New PI technology proposals.

[Carl-NC]

I am not totally agreeing with you. Even the current source can not start instantly.

I know, it's a mathematical exercise. IF you have an ideal current source and IF you switch it on to a coil with a very fast switch, you will get a di/dt and a flyback voltage that looks just like our turn-off.

The time constant of the coil is never changing (unclamped diodes period not taken into account). For both transmit on and transmit off, we have allways same time constants of the coil arrangement.

Oh, no, not at all!

The turn-on time constant is L/Rs where Rs is the series resistance of the coil (and the switch and the battery and whatever else is there, but let's just say the coil for simplicity). Assumes Rd is much much larger.

The turn-off time constant is L/Rd where Rd is the damping resistor. Assumes Rs is much much smaller.

This is why the coil is slow to turn on, and fast to turn off.

Example: L=300uH, Rs=3, Rd=600
turn-on tau = 100us
turn-off tau = 0.5us

This, of course, neglects the effects of C. This is a good exercise for an over-simplified Spice simulation!

- Carl

[Aziz]

Hi Robby,

..
This might be more obvious by the attached picture but I'm sure it will lead to more confusion. You can see though that what occurs during the on period is very important and does play a large part in the outcome.

that's correct. This gives a good opportunity to determine the targets time constant (discrimination) with different pulse width timings.
So the first (short) pulse should be quite low compared to the second (long) pulse. The second pulse should reach a constant current flow through the transmit coil. The first pulse should be at very low TC (0.5..1 TC) to have a higher dI/dt in the transmit coil. Very large TC targets will have considerable effects for processing. But normally, we have low TC targets and omitting the effects during transmit on phase.

I think ML is using such techniques.

Aziz

[gwzd]

Of course, the ultimate reality check is: build it!

- Carl

I couldn't agree more!
Just see how the nanovolt-lock-in-amplifier-laptop-metal-detector turned out!
regards,

[Aziz]

Hi Carl,

"Why is able to change so fast when turn on for the same voltage change is so slow?"

Different time constants (taus). Instead of charging the coil with a voltage source, charge it with a current source. Now you have the same time constants, with a very fast turn-on accompanied by another flyback voltage.

- Carl

I am not totally agreeing with you. Even the current source can not start instantly.

The time constant of the coil is never changing (unclamped diodes period not taken into account). For both transmit on and transmit off, we have allways same time constants of the coil arrangement.
The answer is in the coils self inductance. Any change of the current in the coil will cause self induction. This is at the beginning of the transmit pulse high and we can not instantly increase the coil current therefore.

On the switch-off period, the current change on the coil off is much higher. So we have a higher self induced flyback voltage. Now the damping resistor takes as much as current from the coil, not causing much self induction effects. What would have an effect to a short circuited superconducting coil? Magnetic energy will not be converted to heat and a steady self induction appers (coil ringing). (we are not taking any loss effects into account).

The effect of magnetic field energy conversion into heat in the flyback period is much more efficient due to Kirchhoff's law (P=U*U/Rd) compared to the charge of the coil during on-periods. But there is still self induction during flyback damping period which prevents us to kill the stored magnetic field energy instantly. The critical coil damping says us, you can damp the magnetic field energy with fastest possible speed without ringing the coil.

underdamped coil: ringing ->not convenient for processing, ringing takes too much time, the magnetic field will collaps many many times and will cause many many times of self induction and the magnetic field energy conversion into heat is low.

overdamped coil: two different time constants for discharging magnetic field energy -> difficult for processing due to higher self induction effects

critical coil damping: easiest way of processing

So we need more time for charging the coil to some level of magnetic field energy and we can - thanks to Kirchhoff - damp this stored magnetic field energy with shorter periods even the time constant of the coil is the same.

Aziz

[simonbaker]

The tricky thing is not forgetting about the double-differentiating going on.

Let's say maximum dphi/dt (at flyback peak) creates the maximum current in the target.

But don't forget that it is the dphi/dt from the target that will rock our RX coil! So we need to check: when is the maximum di/dt in the target!? I think that is what Aziz is referring to when noting that during MOSFET clipping, even though the di/dt of the TX coil should be at a steady maximum, and current in the target is maximum, the di/dt in the target will head downwards until flyback comes out of clipping. (Although I haven't completely convinced myself current is constant during MOSFET clipping because of the current through the MOSFET.)

So: when is the di/dt of the target the maximum? -- you need to carefully consider the combination of the forced response from the coil dphi/dt and the natural response of the target.

If the natural response of the target is really fast, then it will respond closely to the forced response, dphi/dt. So if we want maximum response in our RX receive coil, we want a maximum
d(dphi/dt)/dt (second derivitive) in our TX field.

So: we want to ask: how do we maximize the second derivitive of the TX current, not di/dt.

I'm sure that's not the whole story, but an important consideration. Or I'm just smoking bad stuff...

