DD LOOP DISCLOSURES (long post)

[Anonymous]

PREFACE
This post covers a number of technologies related to DD loops. Some of the material disclosed here is prior art which is included for completeness, and is not identified separately. Although some of what I disclose here is new art as far as I know, there may well be prior art I'm unaware of, so at this time I am not making any claim of priority.
Although this disclosure relates primarily to DD loops, it will be recognized that some of the material disclosed here is also relevant to other loop topologies.
Although there are places in this disclosure which describe things in a way that implies a sequence of identical pulses, it will be recognized that the same principles will usually apply to systems in which there is a pulse sequence which comprises different timings and/or waveforms.
DEFINITIONS
"Pulse Induction" will generally refer to transmitter sequences which include an off-time during which signals can be received without interference from the transmitter field. In some contexts, the receive coil may be separate from the transmit coil and may be in induction balance with respect to the transmitter.
"VLF" will refer to methods which rely on induction balance searchcoils and which can potentially demodulate signals during the transmit on-time. In general, single-frequency, multiple frequency frequency domain, and multiple frequency time domain receiving systems are included. The term "VLF" reflects historic metal detector industry usage and does not imply that operation is limited to the VLF frequency range.
It will be recognized that in some cases, there may not be a clear-cut distinction between pulse induction and VLF operation, particularly where continuous-current PI including an induction balance searchcoil is involved.
It will be recognized that where a coil is referred to as a "transmit" or "transmitter" coil, in a PI system this same coil may be used as a receiver.
As is customary in this industry, the words "resistive" and "reactive" may be used rather loosely, the signal being referred to containing both resistive and reactive components. Where the context allows this sloppy usage, the word "resistive" generally refers to signals from electrically conductive materials, which are of the same polarity independently of whether or not the material is ferrous. The word "reactive" in the broad sense, refers to signals which have a polarity which is different for materials having high magnetic permeability (iron metal and iron minerals) than for materials which have near-unity magnetic permeability (salt water, nonferrous metals).
Where the context allows it or makes it necessary, the phrase "transmit on-time" and related phrases may include flyback time.
The words "loop" and "searchcoil" are used interchangeably, indicating an assembly containing one or more individual coils. The word "coil" is used in a more restricted sense.
PART I: DUAL RECEIVER PULSE INDUCTION
Although a two-coil system is described here, it will be recognized that some of what is disclosed is relevant to systems involving more than two coils.
When two coils are furnished, it is possible to energize one or both coils during transmit, and to demodulate separately the received signals from the two coils. The data thus obtained can be combined to obtain certain additional information about the target and its geometric relationship relative to the searchcoil.
In a dual coil (typically in induction balance) arrangement, in many cases the receive coil has a greater number of turns than the transmit coil in order to boost sensitivity. The amplitude and delay characteristics of the receiver then do not match those of the other coil which is used as both transmitter and receiver.
In some cases, it may be desirable to compare matched signals from the two channels. In such a case, equalization may need to be provided. Ordinarily this would be done with an RCL system ahead of the preamp, but in some cases it may be possible to put it elsewhere in the signal chain.
In principle it is possible to store constants in firmware which will perform equalization after demodulation.
PART II: RECEIVER LOOP SEQUENCING
For the sake of manufacturing economy, it may be desirable to connect the preamplifier to one coil, then the other, alternating, rather than providing separate channels for each.
In one embodiment, the switching would be back and forth on alternating off-times.
In another embodiment, the switching would be fast and clean enough to switch back and forth during an individual off-time.
PART III: SIMULTANEOUS COMBINATION PI AND VLF
