PI Power Output

[robby_h]

B^C,
Quite simple really, I just don't know what I'm looking at or where and in what part of a circuit.
Your first pic shows a flat line with small glitches. The third pic looks very similar to the three short pulses in a GP pulse train, ie, ~40usec on and ~80 off. This then gets more confusing in the 4th pic because of the close spacing.

How did you obtain these waveforms?

[robby_h]

Zed,
Sampling during the on-period isn't nigh on impossible as it is routinely done in all CW metal detectors transmitting sinusoidal and rectangular waveforms.

"Do you think the reactive component of the ground response to be a problem world wide".

It depends on what you mean by this. USA goldfields have a substantial reactive component but little R component as is evident when someone says they can set a 2100's GB pots anywhere, something which isn't possible here. This also applies to the TDI where the USA operator can set the GB to set the TC pivot point for iron and coins etc. My local park has areas that require critical ground balance on a GS5 with a ten turn pot so this wouldn't work here either. This USA ground though is often reported to give a ground balancing vlf fits so it's safe to assume these soils have a very substantial reactive component.

"Lets not forget that the reactive ground signal will only ever be a problem while the coil field is in a state of change,with a static coil field we can kiss the reactive component goodbye and say hello magnetic component .
Now sampling during the ontime while the coil field is static.....thats an interesting place to go".

But this then dictates when you can sample and you will miss a lot of targets. There are other ways around the problem if you have been reading the relevant patents.

[B^C]

robby_h,

"Your pictures appear to be what we would see with a balanced RX coil, which is just IB? The time base appears to have been changed in each picture along with the pulse train so it's hard to say what we are supposed to be looking at?"
----------------------------------------------

Your correct in that the time base has changed, this is due to a manually triggered pulse setup.

Not sure how it could be the same as with a balanced RX coil-IB, could you explain why it appears to be the same as the test setup has no similarity to this or anything else for that matter?

I would appreciate your input.

[ZED]

G`day Robby_h

What you say is true,the reactive component of the ground signal during the on time does make the idea of sampling during the on time nigh impossible,....for now! ..well in Australia anyway but ground conditions vary through out the world as you may well know.
Do you think the reactive component of the ground response to be a problem world wide.
Lets not forget that the reactive ground signal will only ever be a problem while the coil field is in a state of change,with a static coil field we can kiss the reactive component goodbye and say hello magnetic component .
Now sampling during the ontime while the coil field is static.....thats an interesting place to go.

Zed

[Qiaozhi]

Hi Tinkerer,
I don't know who suggested we can sample earlier with a balanced coil arrangement but it is clearly wrong. We have to obey certain rules and one important rule is that we must delay sampling until after the TX coil's back emf spike has decayed. There is no way around this.

Your waveforms in post #30 are clearly IB, not pulse induction. You have actually built a high powered IB detector and also given a good example of why we can't sample earlier than at the tx coil when using a balanced rx coil in a pi design.

Your figures in the other thread are L= 325 uH, total R = 2.8 ohm, U = 12v, T = 41usec, therefore the coil current at switch off should be 12(1-exp(-41*2.8/325))/2.8 = 1.27 amps so with this current I think it's safe to assume the tx coil's spike will still be settling at the far right of your screen (~7 usecs) but your sample begins 2.5 usec after switch off and ends 5.2 usecs after switch off so you are looking at a portion of the spike's waveform using a balanced receive coil.
In post #59 on this thread, your conclusions are based on the spike's properties which will be almost identical for the 43 and 100usec pulses so your results are similar because you are sampling the same spike in both cases. Pulse induction would give different results.

The problem with sampling during the pulse on time (or the spike's on time) is that we have to deal with the reactive soil component whereas we don't during a true PI off period.

With pi, we aren't looking at the spike or how the target affects the spike's amplitude, decay, width or settling time, we are looking at the signal the target induces in the rx coil. The two might appear to blend together at the tx coil but they are separate. Some articles, papers and patents may say otherwise but they have it wrong, such as in this paper written by someone who supposedly has a PhD....

http://engnet.anu.edu.au/DEpeople/Salman.Durrani/_teaching/TA5.pdf

Dave Emery's patent is also IB, not PI. He makes the spike resonate with a cappy and he samples the oscillations similar to vlf although he makes no mention of cancelling the ground R component or that it even exists.

