New TX methods

[Tinkerer]

PI metal detectors. The PI stands for PULSE INDUCTION.

It does not say what shape, or length, or duty cycle of pulse.
Square
Rectangular
Sawtooth
Triangular
Sin
Bi-polar
Chirp
Mixes of short and long
Skip
All these names apply to voltage wave forms for pulses.

Now what about the variety of current waveforms?

All these pulses can be used to excite the target.

PULSE INDUCTION does not say what shape, or length, or duty cycle of pulse.

So let's discuss different TX methods. Different ways to excite the target with pulses and different ways to take care of the resulting FLYBACK.

Let's start with the "TWO TAU" METHOD.

Two Tau IB-PI or Sampling during Flyback
Here is a brief description of the method.
Attached is the TX and power supply. The power supply delivers 8 to 15 Volt for the TX, regulated. Also +5/-5V with a common ground for the analog system and a separate +5V for the PIC MCU.
The ICL7662 is sync’d with the switching capacitor voltage doublers and the TX pulse, so that there is no switching noise during the sample times.
The Mosfet is driven with a Mosfet driver to give a high initial rate of change for the TX charge ramp. This rate of change has an influence on the response of the targets. Information about the TC and the presence of FE can be gathered during the charge ramp of the TX, when using an Induction balanced coil IB.
After 100uS charge time the TX is switched OFF. A high Flyback results and is (with traditional methods) eliminated as quickly as possible.
My method instead, controls this Flyback and uses it to excite the target in a CONTROLLED MANNER. I let the Flyback rise to somewhere between 70 and 150V and sample DURING THE FLYBACK. I have used this sample window with a duration from 6 to 10uS.
The same information gleaned from the di/dt during TX, can be found during the Flyback sampling, with the difference that the di/dt is 10 to 20 times as much. The signal amplitude is so many times higher.
It is like using a VLF with low frequency and high frequency. I call it “Two Tau”.
So with each pulse repetition we have the possibility of sampling:
1. During the TX charge ramp
2. During the Flyback
3. After the Flyback has decayed.

This is a sample window during Flyback. The target is a silver ring of 0.55grams. Gain 100,
scope 200mV/div.
Another feature is that the Flyback energy is partly recovered (less dissipation losses) since it is discharged into the +8 to 15V.
These features are possible thanks to the use of an Induction Balanced coil assembly. This coil assembly consists of 3 separate coils.
One TX coil of about 300uH
One RX coil of about 300uH
One Bucking coil of a few turns, wound on the RX coil.
A coil current control circuit makes it possible to precisely control the current in the Bucking coil so that a precise balance can be achieved and also corrected, when an unbalance has been produced by highly mineralized ground or temperature shift or humidity.
Discrimination:
This method produces a natural differentiation between FE and non-magnetic targets.
FE targets produce a negative response.
Non-magnetic targets produce a positive response.

These features are present during TX sampling but are very pronounced DURING THE FLYBACK SAMPLING
There are instances when the X response is equal the R response. In this instance the response is 0, as in the case of the cannon ball at a certain distance. (see picture: cannon ball)

This is the signature of a 6 pound cannon ball at a distance of 30 cm from the coil.
However, the response after the decay of the remnants of the Flyback are unaltered, so that the cannon ball (or FE target) can still be detected as usual.
The RX signal is amplified with a non inverting preamp. Thanks to the FLYBACK CONTROL CIRCUIT, the amplitude of the Flyback is very much reduced, so that the signal can be taken directly from the coil.
The signal amplitude is so high that the target response can actually be seen on the oscilloscope at the INPUT of the preamp.

Ups, the pictures did not come out. I will add them in the next post.

Tinkerer

[Tinkerer]

NEW METHODS

Here is the TX&PS for the 'TWO TAU METHOD"

Does it look a lot different from a traditional PI?

But the signal response amplitude is about 10 times higher.

And guess what, I can sample at 1uS after TX switch OFF.

Definitely "NEW AND IMPROVED"

Tinkerer

Again no picture, so I try again.

[Tinkerer]

NEW METHOD TX&PS

OK here we go again. The TX&PS schematic.

Tinkerer

[Tinkerer]

Controlled Flyback in the "TWO TAU METHOD"

Controlled Flyback in the "TWO TAU METHOD"

In the schematic, we see D2, R17, D3. These components control the Flyback.

Traditional PI tries to get the shortest possible Flyback decay. So the flyback is pushed as high as possible, usually all the way through the avalanche diode of the Mosfet. The energy is turned into heat. Gone up in smoke so to say.

Often this is several mJoules of energy.

The "TWO TAU METHOD" TAKES THIS ENERGY AND MAKES USE OF IT.

