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

[Aziz]

Hi all,

I know now, which foil capacitors I need:
A big bunch of various FKP2 (WIMA) capacitors. Available from 33 pF to 33 nF at different voltage ratings.

And I have found a small mono coil with true HF litz wire in my box for testing (~600 µH, ~1.8 Ohms, 20 cm diameter). It isn't shielded yet however.

The outcome of my brute force mono coil LTspice analysis is:
Let's say, we have 200 g copper wire for the TX coil.
Our reference: Take the wire as is. TX coil will have L inductance, N turns count, R resistance, 200 g copper.
Comparing against
- We take the wire and use it as two strand wire: We get L/4 inductance, N/2 turns count, R/2 resistance, still 200 g copper
- We take half thinner wire, double the turns count 2*N, we get 4*L inductance, 2*2*R = 4*R resistance (thin wire has double resistance), still 200 g copper

For a constant weight of TX coil (x amount of copper), use the wires to increase the coil inductance and resistance instead of lowering the inductance and resistance.
If the TX current gets too much, the (ohmic) losses dominate and the Q decreases and hence less target response. You will have temperature drifts perhaps too (heat).
So use the amount of copper to increase the TX coil inductance.
A TX coil inductance above 300 µH is ok (minimum). Better go for more inductance.

The second argument: Losses on the tuning capacitors
High value capacitors have much higher losses. Particularly on high frequency usage.
If we want to use low value tuning capacitors, we have to use high inductance TX coils. But this is most desired due to first argument.
Another argument for low value FKP2 capacitors is, that they are compact and relative small. Perfect for placing in coil housing. And also perfect for paralleling them.

This is the free lunch.
Aziz

[Aziz]

Hi all,

if we take the energy losses into account (TX coil, capacitors), yep, there is a nice & tasty free lunch available. Indeed.
The brute force method reveals it.
Yummy!

Skin and proximity effects of the TX coil is not taken into account yet.
I am still thinking of either buying the poor man's litz wire (the cheap tinned litz wire) or the true HF litz wire (3 - 5+ times more expensive).
With HF litz wire, the single mono coil would cost me approx. 100 bucks. No go.
Aziz

[Aziz]

Hi all,

I am looking for a free lunch. Free lunch is always yummy. Yummy yummy!
So what is better for a mono coil?

At same copper weight (TX coil wire weight):
Low inductance -> bigger tuning capacitors (bad), low EMI noise induction (good), lower coil resistance (good)
High inductance -> smaller tuning capacitors (good), higher EMI noise induction (bad), higher coil resistance (bad)

More copper to the mono coil: double weight
investing in higher coil inductance -> 4*L, 2R, 2N turns count
investing in lower coil resistance -> R/2, L, N turns


The brute force method with LTspice simulation will reveal this in regards to target response. But I don't know, what the AI will say. Perhaps a lot of nonsense.
Aziz

[Aziz]

I am really short of good & low loss, low capacitance, high voltage rating, small FKP capacitors.
I have to spend at least 100 bucks for a good variation of the required FKP capacitors.

And I have to spend at least 100 bucks for more tinned litz coil wire (1 mm², 0.75 mm² and maybe 2.5 mm² too).
The skin effekt degrades my Q-factor with normal copper litz wire. It can not be neglected anymore.

And the resonant capacitor Ct needs to be multiple parallel connected capacitors (2 - 4) to lower the ESR of the final tuning capacitor further.
Aziz

[Aziz]

Hi all,

AI can do Win32 GDI API functions (Graphics Device Interface Application Programming Interface).
Maybe I should fully specify a function for the AI to code me a nice dI/Q graphics output in C.

Writing the specification (AI prompt) could take almost the same time to rewrite and improve the old quick & dirty dI/Q graphics output however.

---
Now the big answer to the discrimination issue with more VDI's.
No. It won't really improve discrimination of course.
The discrimination would discriminate time constants (TC) of targets only.
The TC is a function of target conductivity, size and orientation. Too many unknown variables to make reliable discrimination.

Aziz

[Aziz]

Ok,

I have looked at the decoded numbers (dI/Q-vectors) at mono coil.
dI/Q-vectors are calculated using two EMA filters (exponential moving average) with long and short time constants.
Long time constant EMA has a smaller alpha (averaging over more time period) and short time constant EMA a bigger alpha (effectively averaging over a few samples only). You also could take the current decoded I/Q vectors instead of short time constant EMA.

The dIQ vector is the I/Q vector subtraction of I/Q(EMA_long) - I/Q(EMA_short). We don't leave the I/Q-vector space. All being done in the 2D vector space.

The dI/Q vectors show at frequency f1 and f2 each consistent collinear X-response.
-> Either the projection method or the Möbius (Cayley) - transform will work fine.
The projection can be made with I/Q vector algebra calculus (using Dot product). So we don't need any phase information nor rotation.

It will work fine. With GB. With two VDI's (for frequency f1 and f2).