Induction balance loops readily lend themselves to the arrangement of using the transmit coil as a PI receiver, and using the other coil as a VLF receiver. The VLF signal may demodulated primarily to extract reactive components for iron discrimination and/or for ground balancing the PI signal; or, the VLF signal may be demodulated to extract a variety of information as is often done in machines which are exclusively VLF.
If the VLF signal is demodulated in the time domain during the transmit off-time, it may be treated as a PI signal.
In some cases it may be desirable in the receiving system to provide two or more amplification channels operating at different gains and/or with and without gating, in order to optimize the receiving system for the type of signals being demodulated. In particular, it may be desirable to provide lower gain for the channel which is demodulated to obtain transmit on-time ("reactive") signal data.
PART IV: SEQUENTIAL PI AND VLF
Because receiver coil/preamp circuit topologies for PI and VLF systems often differ, particularly with respect to voltage clamping and preamp gating, it may be desirable to separate in time the PI and VLF signals. This allows actively configuring the receiver for optimum PI operation during one part of the sequence, and for optimum VLF operation during another part of the sequence.
In one variant of this system, the gain may be switched to low gain during the transmit portion of the sequence, and then to high gain during the off-time or constant-current time of the sequence.
In another variant of this system, different transmit waveforms are transmitted sequentially, one waveform optimized for PI, and the other optimized for VLF.
It is obvious that a machine could be configured so that the operator could select either VLF operation, PI operation, or both simultaneously.
PART V: TARGET DEPTH COMPUTATION
Double-D loops lend themselves readily to real-time reconfiguration into a far-field null configuration, one half being phased opposite that of the other half during the receive part of the sequence. The summing may be done by connecting the two loops together in antiphase, or by processing the signals through separate channels and subtracting the signals during or after demodulation.
The far-field null configuration is well known to minimize electrical interference, an important concern for PI in general, but not our concern here. Another use for the far-field null configuration is to facilitate using a metal detector near large masses of metal such as buildings, pipes, cars, and chain-link fences. We are not directly concerned with that here, although the underlying physical principle is the same.
During the transmit part of the sequence, it is possible to drive both coils simultaneously, in either a series or parallel arrangement, so they act together like one big transmit coil. Although this is not necessary, it is advantageous where other constraints do not prohibit it.
In order to determine the depth of a target, the signals are demodulated two different ways: in the far-field null receiver configuration, and conventionally.
As the loop is swept over the target, the peak signal amplitude from shallow targets will be of the same order of magnitude for both the far-field null and conventional signals. However, the peak amplitude of the far-field null signal will be relatively much lower for deep targets.
The ratio of the two signal amplitudes is indicative of the depth of the target. The amplitudes are preferably compared either over the top of the target, when the first derivative of the conventional signal amplitude is zero and/or the far-field null signal crosses through zero; or, when the first derivative of the far-field null signal is zero as loop is moving away from the target; or, after the loop has left the response area of the target.
PART VI: TARGET POSITION COMPUTATION
In the case of a double-D loop, it is possible to demodulate separately signals from the left and right halves of the loop. Comparing the two signals provides information on the location of the target relative to the center of the loop.
In one embodiment, the polarity of the far-field null signal is used to identify whether the target is to the left or to the right of the center of the target.
In a second embodiment, the ratio of the left and right signals is used to estimate the angle, left or right, of the target position relative to the center of the target. The polarity of the signals indicates whether the angle is to the left or to the right.
In a third embodiment, the searchcoil assembly includes multiple coils which can be arranged to produce two vertical planes of signal strength equality or null, one plane separating left and right (as described above), and another plane separating fore and aft. One or more of the coils within the assembly may be a coil permanently wired in a far-field null configuration, for instance the "figure-8" configuration.
