PCB Coil?

[Andy]

re: Coil Circuit

Bugwhiskers,

You can forget about the coil circuit request, I can figure it out from the photo and the information you posted for BBSailor.

Andy

[bugwhiskers]

Bugwhiskers,

Q:Is the track layer underneath in-between the upper layer tracks? You mentioned this on the previous coils as being the case.
A:Yes the track on the other side is between.

I wanted go through the theory, get a rough idea of the design and expectations, collect the test gear and then methodically proceed to test.

While you're at it can you see if running the tracks very close to the middle of the board is detrimental. There is a school of thought that there needs to be 50% of the diameter free. From my observations the coils with tracks running thru to near the middle work best, maybe that is just because of the extra inductance produced by the extra windings ??


That decay-field [off the coil] is interesting unto itself, it makes me wonder if taking a sample off the coil directly is the best way to get the information.

Some time back I noticed that if a coil was pulsed with a weakish signal (perhaps 1 TC, there was a "sweet spot") and ferrous material was targetted by the coil the back EMF dropped. If copper or alum was targetted then the back EMF rose substantially (20 volts + with close proximity). Maybe you could explore this further.


Q:If your detector is not "private", perhaps you could post just the analog part that is handling the coil.
A:There is nothing unusual with the front end pre-amp.
470 ohm damper
470 ohms to back to back IN914's
220 ohms to the inverting input of the pre-amp.
The coil lead wire you see is 300 ohm TV ribbon cable, it is multistranded copper and I think the insulation is teflon. There is a constant 7mm separation between the conductors.

Q:I am assuming that you are using the PIC just for generating the pulse and everything else is handled by analog circuitry.
A:The ATMEL micro switches the coil MOSFET, switches decay waveform to the integrators via 4053, switches off the SMPS during sampling, monitors the battery voltage, averages the signal before being sent on to the PWM for sound.


Q:It would be nice to know the length of it though, from board-to-board.
A:The test results posted were with 1 foot of seperate copper speaker cable to an alligator clip then 2 feet of 300 ohm ribbon to the coil.

Andy

Yesterday I changed the LM318 for an LF356. The LM318 is damn fast and for my desired targets 1 gram + nuggets with discrimination it was too fast.

To discriminate I change the width of the second sample so the voltage is equal to the first sample when the target is ferrous. Non ferrous metals then produce a "difference" signal. With the LM318 I could detect very fast metals (ie Al foil) too well and in fact the first sample was so strong that I couldnt adjust the second sample to equal it. The LF356 also cuts a lot of noise due to its lower bandwidth.


The next coil I make will be very similar to the one depicted. I am aiming to squeeze another 90 uH out of it with a few more turns and tighter track spacing. Please keep in mind when doing your math my Inductance meter's lowest range is 20 mH so the 310uH could be as bad as +/- 5uH ... all I get to see is 0.31.

I have spent a lot of years programming in AutoCAD. I wrote a routine that creates spiral circular coils and reports the length of track and resistance.
If you can do the math and send me the formulae I could re-write it to include capsule coils and also report on inductance.



regards
bugwhiskers

[Andy]

re: Feed

Bugwhiskers,

If you can, get rid of the speaker cable and directly connect the twin-lead, this will remove any potential for impedance mismatch [current loss], and other unforeseen issues. Best to keep everything consistent.

The recommendation for keeping the center clear was for both inductance and capacitance. Don't forget the field measured by your equipment is going to be different than the actual field [and hence the inductance], under operation. I think BBSailor had the best approach to doing the calculation by inducing a signal [field] with another coil and measuring it. It would be useful to measure with your meter and then do a measurement using the technique BBSailor recommended.

One of the designs I came up with was to use multiple boards with the tracks staggered, as you are doing on a single board. My design has a staged coil, that is, some turns on the first board, then some more turns inside the loop of the first board on a second board, and the same for a third board. By doing it this way I could "plug-and-play" different boards and re-use previous work. Recently I found a Amateur Radio site [Harry Lythall - SM0VPO] that used a similar technique:

http://web.telia.com/~u85920178/info/pcb-coil.htm
http://web.telia.com/~u85930032/pcb-coils.zip

His board is not staggered, in mine, the first board would have for example, 10 turns, and then the next would have 10 turns, but inside the loop of the first, and so forth. The down-side of a layers/staged approach is that it would add to the capacitance, however that could be somewhat compensated by offset placement of the tracks. My goal was to re-use the boards, and to see what I could do about shaping a field pattern.

