Ground Balance Theory

[green]

I plotted more data. Forgot to save the last data so I plotted it again plus data with the coil clamped at 50 volts. Now I'm really confused. Maybe I did something wrong. t 0 = fet off. All data plotted as recorded except for the time corrected plots. Added 1 usec for the 450V data and subtracted 5.5 usec for the 50V data to the time scale before plotting. Don't know why but the corrected plots start at about 5 usec on both plots. The slope seems to be the same at -1.28

[PiTec]

Hi,

here are 2 RAR archives, each with 4 Excel files. Each file contains 4-channel measurement data and an image with the waveforms and channel settings.

www.tb-electronic.de/pi_tech/decay_curves_06_4_channel_100_us_per_div.rar
www.tb-electronic.de/pi_tech/decay_curves_06_4_channel_2_us_per_div.rar

1st measurement: Air signal
2nd measurement: Silver ring + Air signal
3rd measurement: MV soil + Air signal
4th measurement: Hematite + Air signal

Channel 1: RX signal at OPAMP output (2 NE5534 with a total of x200 amplification), 500 mV/div
Channel 2: Trigger signal (microcontroller pin), 10 V/div
Channel 3: TX coil current (100 mR shunt, x10 differential amplifier, i.e. 1A per volt), 5 V/div
Channel 4: TX coil voltage, 50 V/div

All signals are referenced to analog ground which is 2.5 volts above 0
Microcontroller supply is +5V
TX supply is +10V
CH2 and CH4 have an offset of -2.5 volts as they are referenced to analog ground

Timing:
500 µs TXon, first 100 µs (almost) linear ramp up to 5 A, then constant current
10 µs TXoff (flyback), achieved by clamping to approx. 90 V

As the current shunt is placed at the low side (source of the MOSFET), it will not measure the coil current that flows into the clamping circuit correctly.


First set of 4 Excel files has 512 samples per channel (time resolution 2µs), mainly for an overview how the signals look like. Time base is set to 100 µs/div. Start is at -600 µs, trigger at 0, end is at +422 µs.
Second set of 4 Excel files has 8192 samples per channel (time resolution 40 ns). Time base is set to 2 µs/div. Start is at -10 µs, trigger at 0, end is at +317.64 µs.

Additional notes:
The constant current period is a little noisy (simply circuit)
The MV soil shows a strange behavior during the CC period (visible in the 512 sample files). Instead of a decay curve which is visible when testing hematite, it produces an offset in the opposite direction. This must be a side effect of the DC field during the CC period. Magnetite shows exactly the same behavior. Both my MV soil sample and magnetite have a high magnetic permeability.

I hope I did not forget any important information. Just ask if anything is unclear.

Thomas

[PATCHES JUNIOR]

I hope you helped Moodz with his new line of detectors with your WBGB technology.

[Aziz]

Hi Aziz,

if you read the description of my test setup, I never triggered on the falling edge of the flyback pulse, but always on the MOSFET gate switch-off. As you said, this is much more stable.

Then I measured the pulse width of the flyback pulse plus the additional delay needed until the response curve is almost zero (air signal) to determine the point where the data acquisition could start – without knowing if the curves would come out straight. I did it this way to confirm my assumption that t0 (without a correcting 'p') should be at the end of the flyback pulse, without any manual time code correction, just based on real measurements.

Your approach is different, you want to calculate 'p' so that the curves become straight.

For me it is OK to agree that t0 = MOSFET gate switch off. This is the stable trigger point anyway, and the result is the same if the offset is corrected by 'p'. From my experience, 'p' is maily determined by the flyback pulse width, and the slope 'b' seems to be dependent on both pulses (if not time separated), plus of course soil properties and temperature.

As I wrote, you can use the existing measurements for t0 = MOSFET gate switch off by simply adding the times that I mentioned in post #565. Then the calculated 'p' should be close to 9/10/19 µs. Nevertheless, I will try to make a new measurement for you, with TX coil voltage, TX coil current, RX voltage at OPAMP output, and MOSFET gate signal.

BTW, I’m still waiting for some comments concerning the VRM document.

Hi Thomas,

I have built on the correctness of the correct time code of your data. I have been confused to0 much.
I don't trust in your time code anymore unless you can give me the possibility to peer check this.

Do not touch the original oscilloscope data. Never alter or change the time code. Trigger on mosfet switch-off. Sample the flyback voltage on a seperate channel so I can determine the TX system time delays.

That's simple. Isn't it? Now let's try it again.

Aziz

[Aziz]

Easy now /\sif!! No need to cuss in foreign language! We all know how smart you are.

"Himmel, Arsch und Zwirn!"
"Himmel, Arsch und Zwirn!"
"Himmel, Arsch und Zwirn!"
I'm going mad!!!!!!


/\sif

[PATCHES JUNIOR]

You guys need really some math lecture lessons!
"Himmel, Arsch und Zwirn!"
"Himmel, Arsch und Zwirn!"
"Himmel, Arsch und Zwirn!"

Easy now /\sif!! No need to cuss in foreign language! We all know how smart you are.

[PiTec]

Hi Aziz,

if you read the description of my test setup, I never triggered on the falling edge of the flyback pulse, but always on the MOSFET gate switch-off. As you said, this is much more stable.

Then I measured the pulse width of the flyback pulse plus the additional delay needed until the response curve is almost zero (air signal) to determine the point where the data acquisition could start – without knowing if the curves would come out straight. I did it this way to confirm my assumption that t0 (without a correcting 'p') should be at the end of the flyback pulse, without any manual time code correction, just based on real measurements.

Your approach is different, you want to calculate 'p' so that the curves become straight.