The CRO picture shows the resulting Flyback. It peaks around 140V, slowly slopes to 27V and then decays the traditional way.

What is the advantage? Look at the pictures at the following post.

http://www.geotech1.com/forums/showthread.php?t=15441

These sampling windows are taken during the controlled Flyback.

Not only we have a very high signal amplitude, (the amplification is not much at all), but we also have FE discrimination as well as information about the TC of the target.

Amazing what 3 passive components can do!!!!!!!!!!!!!!!!!!

Tinkerer

[Aziz]

Demystifying the Principle

Hi Tinkerer,

do you know, what these 3 parts are doing in the frequency domain?
(Ok, you know it already, we have discussed this on GPOZ forum former.)
You are using slow clamping diodes. Try to make a flat clipped flyback voltage. On the other hand, slow diodes make the flyback voltage more smooth (lower ringing, lower noise).


For others to understand the principle:

Let's assume, we have a constant magnetic impulse energy for two experiments (E=E1=E2). The flyback voltage is clipped to fixed values. To simplifiy the process, we are focusing only to the constant flyback voltage period. The exponential decay will not be taken into account. Usually, the flyback voltage will be clipped due to avalanche breakdown of the mosfet. Let's assume, we control the flyback voltage.

a ) U1=400V flyback voltage: duration 5µs (=dt1)
b ) U2=200V flyback voltage: duration 20µs (=dt2, it takes longer to damp the magnetic field energy at 200V, see proof below)


During the flyback period time, target eddy currents were induced. The version a ) would produce a higher target eddy current, if the skin effect would not present (we have a higher magnetic field energy conversion rate into heat: dE=U²*dt/Rd, thus resulting in a higher dB/dt damping).

But the short time duration dt1 of the flyback pulse has a much higher frequency spectrum, which lowers the targets conductivity due to skin effect. The eddy currents are forced to flow on the thin targets surface. The target seen as an energy storing element (coil), can not store much secondary magnetic field energy: first due to higher resistance (lower conductivity), second due to lower time period or charging (dt1=5µs).

Version b)

Ok, the target eddy current induction will be lower compared to a ) . But the eddy currents can penetrate the target deeper, hence resulting in lowering the targets resistance. More eddy currents can flow and the target can store more secondary magnetic field energy. Secondly, the charge time is longer due to enlarged flyback period (dt2 > dt1). The target eddy current can increase during this period (storing more secondary magnetic field energy).
See http://en.wikipedia.org/wiki/Skin_effect

Conclusions:

To overcome the skin effect, it is necessary to lower the frequency spectrum of the flyback pulse. It might be more advantage, to lower the flyback voltage to make it longer, even the dB/dt is lower. A constant flyback voltage should be better than a changing flyback voltage (nearly flat flyback voltage clipping). A good compromise must be found between the flyback period time and flyback voltage level. It depends on the desired targets to detect.

A technological gift:
(Note: Tinkerer and I will throw bad eggs, if someone tries to patent this idea)

A good detector would control the flyback voltage in each pulse period (multi period, hence each period has different flyback voltage level), to sense the skin effect. This would deliver more features (conductivity of targets, relative permeability of targets), which could be used for discrimination.

BTW, there is a similarity of different pulse width in PI's. These detectors achieve the same result in causing different frequency response of the pulses (different pulse width causes different flyback periods, shorter and longer ones).

So, you can use a fixed pulse width and control the flyback voltage to achieve the same result.

Now the one million dollar question:
Why is ML limitting the flyback voltage to 180V?
Why has ML detectors longer damping period (flyback period)?

Did I win the one million dollar?

Aziz


Proof for longer flyback duration:

The stored magnetic field energy E in the TX coil must be damped. Usually in the damping resistor Rd. The stored magnetic field energy E value can be calculated:
E = 0.5*L*I², L=TX inductivity, I=TX coil current

During the fixed flyback period, the damping resistor sees the flyback voltage and a current is flowing through the daming resistor Rd. It converts the magnetic field energy E into heat. The power is:
P = I²*Rd or P=U²/Rd
The power P is during this period constant as the flyback voltage is constant. The dissipated energy dE in Rd is:
dE = P*dt, dt=5µs or 20µs period
We have two experiments: U1=400V, U2=200V
P1 = U1²/Rd,
P2 = U2²/Rd

It is obvious, that if we take less energy out from an energy storing element (TX coil), we can take out the amount of energy longer in time. But not more than it has stored.
Thus:
dE1 = dE2
P1*dt1 = P2*dt2
(U1²/Rd)*dt1 = (U2²/Rd) *dt2

dt1/dt2 = U2²/U1² (independent of Rd)

dt2/dt1 = U1²/U2², if we put the values in the formula, we get the flyback time in experiment 2 (U2=200V)
dt2 = dt1*U1²/U2² = 5µs * (400V/200V)² = 5µs*2² = 20µs

q.e.d. or not q.e.d., that's the question

Note: We are not taking the exponential decay part into account for simplicity. The flyback voltage is not constant during this time and we have to calculate the dissipated energy in Rd with integrals (dE = P(t)*dt = U(t)²*dt/Rd).

