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

[Aziz]

Hi Flatline,

OMG I FOUND MY PEOPLE! I'm building an android extremely low latency app that receives gpx4500 signal, alerts user to all sorts of customised audio "anomalies" is able to analyse multiple sweeps across a target and display a dynamic target confidence level.

It also plots a heat map of the users path taken. Can be configured to display/overlay approximated coil sweep coverage / audio anomaly detections / estimated EMI/Ground Noise.

If phone mounted on detector shaft , can be configured to monitor coil angle and provides live feedback to user
.
Also logs the amount of sweeps (utilises hardware accelerometer/gyro).

The scope creep has been insane , I just can't help myself.

I NEED DATA! I'm about to start integrating Machine learning (probably going down route of convoluted neural network). I just need trustworthy gpx4500 mono coil recorded audio signals with target details (approx depth, surface area, weight, type - the more the better but bare minimum is audio signal + target type + weight + estimation of depth).

Think of it as "Digital ears". It'll be releasing it free at some point, though really need some legit people to critique and **** on orcas hard as they possibly can (been in IT testing field past 15 years so you'll only be turning me on by finding any sort of flaw). For now it's flaws in my logic , strategy, the entire damn thing could be a pointless waste of time and I've been too deluded to realise. I'm a king of random posts placed in random threads - msg me if anyone has any sort of interest in this proof of concept.

(Might take me a couple of weeks to get back to y'all)

welcome to the forum. I am looking forward to hear more of your project and how it is progressing.
Post processing detector signal hasn't been done yet.
I am also familiar with neural networks.

Cheers,
Aziz

[Flatline]

OMG I FOUND MY PEOPLE! I'm building an android extremely low latency app that receives gpx4500 signal, alerts user to all sorts of customised audio "anomalies" is able to analyse multiple sweeps across a target and display a dynamic target confidence level.

It also plots a heat map of the users path taken. Can be configured to display/overlay approximated coil sweep coverage / audio anomaly detections / estimated EMI/Ground Noise.

If phone mounted on detector shaft , can be configured to monitor coil angle and provides live feedback to user
.
Also logs the amount of sweeps (utilises hardware accelerometer/gyro).

The scope creep has been insane , I just can't help myself.

I NEED DATA! I'm about to start integrating Machine learning (probably going down route of convoluted neural network). I just need trustworthy gpx4500 mono coil recorded audio signals with target details (approx depth, surface area, weight, type - the more the better but bare minimum is audio signal + target type + weight + estimation of depth).

Think of it as "Digital ears". It'll be releasing it free at some point, though really need some legit people to critique and **** on orcas hard as they possibly can (been in IT testing field past 15 years so you'll only be turning me on by finding any sort of flaw). For now it's flaws in my logic , strategy, the entire damn thing could be a pointless waste of time and I've been too deluded to realise. I'm a king of random posts placed in random threads - msg me if anyone has any sort of interest in this proof of concept.

(Might take me a couple of weeks to get back to y'all)

[Aziz]

Hi all,

I am very happy to have NG0/C0G capacitors now and they are really promising to realise the magic High-Q mono coil VLF with built-in capacitors in the coil.
I have to sort out my C0G capacitors yet. Part number does not specify the parts value and voltage rating.
I have measures a few and there Q is >2000.

And I have to improve the sound card LCR measurement further. It's accuracy for Q, D and ESR suffers for low value capacitors.
The LCR calibration data is affected by noise. I have to upgrade the LCR software.

The ribbon mono coil must be rewinded of course. I need a distance spacer (0.5 - 1.0 mm thick) between the windings.
Paper is not convenient as it contains or gets moisture and water is highly polar, which will cause heavy losses.
Adhesive tapes may contain polar chemicals on the adhesive film, which will cause also heavy losses.
If I don't find the right distance spacer, I have to cut thin XPS stripes.

It will take some time and I don't have much.
Aziz

[Aziz]

Hi all,

my NP0/C0G capacitors came just.

The 0603 SMD variants are too small to handle! (fkn tiny!)
0805 is ok, no problems.
1206 is easy.

[Aziz]

Hi all,

I have added j-fet buffers to the LCR breadboard. It is using a very simple rail-splitter for power supply. It works.
I can measure D, Q, ESR much more accurate now.
It is still critical for low value capacitors (< 1nF) : This breadboard is not convenient to achieve more accuracy for D, Q and ESR measurements.

