bipolar pulsing to remove Earth Field

[mickstv]

Hi Altra, good to see you've got it working. What sort of quirks are you seeing with it. The circuit I've got going only uses two FETS to run it and a pair of diodes and the damping resistor.

I've been changing various settings in the processor and a 70us TX looks good. The spacing between TX pulses doesn't seem to change how the EF balance works, once it's set it never needs changing. The spacing tried has been from 50us upto 200us.




Mick

[Davor]

IMHO this is not far from a simple step voltage, where sampling may start right away, and real impedance is constantly near zero. And less quirks electronics-wise.

[Qiaozhi]

If you take a simple high-gain PI, and swing the coil in the usual manner, there will be an unwanted beep at the end of each swing. This is caused by EFE (Earth Field Effect). Apart from bipolar pulsing, the easiest way to cancel EFE is by taking a second sample at a much later time (e.g. at 200us) and subtract this from the main sample. This is the method used in the Hammerhead and Surf PI.

The problem of EFE is real and you can easily confirm it in a simple test.

[Altra]

H-Bridge PI transmitter

Taking the schematic from patent 6130489. I came up with this interpretation. I left out the zenor diodes, which as far as I could figure, had the purpose of snubbing the flyback.
Common n and p channel fets were used in place of bipolar transistor. Supply voltage 12v, 5v switching logic to the 2N7000s. I used a pulse frequency around 2.3khz and pulse width 60us. These could be adjusted up or down. I had timming program already done and just adapted it to this circuit.

The circuit operates by alternating the polarity of the coil using an h-bridge. The
h-bridge is then pulsed on and off with a p-ch fet. One side of the coil is always connected to B-. So this operates like the p-ch Surf-pi with a coil that reverses after each pulse. A differential amplifier is connected accross the coil. The output is true bipolar and centered about +V/2

I need to play with the preamp some more. Lots of parts, but it works on a single supply.
Some people may ask why go to all the trouble of making a bipolar pulse when a mono polar works fine? If you want to play with sampling during "on" times for experiments in PI discrimination. (see Tinkerers PI-IB) A symetrical bipolar receive signal is best. The differential sampling removes DC offsets and synchronous noise. Any comments or improvements welcome, Thanks

[Davor]

I thought of employing a H-bridge in one of my simulations, but ended up with a diode quenched TEM style energy saving ...thing

Anyway, to make your differential Rx really differential you may want to supply each branch an equal impedance, and traditionally it is achieved by duplicating resistances in inverting and noninverting branch. Due to the nature of a noninverting branch to exhibit some interesting properties of current being dependent upon several other things, including the noninverting voltage, in effect the impedance of an inverting branch is lower than the input series resistance. You may fix this by using instrumentation amplifier, or make the opamp truly differential. I may have a solution for that, and I already put my solution on Bandido 2 topic, but here it goes again, a LTspice file with an accompanying picture, and some solutions for various gains and a total input impedance ~1kohm. There are several gains with "nice" resistor values, but many more with "naughty" ones, so have it your way.

You are free to play with this solution as ever you like.

Just a note, CMMR gets spoiled a bit when there is a series resistance of a source added to the calculated values. To maintain the best CMMR it will be OK to introduce Mr. Thevenin to the equation and transform that dumping resistor into two series ones, and substract that value from the opamp ones. Just a thought. (instrumentation amps do that without any math)

[mickstv]

Hi Altra, good circuit but I think it may still have a problem. If you look at your first cro shot and the top green waveform the first and third pulse looks ok but the middle one should be an exact opposite to the other two. Like what mine looks like.



Mick

[Altra]

Mick
That inverted flyback spike is happening in the differential amp. When metal
is near the coil it moves in the right direction. The scope hooked directly to
the coil looks ok. The pulses are not square because I'm using current limiting.

I got excited over your results. So I ripped the h-bridge circuit apart and put togeather
something like you are doing. It works great so far. I used an LM7805 and 2 1000uf caps to form a virtual ground. The ground side of the coil to the LM7805 and the other end to the drains of a p and n ch fets. The fets are switched between 0 and 10v. Am I close?

regards

[mickstv]

Hi Altra, your close but no cigar. I'm only using N-Channel fets to make mine work. The TX is running 0 to 6volts and the flyback is peaking at around 200 volts at about 70us duration.

With the gain mentioned in one of my earlier posts and still only a target channel running I'm detecting a 1 ounce lead weight at about 35-40 cm, hopefully deeper when I finish the rest of the gain stages.



Mick

[Altra]

Thanks, to Mick's outside of the box thinking. I revisited the Dave Johnson halfbridge transmitter. This is decribed in the patent US4,868,504. Dave was way ahead of the times. In my opinion,(fwiw) this patent is miss titled. When you get past the energy
saving aspects of the regenerative halfbridge driver. He then describes
generating different step functions and demodulating fundamental and harmonic frequencies to eliminate resistive and reactive ground minerals.(dual frequency). He also descibes time domain sampling as applied to PIs.

Anyway heres what I came up with, it can be improved with proper fet drives
and I going to replace the LM7805, with an Analog Devices OPA547(ebay) high current op-amp. This will allow the virtual ground to track the supply rails. This circuit can also generate a combination of continuous wave and PI TX signals to create a true hybrid detector using a balanced IB loop and a filter or two.