[Aziz]

Some more thoughts

Hi all,

here are some more ideas for a new detector:

Multiple Flyback Voltage Limitting Periods (more than two):
The flyback voltage level will be regulated individually for each pulse period. This could excite different targets, while detecting all possible of them. So a short TC and a long TC target won't be missed. The number of periods, thus the number of different max. flyback voltage levels can vary.

The TX pulse width can additionally be combined with the method above.


Regarding the one million dollar question above:
Surely, I didn't win the price.
It has benefits of not letting the mosfet avalanche to keep it cool. A temperature stress and change to power switches has an instability effect to the detector. And ML recycles some of the flyback voltage back. The flyback voltage must not exceed the specified capacitor voltage (energy holding cap).

Aziz

[Aziz]

Bug in my proof


Sorry for the mistake in my calculation of the enlarged flyback time above (dt2). It is surely not as large as 20µs. It is lower of course.

Who can correct the proof/calculation?
Aziz

[Tinkerer]

Multiple Flyback Voltage Limiting Periods

Hi all,

here are some more ideas for a new detector:

Multiple Flyback Voltage Limitting Periods (more than two):
The flyback voltage level will be regulated individually for each pulse period. This could excite different targets, while detecting all possible of them. So a short TC and a long TC target won't be missed. The number of periods, thus the number of different max. flyback voltage levels can vary.

The TX pulse width can additionally be combined with the method above.


Regarding the one million dollar question above:
Surely, I didn't win the price.
It has benefits of not letting the mosfet avalanche to keep it cool. A temperature stress and change to power switches has an instability effect to the detector. And ML recycles some of the flyback voltage back. The flyback voltage must not exceed the specified capacitor voltage (energy holding cap).

Aziz

Aziz, thank you for the calculations. You also postulated the base for another good idea.

Multiple Flyback Voltage Limitting Periods (more than two): I twist your words a bit and make it VARIABLE Flyback Voltage Limitting Periods

The Zener D3 blocks the Flyback discharge until it reaches 27V. Then it opens until the voltage has fallen to 27 volts.
R10 then discharges the remainder of the energy stored in the coil.

I have tried to use a Mosfet avalanche diode of higher voltage instead the Zener. The Mosfet has a heatsink and is better equipped to handle the high power.

However, the Mosfets that I tried had noisy avalanche diodes.

But, if we use the Mosfet as a switched variable resistor, we can control it with a MCU or the FPGA and produce nearly infinitely variable Flyback voltage Limiting Periods.

For this we need to chose a Mosfet that has the voltage handling capacity and low noise.
We use soft switching to avoid high switching noise.

What do you think of it?

Tinkerer

[Aziz]

Hi Tinkerer,

a controlled linear coil current ramp (coil current at switch-off) down to 0A would be very nice. The function of the damping resistor can than be replaced by the current ramp controller. So varying the slope (dI/dt) would produce different flyback voltages for different periods.

In my dumb calculation above, there is a mistake:
The proof would be correct, if we would have an ideal (non-lossy) flyback voltage controller. This cannot be made in the real world of course. A lot of the flyback energy would be dissipated in either avalanche mode (mosfet) or in the diodes section rather than in the damping resistor. The damping resistor has then only little contribution to the energy conversion. Therefore, the enlarged period time would never be 20µs.
Just forget the proof. It was my mistake.

Damn, we have a thermal problem. It is causing a lot of stress to the TX part and therefore to the RX part at the end. We need to recycle the energy back in the coil to keep the power parts cool.

Aziz

[ZED]

The way i see it R17 is in parallel with R10 while the flyback is above 27v, R10 is`nt doing anything resulting in R17 and D3 getting hot with higher coil currents because there doing all the work.
Try connecting D2,D3 and R17 to -VB there by forming a voltage divider and spread the resistance a bit more evenly to share the load,also it may help to put a diode on the drain of the transmit fet.

The decay of the coil current at switch off on the m/lab machines is linear.

But putting that aside the build up of coil current when the transmit starts can also be linear but at a slower rate so in effect you would have two frequencys resulting from the build up and decay of the coil current during one transmit.Modifying the relationship of the two would be a better alternative than changing the duration of the flyback alone.

Hmmm i should put this in a provisional patent.


It would be interesting to see if the reactive component of the ground could be subtracted out in the process while leaving the target signal intact.

