PI: Effects of pulse shape and transmit time.

[Teleno]

The transmit current in current PI design have either a triangular or a trapezoidal shape due to the slow rise time of the inductance (V/L * t).

For a given coil, as the TX pulse gets shorter (T1 ... T5), the maximum achievable value of the TX current decreases like this:

Now I've designed a circuit capable of generating current pulses of a rectangular shape down to microsecond scale, like this:



Are there any advantages in a short, rectangular current pulse? for example: ground rejection, small object detection (large object rejection), or anything else?

Would it be worth the effort to go for such a design or is it just a waste of components and time?

[dfbowers]

I would suggest building the Hammerhead. Perfect learning platform for PIs! You can learn first hand what adjusting pulse width can achieve.
Instantaneous current across the coil would be more like this:
Don

[Teleno]

Instantaneous current across the coil would be more like this:
Don

That applies to the turn-off. The tun-on is more linear and is ramp-shaped because it occurs at a constant voltage: V/L * t; if you neglect the coil's resistance. The damping resistor in parallel has no effect on turn-on.

I can go around this and do a sharp turn-on. This allows shorter pulses witout decreasing the peak current, but I don't know if it's a desirable feature or not.

I would suggest building the Hammerhead. Perfect learning platform for PIs! You can learn first hand what adjusting pulse width can achieve.
Instantaneous current across the coil would be more like this:
Don

I apperciate the suggestion, but Hammerhead and all PI's I know use a trapezoidal/triangular current waveform (turn-on is a ramp) with the problem of limiting the peak current available for very short pulses.

My approach allows the peak current to be the same regardless of pulse width. My question: is a short, square pulse at peakcurrent a desirable feature or not? would it improve PI performance for certain applications?

[dfbowers]

Right.. I see your point. But, with PI detectors only a short portion of the decay at the very end is sampled. TX on time or ramp doesn't matter.. so unless you do some sort of VFL/PI hybrid with an IB coil it should not make any difference. Also, peak current will be limited by wire resistance anyway. Or are you suggesting multi period sensing?

[Teleno]

Right.. I see your point. But, with PI detectors only a short portion of the decay at the very end is sampled. TX on time or ramp doesn't matter..

The targets are roughly equivalent to LR circuits. Diferent targets hace different time constants. As the magnetic field turns-on eddy currents are induced in the target. If the field is switched off before the turn-on eddy has had time to decay, both eddy's cancel each other and you dn't get a target response. So the TX time can be used as target filter, shorter TX leaving larger targtes non-responsive. My aim is to detect small gold nuggets, which have shorter decays while avoiding magnetic ground that has longer decay. So I thought making the TX time as short as possible can help... but the turn-on ramp was in the way and I have fixed the problem.

[Teleno]

Also, peak current will be limited by wire resistance anyway. Or are you suggesting multi period sensing?

I want the peak current to stay the same for both long and very short TX times. This is not possible with current designs due to the turn-on current ramp. Currently you can only get lower than peak for short pulses.

I've found a way to eliminate the turn-on ramp, then shorter pulses are possible because the peak current is achieved almost instantly, you don't need to wait for a ramp to complete before turning off.

I believe this way ground response can be reduced and larger objects too (usually iron junk if you're lloking for gold nuggets).

[Silver Dollar]

Well are you going to show us what you propose to do?

I thought the limiting factor to turn on is LRC in the coil.

Some have used a centertapped TX and powered only one
side to good effect.

Others propose that you need to have a long TX to completely
magnify the target so it's response is longest.

And the ultimate shape is a half sine wave...

[Davor]

I want the peak current to stay the same for both long and very short TX times. This is not possible with current designs due to the turn-on current ramp. Currently you can only get lower than peak for short pulses.

I've found a way to eliminate the turn-on ramp, then shorter pulses are possible because the peak current is achieved almost instantly, you don't need to wait for a ramp to complete before turning off.

I believe this way ground response can be reduced and larger objects too (usually iron junk if you're lloking for gold nuggets).

This is equal to a constant current drive, and in effect the same as a current Whites patent signed also by Eric Foster himself, and that was a mild surprise to me. He did not strike me as a patenting kind of guy.
Now, to achieve a constant current from a get-go, you'd have to initiate it by an equivalent of a flyback pulse. There are generally two ways to do it, either by a high voltage source, or by another coil discharge. After initiation a coil is short circuited to maintain a constant current, and avoid power consumption, or much more wastefully by maintaining a constant current by voltage source and a current limiter. In any case, the process is finished by cutting the current altogether to produce a normal flyback, of opposite polarity than a charging pulse, followed by detection.

Now, your question is whether shorter charging pulse has any benefits. In short: no.

