Ground Balance Theory

[moodz]

... Whoops ... Accidentally posted a secret.

[Ferric Toes]

I wrote the following a couple of days ago, and it is relevant to this thread so here goes -

All PI mine detectors that I am aware of sample in the OFF time only, and this effectively decouples the RX signal from the TX. It is decoupled from the Air signal which is at its worst in the case of a mono coil, and it is decoupled from the Ground signal in the cases for normal ground conductivity and susceptibility. This is the case for a first sample delay of 10uS from the initiation of TX switch off. OK, you can sample much sooner with various techniques that have been covered in forum posts, but it is of limited value to do so in a detector for general location of nuggets, coins, rings and many mines. I have a Schiebel AN19/2 which was the primary NATO mine detector from 1980 – 1990. In the absence of detailed technical information, I estimate the sample delay to be about 15uS by observing the comparison in sensitivity to, say, the Deepstar of my design. The Schiebel detector has now been superceded by the Vallon VMH3CS and is often seen in current TV news reports from countries where conflicts are ongoing. I don’t have a complete VMH3CS, but I do have a few search coils that are in working order. It is a simple shielded monocoil in a truncated elliptical shape and has an inductance of 1.5mH. The best sample delay I can get with this coil is 20uS, so unless non-standard technology is being used the small object sensitivity is a bit less than the AN19/2.

Regarding the modulation of the TX current/field this does not seem to be an issue in the practical world, provided certain design steps are taken. Using a soldering iron and bench test equipment, I measured the inductance of the 30mm solenoid coil of my viscosity test gear. It came out at 182.7uH. I then inserted a 10gm sample of Australian ironstone with a susceptibility of 33,000 SI units as measured on a MS2B. The inductance increased to 189.00uH, or 3.4% increase. Provided the TC of the coil circuit is short compared to the width of the TX pulse any modulation of the TX field by susceptible ground will not be noticed, neither will the effects at switch off, provided switch off and damping occur well before sampling. When in Australia, I measured the change in inductance of a 11in 300uH coil when place on ironstone, and it was less than that on the small solenoid.

In some designs, and for other reasons, I have used a constant current source for the TX supply. This in itself would correct any changes in TX current due to magnetic ground effect. However, even in the worst ground conditions there is no evidence that there is a problem in this area.

In the final analysis, even if modulation of the TX field were to take place at higher levels, a good GB system will not respond to magnetic viscosity as the decay law stays the same irrespective of amplitude. The biggest modulation is simply swinging the coil.

Eric.

[Davor]

@moodz, you can observe it like that if you prefer calculus
Think of a transformer ... perfect power transfer happens below the saturation current. Transformers are calculated as per the voltage, and they are thus expressed in voltage ratio. Currents must be calculated only for losses and saturation estimation. Metal detectors are in fact a sort of transformer, so the whole logic behind transformers hold as well.

[Ferric Toes]

Why is it i cannot get some sort of straight to the point answer to this question?.....Why...maybe no one knows.

How does the TX driver influence the ideal Magnetic Field generated in a particular soil? What parameters are important?

What determines the TX driver's characteristics in relation to a target's maximum magnetic field in so and so soils?

Please understand, I am only referring to the Research studies above in those documents.....If it's flawed, then we can all move on.

Cheer's Sid

Hi Sid,

It appears that what is referred to is a single coil. Footnote 9 shows what needs to be done if a separate RX is added. I don't see a reference to the TX current waveform, which seems to me to be important as the shape of the generated magnetic field will be similar. Ideally the coil circuit time constant should be short compared to the pulse width, so that the final current is limited by the dc resistance of the coil circuit. Some PI's do not reach this point, so the current is still growing exponentially and has not reached its final value at the switch off. At the switch off, the drive circuit becomes a high impedance and the voltage across it rises to try and maintain the current and field as it was just prior to the switch off. In fact the TX coil now appears as a closely coupled target of high resistivity and to generate any current at all, 100's or even 1000's of volts are required; hence the high voltage spike. The important thing to remember is that it is the current waveform that determines the shape of the magnetic field pulse and that no high current spike appears at the end of it; just a ramp down, or a fast exponential decay down, depending on any voltage limiting circuit that may be in operation, and the method used to damp any oscillation due to parasitic capacitance.

