Ground Balance Theory

[Aziz]

I ran some decay curves on some clay from the yard awhile back. With a amplifier gain of 160 the slope was about -1.4. With a gain of 1600 the slope was less than -1. Should gain change the slope? I'm not getting much signal from the clay. A strait line for about about 1 decade. Need to improve my setup. I'll post some plots when I get back home in a couple weeks.

Hi green,

if you run the amplifier with a gain of 1600, you reduce the maximum bandwidth of the amplifier (BTW, quite heavily). Of course, you can't get the high frequency content amplified with gain of 1600. Only the low frequency part. That is causing a non-linearity and you get a different exponent.
One have to check the linearity of the amplifier before you can think of measuring the decay curve.
Aziz

[Aziz]

What I used for “ferrite” reference is very large E100 core ordinary soft ferrite made by Kaschke.de, material type K2004, leftover from some commercial stuff. There are zillions of ferrite types out there, high or low permeability, noncrystalline, alloys like metglass (ok. this is not ferrite) etc, some of them can produce weird behavior. Main issue with this is, as always, completely improvised test setup, inadequate equipment and hastily done measurements typical for DIY. At first opportunity, when I have some time (winter) I will retest all this using more decent equipment and publish measured data. This old crappy DSO for example don't have any PC interface and can store only 3 traces, so all “data” is camera picture taken from CRT. More detailed testing of behavior with different, including very short pulses is interesting.

Tepco,

be very careful with the 1/t law. Don't make any claims you can't really prove.
-------
BTW, what happened to the trolling head scratchers? Still scratching their head?
G(t) = a*(t+p)^b


Aziz

[Tepco]

I can't, and not trying to prove anything, especially not with bunch of improvisation I used to measure, proof can be if this is done more seriously. Comment on previous post: at least I have all bandwidth on this world, actually hard time to limit and filter it to some reasonable value, discrete input+LM733 goes over 100MHz, log-amp is good over 1Ghz, only problem was filtering with low order filters, not to screw up transient response. Cautionary words: be careful with high-pas part, time constant must be long enough compared to measurement interval, otherwise “bump” in response will occur, distorting results, I carefully recorded it and almost published (new effect?) In last minute figured out error prevented me from writing complete stupidity.

[Tepco]

One more addition to “proof'”, same result must be obtained using two different setups, built on two different principles, isolating other factors,then I can claim anything.

[Aziz]

Anyone want to provide real data I can process for you?
I'm working very accurate and peer checking everything.
Aziz

[PATCHES JUNIOR]

Anyone want to provide real data I can process for you?
I'm working very accurate and peer checking everything.
Aziz

That is a great offer Aziz!! I know a couple developers who could of used your service for years now!

[Aziz]

That is a great offer Aziz!! I know a couple developers who could of used your service for years now!

You know, my service will bust all the gurus out there. Even patent trolls.

[Tepco]

Anyone want to provide real data I can process for you?
I'm working very accurate and peer checking everything.
Aziz

No way, I’m working with sledge hammer and not checking even what I wrote.


Proposal for relevant test: measuring response from 3 different pulse widths, say 500, 50 and 5uS, generated with same energy release and same field straight, bit tricky but doable. Then measured directly, via log-amp and averaged in one turn, and integrated, over some short period, say 5uS, 5us apart, point by point in another. Should produce same results, within accuracy limits. If matches, this can be considered valid proof, if not, one setup is in trouble. I'm pretty convicted that perfect 1\t response is result of variable pulse width (not generated with same energy release) combined with variable integration time, this will iron out everything to look like 1\t. Any idea or suggestion on this?

[PATCHES JUNIOR]

You know, my service will bust all the gurus out there. Even patent trolls.

That is very nice of ya!! I am sure your offer will help many, many blokes who have been waiting for someone like you to come along!! THANKS for your service!!!

[Aziz]

1A Service by "MadLabs Inc."(c)(tm)(r)

Hi all,

you wanna see the first class service now? And what does it have to do with the head scratchers and G(t)=a*(t+p)^b ?
Ok, read the following posts and the related thread and find the bug.