Allan Westersten's patent US20080224704 is a mix of pi and IB but he doesn't include a means to separate the off-period small target trigger signal from the huge ground signals we experience here in the Oz gold fields.

B^C,
Your pictures appear to be what we would see with a balanced RX coil, which is just IB? The time base appears to have been changed in each picture along with the pulse train so it's hard to say what we are supposed to be looking at?

I suspect you are correct in your statements. That's why I suggested the experiment using a limited number of common targets at different distances from the coil. Any correlation between these targets and the actual target response will most likely vanish once the graphs are plotted.

The only way anyone will prove (one way or the other) if there is a correlation, is to build a prototype and test it in the real world. SPICE simulations are not going to add any benefit at this time.

[Qiaozhi]

I don't know how to represent a target for simulation. I only know how to find the target with a detector.

If you think there is some possibility of adding discrimination to a PI by comparing the response during the ON-time and the OFF-time, then you could try the following with your setup:

Take a limited number of common targets (coins, rings, pulltabs, etc.) and record the samples for each target at different distances from the coil. Then plot these on a graph for comparison.

Although this experiment appears relatively simple , it is complicated by the balancing of the IB coil. For anyone else to duplicate your results, they would need to have their coils set up in the exactly the same way.
This is why simulation is going to be tricky, to say the least. I've had this working successfully in SPICE for a VLF (with the correct phase-shifts) so that it matches reality, but it's a matter of working with both a prototype and the simulator at the same time to fine tune the models. In this particular case, we would also need to take measurements on the real circuit, and then add the findings to the simulator. Many times people get bogged down in simulations, and they forget that it's just another tool. SPICE simulations are extremely useful for getting you in the right ballpark, but eventually you have to build a real circuit. It's then that you find out which of your assumptions were correct, and which were not. Then you can go back and refine the models to match reality, and in the process make other discoveries.

So, what I'm trying to say, is that (in this case) we need a number of real world measurements, before exploring more deeply with the simulator, because of the added complication of using an IB coil, which also has an unknown setup ..... if that's clear?

[robby_h]

This picture was originally drawn by Eric Foster and posted on the PI forum. The cupro nickel's waveform might look dramatically different to the silver ducat's and al ingot's, but note that it is just compressed time wise, ie, just a shorter TC. You might also see why disc is very difficult when ferrite is mixed with a coin, or even worse, a gold nugget.
Some proposed disc methods rely on the coils being critically balanced, such as we might have on the work bench, but this can be rather meaningless if the ground upsets the balance.

Eric's explanation for his waveforms here...

http://www.findmall.com/read.php?34,129762,129778#msg-129778

I hope you realise that ML know these waveforms extremely well and you might eventually see why they favour square or rectangular transmitted waveforms over a PI on-period.

[robby_h]

Hi Tinkerer,
I don't know who suggested we can sample earlier with a balanced coil arrangement but it is clearly wrong. We have to obey certain rules and one important rule is that we must delay sampling until after the TX coil's back emf spike has decayed. There is no way around this.

Your waveforms in post #30 are clearly IB, not pulse induction. You have actually built a high powered IB detector and also given a good example of why we can't sample earlier than at the tx coil when using a balanced rx coil in a pi design.

Your figures in the other thread are L= 325 uH, total R = 2.8 ohm, U = 12v, T = 41usec, therefore the coil current at switch off should be 12(1-exp(-41*2.8/325))/2.8 = 1.27 amps so with this current I think it's safe to assume the tx coil's spike will still be settling at the far right of your screen (~7 usecs) but your sample begins 2.5 usec after switch off and ends 5.2 usecs after switch off so you are looking at a portion of the spike's waveform using a balanced receive coil.
In post #59 on this thread, your conclusions are based on the spike's properties which will be almost identical for the 43 and 100usec pulses so your results are similar because you are sampling the same spike in both cases. Pulse induction would give different results.

The problem with sampling during the pulse on time (or the spike's on time) is that we have to deal with the reactive soil component whereas we don't during a true PI off period.

