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

[Aziz]

How about a figure-8 mono coil?
It gets rid of the EMI-noise by design.

[Aziz]

Google AI for noise prediction and elimination solution.
I haven't tried Claude Code AI yet.

[Aziz]

Next time, I will specify a query for Claude Code AI,
- with the fully specified measurement environment
- describing the Gaussian and Flicker noise behaviour
- to ask it, to remove the unwanted noise
- and combining the cleaned signal with lock-in amplifier output

This is so nice.

[pito]

So lets look at different target responses,

In your program, can generators be replaced by Ux and Uy applied to microcontroller inputs ?

[pito]

Let's look at the pure X-response by placing a ferrite .

should be 90 deg

[Aziz]

Hi pito,

should be 90 deg

Referred to what?
Remember, this isn't a standard IB-coil system, where we can refer to the TX-coil's phase. It's a mono coil VLF and we have only the TX-coil itself. We have to make our own reference system for I/Q demodulation. And we have to process the I and Q changes only.

To make a reference system for the IQ-demodulation, we are internally generating a reference I and Q decoding signal using a lockin-amplifier. By just adjusting the phase of the reference of the I/Q demodulator, we can do the ground balance at once (doing the rotation to the I-axis). The phase angle can be obtained by the angle of the IQ-vector at ground balance procedure. So we are rotating back the demodulated IQ output to the I-axis. But we don't use the IQ output itself. We are using the tiny changes referred to a reference IQ-vector. The change is the dQI vector.
The Q-part (quadrature-phase, imaginary part) of the dIQ-vector is our ground balanced target response (R-Response).
The I-part (in-phase, real part) of the dIQ-vector is our X-Response.
The phase angle of dIQ-vector (0..180 °) can be used for VDI calculation.

I have shown the proof of concept. It is working similar to standard VLF. But it's not the same.
Cheers,
Aziz

[Aziz]

Hi all,

I didn't try the high-power version for the true IB-coil system yet.
I have to order some appropriate ring cores for the choke and low voltage FKP1 capacitors, which are relative small compared to the high voltage ratings. Note, that High-Q requires best foil capacitors with very low losses. Metalized film/foil capacitors won't work well.

The high power version makes use of the High-Q LC-tank. Feeding low current in, to make big current flow in the TX-coil. A Q of 20 - 80 is quite realistic.
A TX-coil current of +/- 500 mA - 1 A should be enough, to get reasonable RX-signal in the RX-coil. We would not require an amplifier yet.

The IB-system would deliver the best sensitivity and depth. If we process both TX and RX channels, we would improve it further.
Aziz

[Marchel]

Hi all,

I didn't try the high-power version for the true IB-coil system yet.
I have to order some appropriate ring cores for the choke and low voltage FKP1 capacitors, which are relative small compared to the high voltage ratings. Note, that High-Q requires best foil capacitors with very low losses. Metalized film/foil capacitors won't work well.

The high power version makes use of the High-Q LC-tank. Feeding low current in, to make big current flow in the TX-coil. A Q of 20 - 80 is quite realistic.
A TX-coil current of +/- 500 mA - 1 A should be enough, to get reasonable RX-signal in the RX-coil. We would not require an amplifier yet.

The IB-system would deliver the best sensitivity and depth. If we process both TX and RX channels, we would improve it further.
Aziz

I have analyzed the coil of the XP Deus 2 metal detector and the TX circuit draws approximately 30-40mA and the range to a small coin is approximately 30-45cm. I think that the current of 500-1000mA is quite excessive.

[Aziz]

Hi Marchel,

I have analyzed the coil of the XP Deus 2 metal detector and the TX circuit draws approximately 30-40mA and the range to a small coin is approximately 30-45cm. I think that the current of 500-1000mA is quite excessive.

yep, it looks like quite excessive.

But we have the magic High-Q feature.
Let's say, we have Q=50 and a TX-coil current I(Tx) of +/-500 mA.
I(in) = I(Tx) / Q = 500 mA / 50 = 10 mA.
We are feeding the LC-circuit with 10 mA only. This is the High-Q-trick. No magic here.

On the otherside, we need as much TX-current as possible to have enough target stimulation on the RX-side. We don't want to use an amplifier at all.

We want to upgrade the High-Q VLF into high-quality VLF.

Cheers,
Aziz

[Aziz]

Hi all,

good for me, not to order the missing parts yet.
Ask the AI for pitfalls, power efficiency and so on before doing this. So I did.

I am working on the high power High-Q transmitter design with the choke.

The appropriate ring core material for the choke for minimum losses (first attempt):
N27 or N87 (Mangan-Zinc ferrite core material, Al approx. 500 - 2000 nH/N²)
Size approx. 20 - 30 mm outer diameter. Better 30+ mm.
Careful design should minimize the choke losses (saturation current, operating voltage = step-up voltage conversion, expected losses)...