Zed

[Tinkerer]

TWO TAU METHOD

The way i see it R17 is in parallel with R10 while the flyback is above 27v, R10 is`nt doing anything resulting in R17 and D3 getting hot with higher coil currents because there doing all the work.
Try connecting D2,D3 and R17 to -VB there by forming a voltage divider and spread the resistance a bit more evenly to share the load,also it may help to put a diode on the drain of the transmit fet.

The decay of the coil current at switch off on the m/lab machines is linear.

But putting that aside the build up of coil current when the transmit starts can also be linear but at a slower rate so in effect you would have two frequencys resulting from the build up and decay of the coil current during one transmit.Modifying the relationship of the two would be a better alternative than changing the duration of the flyback alone.

Hmmm i should put this in a provisional patent.


It would be interesting to see if the reactive component of the ground could be subtracted out in the process while leaving the target signal intact.

Zed

Hi Zed,

thanks for the feedback.

Aziz and you are both right in that there is a thermal problem when the stored energy goes higher than a few mJoules.

The circuit is to show the principle. I can assure you that the response is beyond anything you can imagine.
Look at the posts at....http://www.geotech1.com/forums/showthread.php?t=15441
There you can see that this is not just an untested idea.

These tests were made with this circuit. Look at the signal amplitude (the gain factor is mentioned) and the discrimination. Have you ever seen anything like that with a traditional PI?

Now, to use a TX pulse of 10 Amps, this circuit would not function. Changes have to be made to account for the high power.

What is the function of R10?
It is the damping resistor. It is critical to the damping. More means oscillations. Less means a long decay. At 1k Ohm it is just right for the amount of mJouls it needs to discharge.

Ah, why does it go the the +12V? this way the power that is not dissipated in heat, charges the battery. Not really important.
Another reason is that, with the differential input of the preamp, I do not use the 12V rail as ground for the preamp.

Now, what are D2, D3, R17 for? This is what produces the second TX of the cycle.

The TWO TAU METHOD. 2 different TX in one cycle. 2 TX with different di/dt, therefore each TX optimally excites targets of different TC's.

One TX is negative going. It has a relative slow di/dt, or rate of change. The other TX is positive going, it has a much faster di/dt. A lot of power in few uSeconds. A fast or high rate of change.

But at the same time slow enough to produce deep going eddy currents in large targets.

As I mentioned above, just 3 little passive components give powerful results.

Please bring more questions. It gives me a reason to explain.

Tinkerer

[Aziz]

Hi Zed,

The decay of the coil current at switch off on the m/lab machines is linear.
Zed

Regarding switch-off current ramp:

If you regulate the high flyback voltage to a fixed value, the switch-off current ramp is linear anyway. It is a fundamental fact. So I did not want to mention this as m/lab method. The current ramp can be observed in switch-mode-power supplies (SMPS). Of course, the SMPS tries to regulate to a much lower voltage. But same principle.

Regarding switch-on current ramp:
Could be quite inefficient. More power could be heated up with the current ramp controller.

Aziz

[ZED]

Tinkerer running the diodes to -VB wont effect your preamp,the diode isolates -VB from the opamp input,either way your choice dude.

The transmits arnt going negative or positive,its the growth and collapse of the coil field that are giving you the different eddie currents,but you already know that ?

Yes i have seen the results of sampling during the flyback with my own experiments,see crow shots below,they have been posted before in other threads.Signals taken directly of the coil.



coil current constant after initial rise and then switch off,3 amps.

http://i220.photobucket.com/albums/d...ntwaveform.jpg

Target signal during transmit and during flyback,signal decays while coil current is constant.

http://i220.photobucket.com/albums/d...setocoil-1.jpg


The two scope shots are the result of two different experiments.


Zed

[Aziz]

ZED,

how much is the efficiency of the flat top current limit during TX-on period?

Aziz

[Tinkerer]

Tinkerer running the diodes to -VB wont effect your preamp,the diode isolates -VB from the opamp input,either way your choice dude.

The transmits arnt going negative or positive,its the growth and collapse of the coil field that are giving you the different eddie currents,but you already know that ?

Yes i have seen the results of sampling during the flyback with my own experiments,see crow shots below,they have been posted before in other threads.Signals taken directly of the coil.



coil current constant after initial rise and then switch off,3 amps.

http://i220.photobucket.com/albums/d...ntwaveform.jpg

Target signal during transmit and during flyback,signal decays while coil current is constant.

http://i220.photobucket.com/albums/d...setocoil-1.jpg


The two scope shots are the result of two different experiments.


Zed

Hi Zed,

yes I remember your scope shots of time ago. We should discuss this matter a bit more so that it does not get patented.

Tinkerer