PART VII: COMPUTATION OF SWEEP VELOCITY
It is well known that the duration of the (demodulated) target signal is short for fast sweep speeds and shallow targets, and long for deep targets and at slow sweep speeds. Target duration data does not distinguish between target depth and loop velocity, although the time signature of the demodulated target signal can often provide clues as to the depth of the target within a range of several inches from the loop.
If one knows both the depth of the target and the duration of its response as the searchcoil is swept over the target, the relationship between the two can be used to estimate searchcoil velocity. Call this "method A".
Another way to estimate searchcoil velocity is to measure the ratio of the amplitude of the first (time) derivative of the normal (not far-field nulled) signal with respect to amplitude of the far-field null signal, which is a first derivative in space rather than in time. Call this "method B".
PART VIII: OBTAINING INFORMATION ABOUT TARGET SHAPE
A flat piece of metal in a plane parallel to that of the searchcoil, for instance an idealized coin, produces a narrower response geometrically, and a quicker response in time, than a spherical object or lump. Ferrous metal objects often have a broader "signature" than a nonferrous object of similar shape.
In Part VII, I described two methods for estimating loop velocity, methods "A" and "B". Method "A" is sensitive to target shape, underestimating velocity on lumpy targets and overestimating velocity on flat targets. Method "B" differentiates in both time and distance, yielding an estimate of velocity which is relatively independent of target shape.
By taking the ratio of the velocities computed by methods "A" and "B" respectively, it is possible to obtain a target shape parameter which is at a minimum for a ring-shaped target, and at a (relative) maximum for a sphere or irregular lump. Some shapes may exhibit shape parameter values greater than that of a sphere.
It will be appreciated that one can start with the same raw data and arrive at the same result using alternative algorithms which do not explicitly calculate velocity.
PART IX: OPTIMIZING FILTER RESPONSE
To detect small shallow targets, it is desirable to have a broadband response, even though this increases the baseline noise level. To detect deep targets, it is desirable to have a narrower noise bandwidth in order to reduce the baseline noise level. This is well known in this industry, and is the reason for the existence of "variable SAT" on some products.
The use of microcomputers in metal detectors makes it practical to execute slow and fast filters in parallel. The question is, what filter is currently providing the best information for the target we're sweeping over? There are various ways of answering this question, for instance measuring the ratio of present signal strength to the nominal baseline noise level, or comparing normalized rates of change.
Depth, duration, and velocity data are not available continuously: that information is either not available or of poor quality no sooner than the top of the target. Therefore this data is not suitable for selecting filters or filter parameters in real time. However, most systems which compute target identity, do so during the trailing edge of the target signal or after the pass over the target is complete. If the data used to compute target identity has been stored either from separate real-time filters, or in a form where it can be filtered ex post facto, the velocity-time-depth information can be used to select or produce signals which are optimally filtered to produce the best ID for the target in question.
PART X: COMBINING SHAPE PARAMETER DATA WITH OTHER DATA
Conventionally, target ID information is computed based only on the phase-amplitude or time-domain properties of the received signals, in effect measuring electrical conductivity and (primarily in VLF machines) magnetic permeability. There have been attempts to estimate target inductance, and from that to impute target size. The relationship between target response and frequency/decay time varies not only with apparent material conductivity but with shape.
When determining target ID based only on such properties, it often happens that rather different targets are identified similarly, for instance US zinc pennies and aluminum screw caps. However, using this as an example, the aluminum screw cap is a stronger signal at a given depth; and, if it is intact and not flattened, its "time signature" is broader than that of a penny at a given depth and loop velocity.
Therefore, depth and velocity information can be used to augment other ID information in order to provide improved classification of targets.
--------------- Dave J.

