Announcement

Collapse
No announcement yet.

DEEPER PI DETECTION DEPTH

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Originally posted by Aziz View Post
    Ok guys,

    found another TEM source. In this thread (post #51):
    http://www.geotech1.com/forums/showp...3&postcount=51

    Look at TEM1 and TEM2 variants.

    Aziz
    Great, thanks.

    In a way, it's a lot like buck-boost regulator, sortof kindof.

    -SB

    Comment


    • Originally posted by simonbaker View Post
      Great, thanks.

      In a way, it's a lot like buck-boost regulator, sortof kindof.

      -SB
      This little circuit is simple but genius. It has it's own step-up converter while stimulating the target during charging and recycling phase too. Note, that charging and recycling phase use a single TX-on pulse gate.


      Aziz

      Comment


      • OK, now I remember... thanks!

        Comment


        • So has anyone got any ideas how to implement a ground balance system using the TEM transmitter ? Not that I fully understand it but I get the impression the ground balance system implemented on Whites half sine pulse system relies on taking two samples around the symetrical sine pulse. ie. identical dI\dt rising and falling. I imagine the un-symetrical TEM method is going to complicate that somewhat.

          Comment


          • The Holy Grail

            Hi Midas,

            the ground balance (GB) is the holy grail of the game.

            We have more effects to cope width:
            - induction balance change (coil coupling coefficient change due to ground mineralisation) (if an IB system is used)
            - time constant change of the transmitter/receiver coil (inductivity changes over mineralized ground)
            - static magnetisation M (of the ground/hot rocks, geo-magnetic field included)
            - complex dM/dt (magnetisation change behaviour of the ground/ferrous-target, magnetisation decay and magnetic viscosity effects)
            - TX energy losses due to magnetisation process (magnetic polarisation energy losses)
            .. (forgot probably more effects)

            As we know, the typical basic RLC circuit (i.e., TX circuit) is a two order circuit. It can mathematically be described with second order differential equations. Every energy storing element/part (L and C) contributes an order to the complex equation.

            Energy storing elements:
            L (inductivity) stores current/magnetic field energy (temporary, it can't store it permanently however)
            C (capacitor) stores charge (it can store the energy permanently if we neglect the leakage energy losses)
            Magnetisation process as well (short/long time constant)

            If we add the target (inductivity) to the model, the order increases to three.
            If we add the magnetisation process to the model, the order increases more.

            It's going to be mind blowing now.

            The whole process has a complex system memory (state of each energy storing element in the system, i.e., it's energy content dependent on time t). These dynamic systems can be modelled with differential equations only.

            We want to detect the state of each energy storing element in the system to be able to remove all the disturbing effects at the end.

            As we can not describe the complex system accurate enough, we have to make simplifications (approximations), which give us reasonable results with less cost and effort (but not perfect).

            The art of setting the sampling time position is dependent on the systems state (what has been done with the system previously) and whether a disturbing effect is changing its state. You obviously need at least two samples to detect such a changing state.
            But we have a lot of variables in the system, which require more samples to solve it.

            If a disturbing effect isn't changing it's state at the sampled two time positions, that's easy as we can easily subtract it out. But we can not detect it.

            We have lot's of superposition of effects, which doesn't make it trivial to solve it however.

            You see, it's a big challenge. Even for a guru.


            Cheers,
            Aziz

            Comment


            • BTW,

              a dual-frequency VLF GB system can be simplified (approximated) by a simple linear equation (of form f(x) = a1*x + a0 ) .

              Aziz

              Now, hurry up and run to the patent office patent troll.

              Comment


              • Ground balance

                Originally posted by Midas View Post
                So has anyone got any ideas how to implement a ground balance system using the TEM transmitter ? Not that I fully understand it but I get the impression the ground balance system implemented on Whites half sine pulse system relies on taking two samples around the symetrical sine pulse. ie. identical dI\dt rising and falling. I imagine the un-symetrical TEM method is going to complicate that somewhat.
                In this thread, we have explored the TX methods in search of deeper detection depth.
                another very important factor for deep detection depth, is the ground balance.

                I suggest we move to a new thread for the ground balance. I will name the new thread:

                THE PERFECT GROUND BALANCE

                Tinkerer

                Comment


                • Like double balance coils? I'm concentrating on this possibility.
                  Originally posted by Aziz View Post
                  .. (forgot probably more effects)
                  Maybe Barkhausen effect? With TEM magnetisation it is sure to happen.

