4 quadrant VLF discrimination

[Davor]

I think I have yet another one up my sleeve, but most probably someone else must have found that already. It is the ground balance thing.

In traditional VLFs the Geb channel is used to extract sound as it is basically the same thing as "All metal" one. Ground is balanced in a way that a ferrite and general ground response is quieted. It mostly corresponds to perfect discrimination criteria. However, there are some special cases where salts skew the angle of ground response. In extreme case of salty beaches the traditional ground balancing becomes useless for quieting the ground response.

The easy approach could be the one that uses a separate channel to be used for audio extraction steered independently from the other channels in order to quiet ground response, and simultaneously leave the discrimination criteria untouched. It could be made as an add-on for some of the existing rigs as well. My IGSL has a spare channel already, so I could try that soon.

[Davor]

It seem there is even easier approach to the 4-quadrant: phase inversion. Most of the nowadays IB VLF rigs have some switchers in a front end, and there are only a few different combinations. By observing a polarity of a GEB channel we can drive XORs to invert or not the switching phase of the rest of the channels, and a 2 quadrant rig suddenly becomes a 4 quadrant one. The switcher chip socket can become a "slot" to fit the 4 quadrant upgrade guts for a lot of different existing rigs. Easy as 123.

So instead of inventing all the bits from the beginning, I can invest the effort in a simple upgrade for the benefit of many. Sounds good. Kind of Christmassy.

[Davor]

True, but non motion is more continuous in response anyway, and it makes sense to give it a very different interface.

A variable filter for motion response seem an interesting idea, but IMHO it is an overkill. A main benefit of a HPF filter is to compensate 1/f noise. With typical MD response, it's slope is overwhelmed with the approaching target response. In my IGSL build I squeezed the last drop of the time response within the limits of the existing HPF, and as a consequence I enjoy a nice pinpoint-like performance. I concluded that a somewhat slower motion compensated response would be beneficial, provided proper care is taken of 1/f noise.

Sasho of Bulgaria, also present here on this forum have already experimented on slower motion response.

BTW, I really invented a novel analogue multiplier that would fit nicely to my 4-quadrant rig

[Tinkerer]

When thinking of man-machine interfaces we often put the machine up front, which is wrong. It may be a good marketing, but nothing else. Interfaces should be mutually orthogonal, complementing in information they provide, and not competing with each other.
You may think of vibration as a perfect candidate for non-motion target response interface. You'll have no problem with fast response there, and it will not compete with your audible motion compensated detection or discrimination indication.

A non-motion machine does not necessarily meant it has to be a slow machine. Non-motion means that the ground signal is not motion filtered.

A motion machine uses a filter that, at a given sweep speed, removes the ground response. If you use the motion machine at the wrong sweep speed, you loose the advantage that the motion filter gives.

With a variable speed motion filter, we can use different sweep speeds.

Or we could add a gravity motion sensor or accelerometer, that controls a state variable filter, automatically optimizing the motion filter with the sweep speed.

Tinkerer

[Davor]

I think I have a solution for an analogue multiplier thing. It just dawned on me, a kind of epiphany, so I developed a sort of a DC accurate analogue multiplier.

A Gilbert cell was not an option because it is not DC accurate. To reference it to the ground I'd need some serious level shifting. The other options are either too exotic or too impractical. So I invented my own. Using only garden variety components of course.

One of these days I'll post it here - a world première. Yet another product of my li'l laboratory

[Davor]

When thinking of man-machine interfaces we often put the machine up front, which is wrong. It may be a good marketing, but nothing else. Interfaces should be mutually orthogonal, complementing in information they provide, and not competing with each other.
You may think of vibration as a perfect candidate for non-motion target response interface. You'll have no problem with fast response there, and it will not compete with your audible motion compensated detection or discrimination indication.

[Tinkerer]

True. However, analogue design has it's appeal. Having in mind that you have a vast span of signal levels at hand it is much easier to pre-process things in analogue and then feed it to some easy going micro. Otherwise you'd have to use power hungry ADC + DSP. Or a good ol' analogue design altogether. You only need a reasonable audio output for a good rig anyway. OK, maybe vibration too.

Whatever the means, the ends are the same, including the math that supports it.

So far I learned that you may find good analogue solutions for most of your needs if you only pay a little attention. Otherwise you have to drag some heavy batteries all the time.

I like the idea for a vibrator as target indicator. However, I have not yet found a way to get a fast signal response out of it.
I consider a fast signal response to be lagging only a short time, 50ms, like 2" behind the sweep, at a standard sweep speed of 1 meter and a 10" coil. For good pinpointing.

How do you get that vibrator moving and stopping in this amount of time? At what RPM?

Tinkerer

[Davor]

True. However, analogue design has it's appeal. Having in mind that you have a vast span of signal levels at hand it is much easier to pre-process things in analogue and then feed it to some easy going micro. Otherwise you'd have to use power hungry ADC + DSP. Or a good ol' analogue design altogether. You only need a reasonable audio output for a good rig anyway. OK, maybe vibration too.

Whatever the means, the ends are the same, including the math that supports it.

So far I learned that you may find good analogue solutions for most of your needs if you only pay a little attention. Otherwise you have to drag some heavy batteries all the time.