You could go to a less eliptical pattern and use a more rectangular one [keep the radius on the corners!], this would give you sligthly more inductance in the same space.

In many experiments I noticed the same thing as you did, by supplying a weak signal I was able to get a different response. In some ways it sort of defies theory, seems odd to get a certain range you would feed a small signal identical to what you were looking for. It's probably very explainable once we understand what is happening [hindsight]. I'll definitely watch this effect, again it may be something we can exploit.

I would go back to the LM318, sure detecting foil is a hassle but it also means you are getting a great signal return. You should try to going both ways on the pulse width, making it smaller [faster] and making it longer [slower] and see what the results are with aluminum, Do the same experiments with gold.

To plot a B-H curve you want to have the coil current [really the voltage representing the current] going to the horizontal input of a scope, and the output of the integrator going to the vertical. That coil connection can be tricky, may want to put some precision resistors in there to drop the voltage to a safe level for the scope. Sort of like a high impedance probe. You will see a distorted square [maybe], sometimes it shows up as "eye" pattern, on the scope, and you will see that it distorts in different places depending what's under the coil. Again, watch the voltage off that coil, it can fry your scope input.

Andy

[bugwhiskers]

Bugwhiskers,


The recommendation for keeping the center clear was for both inductance and capacitance.

The clear space is large compared to track spacing so it shouldnt increase capacitance and the extra turns increase the inductance. In a bundled wire coil surely some turns interfere with others.



You could go to a less eliptical pattern and use a more rectangular one [keep the radius on the corners!], this would give you sligthly more inductance in the same space.

It would result in longer traces and higher resistance but the same number of turns. Webers is about number of turns and current and higher webers is surely the goal.

In many experiments I noticed the same thing as you did, by supplying a weak signal I was able to get a different response. In some ways it sort of defies theory, seems odd to get a certain range you would feed a small signal identical to what you were looking for. It's probably very explainable once we understand what is happening [hindsight]. I'll definitely watch this effect, again it may be something we can exploit.

Just a hairbrain idea but as you shine light on objects you perceive colour.
With dim light the colours are dark but increasing the light causes the colours become brighter.... up to a point. Dont know if that analogy could apply here.

I would go back to the LM318, sure detecting foil is a hassle but it also means you are getting a great signal return.

Have gone back to the LM318, the problem I was having was offset ie no decay curve, just a straight sided pedestal. Re-adjusted everything and now getting really good discrimination at distance.

To plot a B-H curve you want to have the coil current [really the voltage representing the current] going to the horizontal input of a scope, and the output of the integrator going to the vertical. That coil connection can be tricky, may want to put some precision resistors in there to drop the voltage to a safe level for the scope. Sort of like a high impedance probe. You will see a distorted square [maybe], sometimes it shows up as "eye" pattern, on the scope, and you will see that it distorts in different places depending what's under the coil. Again, watch the voltage off that coil, it can fry your scope input.

Will try the X Y thing and see what sort of figures I get. Have a 10X probe or 3 and only my old backup CRO has an available Y input.

I still reckon a Spectrum Analyzer view would be interesting. If you can discern differences with different metals/distance/size imagine what boosting the wanted frequencies could do especially on the fastest changing part of the curve which also by chance has the most information about "the good stuff".

Andy

regards

bugwhiskers

[Tinkerer]

spiral wound coils

I have experimented with spiral wound coils, but round ones and wire wound, not PCB. Starting the windings very close to the center, has a strong influence on the pinpointing capability of the coil. The sensitivity of the coil is much enhanced at its center.
The first few turns, being of small diameter, give very little inductance. The outer turns produce significant inductance for each turn. There seems to be a correlation between the amount/length of the wire used for each turn. I suspect that the same applies for capacitance.
Separating the coil windings, reduces the inductance and capacitance of the coil. (I use polypropylene spacers between the wires) With wire wound spiral coils, it is difficult to obtain repeatable high precision in the inductance and capacitance. This problem is easily resolved on PCB coils.
The reason for looking for high repeatable precision is to be able to develop a formula for calculating exactly where the “sweet spot” of the coil is. This is where the capacitance and inductance come to the ideal proportions.
Now, what is that ideal proportion?
I found that different methods of coil windings produce different results in the Q of the coil. A coil can end up with a very high Q that makes it receptive for a very narrow band of frequencies. But a different wound coil, just as fast as the first one, receives a wide band of frequencies. This does not seem to make sense, but looking at the FFT it is obvious. It is also visible on the scope if you know where to look, but the different wavelengths are sometimes hard to see in a short window.
I suspect that information about the targets might be gained at different frequency responses.
I am using high gain on the coil amp for observing the decay curve. When I use targets of different metals and roughly similar size, it is easy to see the difference in the decay curve. However, when the targets are of very different size, this gets difficult.