[mickstv]

Hi Altra, thats a good result, now the only problem left is the P-channel limitations. Here's a couple of cro shots the top one is on the coil showing about 280 volts on either side although my cro wont show it. The tx is running at about 100us. The other shot is just of the cleaned up output of the clamping diodes. I'm still using a 6 volt power supply to run it at the moment but will be making up a 7.2 volt pack and a DC-DC converter to run the opamps etc.

[Altra]

Hi Mick,
That looks really good. Hope you continue to improve on the circuit. What kind of current
are you drawing with the latest version?

I tried 2 n-ch fets in my circuit above and had problems. For now I am happy with my results for a low power pulse. It draws about 100mA with +/-120us pulses and a ne5532 opamp. No dc-dc required. My schematic above has an error the source and drain are reversed on the IRF640. The power opamp mentioned above is made by TI not analog devices.

Best regards
Mark

[mickstv]

Being an former TV tech I happen to have alot of spare parts old boards etc that I have been wrecking for the FET's and other parts, but anyway here is a list of results relating to two slightly different spec'd coils.

FET COIL TURNS DECAY VOLTS AMPS

2SK3911 VER1 17 5us 350v 284ma at 6 volts
2SK3911 VER2 18 10us 280v 219ma at 6 volts


GP70N33 VER1 17 5us 280v 256ma at 6 volts
GP70N33 VER2 18 10us 240v 193ma at 6 volts

Coils are both 25cm and just bundle so nothing special and two different types of wire.



Mick

[mickstv]

I decided to revisit my first design which Altra roughly figured out, whilst I wait for some more parts for the all N channel version.

Below is a picture of the rough prototype with the P/N ch TX.

Two samples on each of the P/N pulses, it has perfect cancelling of Earth Field which was setup using a magnet then waving the coil in air to prove it really works. Not sure if the ground balance is working correctly, the only test I have done is waving a ferrite rod out of an old radio over it and at a certain point the ferrite rod cancels completely whilst other lead targets are still received.

For anyone to answer if I can cancel a ferrite rod out completely does this mean it might work ok in mineralised ground or am I missleading myself.

[Tepco]

For anyone to answer if I can cancel a ferrite rod out completely does this mean it might work ok in mineralised ground or am I missleading myself.

May or may not work, test with ferrite or permanent magnet is inconclusive due to number of reasons. From waveforms posted earlier, it is not clear what minimal sampling time you can achieve (opamp output). If this time is too long, say 20uS or so, detector will not respond to ferrite anyway, fast component will decay in few uS, only with fast sampling, say <10uS or even faster, this component can be problematic.

Permanent magnets are not relevant either, field is much stronger compared to anything that slow magnetic component in any soil can produce, induced voltage may or may not spoil detector operation, depending on design. Magnets itself are problematic, being almost completely saturated, especially stronger ones, they have very small permeability and aside static field they generate, detectors can see them almost like nonmetallic objects(!)

Fighting Earth magnetic field itself is not on agenda, static field whit almost no gradient is harmless, as posted before.

Bipolar pulsing alone will not eliminate fast and slow mineralisation effects and achieve full ground balance. To fight this, something else is needed, well known patents from some guy and some company, having probably more patents than Edison himself.

I know I’m boring now, this may not be completely technically correct, but just simplified version and goes like this (corrections are welcomed, excuses and explanations will come later)

Fast Fe component is generated by absorbing part of PI pulse energy to align magnetic domains rapidly in fast changing field, energy is then released quickly, in few uS, making effective brick wall for fast sampling detector, with very fast sampling it is not possible to distinguish this response from actual target response, secoundary sample will not fix this. Slow component just remain after strong TX pulse, part of material will remain magnetized for at lest some time after it.

Way to combat this is in essence of PI method. Target excited by long enough TX pulse will have some decay time, defined by conductivity and size, usually called L\R ratio. Now received signal will contain target response and response from ground mineralisation combined. But trick works in opposite direction too. Short pulse, containing about the same energy as long one, will almost fully excite mineralisation components, without significantly exciting target, just more time is needed to ramp up some eddy current in conductive targets. In effect, subtracting two samples, one containing target and mineralisation response (long pulse) from another, effectively containing just mineralisation response, almost no target component, target is now weakly excited, is basic way to provide ground balance. In any possible way or incarnation or design used so far.
This is a
bit long writing, sorry, but may be interesting to someone, please don’t nuke me for this oversimplified explanations, maybe topic like “poor man’s DIY ground balance “ can be more appropriate, I will share any data, schematic, test result or scope pics if anyone is interested.

Goal of this is not to make Minelab obsolete, but just to build simple detector, not much more complex than SMPI, able to achieve fast (<8us, or even better, <4) sampling AND maintain ground balance in mineralized soil or salt water simultaneously.

BTW, I made lot of tests using ferrites and real soil samples, but still not figured out how these things are actually affected by salt water, more measurements in simulated conditions must be taken. Any suggestion is welcomed, don’t spit me too much for this, any data,schematic, design etc. is available. Best regard

[mickstv]

Hi Tepco, thanks for the reply, the previous cro pictures were of the all nchannel version, the version in my last picture is a P/N version it's first sample is being taken at about 8us and the second sample is taken approx 35us after TX. The ferrite rod shows up in both primary target and secondary GB sample's. This particular ferrite rod must have a high iron content because magnets are attracted to it and I could balance it out to show no response from the output of the last diff integrator.


I guess the only way to really test it is to finish the audio stage box it up and go out to the goldfields and try it.



Cheers
Mick