How about 4 VDI's in the IB-Coil configuration?
Did you ever see a detector with 4 VDI's?
Would this improve discrimination?
To be or not to be. This is the question.
Aziz

[Aziz]

BTW,

regarding IB-Coil configuration:
it makes fully sense to use a concentric IB coil. The TX coil acts like a full sized round mono coil. So we get good target signals out of the TX coil too. In addition to the IB RX coil.
I don't think, that any VLF detector on the market did this before.

We can get the most powerful and state of the art VLF detector ever.
The simple KISS VLF detector controller, built-in the coil housing, with sound card and processing unit, should outperform the other comparable VLF detectors with ease.

Aziz

[Aziz]

Hi Sergey,

Hi Aziz
> you can't distinguish it from real target<
The difference is easy to spot... by the dominant phase of the search signal, and that is the phase from the ground...

I haven't been able to make the decoded signals as a nice dI/Q graphics yet (due to lack of time).
The easiest way to process the responses on two frequencies f1 and f2 is: (mono coil)
To make two independent dIQ projections to the ground response dIQ-vectors. So individually for each frequency f1 and f2.
And adding two projected R-responses together.
A target response threshold level, which will filter out conductive grounds (all signals below threshold level) and pass the real target responses (above threshold level).
This should work fine.

And for IB-coil type:
4 independent dI/Q projections of course.
And we get the target signals out of the TX coil too.

Well we will see it.
Aziz

[Sergey_P]

Hi Aziz
> you can't distinguish it from real target<
The difference is easy to spot... by the dominant phase of the search signal, and that is the phase from the ground...

[Aziz]

Hi Aziz
>There is absolutely nothing we can do except introducing a threshold level for target response.<
We can do this, and to do this we need to measure the current signal from the ground as accurately as possible and subtract it from the input signal, minimizing all destabilizing factors rather than creating them with a high-Q resonant circuit.

No, any change of the conductivity of the ground will lead to same problems.
The main point is:
If the ground acts like a target, you can't distinguish it from real target.
In this case, real targets must cause more R-response than the ground R-response to be detectable. The threadhold level does this.

[Sergey_P]

Hi Aziz
>There is absolutely nothing we can do except introducing a threshold level for target response.<
We can do this, and to do this we need to measure the current signal from the ground as accurately as possible and subtract it from the input signal, minimizing all destabilizing factors rather than creating them with a high-Q resonant circuit.

[Aziz]

Hi Sergey,

Hi Aziz
A change in the distance between the ground surface and the sensor leads to a change in the quality factor of the resonant oscillatory circuit (if the quality factor of the resonant circuit and the quality factor of the soil are not equal), which, in turn, causes a change in the slope of the phase-frequency characteristic of the resonant circuit and a change in d(I/Q).

The LC-tank with high Q is very very sensitive to magnetic property change of the ground (causing huge X-response). This causes a change in inductance of the coil. And hence a change in resonant frequency. But we are stimulating the TX coil at two different frequencies (f1, f2). The inductance change will either increase f1 response and decrease f2 response or vice versa.

Real target response is different to magnetic property change of the ground.

If the ground is conductive (resistive ground response) it will act like a real target.
There is absolutely nothing we can do except introducing a threshold level for target response.

Aziz

[Sergey_P]

Hi Aziz
A change in the distance between the ground surface and the sensor leads to a change in the quality factor of the resonant oscillatory circuit (if the quality factor of the resonant circuit and the quality factor of the soil are not equal), which, in turn, causes a change in the slope of the phase-frequency characteristic of the resonant circuit and a change in d(I/Q).

[Aziz]

Hi all,

I have made the modification to the latest mono coil VLF bread board. It is working very nice.

DSP path:
Differential input: Signal = right-channel - left-channel (or vice versa)
Demodulation of Signal at frequency f1 -> I/Q vector 1
Demodulation of Signal at frequency f2 -> I/Q vector 2
Demodulation type: Goertzel or Lockin

So a two I/Q vector processing is enough. As there isn't any transmit reference anymore, we are working with changes of I/Q vector (dI/Q).
This detector handles even the heavy ground mineralisation. Very robust operation.

For most sensitivity and detection depth, we need the IB type circuit. We can process the TX coil signals (as mono coil VLF) + RX coil signals. At two frequencies. We can refer to the TX coil signal (phase relation), it has huge TX power, etc., etc., .. and it handles heavy ground mineralisation too.

Aziz

[Aziz]

We have a good IB-VLF base now.

Now back again to pure mono coil VLF:
- It must have high Q.
- Less power drain. High power efficiency.
- Must work with standard line output level (1 V rms) and with every sound card this time.
- KISS & cheap design.
- All parts in the coil housing.

I know, the circuit can be made much more elegant this time.
We need a KISS masterpiece.
AI isn't smart enough to help me.

But I'm working on this right now:

Differential input mono coil. So we are processing both input channels at two frequencies now.

And I have found a real bug in LTspice in AC analysis.
I had to destroy the symmetry of the circuit by adding 1 pF to C1. It helped to override the bug.

Cheers,
Aziz