Tinkerer

[bbsailor]

The consequence of coil characteristics

Tinkerer,

You made an interesting observation below that could be expanded.

"I suspect that information about the targets might be gained at different frequency responses.
I am using high gain on the coil amp for observing the decay curve. When I use targets of different metals and roughly similar size, it is easy to see the difference in the decay curve. However, when the targets are of very different size, this gets difficult."

Once you put a damping resistor (Rd) across the PI coil, it stops responding like a tuned coil except for the value of Rd as it relates to total coil, cable MOSFET and other circuit loading resistances and capacitances. Once you stop coil ringing any further reduction of Rd tends to increase the current fall time.

What remains interesting is what Eric Foster has mentioned. I'll paraphrase: The fall time of the current should be at least five times faster than the total decay of the target's decay to get a good target response. What this suggests is that if you can detect a particular target and then want to gain additional information about that target, you can do the following. Switch to an identification mode where the fall time of the current increases with each sweep of the coil over the target then the response of each sweep is displayed showing a unique response of each type of target. In the PI patent section (on this web site) there is a patent where Rd is replaced by a MOSFET that has a linear region so as to be a good variable resistor that can be controlled by a computer and a D to A converter that will change the TX decay rate in an attempt to extract additional target information.

Those targets having less conductivity would only have earlier responses with faster current decays. However, when target sizes increase or located pretty deep, this method falls apart and requires an additional method such as raising the coil at various increments and seeing the response at different heights. The experienced hunters who want to try to discriminate between junk and coins, for example, use this method to tell coins from cans, espicially on the beach. Deep coins fade out fast as the coil is raised a few inches while cans require a few more inches of coil raising to notice a difference. With ferrous junk, the classic double beep is a tell tale sign when the PI coil is swept in a particular direction in relation to the target.

What you are trying to achieve is to use technology to extract additional target information from both the metal composition with it's associated decay as well as from target size related to how far it can be detected from the search coil. There is an overlapping region of uncentainty that only experience in hunting a specific area can deliver when recovering typical targets as well as the junk.

I would be interested in seeing some scope pictures of your various decay samples and know how fast the TX pulse current is falling to zero. That should raise some interesting discussions.

bbsailor

[Esteban]

In few minutes was created by Corel Draw program:

20 cm diam, 24 turns, 1 mm, spaced 3 mm, aleatory, only for to try the program.

You can print with laser printer in good quality magazine paper and transfer with hot iron.

Here the original 200 dpi:

http://www.mytempdir.com/1092611

[Carl-NC]

I'm dizzy...

[Esteban]

Yes, look for hipnotism!

In used magazine paper or directly in virgin paper for this purpose, this is for magazine, print in laser printer your design. Look the process and the high quality PCB you can make! (but is in Spanish):

http://www.xbot.es/webs/robotika/prensador.htm

[Qiaozhi]

I'm dizzy...

Careful .... watch out for these LRL guys.
Your eyelids are getting heavy .... you're starting to feel vvveeerrryyy sleepy....

[Andy]

re: Exploit weaker pulse

BugWhiskers,

I have been investigating why sending a weak signal first then a stronger one appears to provide better results. This is still speculation on my part, but what appears to be happening is that the first pulse produces some weak eddy currents, and the second pulse adds to the eddy current. The eddy current(s) after the second pulse are now "large" enough to influence the coil-field and affect the decay timing. As I experiment more, and collect more data I'll be able to get a better idea if this is the case.

It appears that sending the same frequency [same pulse repetition] does not lead to this condition, it only happens when a "double" [staggered] pulse is sent out. One pulse [low power] followed immediately by another [high power], and then this is repeated at a non-repetitive rate. I suspect that the first pulse could be a higher-power-one, just need to figure out how to clamp the coil quickly and at what time [time constant], or only energize a portion of the coil [easy to do on a PCB coil].

It makes sense to have two coils to really get this effect. The second coil would one "shorted" until it is needed, so that the activity of the first coil does not influence it. The second pulse would then use the second coil while the first coil is still decaying [first coil is "free wheeling"]. Instead of sampling only the second coil decay, also sample the first coil decay at the point in time that the second is energized. The first coil would pickup the second coil and any eddy currents caused by both coils created in a target, of course the second coil would also pick these up.