With pi, we aren't looking at the spike or how the target affects the spike's amplitude, decay, width or settling time, we are looking at the signal the target induces in the rx coil. The two might appear to blend together at the tx coil but they are separate. Some articles, papers and patents may say otherwise but they have it wrong, such as in this paper written by someone who supposedly has a PhD....

http://engnet.anu.edu.au/DEpeople/Salman.Durrani/_teaching/TA5.pdf

Dave Emery's patent is also IB, not PI. He makes the spike resonate with a cappy and he samples the oscillations similar to vlf although he makes no mention of cancelling the ground R component or that it even exists.

Allan Westersten's patent US20080224704 is a mix of pi and IB but he doesn't include a means to separate the off-period small target trigger signal from the huge ground signals we experience here in the Oz gold fields.

B^C,
Your pictures appear to be what we would see with a balanced RX coil, which is just IB? The time base appears to have been changed in each picture along with the pulse train so it's hard to say what we are supposed to be looking at?

[simonbaker]

Simonbaker,

you lost me there.
I don't know how to represent a target for simulation. I only know how to find the target with a detector.
The Tau I was talking about is the Tau of the sine wave or the ON and OFF transients. Basically how steep the sine is. Like slew rate of an opamp.
Assuming, and I have no proof that this is so, a target with a TC of 5us would reach its peak eddy currents very quickly. A very steep charge curve. And it will discharge at the same rate.

Tinkerer

Right, I think we're talking about the same thing, it's in the equations but not everyone is familiar with the equations. Graphs work well. Seems like Aziz was making some headway with his graphs. I think if we keep making those graphs we'll at least get the answers for idealized targets on many of the questions.

Cheers,

-SB

[Tinkerer]

Ok, Tau (time constant) is a good one. For example, just saying "Tau" implies we are typically talking about a "first order" linear system, because a second order system has two different time constants (unless they are identical). A first order system (if I haven't forgotten) is like an RC or RL circuit - it can't "ring"; a second order system is like an RLC circuit - it can ring, depending on the damping factor. First order systems can make low-pass or high-pass filters. Second order systems can make band-pass filters. Familiar stuff.

So basically Tau completely describes a first-order circuit, which is nice and simple. Now I wouldn't say with VLF detectors that the Tau has a certain value at a particular frequency because it really is just a property of the circuit. But the Tau will determine the phase shift at a particular frequency, which is probably what you are getting at.

So if we assume that different targets have different Taus, and we want to discriminate targets, then what we're really trying to do is estimate the Tau of our target by measuring something that Tau affects, like phase shift or the decay of the current in the target.

I'm not saying Tau really is a good way to discriminate targets, but we can try using it, and we can estimate it in various ways.

If we can model our targets as "linear" systems, we're home free because so much is available to predict exactly how linear systems respond to stimuli. There are the terms "forced response" and "natural response", and we really need to be aware of those concepts to discuss the pros and cons of using the TX turn-on versus the TX turn-off, because that's what were dealing with.

I mean it's really all there, the answers to how linear systems respond to any kind of stimulus. I'm just saying this because it seems we're beating around the bush about subjects that are extremely well known, such as whether a slower or faster forcing function will create this or that response etc.

The real question probably is just how to model a real target. For instance, is Aziz's linear RLC circuit realistic, or is something very important being left out.

Other than that, if we can assume the target is a first or second order linear system, then if anyone asks the question clearly enough, the answer should be available with a little calculation or LTSpice simulation. Getting those answers would be a good way to sharpen our intuition and get us all on the same page so to speak. I think Aziz made a couple of good stabs at it, we should keep pushing along those lines.

Regards,

-SB

Simonbaker,

you lost me there.
I don't know how to represent a target for simulation. I only know how to find the target with a detector.
The Tau I was talking about is the Tau of the sine wave or the ON and OFF transients. Basically how steep the sine is. Like slew rate of an opamp.
Assuming, and I have no proof that this is so, a target with a TC of 5us would reach its peak eddy currents very quickly. A very steep charge curve. And it will discharge at the same rate.

Tinkerer

[simonbaker]

You are right. To be able to work together on the subject, we need to agree on the terms.
To start with "saturation", lets banish this word. The meaning is the moment of the maximum response or the maximum generated eddy currents.

Is skin effect linear? I don't know. We need to research and dig up the information.

There is a wealth of information available relating to VLF. This is the same information that is applicable for PI. we just have to see where it fits in.