 

[Anonymous]

Hi Dave,
Thought you were quiet on Sunday. Now we know why. Many thanks for all the good information. What we want is a software program so that the fields and responses can be modelled, and coils can be optimised for different types of object. A few questions and coil alternatives come to mind, so I will be coming back to you with some questions shortly.
Eric.

 

[Anonymous]

Commercial magnetic/electromagnetic modelling software exists which is probably up to the job, but it's expensive and I don't know what kind of learning curve it's got. Nor do I know what's good and what's not.
--Dave J.

 

[Anonymous]

Hi Dave,
Remember seeing a mine detector patent that had an accelerometer mounted on the search coil. Analog Devices make some little integrated beam rebalance parts, don't know if the transmitter will mess them up or not. As you know the intergal of acceleration is velocity, intergal of that is position. The search coil is always accelerating and stopping at the end of each sweep (zero velocity) with max velocity occuring at the center.
Unless you are spinning around and around in which case you get dizzy.
Then can have dynamic filtering to match the swinging (which motion detectors need bad). Most people when they get a beep, go back and swing back and forth real fast to find it again, funny to watch.
Make a good non-motion detector and I am happy.
VLF and PI combination is interesting, sure it would sell to those who can't make up their minds between VLF or PI. Just don't end up like the combination street and dirt tires that they have for motorcycles. The tires don't work good on the street or in the dirt. <IMG SRC="/forums/images/smile.gif" BORDER=0 ALT="">
JC

 

[Anonymous]

JC:
I've been keeping up with accelerometry for metal detectors for a few years now, waiting for the accelerometers themselves to become suitable. I found some prior art maybe a year or so ago, and kept it on file to refute any noxious patents. Meanwhile had to keep my mouth shut.
I figured the mention of "velocity" would produce some posts on accelerometers. Well, now you've spilled the beans, so I can at least say the word "accelerometer". Thank you.
Then there's the question-- how do you make it worth the trouble? The obvious answer is "to compute velocity". My answer to that same question is unconventional and for now I'm hanging onto it as a trade secret.
"Make a good non-motion detector and I am happy." You can't believe how much time and money has been poured down that rathole. I recommend being happy with motion machines, at least with autotune/SAT.
This business of combining PI with VLF isn't about selling to people who can't make up their minds between one and the other. It's about taking advantage of the capabilities of both technologies simultaneously to achieve what cannot be done with either alone. The customer doesn't care whether the technology is called PI or VLF or both or neither.
--Dave J.

 

[Anonymous]

Here's some links for EM software.
I have played with Quickfield, the student version
is free, but you have to tell em about yourself.
Scroll down and find it. Looks like there is a shareware version of some other software from a company in England, don't know nuthing about it.
The Quickfield version is full feature but limited in mesh nodes.
JC

 

[Anonymous]

Pasted in the wrong slot. Got to start instructions one of these days.
JC

 

[Anonymous]

I often get asked questions like this one, "How about some software with fuzzy logic to give either a dig / no dig message?". My answer is that the detector does not have as much information as the operator does. You can swing the coil over a target a few times to gather a lot of information about it. For example, you might swing over it in one direction, then rotate the direction of swing 90 degrees. Or you might swing over the center and then off to the side. Or you might swing slow and then fast, or low and high. Or you might switch from Disc to All Metal to check size or depth. The problem is that the detector does not know if these signals are all coming from the same target or if you have swung over a dozen different targets.
So I have been thinking for a couple of years about using accelerometers for a low quality inertial measurement system. It would only have to be accurate enough to tell which swings in the last minute or so have been over the same spot and which have been over different spots. I don't think I would actually like to use a detector like this, but I bet the marketing department would love it.
For coil velocity you do not need a linear accelerometer at the coil, an angular accelerometer in the box would be enough as long as you tell it how long the rod is.
Robert

 

[Anonymous]

Maybe one could measure the earth field induction as one swings the detector. Make this work for us for a change. Have to dial in lattitude though, might not work at the equator.
Still people who believe the earth is a giant magnet. <IMG SRC="/forums/images/smile.gif" BORDER=0 ALT="">
I'm afraid that alot of people don't know the difference between VLF and PI, sometimes I don't, and would think they were getting the answer to the "best detector". Could be used as a gimick for the unsuspecting by unreputable companies (Not that there are any).
Now if it really would switch between two modes, and work good in both, that might really be the "best detector". And worth the extra cost.
Yea, I know, I'm still hung up on non-motion detectors, but there are alot and I mean alot of places, when prospecting gullies and arroyos that you cannot swing a detector. Vegetation everywhere else. A non-motion detector can be built with less sensitivity, or same sensitivity but with drift.
And the drift can be lower with modern op amps.
Coin shooting the parks, or beaches this is not a requirement.
JC