Because charging pulse and flyback pulses are of opposite polarity, if close to each other, some of target activation is lost, especially of longer tau. If charge and flyback pulses are further apart than a target tau, you will not be able to notice any difference. Somewhat more dramatic difference will happen with GB response that has a knee at a time after flyback equivalent to the charging period, and you'd wish it gone if you are after a perfect GB solution.

Duration of flyback is often debated, but with targets in mind, you can say that flyback duration is of negligible importance as long as it is shorter than the shortest tau, and it is as energetic as the integral of its shape. Taming a flyback makes sense as avalanche effects tend to be noisy, but that's another story.

[Teleno]

Now, your question is whether shorter charging pulse has any benefits. In short: no.

Because charging pulse and flyback pulses are of opposite polarity, if close to each other, some of target activation is lost, especially of longer tau.

So it can actually be used to silence targets of longer tau. I'm thinking ferromagnetic.

Somewhat more dramatic difference will happen with GB response that has a knee at a time after flyback equivalent to the charging period, and you'd wish it gone if you are after a perfect GB solution.

As I understand, ground response would also be quieter but not gone. It sounds like an improvement. Software could get rid of the rest, or two-level pulsing.

Since I've got nuggets in mind, it seems the constant current drive can be of advantage.

Thank you for the informative response!

P.S. I just saw the patent: http://www.google.com/patents/US8749240 Eric Foster gave me the exact answers I needed. Thanks again for the tip.

[Davor]

I think you got some ideas incorrectly. The longer tau targets will not be completely obliterated by short charging, and in terms of sensitivity, you'd not lose much. The biggest impact of this effect is perhaps in marketing of the devices with constant current pulses. Apparently many people suck those and spend bags of money on things that are not there.

As for GB, all nowadays GB solutions in PI rely on a ~1/t law of ground response decay which is linear in log-log scale. Introducing a knee in a perfect ground response curve would spoil your signal arithmetic, and hence separating charging pulse from flyback improves your odds of eliminating ground. In case of a rig without any sort of GB there is no difference, and ground is not quieter at all.

[Teleno]

Davor, do you have any pointers to the active damping circuits published in the forum? I'm having a hard time reading all of moodz threads and don't seem to see any implementation.

[Davor]

You'll hardly find any. Moodz was secretive about it during a process of obtaining a patent, and then he simply vanished from here disgruntled by a crop of trolls of the time and mishandling of the forum contributors. Lately he was convinced he was banned from here.
He is still secretive about his advances.

[Carl-NC]

Teleno, using a constant current TX with different pulse widths is a very worthwhile approach. Due to my involvement with prior and current employers I cannot offer much more detail, other than to suggest you should pursue this. Minelab has filed a couple of patents on CCPI, White's has one and should have another in the works, so CCPI methods are being commercially developed.

[Teleno]

Thanks Carl. I came up with the idea not knowing the big players were doing it already and calling it CCPI. Anyway I've designed and optimized my own circuit and I'm satisfied with it.

I'm also developing a new approach to cancelling magnetic ground response. It's based on a specially shaped pulse capable of exciting magnetic domains while ignoring conductive targets. I've designed such a pulse using convolution. All I need now is a pound or so of hot rock to start experimenting with. In my neck of the woods we only got common sand (The Netherlands).Perhaps some forum member from Down Under would be so kind as to provide me with samples typical of the gold fields, shipping costs on me... Hello! any Aussies reading this?

...And the ultimate shape is a half sine wave...

A half cosine wave is how the moodz active damping discharges the TX coil. I've done the SPICE simulation with this:

L1 1 0  20uH IC=1A 
C1 1 0  20pF IC=9V 
R_VAR 1 0 R='V(1) >= 0? 100000R : 1R + abs(V(1)*1.7)'

Command: "tran 1n 200n uic"

Where R_VAR is the current sink that extracts a constant current from the resonant circuit (moodz used a MOSFET polarized as a current sink).

This is what the discharge looks like when the MOSFET is properly biased:

[Qiaozhi]

Unfortunately LTSpice does not support the ternary operator. I tried to fudge the same functionality using a behavioural source in conjunction with the "if" function, but could not get exactly the same results. However, I understand how it's supposed to work, and your results look correct. In the end I gave up, as I'd already spent more time on it than I should.

Essentially (for those wondering what the ternary operator does) it's basically an IF-THEN-ELSE statement. In this case: If the coil voltage is greater than or equal to zero then R=100k, else it equals the product of the absolute value of the coil voltage and 1.7, plus 1. This produces a resistor that changes its value, when the coil voltage is negative, such that it sinks a constant current of approximately 600mA. If the coil voltage is zero or positive, the value is fixed to 100k.