Here is an example:

Maybe you would be so kind to word it better for me.

With VLF we have a sine wave with a certain frequency. At a certain frequency the Tau has a certain value.

With PI we have a transient that has a Tau with a certain value. The effect is the same. The method of reading is different.

Tinkerer

Ok, Tau (time constant) is a good one. For example, just saying "Tau" implies we are typically talking about a "first order" linear system, because a second order system has two different time constants (unless they are identical). A first order system (if I haven't forgotten) is like an RC or RL circuit - it can't "ring"; a second order system is like an RLC circuit - it can ring, depending on the damping factor. First order systems can make low-pass or high-pass filters. Second order systems can make band-pass filters. Familiar stuff.

So basically Tau completely describes a first-order circuit, which is nice and simple. Now I wouldn't say with VLF detectors that the Tau has a certain value at a particular frequency because it really is just a property of the circuit. But the Tau will determine the phase shift at a particular frequency, which is probably what you are getting at.

So if we assume that different targets have different Taus, and we want to discriminate targets, then what we're really trying to do is estimate the Tau of our target by measuring something that Tau affects, like phase shift or the decay of the current in the target.

I'm not saying Tau really is a good way to discriminate targets, but we can try using it, and we can estimate it in various ways.

If we can model our targets as "linear" systems, we're home free because so much is available to predict exactly how linear systems respond to stimuli. There are the terms "forced response" and "natural response", and we really need to be aware of those concepts to discuss the pros and cons of using the TX turn-on versus the TX turn-off, because that's what were dealing with.

I mean it's really all there, the answers to how linear systems respond to any kind of stimulus. I'm just saying this because it seems we're beating around the bush about subjects that are extremely well known, such as whether a slower or faster forcing function will create this or that response etc.

The real question probably is just how to model a real target. For instance, is Aziz's linear RLC circuit realistic, or is something very important being left out.

Other than that, if we can assume the target is a first or second order linear system, then if anyone asks the question clearly enough, the answer should be available with a little calculation or LTSpice simulation. Getting those answers would be a good way to sharpen our intuition and get us all on the same page so to speak. I think Aziz made a couple of good stabs at it, we should keep pushing along those lines.

Regards,

-SB

[B^C]

Gday Zed,

Thanks for the replies, things will be more understood if there is more conversation.
The problem i have is that there are NO Fets, no MD circuit at all, all this was avoided for the reasons you mentioned.
The Target reading device is 1mtr away from the inductor.

A bigger Target has no flat top as you put it on the signal but more peaks to a point & then decays but small targets do indeed have a flat top signal as mentioned, it does appear that they have reached there maximum.

Increases in voltage on the same smaller targets deliver the same signal results, but on larger targets there is a marked difference in the signal readings.

I get what your saying about saturation being the end etc, but this depends on how one looks at things i guess. Remember only yesterday when we all thought the world was flat?

I would explain more but i need a friendly response hahaha!

Any thought's would be appreciated.

[Tinkerer]

This is a very good point. Unless we are all singing from the same hymn book (so to speak) we'll find ourselves just going round in circles.

I would be very happy if you would write the hymn book for us. My vocabulary is very restricted in English as well as in Electronics.
The Google translator helps, but some things always get lost in translation, so we end up comparing apples with beets.

Tinkerer

[Tinkerer]

Hi Tinkerer - I'd like to understand better your diagrams.

First, I'm not sure what you are demonstrating. Are you demonstrating that TX turn-on can have stronger target response than TX turn-off? Or are you illustrating two different TC targets' response to TX turn-on? I'm confused because graphs b and c seem to be about TX turn-off, but I don't see how they tie in. Would you explain b and c a little more?

Looking at graph e, are you showing that a longer TX pulse allows the slower target current to grow more and therefore maybe there is an advantage to the slow turn-on phase compared to fast turn-off?

I have a couple of thoughts about that, which if I were smarter maybe I could show with equations. The first thought is that when we detect a target, our detector coil responds not to the peak current of the target but rather to the peak rate-of-change of the target current (same physics as the target responding to the rate-of-change of the TX current). So really in graph e it is the fast up slope and fast down slope that makes the received "signal". I'm not sure how that affects your point, but may be relevant.

