something?

Green,

When you did the samples at work, did you identify the source of the noise? It could be any of the following things.
1. Excess gain picking up electronic noise inside your building.
2. Noise from the ground matter that you put your target in.
3. Your indoor simulation possibly not representing a real outside test.

In doing research you need to build a series of small coil sizes to see what coil diameter is optimum for your #9 lead shot.

The reason why they make a ground balance type metal detector is to deal with soil problems so you can still detect a desired target. Modeling is a good way to plan what you might want to prototype but you still need to build prototypes, collect data and compare your collected data to the model data.

Since you want to detect a #9 lead shot at 2 inches, you might want to prototype some coils in the 2 to 3 inch diameter and collect some real data and compare it to larger coil prototype data.

The TX pulse length, pulse discharge slope necessary to fully stimulate the desired target and the delay to catch target signals before they fully decay will all interact with the coil diameter, sweep speed and the number of samples that can be integrated to improve the target response. This requires doing prototypes and looking for a noise free location to take measurements.

If you can reduce about 90pf of capacitance from the coil circuit due to cable capacitance, you have the potential to reduce the delay. However, this means putting the active circuit pieces close enough to the coil to allow this to happen. This is where doing some creative thinking is necessary to achieve your ultimate goal of detecting #9 lead shot at 2 inches depth. If you need to change your coil style to a DD, DOD, or figure eight type to deal with problem soil, you now need to evaluate the sweep speed and number of detected samples to integrate with this new coil style.

This forum has a lot of stuff related to what you want to do. Find something with potential, prototype it, measure it, compare what you measure to simulated data and find out why the difference, if any. This is all part of the learning process that goes on during doing the research for doing something new.

Joseph J. Rogowski

[green]

When you did the samples at work, did you identify the source of the noise? It could be any of the following things.
1. Excess gain picking up electronic noise inside your building.
2. Noise from the ground matter that you put your target in.
3. Your indoor simulation possibly not representing a real outside test. We recorded torque, speed, temperature, strain, etc in the lab and in the field. Metal detectors are brain exorcise. Sampling is a different problem. We determined the number of samples required to define the desired signals and filtered before sampling to prevent aliasing. Noise wasn't a problem most of the time.

In doing research you need to build a series of small coil sizes to see what coil diameter is optimum for your #9 lead shot. Since you want to detect a #9 lead shot at 2 inches, you might want to prototype some coils in the 2 to 3 inch diameter and collect some real data and compare it to larger coil prototype data. I started a thread awhile back on selecting coil diameter. Was told I wasn't asking the right questions. I'm thinking the coil diameter that gives the highest S/N is somewhere between the target depth and 2 times the target depth. Doesn't matter what the target is, #9 lead shot or a US nickel. Don't know what the desired distance for a piece of #9 lead should be for a PI, 2 or 3 inches was a guess. Thinking a 4 to 6 inch diameter coil would be a good place to start(2 times the desired depth). If the target was at 1 meter, a 1meter diameter might be a good place to start. Don't know if it makes sense.

[green]

Been having fun learning bipolar Tx. Wondering if I compare S/N, unipolar vs bipolar. Same integrator circuit, coil, Tx on time, Tx pulses/sec. Would bipolar be better?

[Ferric Toes]

Joe Rogowsky made this point 'If you can reduce about 90pf of capacitance from the coil circuit due to cable capacitance, you have the potential to reduce the delay. However, this means putting the active circuit pieces close enough to the coil to allow this to happen. This is where doing some creative thinking is necessary to achieve your ultimate goal of detecting #9 lead shot at 2 inches depth'. Here is some creative thinking-------

Take a large pot core half, say 150mm diameter, and wind the coil on a bobbin to fit. Cut an aluminium disc 150mm and 2-3mm thickness. Mount the electronics on the aluminium plate which is then secured to the back face of the core. The leads from the coil pass through a hole in the plate, straight to the electronics with no intervening lead.

With the short Tx pulse of no more that 10uS, any leakage flux around the core will not give rise to any significant eddy currents in the plate. Use a delay of say 2uS and a rep rate of >10^4 pps at a low peak current and you should get close, or even better. This works!

