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

[Aziz]

Hi all,

I have wound the signal transformer finally. It was quite difficult.
Voltage step up of 1:10 (16 turns -> 160 turns).
I have a very good coupling coefficient k and very low stray inductance (1.7 µH, k=0.9997). This is excellent.
I have divided the secondary coil windings into 8 sections with 2 mm gap. 4 mm gap for secondary coil endings. This should make the coil capacitance low and the bandwidth of the signal transformer high.

But I have to test it yet.
Aziz

[Aziz]

Hi all,

if the signal transformer does cause too much problems (temp drift, low bandwidth, low voltage transforming, etc..), I will go to the second best option without any transformer.
Doing the chancellation with 6 V only. The TX coil voltage will be limitted to 30 - 60 V. So the mixer would attenuate the target signal 5 to 10 times.
And we have the j-fet amplifier to compensate the target signal attenuation. A gain up to 10 times would be enough.

I am sure, it isn't really required to wind the signal transformer. It is difficult and expensive part.
Anyway, I'll test both options.

Let's tweak the j-fet amplifier and let the AI expressing its opinion. Claude AI does support LTspice files too.
I have to build the j-fet amplifier part next time.
Aziz

[Sergey_P]

Hi, Azis!
You were looking for something unusual in this thread—and I happen to have exactly what you need: inductive balancing devices + a bipolar square-wave AC signal + real-time target signal analysis.
The device features low power consumption, is independent of (or depends only negligibly on) the input signal level, enables target signal detection against a noise background, and boasts high resolution (dT = 1 µs).
The signal received from the target "colors" the noise of the broadband amplifier (which ensures the 1 µs resolution); subsequently, statistical methods and a correlator are used to extract and display the probability of the useful signal's presence. In this prototype—unlike the background noise—the useful signal is encoded using a phase-modulated M-sequence 255 bits in length. At present, no analogues to this type of detector are known to exist.

Top to bottom (time scale: 1 µs to 64 µs):
1. Sensor
2. Aluminum
3. Copper
4. Ferrite
5. Iron
6. No targets present
Distance to target: ~15 cm

[Aziz]

Hi all,

my 1:10 signal transformer is working fine (above 3 kHz). Below 3 kHz, signal distortion and high current protection circuit triggering happens (connected it directly to the line output without any current limitting feature).
Bandwidth is high enough: Tested up to 86 kHz with a 220 pF capacitive load without any distortion or resonance effects.
Above 86 kHz, the sound card produces more phase noise and low output voltage.
The output voltage at high-gain mode is 70 V peak with 220 pF load (+/- 70 V). Input is 7 Vp (+1 V more ).
So we can go with the TX voltage up to 100 Vp or even more.

I will do the simple primary current limitting resistor (10 - 47 Ohm) in series with the signal transformer to be in the save region. This is the cheapest solution.

I should test the signal transformer with 8 turns for the primary coil for 1:20 voltage boost. Let's look, what happens. (Best case: TX voltage up to 200 Vp).
And I should measure the ferrite core loss / magnetisation current.

The mixer capacitors need to be FKP1 and high voltage (min. 400 V=).

Cheers,
Aziz

[Aziz]

Yeah,

it also works perfect with 8 turns for the primary coil. Each coil section has 1 turn primary coil winding now.
7 Vp input voltage is being transformed into 140 Vp (280 V span!). 1:20

But for above 12 kHz.

Below 12 kHz, the distortion happens (delivering not enough primary coil current). Too low primary coil inductance too.
I will use frequencies between 24 - 48 kHz and this isn't going to be a serious problem.

It seems, that we can get almost all of the target response out of the mono coil in VLF mode.

I have to test the thermal conditions yet. But I think, it will cause temp drifts at extreme levels.
We will see it.

[Aziz]

Oh man!,

Google AI (Gemini) does make so much serious errors in schematics analysis and design.
Claude AI didn't give me any answer and has reached the time limit.