[Aziz]

Hey Davor,

don't you think, that changing into the digital domain (DSP) would be much easier?
You could do all this stuff in software very very elegant.
Aziz

[Davor]

So I wanted to have some nice audio masking function that would provide me with both classic GEB that cancels ferrites, plus another notch that would cancel salts. These are ~90° apart. In addition to this mask, I'd have one (like IDX) or more discrimination channels working in TRUE/FALSE way to establish 4-quadrant discrimination criteria. Tone is produced as TRUExMASKxCARRIER, where MASK is a level obtained from the audio masking function, and CARRIER is an arbitrary audible tone, say output from 555 or better.

My first idea was using analogue multiplication where a=GEB and b=Disc and Disc is ~90° from GEB. However, ab is not the best choice because a~b -> ab~a^2 hence for the small target response the multiplier response is ever so smaller. At Wolfram Alpha it is presented in 2D as this:



There is a huge plateau with nothing happening around the saddle point in the middle, and that means poor small signal response. So there is no point showing off with my super sexy and novel Armstrong modulator multiplier circuit with PWM output

There is an interesting twist to the ab multiplication. It can be expressed as: (a+b)^2 - (a-b)^2 = 4ab
It is incredibly similar to the following: abs(a+b)-abs(a-b) and while it has all the zeroes and maxima at the very same spots, it also has incredibly better small signal response. A bit pointy, but better. See Wolfram 2D representation, it has a distinctive diagonals:



A step forward is providing some compression to account for huge span of signal levels received from targets. A simple approach is using a limiter with antiparallel diodes, and it's transfer function would be like this:



Compressed abs(...) function thus has a formula and in 2D looks like this:



... while the compressed ab function has expression and looks like this:



The ab function after compression looks pretty, but on close inspection it still lacks a proper response for small signal. It could be abused as a sort of anti-chatter function.

In analogue world the ab function is obtained using the analogue multiplier, and the abs function is obtained by linearised full wave rectifier..

[Davor]

Hi all, just a small update.
I found two candidates for an analogue multiplier thing, one real, and the other one not so much, but with a great potential.

I first thought of a PWM modulator thing that would be a nice one to manipulate by logic ICs further on, and in a process I came up to a real treat, a direct way of producing PWM by means of the Armstrong PM modulator. Sounds like a rocket science, but it is not. I just wonder how come no one else didn't think of it before me. So what I get is a signal that produces almost perfect 4-quadrant product of two input signals using no log-antilog devices anywhere. Just analogue switches, comparators and EXORs. The output is in fact binary, but all you need to do to get analogue out is filter it. And instead of analogue switches I can do all multiplications by CMOS EXORs. Power consumption? Negligible.

However, a real treat is in similarity of these formulas that explain the second approach:

(a+b)^2 - (a-b)^2 = 4ab ...(1)
abs(a+b) - abs(a-b) = √((a+b)^2) - √((a-b)^2) ...(2)

Both these functions share the same behaviour at zero crossing of either a or b variables. To implement the second formula I only need some precision rectifiers configured by a few opamps, and the output is analogue.


I'll soon post some schematics of these, most probably in a form of "ABM blocks" just to show what it does. So far I prefer the latter approach because it preserves the low signals - the output is under the square root law - ideal for metal detecting.

[Davor]

Yes, that one might work as well. I'm not certain if the radii calculation is accurate enough because field concentrates in proximity of wires (TX wires that is), and not dipole-like as in most of the calculations. I'm not experienced enough in such calculations, so - why not? It does not reverse phases as figure 8 coil does, at least not in the centre of the coil.

Maybe Aziz would care to test this coil in his nice software.

[Sergey_P]

A differential ... - ... coaxial coil

http://www.md4u.ru/download/file.php?id=5035
4/3*r0=r2-r1
( http://www.md4u.ru/viewtopic.php?f=77&t=3866 )

[Davor]

A differential coil I'm thinking about is one with figure 8 Rx coil placed inside an O shaped Tx, or a bicycle style in-phase/anti-phase coils. Balance is incredibly simple to achieve, but a target under one Rx coil side gives response that is phase reversed under the other. So if you use a "normal" 2-quadrant VLF metal detector only one of the loops gives sensible response. Normal detectors do their thing only within 0-180° span. Using the counter phase as well would make these differential coils do superb job. Beside the far field EMI, EF, and what not, a ground response is cancelled as well. So it is a good thing.

The coaxial coil suffers from losing balance in proximity of ferrite laden ground. Coil closer to the ground gains in inductivity more than the one on top of it. That's why I expect much more from figure 8 configuration.

[Ferric Toes]

Back to coils for IB and PI, many years ago some IB detectors had stacked coaxial coils, i.e. TX sandwiched between two antiphase RX coils. I tried this on a PI discriminator back in the roaring '80s and it worked well on discrimination but failed if there was any ferrite mineralisation. The nice thing about this configuration is that the responses are simple with no negative going signals under the coil. Good emi cancellation too. Only downside is that the coil is considerably thicker. Garrett once used this type of coil for VLF units. In fact I used to buy formers and outer shells from him. Would this be worth looking at again in the light of the new IB technology being discussed.

Ferric.