Eric.

[green]

Joe Rogowsky made this point 'If you can reduce about 90pf of capacitance from the coil circuit due to cable capacitance, you have the potential to reduce the delay. However, this means putting the active circuit pieces close enough to the coil to allow this to happen. This is where doing some creative thinking is necessary to achieve your ultimate goal of detecting #9 lead shot at 2 inches depth'. Here is some creative thinking-------

Take a large pot core half, say 150mm diameter, and wind the coil on a bobbin to fit. Cut an aluminium disc 150mm and 2-3mm thickness. Mount the electronics on the aluminium plate which is then secured to the back face of the core. The leads from the coil pass through a hole in the plate, straight to the electronics with no intervening lead.

With the short Tx pulse of no more that 10uS, any leakage flux around the core will not give rise to any significant eddy currents in the plate. Use a delay of say 2uS and a rep rate of >10^4 pps at a low peak current and you should get close, or even better. This works!

Eric.

Digi-Key has a 150mm pot core half for $307 each if buy 4($1229 total), not in stock. Wondering if the one you are using is the same or similar. Did you measure circuit resonance with no damping? I've been wondering what parameter controls minimum delay and if there is a formula to approximate minimum delay.

[Ferric Toes]

Digi-Key has a 150mm pot core half for $307 each if buy 4($1229 total), not in stock. Wondering if the one you are using is the same or similar. Did you measure circuit resonance with no damping? I've been wondering what parameter controls minimum delay and if there is a formula to approximate minimum delay.

This is not something I am currently doing. The last time I was working with a pot core was in 1998 when I obtained some 70mm diameter ones. Prior to that in 1989, I modified a PI frisker design to run at very short delays, using a 46mm pot core.

I was surprised to find these huge potcores such as 150mm and even up to 300mm by a German manufacturer called BLINZINGER. Hate to think what a 300mm would cost. No way am I going to fork out that sort of money.

Eric.

[6666]

This is not something I am currently doing. The last time I was working with a pot core was in 1998 when I obtained some 70mm diameter ones. Prior to that in 1989, I modified a PI frisker design to run at very short delays, using a 46mm pot core.

I was surprised to find these huge potcores such as 150mm and even up to 300mm by a German manufacturer called BLINZINGER. Hate to think what a 300mm would cost. No way am I going to fork out that sort of money.

Eric.
[ATTACH]54412[/ATTACH] [ATTACH]54413[/ATTACH] [ATTACH]54414[/ATTACH]

Eric Invalid Attachment's

[Ferric Toes]

This is not something I am currently doing. The last time I was working with a pot core was in 1998 when I obtained some 70mm diameter ones. Prior to that in 1989, I modified a PI frisker design to run at very short delays, using a 46mm pot core.

I was surprised to find these huge potcores such as 150mm and even up to 300mm by a German manufacturer called BLINZINGER. Hate to think what a 300mm would cost. No way am I going to fork out that sort of money.

Eric.
[ATTACH]54412[/ATTACH] [ATTACH]54413[/ATTACH] [ATTACH]54414[/ATTACH]

If I click on them they work. They should have appeared as .jpg's so will load them again.

[6666]

If I click on them they work. They should have appeared as .jpg's so will load them again.[ATTACH]54418[/ATTACH] [ATTACH]54419[/ATTACH] [ATTACH]54420[/ATTACH]

Eric they are working now thanks.

[green]

Charted signal vs distance for some targets. Think it's more repeatable. Noise level effects detection distance. The lead shot cover .5 to 1.5us TC targets(small nuggets). Nickel(10us TC) and Quarter(145us TC)targets to see what effect changing Tx and sample times parameters have on higher TC targets.

The quarter looked different. Changing Tx Rd effects the higher amplitude(lower distance)points. Under damped reads higher volts. Any thoughts why?

Thinking of trying Tx 40 and 80us, sample time 5 and 10us and first delay 3.5, 6 and 10us. Any thoughts?

[eclipse]

Zapped a couple fet drivers yesterday.

Ops same story here. The dual output mosfet driver that turns the N channel fets burned twice when coil is connected. This is tougher than LTspice suggests.