The j-fet amplifier design must be finished by hand.
I will raise the drain current to 3 - 4 mA as it will operate in a lower noise mode and the power consumption is not much for a 9V block battery (it is a compromise only).
Gain of 4 - 8 would be enough.
Aziz

[alex.drozdov]

Удельное сопротивление Грунта (чернозем, суглинок...) в зависимости от влажности меняется в очень широких пределах(10...150ом/м). Примем что-то среднее =75ом/м

На катушечном калькуляторе с размером катушки 150мм имеем примерно
137витков, провода 0,25мм, Rкат=24ом>>

Rac changing with frequency, 75ohm/m and 24ohm can used only for DC.

[Sergey_P]

Hi, alex
Changing the signal frequency changes the impedance due to changes in the reactive component's resistance. The active component (R) is independent of frequency.
Energy is expended on the active component (R), so matching between the source and load occurs at the R level.

[kbsurmeli]

Aziz abi merhabalar,

Amatör bir dedektör hobicisi olarak, bu işin donanım tarafına hakim insanlarla sohbet etmek ve sorularımı yöneltmek istiyordum. Sizle daha kolay konuşabilmek adına bir iletişim adresiniz var mıdır? Numara ya da mail.

Teşekkürler.

[Qiaozhi]

Aziz abi merhabalar,

Amatör bir dedektör hobicisi olarak, bu işin donanım tarafına hakim insanlarla sohbet etmek ve sorularımı yöneltmek istiyordum. Sizle daha kolay konuşabilmek adına bir iletişim adresiniz var mıdır? Numara ya da mail.

Teşekkürler.

Please read the forum rules -> Basic Rules of the Forums
and make your posts in English.

[Aziz]

Hi all,

I am quite confused.

I have found a serious bug either in LTspice XVII(x64) (17.0.37.0) or in my coil software.

Following situation:
Coil simulation for target response: TX -> Target -> RX
Simple circuit simulation: giving L1 (TX), L2 (Target), L3 (RX) = L1 and coupling coefficient k (L1, L2), same k for (L2, L3)

The interesting fact is, that the AI and my coil software delivers same results.
My coil software must be halucinating too. Maybe I am hulucinating much more.

LTspice simulation is delivering exactly 100 times less target response. WTF!, what is going on?


I have to fix this issue.

[Aziz]

Ok, I have found the bug in the system.

No real bugs! LTspice, AI and my coil software is working fine.

Wrong parameter. I have forgotten to set the proper target coils resistance.

Everything is fine now.

I can simulate a mono coil VLF model to check, what we can expect target responses. This will help to set the j-fet amplifiers gain.
Aziz

[alex.drozdov]

Hi, alex
Changing the signal frequency changes the impedance due to changes in the reactive component's resistance. The active component (R) is independent of frequency.
Energy is expended on the active component (R), so matching between the source and load occurs at the R level.

Hi, Sergey_P!
As i know Skin and proximity effects increase active resistance with increase signal frequency.
Classic MD coil is tight wounded with many turns so this coil should have big difference of active R on DC and AC on khzs , for freqs around 20khz skin effect not too much but proximity effect should give great increasing of active R. On 50khz skin and proximity should give great increase of R.
More active R - more voltage need to give enough current for enough magnet flux, more voltage is sign of increasing real impedance of TX coil and electricity part of EM field generated more in this coil, if compare with some current but on low acitve R coil..
Last coils from minelab equinox e.t.c. has very low turns coil and with very low active R (litz on rx and tx ) I think in minelab it did for good matching coils with low impedance soil (salted soils, highly minerilized soils in australia ?)

skin and proximity effects should be applicable for soil too.

[Sergey_P]

Hi, Alex
There are changes, but not at the detector frequencies

[Aziz]

Hi all,

after making extensive circuit simulations, the active nulling on mono coil vlf won't work on heavy ground mineralization.
The ground mineralisation response would be of order 2, 3, or 4 and more (order = 10^x) compared to target response.
This would overload the pre-amp input. Active nulling does not make sense anymore.

It would be possible to realise the idea on mild ground however. To some level of ground mineralisation of course.

3 lecture lessons learned:
- How to design and wind very good signal transformers
- Checked my coil software: It is calculating correct responses.
- Need to do more circuit simulations: It saves a lot of time.

Aziz