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

[Aziz]

BTW ivica,

the Behringer U-Phoria UMC202HD USB Audio Interface is not that bad. It is even better than my Creative's G6. And cheaper.

Found some measurements diagrams on this site:

https://www.diyaudio.com/community/threads/behringer-umc-202hd-for-measurements.341309/page-15

It can do 192 kHz SR at 24 bit resolution.
I will search for alternatives for my G6 sh!t.
Aziz

[Aziz]

Hi ivica,

Unfortunately, I can't compete with you.
My card is 44.1kHz... too weak!
I have analyzed somewhere before; you need about 500ksps to comfortably play with those ideas.
This is the closest I can get on local ads:
https://www.kupujemprodajem.com/muzi...rId=8382013258

no problem. Even a sample rate (SR) of 44.1 kHz is enough. You just miss small time constant (TC) target responses.
Important is: low noise and high resolution (24 bit).

Look at this nice project. Fully balanced inputs. Very quiet. High sampling rate. Good bandwidth. USB. With schematics.
Link: https://www.mvaudiolabs.com/digital/...ent-interface/
No, I'm not interested in building one.

What I want is to use mass product USB sound cards or interfaces. And I want the sound card engineers not to cheat.

Creative engineers for instance (the developer of Sound Blaster products) do cheat on G6 (my sound card).
I'm sure, that the high frequency noise is coming from the sound card itself. They don't or can't manage it to make it quiet.
Directly visible on the line input, output and loopback frequency response.
The input line front-end amplifiers band limit the high frequency response. They want to hide the huge amount of noise in the upper half of the frequency spectrum.

Anyway, there are other better products available. I have not found the right one as all reviews, specs and diagrams apply for the audio band only (20 kHz BW).
Aziz

[ivconic]

Unfortunately, I can't compete with you.
My card is 44.1kHz... too weak!
I have analyzed somewhere before; you need about 500ksps to comfortably play with those ideas.
This is the closest I can get on local ads:
https://www.kupujemprodajem.com/muzi...rId=8382013258

[Aziz]

Hi all,

I have finally measured the real frequency response of the line input and output of my sound card G6 independently.
It is nearly flat for the audio region (0 - 20 kHz).
Line output: -3 dB at 85 kHz
Line output intermodulation distortion happens above 84 kHz (slowly rising to the nyquist frequency)
Line output is not critical because I will use it in the low frequency region for pulse timing control for the next sound card interface measurement board.

Line input: almost flat till 55 kHz
Line input: -3 dB at 78 kHz
Line input: -6.7 dB at 90 kHz
Line input: -8.5 dB at 93 kHz

I am going to lose a bit the high frequency response. But the high frequency region above 50 kHz is noisy and rising to the nyquist frequency steadily. Up to 25 dB more noise.
Nevertheless, the G6 input could be used up to nyquist frequency (96 kHz). So a flyback period of 5-10 µs could be done.

I don't know yet, where the high frequency noise is coming from.
Aziz

[Aziz]

to be continued..

That's the reason for, why I am looking at response regions, where the X-response dominates over R-response. The X/R-separation will lose less target response. It's a GB subtraction method.
In the IB system measurement above, which is heavy band limitted to old sound cards (48 kHz SR) and TX response band width limitted too, I have the biggest X-response at the pulse frequency (1500 Hz) with lowest R-response in relation.

Pure R response has distinct frequency response. Once you have removed the X-part from the signal, the R-part remains.
And this could be a gold nugget in high mineral ground or iron in mild ground.

PS: The signal frequency response (curve) need not to be normalized of course. Plain signal frequency response change out of the RX coil.

[Aziz]

Hi all,

I did the iron response X and R separation many years ago. This was the part of the GB system in multi frequency analysis using only the FFT magnitude part. I have processed only the change of FFT magnitudes for frequencies up to 18 kHz. No phase info was required or was not available at that time (I had no reference from the TX coil).


The typical pure X response (absent of any R-response) change curve in an IB system can be measured and normalized. You can see, that a mineralized ground has the most response at the lowest frequency. Targets may have higher response at higher frequencies depending on the TC of target. The special case iron produces both X and R. Now imagine, there is no iron in the ground but a gold nugget in high mineral ground. This will look like an iron target. And hence can not be really discriminated.

Look at the Excel-Table, how I do the X/R - separation except for the fundamental frequency of 1500 Hz. I am assuming that the R-part of 1500 Hz response to be X-part too. So assuming all response of 1500 Hz is X response. Then calculating according to the pure normalized X-response the X and R part of the frequency. Integrating of the values gives the total X and R response.

This is nothing different to solving the problem using the linear combination factors.
Using only magnitude response will not solve the discrimination problem. You never know, whether there is iron on mild ground or gold nugget in mineralized ground.

Note, that VRM effects and conducting grounds (salty ground) can not be solved as well. They will look like R-response. So you have to additionally manually ground balance too.

Aziz

[Aziz]

Hi all,

I have tested my LCR meter on the Sound BlasterX G6 at 192 kHz SR. Well, the upper half frequency region (>48 kHz) produces too much phase errors and the LCR measurement suffers and gets quite inaccurate. I know I have a bad USB powering issue on my PC (producing too much noise). I hope, I can solve this issue with an external USB Hub with extra power supply. I have ordered at Amazon a cheap one and additionaly an USB EMI Kit for filtering USB power noise. I can put an extra filter for the power supply if necessary.