For me it is OK to agree that t0 = MOSFET gate switch off. This is the stable trigger point anyway, and the result is the same if the offset is corrected by 'p'. From my experience, 'p' is maily determined by the flyback pulse width, and the slope 'b' seems to be dependent on both pulses (if not time separated), plus of course soil properties and temperature.

As I wrote, you can use the existing measurements for t0 = MOSFET gate switch off by simply adding the times that I mentioned in post #565. Then the calculated 'p' should be close to 9/10/19 µs. Nevertheless, I will try to make a new measurement for you, with TX coil voltage, TX coil current, RX voltage at OPAMP output, and MOSFET gate signal.

BTW, I’m still waiting for some comments concerning the VRM document.

[Aziz]

You guys need really some math lecture lessons!
"Himmel, Arsch und Zwirn!"
"Himmel, Arsch und Zwirn!"
"Himmel, Arsch und Zwirn!"

[Aziz]

Hi Deemon,

I'm not demanding too much. I'm trying to make the conditions for comparable results with the scientific publication only.
We will see, whether we need a much better VRM formula at the end.

"Himmel, Arsch und Zwirn!"

Aziz

[deemon]

"Himmel, Arsch und Zwirn nochmal!"

Hey guys, I must insist on t0 = mosfet switch-off time. This is the standard reference time in the VRM science.

If you take t0 at a different time p, you are just implementing G(t) = (t+p)^b.
And you get different exponents b and they won't be comparable to the parameters of the standard VRM science (exponent b around -1.0).

The VRM response is dependent on the period prior to t0 (=switch-off), i.e. the TX pulse history (shape, duration, etc.) and flyback period (the shape, duration, etc. ) .

It is quite difficult to determine/measure the end of the flyback period. But it is more precise to take the mosfet switch-off time as t0.

"Himmel, Arsch und Zwirn nochmal!"
"Himmel, Arsch und Zwirn nochmal!"
"Himmel, Arsch und Zwirn nochmal!"
Aziz

By the way , Aziz , I think that you want too much I mean that you are trying to find "too precise law" . You see , even in the laboratory measurements you find some inaccuracy and inconsistency , but what can you expect on the real ground , in the real field ? Even if you find an ideal law in ideal conditions - in the real field you'll meet some disturbing factors ( temperature , magnetic fields , etc ) that will certainly spoil your ground balance My point is that if you want a reliability - you cannot demand too high precision , it's a kind of a "natural law" For example , in "general electronics" I noticed this many times - if some circuit needs too complicated calculations , it's better not to use this circuit . It can work on the bench , but in the field it will fail anyhow . The dilemma is simple - if we have a circuit and some very complicated formula with many variable parameters that describes its behavior , there are only two variants :

1. This formula doesn't describe it precisely , so we don't need this formula .

2. This formula does describe it precisely , so we don't need this circuit .

And with this viscous ground , as I think , we must consider the law as "something like 1/T" , and try to find another features of the soil that will help us to distinguish it from metal .... it can be its magnetic permeability , or maybe this dependence on the previous "pulse history" that you told about here ... but if we try to base the ground balance only on the "super-duper ultra-precision" formula - it will be a mistake .

[Aziz]

Hi Thomas,

could we make an agreement, that t0=switch-off time? Once and for ever.

It doesn't matter what happens between t0 and end of the flyback time. The measured data later will take the effects of the flyback period into account. But we can trigger on t0 much better (because, it's source is logic and very stable) instead of waiting for the end of the flyback.

Would you be so kind and make the measurements once again with t0=mosfet switch-off time.

Cheers,
Aziz

[PiTec]

Hi Aziz,

no problem to use t0 = MOSFET gate switch off – then 'p' has to be adjusted to compensate the flyback pulse width. 'p' may also be influenced slightly by the driving pulse width as long as the pulses are not separated in time.

The VRM response is dependent on the period prior to t0 (=switch-off), i.e. the TX pulse history (shape, duration, etc.) and flyback period (the shape, duration, etc. ) .

Yes, fully agreed. The problem is that the guys in the VRM document that you linked in post #531 completely ignore the flyback pulse, and only use the pulse width of the driving pulse to derive their equations.

Again, please read my comments to this in post #545.

[Aziz]

"Himmel, Arsch und Zwirn nochmal!"

Hey guys, I must insist on t0 = mosfet switch-off time. This is the standard reference time in the VRM science.

If you take t0 at a different time p, you are just implementing G(t) = (t+p)^b.
And you get different exponents b and they won't be comparable to the parameters of the standard VRM science (exponent b around -1.0).

The VRM response is dependent on the period prior to t0 (=switch-off), i.e. the TX pulse history (shape, duration, etc.) and flyback period (the shape, duration, etc. ) .

It is quite difficult to determine/measure the end of the flyback period. But it is more precise to take the mosfet switch-off time as t0.

"Himmel, Arsch und Zwirn nochmal!"
"Himmel, Arsch und Zwirn nochmal!"
"Himmel, Arsch und Zwirn nochmal!"
Aziz

[PiTec]

Thanks, green. Try to extend the flyback pulse width by clamping the flyback voltage to 100 volts or less, then the difference between the two trigger points becomes more obvious. Also, the steepness of the slope should decrease.

[green]

Plots not as clean as PiTec, but show a trend. In the past I've used gate turn off as t 0. Didn't matter for log-lin plots. The log-log plot looks better if I use coil turn off as t 0 same as PiTec. The clay from the yard doesn't give much response. I'll try to add more gain and try to get a better curve. The scope was triggered at gate turn off,t 0. A line with a -1.28 slope was added to both plots.