PI TX Settling

[Ferric Toes]

Hi Teleno. I think we were misunderstanding each other as what you have now plotted in I(L2) is exactly what I meant. As you say, a balanced coil is needed to look at the on-time signal and you would need to be sure that the current and field had settled to a constant value before sampling. If still changing, the the target will unbalance the coil and the resulting signal will be a composite of the derivative of the target eddy current and the unbalance waveform of the tail of the current growth. This makes for a good discriminator as the balance goes one way for ferrous and the other way for non-ferrous.

Eric.

[Teleno]

Hi Teleno. I think we were misunderstanding each other as what you have now plotted is exactly what I meant. As you say, a balanced coil is needed to look at the on-time signal and you would need to be sure that the current and field had settled to a constant value before sampling. If still changing the the target will unbalance the coil and the resulting signal will be a composite of the derivative of the target eddy current and the unbalance waveform of the tail of the current growth. This makes for a good discriminator as the balance goes one way for ferrous and the other way for non-ferrous.

The design phase of this on-time metal detector is in an advanced stage. I expect to build a prototype by the end of the year. The settling time of the on-current will be fast enough, 1-2us.

The main advantage is a 2 x orders of magnitude increase in the the tau bandwith, since the amplitude of the targets is limited by the off-time (1ms), not the on-time (20us).

[Qiaozhi]

The design phase of this on-time metal detector is in an advanced stage. I expect to build a prototype by the end of the year. The settling time of the on-current will be fast enough, 1-2us.

The main advantage is a 2 x orders of magnitude increase in the the tau bandwith, since the amplitude of the targets is limited by the off-time (1ms), not the on-time (20us).

That will be interesting.

[Qiaozhi]

Here's a variation on the previous simulation, with each TX pulse ending at the same time, instead of starting at the same time.
To make the display less crowded, I'm only showing the results for a pulse width of 50us and 500us. Although in both cases the coil contains the same amount of energy, it is obvious that you need to wait for 5 target TCs to allow the reverse eddy currents to dissipate. Otherwise the reverse eddy currents will cancel some of the eddy currents generated when the TX turns off. For the 50us pulse width, this gives a significant degradation in the target signal. In fact, more than 50% of the target signal is lost in that particular case.

The 50us pulse width is shown in green, and the 500us is blue.

[Teleno]

For the 50us pulse width, this gives a significant degradation in the target signal.

You can also see how the on-time target signal does not degrade at all. Long taus can be measured at full sensitivity with very short pulses.

[Qiaozhi]

You can also see how the on-time target signal does not degrade at all. Long taus can be measured at full sensitivity with very short pulses.

Yes, I can see what you're getting at.
This will be true (of course) because the TX-on pulse doesn't have to fight with any reverse eddy currents in the target.

[green]

What is the difference or advantage of sampling a PI during Tx on time vs using a VLF detector?

[Teleno]

What is the difference or advantage of sampling a PI during Tx on time vs using a VLF detector?

All the advantages of a classic PI vs VLF except the need of a balanced coil.
Added advantages vs classic PI:

1. faster transient and settling time,
2. increased range of targets detectable (no attenuation of long taus, earlier sampling of shorter taus),
3. power savings (shorter on-time),
4. full ground response (1/t) for accurate ground modeling and balance.

[Sean_Goddard]

For what my two penneths worth might offer, I used a circuit which measured when the coil di/dt was zero then switched off, meaning the coil I was always optimum. THis seemed to give much better performance noise and depth wise. I have the circuit comewhere. Will post it if I ever find it.

This gives me a thought that if results for target response are best when the target tau matches the Tx tau then with a series of variable width pulses, matching tau to the target will produce a bell curve of results for a swept Tx tau. If these could be correlated to known profiles this might be a form of discrimination. Thoughts?

[Carl-NC]

What's being discussed is what I call "constant current PI" (CCPI), where the turn-on time is roughly as fast as the turn-off time. One way to do CCPI is to pulse the coil on & off with a dead time in between:



I did some work on this approach at White's. Another way is to continuously slam the coil between a positive & negative current:



This is what the Minelab GPZ does.

My personal preference is for the dead time approach. First, it avoids Minelab IP . Second, it's easier to implement. Third, the dead time is very useful. Fourth, although I haven't pursued this approach, I expect it would be useful to have the ability to control the turn-on slew rate. This is also easy to do.

And, yes, I'm still pursuing CCPI, "in my spare time."

[Brian Deese]

Here's a variation on the previous simulation, with each TX pulse ending at the same time, instead of starting at the same time.
To make the display less crowded, I'm only showing the results for a pulse width of 50us and 500us. Although in both cases the coil contains the same amount of energy, it is obvious that you need to wait for 5 target TCs to allow the reverse eddy currents to dissipate. Otherwise the reverse eddy currents will cancel some of the eddy currents generated when the TX turns off. For the 50us pulse width, this gives a significant degradation in the target signal. In fact, more than 50% of the target signal is lost in that particular case.

The 50us pulse width is shown in green, and the 500us is blue.

So what I am taking from this is we need to flat top our tx current,and hold it there for the full 5tc of our desired target so the charging eddys will die out and not be counter productive to the rx signal strength. What is the method to get constant current in the coil other than series resistance?

[Qiaozhi]

So what I am taking from this is we need to flat top our tx current,and hold it there for the full 5tc of our desired target so the charging eddys will die out and not be counter productive to the rx signal strength. What is the method to get constant current in the coil other than series resistance?

According to Eric and Carl, you can wait for 3TC of the desired target, and it will close enough without wasting power to get the last few percent of target signal.
The simple method is to use a series resistor, but that fixes the TX pulse width, which is not adjustable without also changing the resistor value. The more complicated method is to use a constant current source as the power supply. I fudged it in the simulation by using an ideal constant current source with a zener diode in parallel to stop the voltage rising above approximately 12V.

Teleno presented a solution here -> http://www.geotech1.com/forums/showt...179#post214179

[Brian Deese]

Right using higher Rs will set the coil tau lower R=L/t so the current rises to peak in 5tc,so since you set the coil to lower tau the current rise time will be relatively short compared to your long tx on time. If we are going to have constant current in our coil for our targets 5tc,then we cant have exponentially rising coil current matched to out targets 5tc. One or the other,but not both. And will a three terminal reg like a lm317 make a fit constant current supply or is there a better choice?

[deemon]

First, it avoids Minelab IP .

Proprietary IP isn't a problem here , because my CCPI ( square wave pulse induction ) project is a kind of "open source"
All theory , pictures and circuits are in public domain from the very beginning , and everybody can contribute ...

[green]

According to Eric and Carl, you can wait for 3TC of the desired target, and it will close enough without wasting power to get the last few percent of target signal.
The simple method is to use a series resistor, but that fixes the TX pulse width, which is not adjustable without also changing the resistor value. The more complicated method is to use a constant current source as the power supply. I fudged it in the simulation by using an ideal constant current source with a zener diode in parallel to stop the voltage rising above approximately 12V.

Teleno presented a solution here -> http://www.geotech1.com/forums/showt...179#post214179

Another attempt at measuring the TC of a US quarter and a 1 troy ounce 99.9% pure copper coin(guessing it would be close to a US silver dollar). US quarter TC(at least 125usec), copper coin TC(at least 380usec). Are you suggesting a 1200usec flat TX when looking for a US silver dollar?