[Aziz]

The shipping company is saying:

Due to recent adverse weather conditions and the resulting backlogs, there may be delays in the delivery of your shipment(s). We ask for your understanding.

Anyway, I can't find my other LCR measurement breadboard anymore.

So I have to build a new version.
But I have found an OPA2134 (dual j-fet op-amp) in my box, which should do a good job for the voltage buffers.

[Aziz]

Hi all,

I have to find my old active j-fet op-amp buffer LCR-meter board for Sound card LCR measurements. Or I have to build a new one.
This is absolutely required to measure D, Q, ESR more accurately. My PeakTech 2170 LCR meter is not accurate enough.
The NP0 capacitors have not yet arrived. Waiting, waiting..

[Aziz]

Google AI summary says:

All three coils are highly imperfect.
Coil 2 (HF-Litz) provides the best magnetic performance up to (100 kHz) due to its stable AC resistance, but is bottlenecked by its tight bundle topology.
Coil 1 achieves a deceptive peak Q due to its spiral geometry reducing proximity effects, but will fail rapidly at higher frequencies due to severe PVC dielectric absorption.
Coil 3 is fundamentally flawed for high-frequency use due to its uninsulated strands.

The AI analysis is really very good.

[Aziz]

Hi all,

I have made the 3 air core coil analysis by using Claude Code.
I'll put the coil data files, AI prompt and the output of the analysis.

This is really interesting.
Aziz

[Aziz]

Hi all,

this is funny. The Google AI can interpret and analyse the coil measurements data stored in an Excel file (as .csv text file).
I will make a nice Excel-file with all the measured coil parameters and let the AI analyse the coil for me.

The AI is saying, that my ribbon coil is still just fine!
Aziz

[Aziz]

Hi all,

I will show you some LCR-Meter measured parameters for 3 different coils soon. This is quite interesting to interpret the occuring effects and "wrong" measurements.
The ribbon cable coil, standard copper litz wire (it's not a HF-Litz) and a true HF-Litz wire coil.

My ribbon cable coil:
I will rewind the coil using one or two paper stripes between the coil windings to minimize the proximity and capacitive effects a bit later.
And I can also use the coil in low frequency mode 5 - 10 kHz to minimize the high frequency effects too. It can still be operated as a nice mono coil.

For real high performance coils, I have to buy 50 m true HF-Litz wire yet.

My ordered NP0/C0G capacitors will arrive tomorrow.
Aziz

[Aziz]

Hi all,

I have made a PDF-Table on LCR-Measurement with AI (using python code).
It sums up the effects occuring in low / high frequency when using an LCR-Meter for coils.
(Note: I don't know, whether the AI is halucination or lying.)

[Aziz]

Hi all,

regarding my sad ribbon mono coil:

Here is a good explanation from the AI, why this coil with ribbon cable will degrade the Q of the LC tank.
Ribbon cable consists of PVC isolation. There is much more PVC than copper wire in it. It's permittivity is quite large (increasing the coil capacitance further more).
PVC is polar, so it will (the PVC molecules) polarize in an electric field. This is frequency dependent. At high frequencies, the polarisation character
will cause heavy losses (reducing radically Q of the coil).
In conjuction with the increased coil capacitance, a relative big portion of the capacitance will contribute to the LC-tanks capacitor.
So a very "bad capacitor" with high losses is connected inherently parallel to the LC-tank.

These losses can exceed the proximitiy effect losses. Potentially several order magnitude (10^x).

A plain magnet wire woudn't have much losses compared to the PVC wire cable.

So we must avoid PVC cables. (Except for very low frequencies).
Better isolation would be Teflon (PTFE). Very expensive. Difficult to get the wire.
Next better PP (Polypropylene). Difficult to get the wire.
Next better PE (Polyethylene). Cheap, still difficult to get the wire.
Next better PU (Polyurethane, thin magnet wire isolation).

Unfortunately, almost all standard cables are made of PVC. It is cheap and good temperature resistive.

I have a nice tinned PVC litz cable. Pealing off the fckn PVC isolation and isolating the wire thin PE stripes?
Aziz

[Aziz]

Now for a three row basket weave mono coil with N1, N2 (N1=N2) and N3 turns count:
N1 = 0.866*N3
N2 = 0.866*N3
This time N3 > N1, N2. This is good.



The B=0 zone is approximately (averaged over the large gap area) in the middle.

[Aziz]

Oops,

my fault. Looked at wrong figures.

N2 = 0.866*N1.
The outer coil will be slightly lower turns count. This is bad.