 

[Anonymous]

For a non-motion detector try 3 coils arranged as below, two transmit windings and one receive. Slowly modulate the transmit signals so the sum of the fields from the two coils is constant but the current gradually shifts from one coil to the other. The coils can be held still and the field sweeps left and right.
Robert

 

[Anonymous]

Robert:
Distance is the second integral of acceleration. For a metal detector to know where the searchcoil has been for the last 60 seconds is beyond the capability of accelerometers which could be afforded for a consumer metal detector.
Accelerometer technology is advancing rapidly these days, and they may soon reach the point where they're usable as you describe. The question will then be, "is it worth it?"
Ah, yes, the problem of knowing to what length the rod has been extended. Ahem and amen. There are ways to do it, but for now, they seem to fall in the category of things which could be done, but shouldn't. Tomorrow I might have a different opinion, though.
Regarding fuzzy logic, "no comment".
-Dave J.

 

[Anonymous]

Robert:
The scheme you describe is one of those things that looks good at first glance. It's one of those things that I occasionally come back and revisit; I suppose that most other engineers in the industry do the same.
Unfortunately, it doesn't solve the problem of loop offset (imbalance) drift. And, the very limited apparent motion produced by the transmitter wobbling will produce a poor response to deep targets.
If one abandons the idea that the purpose of the coil arrangement is to provide fully static (non-motion) operation, then it is more or less equivalent to several schemes which I disclosed, except that the roles of transmitter and receiver have been reversed. In general, it is better to use a large transmitter and a small receiver, as this reduces susceptibility to electrical interference.
If you call the two smaller coils the receivers, then by switching back and forth between them, you can do things like determine relative target position and cancel the first order ground signal, as described in my earlier posts.
Thank you for your post. Maybe someone will find something uniquely useful about it that I have missed, and in any case the design merits posting for the sake of shedding further insight into the nature of searchcoils.
--Dave J.

 

[Anonymous]

Dave... What's the advantage of using a printed circuit coil (SandShark)? Is it cheaper to make?

 

[Anonymous]

Hi Robert,
I like it! May not be perfect but I like the thinking. I should probably shut up about my special problem of not being able to sweep the coil everywhere I want to search. One gives up alot of information having to do this static, and there is not alot to work with to start with. Just two or three coils on the end of a stick.
But I do like your ability to get around the problem and still get the information. I'm impressed. Built some VLF devices for a company a long time ago, which because of the application, had to be "static". Already had a microprocessor with A/D and the D/A wasn't being used. Had to servo on the drift at the output of the integrator, measuring the drift and then feeding back with the D/A to the input of the integrator to null it. Was able to get all the sensitivity of the detector AC coupled (also a requirement, total gain remain the same) but worked static.
JC

 

[Anonymous]