I'll deal with the effect of the ground later. Busy day today.

Eric.

[moodz]

with given L relevant is the L(di/dt) which equals to the coil voltage. With that in mind, the coil voltage spectrum should have something to do with frequency response of the fastest targets resolvable by the receiver (and not faster than that), which is not necessarily the case with typical PI.

...hint hint .... you might want to rethink that ... magnetic fields are generated by currents not voltages. The coil is not a resistor ... it is a reactive component .... coil voltage is not the same as current ... for instance it is possible to hold the coil voltage constant for a period of time whilst changing the current and thus the magnetic field. Measuring the voltage at this time might lead you to an incorrect analysis.

[Davor]

What parameters are important?

with given L relevant is the L(di/dt) which equals to the coil voltage. With that in mind, the coil voltage spectrum should have something to do with frequency response of the fastest targets resolvable by the receiver (and not faster than that), which is not necessarily the case with typical PI.

[sido]

And you are referring to PI Tx?

Yes

Regards Sid

[Davor]

And you are referring to PI Tx?

[sido]

Why is it i cannot get some sort of straight to the point answer to this question?.....Why...maybe no one knows.

How does the TX driver influence the ideal Magnetic Field generated in a particular soil? What parameters are important?

What determines the TX driver's characteristics in relation to a target's maximum magnetic field in so and so soils?

Please understand, I am only referring to the Research studies above in those documents.....If it's flawed, then we can all move on.

Cheer's Sid

[Davor]

Sure, it makes sense. There is still much room for improvements in Tx layout and coil balance, because other approaches tend to fall into some of the contradictory traps. Division by Tx envelope would fix all the cyclo-dynamic components of Tx AM errors, but only if applied prior to motion compensation filtering, hence at full dynamic range with "air signal" and all. E.g. it would not help my QSD chopper approach with motion compensation, as cyclo-stationary component is not passed through. (I could devise some kind of signal sink that would perform such division ... worth a try)

Funny thing is that once you are finished observing PI and CW as different and separate systems many solutions become interchangeable

[Aziz]

Now you guys can imagine, why I am sampling and processing the TX coil voltage as well (either PI or CW PI/VLF/LF).

Aziz

[Davor]

Completely true, except maybe formulation of

2. Each delay in TD determines a frequency in FD.

...which in fact is a frequency band from DC to a frequency corner set by the Nyquist–Shannon sampling theorem.

Both CW and PI Tx-es suffer from poor PSRR, and your reasoning viz. the air signal error vector is very relevant so power supply stabilisation and filtering is essential. There is however one catch: air signal is reduced by IB coils, where both cyclo-stationary component due to the power supply DC component, and the AM error vector due to the power supply ripple get reduced proportionally. The AGC schemes are mostly a mild let down because vast majority of them are referred to a power rail, hence not helping with PSRR at all, and in the above Garrett example gloriously miss the whole point of regulation.

Otherwise there are little differences between PI and CW.
At the moment it seem as if PI has some advantage over CW, but that's only because of CW development stopped at about 1990, with only a few stray designs that did not help. Only recently CW Tx coils "discovered" low inductance as the approach to stronger field, and Rx-es are still catching their own tails.

Congratulations on your new job

[mikebg]

This discussion seems to have shifted from what happens in the time domain (pulsed TX), to what happens in the frequency domain (CW oscillator TX).

Due to the decoupling of TX and RX by the sample delay and with the correct TC's for the coil /TX and damping circuit, ground modulation of TX field has no effect on the detector operation at the sample delays that are suitable for general use.