-SB

[Aziz]

Hi Qiaozhi,
That's correct. The maximum rate-of-change of current (target eddy current) occurs at the peak of the "flyback" voltage. At this point in the transmit coil, also the maximum change of dB(t)/dt occurs due to maximum magnetic field energy conversion into heat is possible.

Now consider the following:
We have charged the coil with lots of energy and lets say, we would have 2 kV flyback voltage. The mosfet is clipping at 400 V. Now the clipping period hasn't a change of current in the transmit coil anymore. As the flyback voltage is allways 400 V in this period, we have a constant current draw in the transmit coil (through the damping resistor). Looking at the same time to the target eddy current, we have despite of the not changing current in the transmit coil a constant high (at max) eddy current induction.
Why?
Because, we have reached a maximum magnetic energy conversion rate (dB/dt) and this rate is now constant, until the flyback voltage decays below the breakdown voltage. It is therefore misleading taking allways the dI/dt. We have to make some energy conversion balance sheet. Only dB/dt matters for target eddy currents.


Now to all:

Now the time constant of the transmit coil (T=L/R).
What is saying this (without taking coils capacitance into account)?
The inductor is an energy storing device like a capacitor. You can not store into and get out from energy as fast, as long and as much as you want. It has an "inner resistance" (coils resistance) and "capacitiy" (coil inductivity) which describes the behaviour.
What is happening to a voltage source if you try to get more current (energy) out? It is answering with voltage drop off. Same occurs at the flyback voltage decay if there isn't much energy stored. The inductor has also a self inductance, which describes the time behaviour. The critical coil damping gives the fastest discharge of the coil, without causing much self inductance, which will fight against our intention. As the current flow through the coil matters (self inductivity), we can change (increase) the flyback voltage level to make the energy conversion faster.


Aziz

PS: The rate of charging and discharging (coil current) of the coil is the same for transmit pulse on and flyback period. This is not changing. The time behavoir stays equal.

Carl,
I think you will find the implications are very significant for ground canceling pi designs, multiple pulse length pi designs, MP rectangular waveforms and off period time constant target discrimination.

unsubstantiated, speculation!

[Carl-NC]

"Why is able to change so fast when turn on for the same voltage change is so slow?"

Different time constants (taus). Instead of charging the coil with a voltage source, charge it with a current source. Now you have the same time constants, with a very fast turn-on accompanied by another flyback voltage.

- Carl

Good talk

Hi All, good talk, yes it is important to understand

what is actually going on. nice pic robby h.

limitations of Spice ? there aren't many, Spice is a

math program written usually by the University of Berkley.

It gives correct answers to electronic problems except for

a few bugs, but those are way above what this is.

Carl you got it right the first time, the models are wrong.

bad simulations as you posted above. An inductor in Spice,

without a series R is an infinite Q device. The capacitors are

also infinite Q devices without correct and more modeling.

The coil should be distributed RLC sections, the more the better.

Sometimes correct modeling can get quite complex, but will yield

much more accurate answers.

Now here is fun.

On turn on of the coil a voltage say 12 volts is placed on the coil.

The voltage is switched onto the coil fast say 10 nS. from zero to 12 volts,

But the current builds up slow because of the inductance of the coil,

say 300 uH, say it grows to 2 Amps. The rate of change of current

is slow, so not a great amount of eddy current is introduced into the target.

And given a fair amount of time at steady state 2 Amps the target

eddy current decays to close to zero. Now the same voltage is switched off,

from 12 volts to zero, turn off, in say 10 nS.

Does the 2 Amps current flowing in the inductor go to zero in 10 nS ?

Even if it takes longer, it is still much faster than turn on, for the same voltage.

Why is able to change so fast when turn on for the same voltage change is

so slow?

[Carl-NC]

Now, I'll ask this question without thinking about it -- is it possible to have two different coils that produce the same "flyback" voltage but have different dphi/dt affecting the target?

Sure... you have a system with 2 variables: Io (coil current at turn-off), and any one of R, L, and C. So let's say you decrease L... you have to increase C (maintain the same tau), then adjust R for critical damping... all you need to do now is adjust Io to get back the same v(t) flyback curve. Different L, same flyback.

I guess what we're all debating is: is there some optimization we can still do to get more out of these coils and circuits, or are the laws of physics and math trying to tell us to stop trying to make gold out of lead? Since it's impossible to solve exactly, the analogies and approximations in these discussions give some interesting food for thought.

I'm all for trying out new coil ideas, but I think there are 2 things that are critically important. #1: A good understanding of what is going on with pulsed currents, magnetic fields, and target eddies. And, #2, a good understanding of the limitations of Spice. It's really easy to get deceived by bad simulations of something you didn't quite understand in the first place.

Of course, the ultimate reality check is: build it!

- Carl

[robby_h]

I think it's only a factor with some of the high pulse frequency / low coil current designs Eric does.

Carl,
I think you will find the implications are very significant for ground canceling pi designs, multiple pulse length pi designs, MP rectangular waveforms and off period time constant target discrimination.

[robby_h]

Your post has left me a little confused.