I will study the AI proposals for best power efficiency further. It can calculate all your design parameters. What a nice feature.
We can achieve even a Q-factor up to 200. Or more.
The High-Q transmitter is damn powerful. But requires careful design to tame its extreme power.

But we are dealing with more issues: Temperature drift, shielding, etc.
Did you know, that even the LC-tank capacitors need a proper shielding? We would require a shielded housing for the transmitter. All cables need a shielding too.

[Aziz]

Hi all,

now the fitfall with High-Q: The SNR.
If your source has some noise, the noise will be amplified by the Q-factor. No free lunch. No gain.
The line-output of the sound card is quite noisy.
How to overcome this source issue and improve the SNR on this side?
Pulse the source with a 50% duty-cycle pure clean and digital clock source (rectangular pulse), where it's frequency is controlled by a crystal oscillator and its phase and amplitude noise is low.
This is a good job for an embedded microcontroller PWM-output.

The TX-coil will pick-up EMI noise. This noise will be amplified by the Q-factor. Again no free lunch, no gain.
The figure-8 TX coil will help a bit, but will effectively reduce the coil size and detection depth. No gain.

We have to realise a compromise: Lower Q-factor (4 - 10), pushing more TX-current to get more from the RX-coil side.
You see, that the design concept is moving towards the standard the IB-coil configuration.
The pure mono coil VLF design has reached it's limitation now.
No matter what I do, there is no free lunch with my sound card solution. No gain.

Cheers,
Aziz

[Aziz]

Hi all,

I am still thinking of to improve the SNR on mono coil VLF (single TX-coil solution). Particularly to reduce the line-output noise of the sound card.

As the high Q-factor is very sensitive to line-output noise of the sound card, there is still a way to reduce the source noise contribution.
A second Lock-in demodulator from the line-out to the other line-input channel. So the pure source reference signal. The noise contribution can be detected and removed from the TXref signal.
There is a SNR improvement of 40 dB possible (100 times lower source noise).


But the picked up EMI noise is still there. Only the TX-current increase will improve the SNR.
Aziz

[Aziz]

Hi all,

I have detected a bug in my software but not fixed it yet. This is also causing huge "noise contribution". No wonder, if I am still doing quick & dirty implementation. A bug is made quickly.

I have realised, that the phase noise is quite low compared to the amplitude noise of the line output. We can make a rectangular pulse stimulation from a clean source supply. So the amplitude is much more clean and stable compared to the line-output of the sound card. This is an option to consider.

[Aziz]

Hi all,

best strategy to find a possible bug is to replace some old implementations with new Claude AI code.
This is an example for circular FIFO buffer implementation in Ansi C.
The Claude Code query (translated to english):

Implement a FIFO buffer in ANSI C for floating-point numbers. FIFO stands for First In First Out buffer and can be of arbitrary length. It will be written to and read from in blocks. No single values ​​will be handled.
The floating-point numbers should be typecast as REAL.
Create a FIFO type for internal data.

A function for `int FIFO_Init(FIFO *pFF, int nBuffer)`, where `pFF` is a pointer to the FIFO object and `nBuffer` is the number of floating-point numbers (buffer length).

A function for `int FIFO_Reset(FIFO *pFF)` to reset the buffer when no data is present.

A function for `int FIFO_GetSize(FIFO *pFF)` to determine the number of data items in the FIFO buffer. The return value is an integer.

A function for `int FIFO_Write(FIFO *pFF, int nBuffer, REAL *pBuf)` to write data to the FIFO buffer, where `nBuffer` is the number of data items to write and `pBuf` is the pointer to the floating-point data. The return value should be the number of data items written.

A function for `int FIFO_Read(FIFO *pFF, int nBuffer, REAL *pBuf)` to read data from the FIFO buffer, where `nBuffer` is the number of data items to read and `pBuf` is the pointer to the REAL buffer. The return value is the actual number of data items read.

A function for `int FIFO_Destroy(FIFO *pFF)` to release all internal resources.

Error codes can be included in the return values. Create an include file with type definitions, error codes, and function prototypes.

Create an implementation file. Review the implementation, correct any errors, and optimize it for speed.​

The implementation of the AI query is attached.
Nice isn't it?

Make use of AI. Just do the same as I did. Create your own query. Try it. Or approch your needings step by step.
Aziz

[Aziz]

Hi all,

no, the phase noise of the sound card is not lower than the amplite noise. It was a software bug.
I have checked this with reshaping of the transmit frequency and converting it into a digital clock with a comparator from a clean battery powered source and fed the transmitter with rectangular pulses. This option failed.

Next option for SNR increase: Second lock-in amplifier.
Aziz