The second thought is: is what you are getting at like the concept of "tuning", where the target responds better to a particular wave shape than others? For instance, suppose our target is underdamped and would "ring" in response to a pulse. Then it would make sense to create an on-off pulse that turns on and off at the ringing frequency so as the magnetic field reversal kicks the target in both directions synchronously with it's natural oscillation. I could buy that, although it would mean only certain targets would benefit from a given pulse width. Perhaps there is an analog to underdamped targets as well.

Ok, enough for today.

Cheers,

-SB

For certain targets, we can read a higher signal amplitude response during TX on than after TX OFF.
Consider a target with a TC of 5uS. Sampling at the very moment of maximum response, during TX you can read this maximum value.
Sampling the same target 15uS after switch OFF, you will get at best 5% of its maximum response.

For graphs b) and c) please refer to post #43. There you can see how the response of the target lags behind the excitation.

The same happens with the ON transient. The response lags behind the excitation.
I agree that the strongest excitation happens with the moment of highest di/dt, but there is a time lag between the "charge " of the target. Commonly we represent the target with LCR. So we can expect an exponential charge curve.
The unknown factor is the values for LCR, but we have the known factors of di/dt and the total time of the exposure of the target to this factor.

What we are trying to do with discrimination is to calculate the values of the target's LCR.

My graph e) shows the signal generated by 2 different targets, during TX only.
Now, if you use both the responses generated by the ON transient and the OFF transient, you have the information that would be generated by a low frequency and a high frequency VLF.

About "target ringing" I have no information yet. During my experiments I often found this "invisible sine wave" and was wondering if this represented a ringing target. After all, the eddy currents expand throughout the target, but where do they go when they have reached the far end? are they reflected back? If yes, ringing could be possible. Just speculating.

Now I believe that the PIVOT as a part of the "invisible sine wave" is simply a different Tau.

Tinkerer

[Tinkerer]

It would be really useful to state again precisely exactly what your theory or theories are and what any disagreeing theories are. It gets hard to follow this because we all use different language and analogies -- there might be more agreement than we think except our words make it seem different.

My feeling is that if we had a really decent math model of a target, we should be able to answer a lot of these interesting questions. I think the coil magnetic field is pretty well understood because we can measure the voltage and current fairly well and it's pretty linear. But the target is harder to measure.

First of all, is a target "linear" or non-linear? Is skin effect non-linear? I ask the question because non-linear systems are really hard to get a handle on -- hitting them hard is different from hitting them soft, they don't scale.

But my basic question is: can't we treat this discussion like any other electronic circuit -- write a differential equation for the response to a stimulus of our choice and run it out? We throw around ideas like "saturation" etc., but these should show up in our equations.

If targets are "linear" in terms of response, then a lot of stuff should be well understood. For instance, response to an impulse (PI TX pulse) and to a sine wave (VLF TX signal ) are completely related - you know one, you know the other (of course you need all sine waves...). And it's no big deal to take a specific pulse shape and run it through a computer program and see the response and answer the question of how best to sample the PI response.

So it seems like we need to get a really good model of some targets and get running some simulations. Aziz, are you available?

Now if it doesn't match real life, good. More work is need on the models. Once you can match reality, you can understand what's going on.

I suggest this because one man's saturation may be another man's steady state -- loose terms that are misleading unless we tie them to an equation.

I'm not saying stop talking with analogies, it's fun, I'll keep doing it too. But it seems we go in circles because we don't have a common language to nail down conclusions. Unfortunately, that language is mathematical models.

I'd like to start with the concept of "saturation" of targets, and ask what is really meant by that and what analogy is behind it?

Cheers,

-SB

You are right. To be able to work together on the subject, we need to agree on the terms.
To start with "saturation", lets banish this word. The meaning is the moment of the maximum response or the maximum generated eddy currents.

Is skin effect linear? I don't know. We need to research and dig up the information.

There is a wealth of information available relating to VLF. This is the same information that is applicable for PI. we just have to see where it fits in.

Here is an example:

Maybe you would be so kind to word it better for me.

With VLF we have a sine wave with a certain frequency. At a certain frequency the Tau has a certain value.

With PI we have a transient that has a Tau with a certain value. The effect is the same. The method of reading is different.

Tinkerer