The reason behind this approach [twin "active" coils] is that by the time the second coil decays, the induced eddy currents may be gone, however because the first coil is present and has an active field, it is influenced by the eddy currents before the eddy current field has faded away. This approach will only work if both coils are active, it will not work if one is used as a transmitter and the other is used as a receiver. The key to making this work would be the amount of time between when the first pulse is fired and when the second pulse is fired. Intutition would say fire them as close as possible so that the induced eddy current [in the target] by the first coil is re-inforced by the second coil and allows the first coil to "pickup" the [target] eddy current. It is likely that the second coil will never see the "weaker" eddy currents such as those produced by gold nuggets, but the first coil should/will always see it.

My thinking is to have the two coils be different sizes, one larger than the other, as in, one inside the other. Using PCB coils this would be easy, just break the track at some point, the other side of the break would then be the start the second coil.

Sigh, I picked up metal detecting because it was a great way to get outside and avoid "work", looks like it's now more work than fun.

Well if this pans out, we may have discovered a way to detect smaller and deeper targets. My gut feeling is that we are on the right approach using a PCB coil, not much room for any deviations. We should also think about current monitoring circuitry for both coils, and a very accurate time-base [perhaps nanoseconds] to measure decays. Between these two [currents and timing] we should be able to map out what type of target is under the coil(s).

Andy

[Tinkerer]

Bbsailor,

I am setting up a test board to tryout your suggestions. Here are some more of my observations:
I do my testing in a very noisy environment. When observed on the scope, noise levels are several volts of 60Hz as well as many other frequencies up to several Mhz at lower amplitudes. At first I tried to eliminate some of that noise, then I decided to rather design my PI such, that it is capable to function in spite of all that noise.
Now, when I hook the coil up to the scope or FFT, I first do it without shield to look at the magnitude and frequencies of the noise that the coil receives. Interestingly, the 60Hz noise is not picked up well by the coils, rather, the output is dominated by the self resonant frequency of the coil. Then I shield the coil and again look at what noise is still left. I feel the shield is good, when the noise level is very small and does not change when my hand is on the coil.
Now, when I run the Tx-Rx and do the same test, with shield and without shield, I can clearly see that the noise of the unshielded coil adds significantly to the saturation of the coil amplifier, extending the decay curve/sample delay.
My interpretation of this, is that the decay curve is made up of many different frequencies.
My next goal is to try to find the source of each of these frequencies.

Those targets having less conductivity would only have earlier responses with faster current decays. However, when target sizes increase or located pretty deep, this method falls apart and requires an additional method such as raising the coil at various increments and seeing the response at different heights. The experienced hunters who want to try to discriminate between junk and coins, for example, use this method to tell coins from cans, espicially on the beach. Deep coins fade out fast as the coil is raised a few inches while cans require a few more inches of coil raising to notice a difference. With ferrous junk, the classic double beep is a tell tale sign when the PI coil is swept in a particular direction in relation to the target.


By designing a detector to give information on the time of the signal, instead of its amplitude, this problem can be resolved to some extent. I think I saw an Australian Patent that proposed a similar solution.

I would be interested in seeing some scope pictures of your various decay samples and know how fast the TX pulse current is falling to zero. That should raise some interesting discussions.

When I have my new test board ready, I will send you some pictures. Please remind me of it, since my memory hardly lasts 24 hours. I would also appreciate if you could suggest a good way to produce repeatable decay samples. Right off the bat, I can think of using a set of US coins, 1C, 5C, 10C, 25C, as well as maybe steel, stainless, brass and copper washers of about quarter size.
I also have a 0.6 gram gold nugget, as the smallest sample. To be able to see its influence on the decay curve on the coil amp, it needs to be quite near the coil.
On the other hand, the other samples mentioned above, should maybe be sampled at a further distance.
The sampling will have to be static, it gets too complicated to build a machine to wave a sample across the coil at repeatable speed. I have been thinking of using a pendulum for that purpose, I might still do it some day.
Tinkerer

[bugwhiskers]

Reply for Andy

[quote=Andy;47845]BugWhiskers,

I have been investigating why sending a weak signal first then a stronger one appears to provide better results. This is still speculation on my part, but what appears to be happening is that the first pulse produces some weak eddy currents, and the second pulse adds to the eddy current. The eddy current(s) after the second pulse are now "large" enough to influence the coil-field and affect the decay timing. As I experiment more, and collect more data I'll be able to get a better idea if this is the case.