I have to solve this issue anyway. For the multi frequency response measurements upto 96 kHz. And I want to know, whether my G6 is good or not. I'm not really happy with it at the moment.

[boilcoil]

To make a categorical discrimination between ferro and non-ferro in an impulse balanced system of PI metal detector, (as we call it here IB_HYBRID), a minimum of two samples are required in the TX-ON period.
The first should be in the initial ON interval, where the magnetic field is weak, but the rate of its change is the greatest and there are the main changes concerning conductive targets (eddy currents).
The second sample should be towards the end of the pulse, where the magnetic field is high, but the rate of change is low. Here the magnetic permeability of the target mainly affects.
Through an elementary analysis of the two samples, it is possible to make a ferro/non-ferro separation, almost equal to that of VLF detectors.
If there is a possibility for a third, fourth, etc. samples, the accuracy can be 100% like the phase discrimination of VLFs.

[ivconic]

Both points are correct.
In the end, everything will come down to multiperiod sampling and some kind of hybrid.
I would not dwell too much on the coil, because there you have a heavy tradeoff of losing performance at the expense of coil immunization from the mentioned problems.
The eddy currents in the target are of opposite phase, it might be a good idea to introduce phase processing in the PI detector as well! Ha!
Or has someone already tried it? (I don't doubt ML at all)

[Carl-NC]

But I have made interesting observations on some (1) non-conducting magnetic materials (Australian hotrocks and my special iron oxide) and (2) piece of iron.
(1) Either the magnetic viscosity or big magnetic hysteresis of the material, which causes losses to the magnetic field is behaving like a target.

Correct, VRM creates an elevated loss angle that (compared to the ground reference) looks like a target.

(2) Iron behaves like a magnetic material and target. It may compensate the inductance change due to eddy currents and magnetic properties at the same time.

Correct also. Iron increases the permeability of the coil and raises the inductance, while at the same time eddy currents induced in the iron steal TX energy and lowers the Q. Either one can dominate depending on frequency, depth, and other factors so a given iron target can appear ferrous or non-ferrous.

[ivconic]

This program is just an tip of the iceberg, I am working on something more complex and more interesting.
later...

https://www.geotech1.com/forums/forum/technology/programming/444334-inductolabpro#post444334

[Aziz]

Hi all,

ok, I could not resist to implement the relative and percentage change output of the sound card LCR-meter. This is very comfortable for me.
The PeakTech LCR-meter is again totally useless (has also Rel % function). I can kick it. Pure waste of money.

But I have made interesting observations on some (1) non-conducting magnetic materials (Australian hotrocks and my special iron oxide) and (2) piece of iron.
(1) Either the magnetic viscosity or big magnetic hysteresis of the material, which causes losses to the magnetic field is behaving like a target.
(2) Iron behaves like a magnetic material and target. It may compensate the inductance change due to eddy currents and magnetic properties at the same time.
This is very interesting.

Anyway, I will make the accurate measurements soon, put the results in an Excel-table and post it here. Then you can see, what I am talking about.
I am expecting very small changes of inductance as far as I can observe during testing of the LCR-meter:
- mild ground (0.0 .. up to 0.05%)
- mineral ground (up to 0.2 %)
- hot ground (up to 0.5 %)
- super hot ground (up to 1 %)
- standing on pure iron ore ground (above 1 %)
Aziz

[Aziz]

Hi Carl,

In real use just height variation of the coil will cause the inductance to vary a fair amount, so getting 1nH or even 10nH resolution probably won't buy you anything. In any case, once you decide you'd like to monitor the inductance in real time, how will you do it?

No, it has nothing to do with metal detecting yet. It relates the performance of the sound card based LCR meter. A maybe 5 bucks adapter board can outperform an expensive LCR meter. And it's accuracy can be managed easily. Without using Kelvin clips.
The LCR meter is only required to get the maximum limits on various ground conditions. This will answer my question, whether the IB detector will work on severe ground conditions (heavy iron ore gold fields).

Inductance change monitoring on the TX side should deliver more information about ground conditions (and targets). BTW, I do not monitor the real inductance change but the effects of it to the transmitter voltage at node Vc in the transmitter circuit above. The transmitter is optimized for TX coil inductance changes.
I am looking for a signal in the frequency spectrum, where the ground effects dominate over target effects. The ideal situation would be a pure ground signal, which is not superimposed with target response. This situation doesnt exist in real world however but a signal, where ground effects dominate over target effect would be fine.

This is an experiment or a proof-of-concept. It will fail or win. We will see it. Processing additional the TX-side is definitely a win-win situation.
Aziz

[Carl-NC]

In real use just height variation of the coil will cause the inductance to vary a fair amount, so getting 1nH or even 10nH resolution probably won't buy you anything. In any case, once you decide you'd like to monitor the inductance in real time, how will you do it?

[Aziz]

Hi all,

my sound card LCR meter is really very very sensitive. I can detect even 0.01 µH coil inductance changes (even better up to 0.001 µH). Found a nice feature: short DUT calibration on connected DUT delivers inductances changes including the phase angle phi changes as they will be zerod by the calibration. Any hot rock or hot ground nearby the TX coil can be detected at far distances.

I should implement a new feature maybe. Relative and % changes of all parameters. I will think of this.