Targets frequency response

[TheWizard]

Aziz,

In SD200 the demodulators is not Qadrature Demodulators
This is like deferential integrator for good SNR...
If you want to make Digital PI (true), first you need very fast ADC
I make before tests with 12 bits, 400 K samples per sec(my avatar is PCB board) - results is good but not the best...

[Aziz]

Aziz,

Please check your email, i sent you something, let me know if it helps out?

Hi B^C,

thanks for your email. But I will make a general simulation, which is based neighter on the last shown timings nor to a particular design. It will be a general sight of eliminating noise from unwanted frequency spectrum to achieve more SNR. So it is not necessary to compare detecting technologies. But it might be helpful, to make a new PI design using lock-in amps in general. I am not sure, whether this is elegant possible due to some limiting factors.
Regards,
Aziz

[mikebg]

Excuse and supplication addressed to B^C

Dear Bruce,
In front of the whole world (because this is an excellent worlds forum), I beg your pardon for my bad words comcerning your knowledge, education and your old Bible with technical articles. Too soon the fate has showed that I'm wrong and willed me to look for help from physicists and from old technology Bibles for solving problems, which an amateur (ham radio) designer, like me can't solve using all WEB resources and whole community of amateurs.
I will explain these problems and my requests for help in other postigs.
Submissive yours Mike.

[B^C]

Aziz,

Please check your email, i sent you something, let me know if it helps out?

[Aziz]

But the simulation results could take me some weeks. I have to implement much changes on my MD application and also some tidy up the old code. So I can implement and simulate different types of measurements and analysing methods. A noise analysis will be taken into account.
Also the dumping process of the coil with more stages could be interesting when taking the parameters dumping resistor, coil capacitance, diode states, etc. into account.

Aziz

[Aziz]

Hi bbsailor,

yes, I found the timings and there is definitely room for enhancements! As I can see, the SD2k does not really implements a true lock-in amp. It is an approximation process to lock-in amp only.

On channel two with four sampling windows, this gets more like a lock-in amp. This is necessary, to increase the SNR and making the detector stable.
Channel two is very likely for short time constant targets (gold) and channel one for long term (iron). The SNR can be increased by increasing the number of sampling windows.

I will implement a simulation on the timings (see below) and will compare this to a real lock-in amp.

There are some lack of information about:
- the real signal(t) function (inverted exp()-function and parameters)
- the signal strength decay of t (from ~Umax .. down to µV or nV)
- the amount of noise(t)
- etc.

Anyway, I can start with the coding and if I know the missing parameters, I can take this into account later.
Aziz

[bbsailor]

Aziz,

I don't believe that Bruce Candy mentions all the timings in the patent but I believe that the long pulse on figure 2A is 240us long (Bruce states this and .25ms) and each of the four short pulses is 60us long.
Figure 2B is 50us wide or maybe 60us?
Figure 2C is 60us wide
Figure 2D is 60us wide
Figure 2E pulses are each 15us wide aligned with the TX flyback pulse.
Figure 2F pulses are each 15us wide and delayed until after the TX pulse.

You should be able to figure out the timings between the pulses.

Go to the following link to see a schematic of the SD2000 which looks like an implementation of the patent. http://thunting.com/geotech/forums/s...ad.php?t=14001

bbsailor

[Aziz]

Hi bbsailor,

your proposals needs to be investigated deeply. And I see therein much potential to increase the signal-noise-ratio (SNR) and the stability of detecting.

I have allready planned a computer simulation about the amplifier and demodulator stage of the PI and the implementation of the lock-in amp.

Do you know the timings from gate 52 to 57 (on page 1)?

window 1:
gate for non-inverting signal: 52
gate for inverting signal: 53

window 2:
gate for non-inverting signal: 54
gate for inverting signal: 55

window 3:
gate for non-inverting signal: 56
gate for inverting signal: 57

Unless I don't know the timing behaviour, I can not say, if there is more potential to get. This is necessary for me to analyse, how the "lock-in" in the demodulator stage is implemented and whether there is room for further enhancements.

Aziz

[bbsailor]

Mikebg and Aziz

Look at a Burce Candy patent number 5506506 at this link http://geotech.thunting.com/cgi-bin/...le=patents.dat.

Bruce Candy does a methematical analysis on page 9 to describe how he can separate the time constants of various targets to attempt to identify them even in the presence of highly mineralized ground. See his figure 3 where he graphs the responses of various long and short TC targets. I believe the vertical axis of the graph has an error where the upper -1 should be a +1.

He states on the bottom of page 14 that his analysis works for PI as well as continuous wave detectors. This is related to your recent discussion.

Look at the figure 4 where he labels the timed switches as "synchronous demodulators", which is esentially the function of the lock-in amplifier.

He uses three sets of synchronous demodulators to separate and analyze the long and short TCs of targets and filters out the relatively slowly changing ground to recognize and indentify target signatures with a microprocessor constantly adjusting and cancelling out the ground and only responding to the faster target TCs.