It appears that sending the same frequency [same pulse repetition] does not lead to this condition, it only happens when a "double" [staggered] pulse is sent out. One pulse [low power] followed immediately by another [high power], and then this is repeated at a non-repetitive rate. I suspect that the first pulse could be a higher-power-one, just need to figure out how to clamp the coil quickly and at what time [time constant], or only energize a portion of the coil [easy to do on a PCB coil].

I wonder if this phenomenon is like ocean waves, if the collide at the right moment they can create an even greater wave and at the wrong time anhilate each other. This could also explain the timing mystery.

It makes sense to have two coils to really get this effect. The second coil would one "shorted".......
A shorted coil in close proximity will absorb and waste all the energy of the pulse.



The reason behind this approach [twin "active" coils] is that by the time the second coil decays, the induced eddy currents may be gone, however because the first coil is present and has an active field, it is influenced by the eddy currents before the eddy current field has faded away. This approach will only work if both coils are active, it will not work if one is used as a transmitter and the other is used as a receiver. The key to making this work would be the amount of time between when the first pulse is fired and when the second pulse is fired. Intutition would say fire them as close as possible so that the induced eddy current [in the target] by the first coil is re-inforced by the second coil and allows the first coil to "pickup" the [target] eddy current. It is likely that the second coil will never see the "weaker" eddy currents such as those produced by gold nuggets, but the first coil should/will always see it.

Using a pot to vary the time between weak then strong pulse while monitoring the decay curve should give some clues.


My thinking is to have the two coils be different sizes, one larger than the other, as in, one inside the other. Using PCB coils this would be easy, just break the track at some point, the other side of the break would then be the start the second coil.

Why not just vary the on time to a single coil to achieve the same result ?

Sigh, I picked up metal detecting because it was a great way to get outside and avoid "work", looks like it's now more work than fun.

I hope you like worms because you have opened up a few cans full of them

Well if this pans out, we may have discovered a way to detect smaller and deeper targets. My gut feeling is that we are on the right approach using a PCB coil, not much room for any deviations. We should also think about current monitoring circuitry for both coils, and a very accurate time-base [perhaps nanoseconds] to measure decays. Between these two [currents and timing] we should be able to map out what type of target is under the coil(s).

My early circuit experiments that produced good discrimination was
passing the back emf spike (limited to 30V) thru to a comparator with an adjustable threshold. Ferrous material caused a negative response and the reverse for non-ferrous. With higer back EMF's I saw as much as 20+ volts signal difference with a bar of aluminium (60*100*10) in close proximity.
I believe the behaviour is typical of loosely coupled transformers of which PI is also a form.

I hooked up the XY inputs of the old CRO to the coil current signal and integrator output but didn't see anything startling. My MOSFET is a PNP so the bottom of the coil is grounded which made hooking up a lot easier.

Have you considered using lower voltage MOSFETS. The lower voltage internal avalanch diode kills off the back EMF quicker allowing faster sampling. The PCB coil helps here also as the lower capacitance means less juice to heat up the MOSFET and damping resistor.


regards

Bugwhiskers

[bugwhiskers]

Reply to Tinkerer

[quote=Tinkerer;47846]

My interpretation of this, is that the decay curve is made up of many different frequencies.
My next goal is to try to find the source of each of these frequencies.


Did you see my post titled "Are we going round in circles ?"

If the pulse repetition rate is set to say 500Hz (close to the ears most sensitive frequency (human voice)) and "wanted" frequencies (good targets) are boosted then the ear would hear the differences between targets and learn what various targets sound like.

Just need someone with a Spectrum Analyzer to find the frequencies of interest and a few filters (gyrators).

Imagine a machine with a graphic equalizer type control panel where you dial in "wanted targets".

regards

bugwhiskers

[bugwhiskers]

Reply to Tinkerer

Bbsailor,

Interestingly, the 60Hz noise is not picked up well by the coils, rather, the output is dominated by the self resonant frequency of the coil. Tinkerer

Hi Tinkerer,

Remember the trick bbsailor told us to test a coils TC, ie putting the test coil near another coil while it is pulsing and monitoring the test coil with a CRO.

I wonder if either your coil is under-damped or if it is responding to a very strong early eddy current causing it to self resonate briefly??


regards
bugwhiskers