It's Jack Gifford's design, so it would be nice to hear him tell it. There may well be aspects of this that I've missed.
The design is a flat spiral, like a phonograph record. I believe some of his advertising has stated this. In any case, it's not a trade secret: anyone can pry the thing open and see for themselves.
Here are my opinions.
Since a wire mono PI coil (it's just the coil we're talking about here, not the whole assembly) doesn't cost much, I doubt that the PC board approach represents a significant cost savings if any.
The low Q is acceptable in a conventional PI, which the Tesoro PI's are widely reported to be.
It reduces distributed capacitance within the coil itself, but it increases capacitance to the loop shield. Cable capacitance is the same either way. So, from the standpoint of total distributed capacitance, it's probably a wash.
Jack's advertising says that the flat arrangement makes it more sensitive to the tiny stuff. I disagree. A regular coil is highly sensitive near the coil winding itself because all the windings are there right together in (almost) one place. If you go over tiny target, a regular coil may be dead in the middle but you pick the target up on the coil edges. The flat spiral doesn't concentrate the field anywhere, so you're more likely to miss small targets. But... the situation may not be quite that bad: see next paragraph.
For a given number of turns, you get more inductance by bundling the turns together, than if you spread them out. Looking at it through the other end of the telescope, if the circuit is designed for a particular inductance, then spreading the turns out allows you to have more turns. ....... For a given circuit, resistance, and inductance, the current is also a given. Therefore: More turns (of the same subtended area, and if total resistance remains the same) means a stronger transmit field. It also means higher receive sensitivity. But.... starting from the outside, where a wire coil would be, the turns on the printed circuit get progressively smaller and smaller. The turns near the center don't add much field, but they don't add much inductance or much resistance either, so they don't play an important role.
In my opinion, the idea of spreading turns out to get more turns per unit inductance deserves some attention. Some of the common coil formulas lend themselves to quick (slide rule!) estimates. However, I doubt that very much can be achieved in this regard. First, the exterior dimensions of the loop housing are usually a given: you can make transmit windings smaller, but not larger-- and smaller turns have less magnetic dipole moment per turn. Second, more turns increases resistance, so in applications were Q matters, you're going to have to fatten up the wire, but there's a good chance you were already minimizing wire size in order to keep loop weight down. So.... although this aspect of loop manufacture bears some investigation, it doesn't appear to offer major improvements, and in the end, manufacturing convenience will probably win. Reducing coil distributed capacitance is probably the greatest benefit to result from the small amount of spreading that can readily be achieved.
Sometimes a company will do something different which is not necessarily "better" overall, just different. This gives them something to advertise and draw attention to their product, and helps maintain brand identity. Although from an engineering point of view this may seem silly, this is good marketing practice, and if marketing doesn't sell stuff, engineers go hungry, so..... maybe marketing's needs aren't so irrelevant after all.
--Dave J.

 

[Anonymous]

Hi Dave on your comments that spiral coils,seem to be less sensitive to small targets,would this design lend itself for not picking up so much ground noise and also for deeper targets.
Regards Frank Wallis

 

[Anonymous]

Frank,
Regarding deep targets, without considering interference from ground, for given current that's mainly a matter of number of turns and their area (i.e., total dipole moment), as alluded to in my prior post.
It doesn't seem to me that the system would provide an improvement in ground pickup, but haven't thought it through.
I went to Tesoro's website looking for Tesoro's "how to contact us" email address, and didn't find one there (except for ordering catalogs), so I posted an invite on the Tesoro official factory forum.
My guess is that they probably won't respond (for pretty much the same reasons that factory people in general stay out of this forum), but Jack doesn't always do things the way everybody else does. Unlike most factories, Tesoro plays an active role in their own forum, so we just might find a reply either here or on the Tesoro forum.
--Dave J.

 

[Anonymous]

Hi Dave
Why I thought that ground mineralization factor would not be as great is your comment that the spiral coil tended not to be sensitive to smaller targets,and with extra turns a better drive signal.
Regards Frank Wallis

 

[Anonymous]

Thanks JC, that is a very comprehensive list. I have the Gemini by Infolytica, but it is only of limited use. I know the Vector Fields Opera 2D is used by a US university looking at PI for mine detection. Way out of my price range though. One that does look interesting is Pulse 4.0 by Field Precision.
Eric.

 

[Anonymous]

Strange, how there seem to be a lot of people coming out with ideas the same as mine, well, great minds think alike, and all that!
Dave, yes the answer IS unconventional, and I've put it to good use on the dynamic filtering system on the my MFX machine.
Looking through this forum, it's amazing how many GOOD ideas people are having. the only thing is, to be able to put ALL the realistic ones into the same machine will cost someone a fortune.
I have what i consider the ULTIMATE machine, which I designed back in 1991 which incorporates most, if not ALL of the suggestions i see on this forum.
Target development cost to date,