You will notice that all targets do not produce an absolute response as their orientation in the ground may alter their response. That is why it is important to get a good understanding of the range of various target responses in all common orientations. Unfortunately, sometimes the desired and undesired target responses may overlap under certain orientation conditions or ground conditions.

Here is an idea that may help your analysis. Most targets are initially located while the coil is in motion over the ground. What if the coil were stopped over a target and placed firmly on the ground where a push button is activated to enter another target ID mode where the last ground signal, held in memory, is subtracted from the target signal? Then, the ID mode analyzes the target response during the TX period using the target to couple the RX coil to the TX coil, as in a loosley coupled transformer, and analyze the target with another set (along with the target parameters stored while in motion) of target ID parameters to minimize any target ambiguity due to target orientation in the ground.

The big tradeoff of lock-in amplifiers is the time needed to extract the signal from the noise. Once the coil is stopped over the target, the coil motion speed is no longer an issue, and the synchronus demodulators can collect many thousand samples to better identify the target.

You should think about optimization in two modes. (1) The motion mode where the design is optimized for target acquisition. (2) The still mode where the design is optimized for target identification.

Here is where your target analysis and coil analysis could lead to some interesting new design opportunities.

bbsailor

[Aziz]

Hi Aziz,
You have used two times the incorrect notation
Z=R+jXL
The correct notation is
Z=R+jX.
Please read [2] Chapter 30 to understand why.
In our case L is an irreal imported inductance, because is negative and transient (depends on time or on frequency). As you knows, a real coil has positive inductance which is almost independent on frequency. This will become visible when we start exercises with SPICE simulator.

Mike,

thanks, I meant with XL (w*L) the reactance of the coil of course.
I will write only X in the future.
Aziz

[mikebg]

Z=R+jX

Hi Aziz,
You have used two times the incorrect notation
Z=R+jXL
The correct notation is
Z=R+jX.
Please read [2] Chapter 30 to understand why.
In our case L is an irreal imported inductance, because is negative and transient (depends on time or on frequency). As you knows, a real coil has positive inductance which is almost independent on frequency. This will become visible when we start exercises with SPICE simulator.

[mikebg]

Exercise No 2

Exercise 2 "Normalized Transfer Impedance of conductive nonferrous object "
Attached FIG. 6 shows transfer impedance of a conductive nonmagnetic object in which flow whole set of eddy currents. You can see, in above mentioned WEB site, an animation - how moves point M when we increase frequency and how - if there is different cracks in the structure. The shown impedanceform is valid for most nonferrous targets and for halfspace (salty water and conductive soils without ferrous mineralisation). Only metal foil has slight different impedanceform because skin effect is inessential at low frequencies (the whole thickness of foil is used by eddy currents).
Because this impedanceform is normalized, it is independent on:
- diameter of target,
- diameter of TX loop,
- conductivity of metal or of halfspace,
- distance between sensor and object,
- ampereturns of TX current,
- turns of RX winding.
NOTE the following regularities:
- The impedanceform at low frequencies seems like those of an object having single eddy current.
- The cutoff frequency fc seems as if no difference with such one object.
- The magnitude of signal at high frequencies is remarkable greater (approximatelly 30%).
IMPORTANT: It follows that if we design a wide band metal detector displaying impedanceform in complex plane, we can discern the shape of nonferrous targets. If the impedanceform is half-circumference, the target is like bracelet or ring. If the impedanceform is stretched at high frequencies, the object has no large opening.
Using this normalized impedanceform, you can solve following problems and answer to following questions:
1. Plot in normalized form frequency functions in Cartesian coordinates Re(f) andIm(f).
2. Plot in normalized form frequency functions in polar coordinates: Magnitude vs. frequency and Phase vs. frequency:
a) in linear scale
b) in log scale,
c) approximate log scale as Bode plot.
3. Explain what is the difference in comparison with some frequency functions of single eddy current.
4. Are there errors in the figure on the WEB site?
HINT: Pay attention to inscriptions on axes; they are different from our exercises. Note that at nonferrous conductivity, imported inductance (coordinate Im or X) is always negative.

[mikebg]

Re: German site

Thanks Aziz,
This site has english version. To select, goto home page.

[Aziz]

I will investigate more on the eddy current effects on targets in the future. Since I haven't a measuring instrument for this yet, I couldn't check the formulas and my simulation results.

Now I am able to measure small AC voltages and currents using lock-in amps.
I also have to look for some physics background.

I found the typical eddy current measurements by the industry are based on the change of impedance Z=R+jXL of the primary air cored coil influenced by targets. According to the following web-site, both R and XL are changing.

http://www.uni-magdeburg.de/iwfzfp/EddyCation/dt/EddyCation.htm

Sorry, it's in german. Look particularly to the impedance graph (right bottom).
Regards,
Aziz

[mikebg]

Education

Thanks Bruce,
Here a saw of Bulgarian amateur designers:
IF YOU CAN'T DESIGN IT SIMPLE, YOU SIMPLY NEED EDUCATION.