Eric.

Eric, the CW oscillator TX is simply a specific case of excitation with periodic TX current. When something happens in the Frequency domain, it shows what is happened in the time domain. The Frequency domain is similar to a tool for seeing through the fog. The Time domain is filled with a fog. Fourier transform shows things that will turn your existing concept of PI technology. The conventional PI is other specific case of wideband technology with low energy efficiency. For high efficiency, we need ASinc "ringing" of TX tank circuit instead damping.

http://www.geotech1.com/forums/showthread.php?15518-Missing-Link-between-pulse-and-continuous-waveform-detectors&p=118177#post118177

1. Frequency domain shows that there is no decoupling of TX and RX by sample delay. There is convolution with AIR signal, despite TX current is zero. The conventional PI is not a radar.

2. Each delay in TD determines a frequency in FD. The conventional PI makes discrete signal processing with short samples. However we know even your designs with continous signal processing - an integrating window instead series of samples.

3. The value of each short sample determines the response at a frequency. This is low noise narrow band process.

4. We can make a PI detector to operate as a narrow band CW detector, however with bad TX efficiency because magnetic field contains useless frequencies. It has even a component with zero frequency (that means DC).

5. Frequency domain shows that flyback voltage is simply the AIR signal received when search head is lifted. This is frequency spectrum that provokes the eddy currents in environment.

If the AIR signal is modulated by ground, the synchronous demodulator sees the modulated carrier wave as an additional ground sugnal because the modulating frequency band is the same (for example 0.2 ... 0.8Hz). However the AIR signal has different frequency properties than the genuine GND signal. The AIR signal is direct proportional to frequency. The GND signal has ferrite component also direct proportional and conductive component proportional to frequency squared.

Oops, I speak in Frequency domain only. In Time domain, the AIR signal is proportional to derivative of TX current, but conductive groung generates GND signal proportional to the second derivative.

Fortunately, since the beginning of this month I am no longer unemployed, but this drastically reduces my free time for hobby. I will try to participate in this forum because I have ready drawings. Unfortunately I have no prepared text for the drawings. Attached is the drawing explaining the method for measuring the ground modulation of TX (modulation of AIR signal). The method is valid for all types of metal detectors. I do not remember if years ago I published in the forum description of the method. I'll search the forum and if I can not find the description, I will write new.

[Ferric Toes]

OK, let's have it your way, can you tell me what order is the regulation network in this Garrett oscillator, and why it's regulation overshoots?

I thought of a lengthy answer explaining all that's happened between the 1998 cornerstone paper by A. Hajimiri and T. H. Lee, "A general theory of phase noise in electrical oscillators" which led to better oscillators of nowadays, but let's hear your arguments.

This discussion seems to have shifted from what happens in the time domain (pulsed TX), to what happens in the frequency domain (CW oscillator TX).

Due to the decoupling of TX and RX by the sample delay and with the correct TC's for the coil /TX and damping circuit, ground modulation of TX field has no effect on the detector operation at the sample delays that are suitable for general use.

Eric.

[Davor]

Davor, why you hate to copy the best solutions? Do you know more than the world known masters?
GARRETT uses P-I controller in their TXs more than 30 years and MINELAB - more than 15 years.

Please fill a plastic bowl with salt water and measure the modulation index of AIR signal pumping with search head. You'll get a much larger modulation index of AIR signal than does this the target. The SPICE can't show such small index, but the synchronous demodulator senses it.
Then do the same with a P-I controlled TX to realize that you're wrong.

OK, let's have it your way, can you tell me what order is the regulation network in this Garrett oscillator, and why it's regulation overshoots?

I thought of a lengthy answer explaining all that's happened between the 1998 cornerstone paper by A. Hajimiri and T. H. Lee, "A general theory of phase noise in electrical oscillators" which led to better oscillators of nowadays, but let's hear your arguments.