                  Comment


                  • Originally posted by Davor View Post
                    Like double balance coils? I'm concentrating on this possibility.
                    Maybe Barkhausen effect? With TEM magnetisation it is sure to happen.
                    With 4 coils and 3 different value inductance interacting and having different coupling coefficients, it is already quite complex. But this is not where the major problems with ground balance are.

                    Tinkerer

                    Comment


                    • Well. that would be a whole idea of ballance. If a same mechanism works here and the same mechanism works there and everywhere, it is balanced out. Only in case of various perversions, e.g. golden nuggets, you'll get a broken balance. That's my idea.
                      Single balance should work with homogenous medium or a decent discontinuity with no gradient. Double balance should work with a linear gradient too. I'm afraid nothing will ever work with non-linear gradients, Barkhausen, saturation, or real targets. No way you will ever cancel that.

                      BTW I noticed that the majority of IB MD-s have unbalanced front ends (bad), and some of them have shielding (better). Also majority of IB MDs have asymmetric Tx with THD controlled only by the tank Q and samples timing is derived from there, hence it is prone to PWM and false ground problems - due to the proximity a tank Q is changed, oscillation average value and THD is changed and consequently timing criteria are shifted (PWM style). Knock-knock, the ground (proximity).

                      PI designs tend to suffer from high Z problems that shorten useable sampling period, and too much gain at front end that prolongs recovery. Frankly, you are orders of magnitude beyond the ground effects anyway.

                      I'd say that many design problems are attributed to ground and a whole story is blown up beyond its real proportions. Once you fix a first approximation problem, e.g. balance your coils properly, only then you'll be able to grasp the second. IMHO ground proximity is a first approximation to GB.

                      Comment


                      • Hi all,

                        I do not want to demotivate you but you have to cope with other facts too:
                        The non-linearity produced by the processing electronics:
                        - (ADC/DAC)
                        - (pre-)amplifier
                        - demodulator
                        ...(and so on)

                        Consider the following in the amplifier stage only:
                        Clipping/saturation/overloading of the amplifier is distorting your signal and hence producing harmonic distortion to your signal (spectral blur). Slow amplifiers causing distortion and low-pass filter. Fast amplifiers tend to resonate (oscillator).
                        Even if you have a perfect ground balance formula, it's worth nothing without compensations to the non-linearities occuring.

                        Dream on guys. There ain't perfect ground balance. But a better GB (a better approximation solution) is available.

                        Cheers,
                        Aziz

                        Comment


                        • What should a perfect pre-amp stage look like?

                          Originally posted by Aziz View Post
                          Hi all,

                          I do not want to demotivate you but you have to cope with other facts too:
                          The non-linearity produced by the processing electronics:
                          - (ADC/DAC)
                          - (pre-)amplifier
                          - demodulator
                          ...(and so on)

                          Consider the following in the amplifier stage only:
                          Clipping/saturation/overloading of the amplifier is distorting your signal and hence producing harmonic distortion to your signal (spectral blur). Slow amplifiers causing distortion and low-pass filter. Fast amplifiers tend to resonate (oscillator).
                          Even if you have a perfect ground balance formula, it's worth nothing without compensations to the non-linearities occuring.

                          Dream on guys. There ain't perfect ground balance. But a better GB (a better approximation solution) is available.

                          Cheers,
                          Aziz
                          Aziz, Davor, Midas, Moodz and many other voices point out the possibility that the preamp stage could be improved a lot. It is time to do it.

                          We do NOT WANT: Clipping
                          Saturation
                          Overloading of the pre-amp.

                          We want to use an opamp that is fast enough, but not too fast.

                          We want to use the maximum amplification, while staying within these parameters.

                          Would using a +/- 12V supply instead of +/-5V, help? More amplification, without clipping and saturation.

                          Limiting the bandwidth to the actual needed one, would help.

                          What is the needed bandwidth anyway?

                          If we want to resolve a 1us target, we should aim for 1MHz. But, can we build a coil with a self resonating frequency of 1MHz? If the coil can only resolve 250kHz, we can also reduce the pre-amp bandwidth to the same frequency.

                          Reducing the bandwidth, increases the dynamic range.

                          What else can we do???

                          Attached is a proposal for a differential input pre-amp, to go with the TEM TX shown above.