Post #47:
http://www.geotech1.com/forums/showt...720#post167720

Post #69:
http://www.geotech1.com/forums/showt...794#post167794

There are some real measurement data on this forum (decay_curves.rar in post #69). Download it and process it. And find the bug or scratch your head until I reveal it. I also want to revisit the topic.

Aziz,
the ultimate data processor *LOL*

[Aziz]

Solution and Solved Issues

Hi all,

I think you have scratched your head long enough. It's high time to reveal it.
VRM decay measurements require precise and accurate timing records.

I'm referring to the data provided by Thomas mentioned in the post above (see link there). And why the hell it hasn't been a straight line log-log decay?

In the post #47 and we can obtain the following infos:
Pulse on period: 50 µs, flyback period: approx. 2 µs, "Decay curves start approx. 5µs after the 2µs pulse". And we have the time code of the samples starting from 2.4 µs.
One of the VRM measurement was "adjusted" to have same initial decay value (approx. 456 mV). It is very obvious, that the time base for that VRM measurement was time shifted. But recorded at the same time code with the other VRM measurement.

So we have a totally unknown and imprecise time base for the VRM measurements.
time code = real time + offset time

And now comes my magnificiant VRM formula:
G(t) = a*(t+p)^b

This formula can detect such unknown offset time shifts in the measurement data. Provided that, there is no significant physical effect in the early decay times. We haven't figured this out yet and we would require very precise data for this issue to solve. But it is very likely, that p either doesn't exist or can be totally neglected due to not significant effects (p < 1 µs). Note that every TX and RX combination of the front-end may cause p to exist even there might no physical effect.

Well, if the time code in the VRM data was correct we would see nice straight lines in the log-log plots. Indeed, that's going to happen (see below). We have for the hematite sample a decay exponent of b = -1.3 and for the MV soil b = -1.21. BTW, it has been already found that the exponent is dependend on the pulse length (see Paltoglou and other scientific publications).

G(t) = a*(t+p)^b reduces with the correct time code into
G(t) = a*t^b

And that is the good old exponential decay (not to confuse with the exponential decay of the target responses Tgt(t)=a*Exp(-t/TC) ). The 1/t law requires conditions to be valid, whereas the exponential decay is always valid.

The pictures and the Excel file to peer review for you (that's very important of course).







The Excel file:
DecayCurvesCorrection-01.zip

Am I Sherlock Holmes or Nick Knatterton or an evil man? *LOL*



Kombiniere: I'm an evil man!
(in teasing the forum trolls *LOL*)

BTW, single time constant (TC) target responses in the log/lin plot can be shifted in time without deforming the log/lin straight line.
If you have some more precise data then let me process them for you. There are so many mistakes and pitfalls you can make. But forget it, if you can't meet the requirements for accurate VRM measurements!

Cheers,
Aziz

PS: Forgot to add something:
The corrected time code is calculated as follows:
corrected time code = time code t + p

[Aziz]

Wasn't that a premium 1A service?


Well, it goes so far, that I'm not trusting anyone providing data or making claims. I do not trust in patents and scientific publications. I do not trust in scientists and their work. As I have mentioned, everyone can make mistakes and fall into pitfalls.

One have to work very very accurate to avoid mistakes. And one have to give the oppurtunity to peer review the claims.
I like to prove it by myself (the peer reviewer).

Cheers,
Aziz

[WM6]

I saved this forum page for PI history.

With many thanks and respect to the "MadLabs Inc."(c)(tm)(r).

[Aziz]

BTW guys,

I'm using the Excel 2010 software and there is a bug in the software. I save the table in an old Excel 97 format to give other people with an old Excel software to load and open the file.
Well, the Excel 2010 software loses the manual max. time range in the log-log plot. You have to change it to the original value (200 µs) everytime when you face the software bug.

Pity, I can't afford the expensive accurate measurement instruments (>1,000 EUR). But I'm thinking of making a very precise PI VRM measurement tool using the USB sound card solution. Once I have it, I'll bust all scientists and gurus!!!! *LOL*

Cheers,
Aziz

[WM6]

Interesting VRM experiment:

http://jnaudin.free.fr/dlenz/DLE21en.htm

But as seems VRM is not constant (tend to decrease in time on test sample):

http://www.youtube.com/watch?v=oRcK6jE42PA