Induction Balance Stuff - Single/Multi Frequency Response, GB, Disc, Measurements, Ideas, Fun, etc.

[Aziz]

Hi all,

there might be a very tricky solution to the huge ground reponse on mono coil vlf with active nulling however.

If we would expect 1 .. 2 % TX inductance increase due to heavy ground mineralisation, a chirp modulation of the TX driving and nulling signal from Frequency f1 to f2,
where f2 is the resonant frequency with 0% TX inductance change and f1 the resonant frequency with maximum expected TX inductance change (1 .. 2%).

It could be worth to test this idea. This is the last chance for mono coil vlf with improved performance.
Aziz

[Aziz]

Well ok,

the nulling won't work on chirp modulation. Too complex issue.
The TX voltage varies still too much for 0%, 1% and 2% TX inductance increase.

It is really nice, that LTspice has a modulate module to make chirp modulation possible. The AI told me how to use it.

Aziz

[Aziz]

Hi all,

well ok. The problem with the high impedance signal path has been solved and there is no j-fet amplifier required anymore.
Also the mixer has been radically simplified. It's really a KISS masterpiece now.
The mixer has been reduced into one resistor (R1).

Provided that, the ground mineralisation isn't much, the following basic LC-tank with active nulling can be used and the signal source can be directly fed to the input of the sound card.
It is now a quite low impedance path (impedance aprox. R1/2, so much lower than 10 k Ohm).

Aziz

PS: And no external battery required anymore too.

[Aziz]

BTW,

the last mono coil VLF with active nulling has some advantages:
- Transmit energy balanced to two channel outputs (left, right).
- Transmit power doubles (both channels used) -> high coil current, high coil voltage. We can push up the transmit power if required by increasing capacitor value Cs (1 nF).
- Easy to find resonance frequency and easy to null:
1. Set nulling VL channel to 0 V (no signal), transmitter works with half power
2. Adjust VR channel frequency for resonance, max Lin-Signal -> resonance frequency found
3. Set the found resonance frequency to VL channel and adjust voltage level and phase shift (180 + x) until Lin-Signal is minimum (near 0 V). Some residual signal level is required. No need to perfectly match to 0.000000 V.
4. Ready to go.

I have to build a new bread board to test the circuit. I may add a voltage divider for the TX reference signal (Rin).

Aziz

[Aziz]

Hi all,

this is becoming very very interesting now.

A true two channel VLF processing is possible in conjuction with high TX current (up to 500 mA peak) and IB-coil (with IB RX coil).
Channel 1: TX coil processing (at the nulling node, with some residual signal level).
Channel 2: RX coil processing.
As we can't take the TXref signal, we are referencing to the internal generated reference.
There should be enough ground balance information to cope with high ground mineralisation. Well, I hope.

[Aziz]

...
and this will allow us to operate the detector at different modes:

Mono coil plugged:
-> mono coil VLF

IB-coil plugged:
-> Mode 1: mono coil VLF (TX only)
-> Mode 2: IB-coil VLF (RX only)
-> Mode 3: mono + IB-coil VLF (TX, RX, dual channel)

[Aziz]

Hi all,

I have replaced R1 with a capacitor C1. It is more power efficient and should not produce temp drifts.
Same principle.

[moodz]

Aziz ... you are chasing the wrong method ...read the relevant minelab patents. tempco is not the caps its the coil itself.

The static null holds only at one operating point because Vn and Ph are open-loop constants. Ground loading perturbs the coil in two ways at once: soil susceptibility adds reactive ΔL (shifts tank resonance) and eddy-current loss adds ΔR (drops Q). Both move the TX-current phasor that the null was trimmed against, and nothing in the circuit chases it.

The damaging part is the Q multiplication. For a fractional inductance change δ = ΔL/L, the resonance moves by ½δ in frequency, and near resonance the phase slope is ~2Q, so the TX current phase rotates by roughly

Δφ ≈ Q · (ΔL/L)

Your .step p list 1.0 1.01 1.02 is precisely δ = 0%, 1%, 2%. At Q = 100 a 1% ground-induced ΔL is already ~1 rad ≈ 57° of phase walk on the feedthrough — the null is simply gone. Even a modest Q = 20 gives ~11°, far beyond what any fixed Vn/Ph can hold. So the very thing that makes the tank efficient (high Q) is what makes the static null fragile against ground.

Tempco hits both axes independently: Ct and the former's CTE drift the resonance (so more Δφ), and copper's +0.39%/°C drifts R, hence Q, hence the feedthrough amplitude — ~12% over a 30°C swing, directly an amplitude null error since Vn is fixed.

[Aziz]

Hi Paul,

Aziz ... you are chasing the wrong method ...read the relevant minelab patents. tempco is not the caps its the coil itself.

The static null holds only at one operating point because Vn and Ph are open-loop constants. Ground loading perturbs the coil in two ways at once: soil susceptibility adds reactive ΔL (shifts tank resonance) and eddy-current loss adds ΔR (drops Q). Both move the TX-current phasor that the null was trimmed against, and nothing in the circuit chases it.