                          Can it be improved? For sure it can, a lot.

                          Suggestions?????????????

                          Tinkerer
                          Attached Files

                          Comment


                          • Originally posted by Tinkerer View Post
                            We do NOT WANT: Clipping
                            Saturation
                            Overloading of the pre-amp.
                            The biggest point is that you actually don't have a 1us target response, only a concept that forces you to think into that direction. By my standards you already have far too much amplification in any "classical" PI design. Noise level is at, say uV levels, and signal is near mV levels, hence by applying extra gain you just decrease your dynamic range.

                            What actually is bothering you much more than everything else is a floating ground, and that is your main enemy with opamp in frontend.

                            There is a simple solution to many problems, and much more, if you decide to drop the high gain opamp. It is a chopper. Ready-to-wear chopper stabilised opamps are slow and expensive, but it is not a big deal to make your own implementation using analog switches. You don't have to be extremely accurate, so your design may be fast enough to properly sample a PI signal. So instead of one or two windows as for discrimination purposes, you could have, say 10 windows, each supplied to an alternating opamp input, or keep them in SAH-style analog buffer for bucket-brigading into something slow and accurate.
                            If you are only into high amplification, then just chopper-sample into an alternating signal, amplify, and anti-chopper into a perfectly windowed and perfectly amplified PI Rx windowed signal. Or something in between. Say, two perfectly amplified PI Rx windows for discrimination purposes. You can even apply some weighting function to samples to make them even more interesting.


                            What is the needed bandwidth anyway?

                            That would be related to your integration window. Incredibly fast pulses (chopper) can be successfully integrated by incredibly slow integrators. In case you apply some averaging scheme using samples from several PI pulses, then it goes longer and loooonger. You ca'nt make it too long because metal detecting is a somewhat physically active thing, but I'd say that ~20 Hz could become practical. Same as with VLF IB.

                            Comment


                            • Maybe I was not too straightforward in my explanation, so I suggest exploring a Tayloe mixer as a perfect example what a chopper can do.

                              A Tayloe mixer uses successive sampling of typically 4 pulses in a row at 90° intervals of a RF signal. 0 and 180° are used for I and 90° 270° are used for Q. It is also possible to use any number of samples, as long as they are used with respective phases.

                              In PI signal decomposition you don't have phases but time deltas, and your sampling will run a finite number of samples after each PI pulse, thus in effect running cyclically.

                              Your design already has a balanced input and a ground reference (centre tap), hence your choppers do not need to alternate between signal and ground, so it is a perfect candidate for Tayloe style sampler. Tayloe is usually realised with bifilar coiled transformer with centre tap (double balanced). With Tayloe, samples are stored in capacitors that - in a succession of many samples - act as a LPF. I'm thinking of ~20Hz range. Any garden variety opamp can deal with that. They can also be anti-Tayloed at a pace of your need, say ADC clock, and you can do even more magic with them in DSP. Or just do some plain old boring math and spit out indications for size and a kind of metal.

                              I think you could as well do with only one sampling window, but why not more?! In case of the above piece of schematic, just replace R7 and R9 with analog switches gated at PI window sample, and a C between them replace with two signal ground referenced capacitors dimensioned for ~20Hz LPF.

                              PI signal is very orthogonal against the Tx pulse by virtue of non-concurrency. Even with monocoil. It can be fixed to do much more and better than IB VLF. I just know that.

                              Comment


                              • It seem only appropriate to continue here where we left some time ago. Now this is big. I present you a LTspice simulation of a logarithmic weighted PI Rx with dead time of less than 2 microseconds. Furthermore, its output is logarithmically compressed so there will be much fewer bits required for ADC, where required.

                                There are some additional repercussions of this approach that enable some very interesting solutions to discrimination (say, you may play with C3 and observe consequences), but finally it is THE weighted solution I was mumbling incoherently for the past few months. Also, it is an inverting solution and thus the signal decays opposite to the "Surf" solution.

                                On the schematic you'll find two switches that are used to divert signal to a common node, and prevent two Rx-es to interfere with each other, and to draw output in nice colours. The opamp with a bunch of diodes is a log detector: nothing fancy, but works as a horse. I'm aware this is going to perform a bit differently with a real world opamp, but here you go - a place to start experimenting. This can sample very early.

                                Patent trolls - keep off!!!
                                Attached Files

                                Comment

                                Working...
                                X