In my attempts to investigate the claims (made by Aziz) that a center-tapped coil has advantages over a mono coil, when it comes to settling time, I did not include any models for the target. Also, the latter simulations were extreme simplifications to show the relationship between voltage and current in the search coil. In these particular simulations, I purposely excluded the coil capacitance and damping resistor. It was basically stripped back to the minimum parts required to demonstrate what's going on. Only Aziz's simulations included a target model.

So I'm not sure whether you're actually agreeing with me, or not.

The misunderstanding that some people have about PIs, is what actually "kicks" the target, thus creating eddy currents to flow. As Carl said earlier - the back-emf (or "flyback", if you prefer) voltage is simply an artifact. In reality it is the di/dt (rate of change of current with time) generated by the collapsing magnetic field (when the switch is opened) that creates the "kick". There is often a misconception that the ON-current somehow charges the target. This is incorrect.

My first post was an attempt to thrash out a model to explain what occurs at switch off and the part the spike plays in the collapse of the magnetic field. It wasn't an attempt to explain what "kicks" the target.

Your reply: "I think you are attempting to prove that the back-emf is not a negative voltage. This is incorrect".
But it is correct if I'm using a conventional circuit with the top end of a coil tied to the positive rail and the bottom end switched to the negative rail. In this case a negative spike would just switch the coil on harder for a period at switch off.

"If you monitor the voltage across the coil you will see that the amplitude is positive when the switch is closed. But when the switch is opened the collapsing magnetic field generates a large negative voltage (back-emf) in the coil".
Yes but I assume you must be switching the positive end of the coil. It is a large positive voltage (back-emf) in my case.

So I suppose I'm agreeing but also disagreeing.

As I see it...
The abrupt magnetic step at switch on and switch off "kicks" a target and both periods are made use of in PI IB designs but we are only concerned with switch-off for now. It might be best to just assume the pulse is extremely long and the field has been static for a very long time. The step at switch off occurs at the beginning of the leading edge of the spike and eddy currents immediately begin to flow in the target but the spike doesn't kick the target and the net coil current is effectively zero. If we have settled on a design then there is no further direct interference from the spike's clamp and settling time.
The duration of the spike, it's settling time or the duration of the magnetic step itself doesn't determine when target eddy currents reach maximum. The target determines this when trying to cope with the massive step from one state to another and the length of time it takes to cope.
This might be more obvious by the attached picture but I'm sure it will lead to more confusion. You can see though that what occurs during the on period is very important and does play a large part in the outcome.

[Qiaozhi]

Hi Aziz,

I think you're alone with this one.

Try post-processing your simulation results, and look at the plot of di/dt. You will see that Carl is correct. The maximum rate-of-change of current with time occurs at the peak of the "flyback" voltage.

[simonbaker]

Hi Carl,

I don't like the naming "artifact". It is an Uind = -L*dI(t)/dt (Lenz's and Faraday's rule). So dI(t)/dt in the coil end up to dB(t)/dt where the magnetic field energy will be converted to heat during the flyback process. This seems the fastest way of killing the B field. The target is contributing to this process more or less.

If the induced voltage is not dissipated through the damping resistor, it will ring and eddy currents will still be produced on the targets (the magnetic field energy will then be dissipated on targets and conductive objects nearby the coil). But we are not taking VLF and PI process to the same level. We have the damping process, which we focus on. And I still state, this is the most important process of a PI, which kicks the target on PI method.

Anyway, we are almost describing the same. We have only different names.

Aziz

In the interest of procrastinating from doing something more productive like making my TGSL PCB, I'll join in!

I think people call the "flyback" voltage an artifact to make sure no one thinks the voltage is somehow jumping through space and motivating electrons in the target.

Now, I'll ask this question without thinking about it -- is it possible to have two different coils that produce the same "flyback" voltage but have different dphi/dt affecting the target? If so, then the flyback voltage would truly be misleading for judging target response and maybe useful to call it an "artifact". If, however, the flyback voltage is tied to dphi/dt as concretely as V, I, R in ohms law in a DC circuit, then, well, it's another way of "measuring" the phenomena that affects the target.

It seems at least for a particular coil the "flyback" voltage is darn connected to the di/dt, and because we've got non-linear elements (MOSFET) that "clip" this voltage, it's a highly useful parameter to monitor and control.

Also, let's not forget that according to Einstein and Maxwell, there really is no such thing as a magnetic field. It's just a relativistic viewpoint of an electric field, and the only thing that moves electrons is an electric field. So in some respects, a voltage (electric field) really is jumping across space! (way out on a limb here...)

I guess what we're all debating is: is there some optimization we can still do to get more out of these coils and circuits, or are the laws of physics and math trying to tell us to stop trying to make gold out of lead? Since it's impossible to solve exactly, the analogies and approximations in these discussions give some interesting food for thought.

Cheers!

-SB

[KingJL]

Artifact: An inaccurate observation, effect, or result, especially one resulting from the technology used in scientific investigation or from experimental error: The apparent pattern in the data was an artifact of the collection method.

I'm with Carl... artifact.