The damaging part is the Q multiplication. For a fractional inductance change δ = ΔL/L, the resonance moves by ½δ in frequency, and near resonance the phase slope is ~2Q, so the TX current phase rotates by roughly

Δφ ≈ Q · (ΔL/L)

Your .step p list 1.0 1.01 1.02 is precisely δ = 0%, 1%, 2%. At Q = 100 a 1% ground-induced ΔL is already ~1 rad ≈ 57° of phase walk on the feedthrough — the null is simply gone. Even a modest Q = 20 gives ~11°, far beyond what any fixed Vn/Ph can hold. So the very thing that makes the tank efficient (high Q) is what makes the static null fragile against ground.

Tempco hits both axes independently: Ct and the former's CTE drift the resonance (so more Δφ), and copper's +0.39%/°C drifts R, hence Q, hence the feedthrough amplitude — ~12% over a 30°C swing, directly an amplitude null error since Vn is fixed.

I know, I will get huge headache with it.
But I am planning to look at a narrow bandwidth chirp modulation and see what happens. Even I can't null it.
So away from single frequency to multi-frequency processing.
I am curious.
Aziz

[Aziz]

Hi all,

my new breadboard is ready to fire it up.



On a single frequency resonance, it is the most sensitive hot ground/ferrite detector ever.
The circuit simulation did tell it.
The nulling works perfect and I can achive 0.000000 V on my virtual ground node.

Instead of nulling, a dual frequency mode is more interesting. BTW, there is no need for chirp modulation. Two frequencies are enough.
Left channel: Frequency f1
Right channel: Frequency f2
f2-f1 = 100 Hz (or so)

The two frequencies deliver enough information to cope with ground balance and target response processing.
The sensitivity is less than compared to single frequency response (the transmit spectral energy is distributed on two frequencies).

Not bad for the simplicity of the design (total 3 capacitors and a mono coil TX, I'm using two 15 nF parallel resonant capacitors in the picture above).
The dual-frequency detector can be realised on a mono coil. But it can't concur with other detectors, which avoid much larger x-response.

Aziz

[Aziz]

Hi all,

so lets go to the ultimate ground breaking dual-frequency KISS detector solution.
You will be satisfied with the detection depth and high mineral ground handling.

Instead of operating the detector with a mono coil, we are going to use an IB-coil.
Operated on two frequencies f1 and f2.
We have 4 I/Q vectors for processing:
- IQ_TXf1,
- IQ_TXf2,
- IQ_RXf1,
- IQ_RXf2
We just add two protection diodes for the RX coil.

This is the final schematics:

You will really get a good detection sensitivity and depth with it.
And there is no need for an amplifier. There is enough target signal there.

But the detector software is tricky. All the power lies in the coding.

Aziz

[Aziz]

LTspice file (zipped)..

[Aziz]

Hi all,

the circuit needs to be optimized further.
In order not to attenuate the line input level in the sound card much (due to big TXref and small RX coil signals), the TXref signal needs to be divided by a simple resistor voltage divider. So two resistors will be added further. Thats all.

This will be my ultimate ground breaking VLF-IB detector.
It will even operate on pure iron ore grounds (severe ground mineralisation).
Note, that we are not fixed to the in-air resonant frequency. We can adapt it for severe ground types. And the frequency distance (df = f2 - f1).
So it will cover some ground mineralisation variation too.

And we can push up the TX power as we have the voltage divider for the TXref.
The Sound BlasterX G6 can deliver at least up to 100 mA current @ 7 Vp output for each channel (but it gets hot).
We can have way above 100 Vp TX voltage. But this is really not required.

A source current of 20 - 50 mA is enough. Multiply this with Q and we have the TX coil current peak.

Aziz

[Aziz]

Hi all,

BTW, the high sensitivity of the TX LC-tank to ferrite/hot ground is fully intended. It is not a bug, it is a feature! We get very valuable GB info.

So how can we still use a mono coil TX? Without induction balance?

This is really trivial.
We could use a quasi mono coil design.
Round main mono coil as always used before + a few turns smaller concentric co-planar secondary coil, fed into the other channel.
The second coil has less coupling coefficient (k < 0.1) and hence will induce less voltage. But enough to process the signals.

The hot ground mineralisation will modify (increase) the coupling coefficient k. There we get the GB info. We can remove the ground response content in the main mono coil signal.
Of course in dual-frequency mode. 4 I/Q vector processing.

The quasi mono coil is quite easy to build and is not critical.
Aziz

[Aziz]

Hi all,

I have now the decoded 4 I/Q vectors visible. I'm using the Goertzel as it is the most efficient.

All the ground and target information is there now.


I have observed a nice and distinc response change on the intermodulation frequency (in my case 100 Hz).
Especially on the envelope curve of the intermodulation frequency.
The X-response differs from R-response.

One possible way to process a single channel:
Hilbert Transform -> Envelope -> FFT of Envelope

FFT delivers the frequency spectrum of the envelope. Note, that the FFT is being used for the very narrow band intermodulation frequency. We get a much finer frequency resolution. Whole spectrum response in the 100 Hz bandwidth is visible.
Using only the magnitude spectrum of it, we can use the linear combination method to get rid of the ground response (simple processing variant).
We also could use phase spectrum too. Or an array of complex I/Q vectors.

But this isn't really required.
The decoded I/Q vectors deliver enough information for simple processing too.

Aziz