DEEPER PI DETECTION DEPTH

[Tinkerer]

Another possible conclusion is that a "pulse" is not the ideal TX signal, because a lot of the energy is spread in the lower frequencies. Perhaps we need a "high frequency pulse" (whatever that is) that concentrates all the energy up in the higher frequencies. Well, that is probably some kind of sync function that moodz and others were toying with. Or heck -- just make a continuous wave hi-freq MD. Oops -- I forgot. The phase difference seems to disappear at higher frequencies, making discrimination harder. But wait... maybe a tuned RX coil would help (probably not much). Well, maybe we have a trade-off to work on.

We'll see more interesting designs I'm sure.

-SB

Logic deductions get you there. It is the TEM method. A continuous wave TX, with 2 different wide spectrum bands of high power stimulating frequencies.
Just scroll up a bit to see the wave forms.

Tinkerer

[simonbaker]

Hi guys,

I am happy now, that you all are dealing with and talking about the frequency domain to understand, what is happening there.


Regarding the TX coil current and the 20 dB roll-off of it:
The higher the (stimulation) frequency, the lower the TX coil currents gets due to the impedance (reactance XL) of the TX coil. But this doesn't matter as the dI/dt increases proportional with the frequency and the induction law compensates this in the RX coil. This is the reason, why the RX response tends to go to the same flat level (see the AC response I'm still talking about). High TC targets go in the low frequency region to this level and low TC targets go in the high frequency region.
Interesting to know, that in the high frequency region, all targets respond with the same level of response. Whereas in the low frequency region, low TC targets will be missed due to low response.

So if you want to detect the low TC targets, you need some higher bandwidth (covering the high frequency region). Higher bandwidth is achieved with fast amplifiers and fast coils (see bbsailor's article).

Aziz

Another possible conclusion is that a "pulse" is not the ideal TX signal, because a lot of the energy is spread in the lower frequencies. Perhaps we need a "high frequency pulse" (whatever that is) that concentrates all the energy up in the higher frequencies. Well, that is probably some kind of sync function that moodz and others were toying with. Or heck -- just make a continuous wave hi-freq MD. Oops -- I forgot. The phase difference seems to disappear at higher frequencies, making discrimination harder. But wait... maybe a tuned RX coil would help (probably not much). Well, maybe we have a trade-off to work on.

We'll see more interesting designs I'm sure.

-SB

[Aziz]

Hi guys,

I am happy now, that you all are dealing with and talking about the frequency domain to understand, what is happening there.


Regarding the TX coil current and the 20 dB roll-off of it:
The higher the (stimulation) frequency, the lower the TX coil currents gets due to the impedance (reactance XL) of the TX coil. But this doesn't matter as the dI/dt increases proportional with the frequency and the induction law compensates this in the RX coil. This is the reason, why the RX response tends to go to the same flat level (see the AC response I'm still talking about). High TC targets go in the low frequency region to this level and low TC targets go in the high frequency region.
Interesting to know, that in the high frequency region, all targets respond with the same level of response. Whereas in the low frequency region, low TC targets will be missed due to low response.

So if you want to detect the low TC targets, you need some higher bandwidth (covering the high frequency region). Higher bandwidth is achieved with fast amplifiers and fast coils (see bbsailor's article).

Aziz

[simonbaker]

The way I see it, it is if little use to stimulate the targets with higher than 300KHz. A traditional PI is not capable of detecting targets with a TC if 1us.

With the TEM method I can sample any time so there is maybe an advantage with higher frequency stimulation.

The FFT shows 2 stimulation frequency bands.

Low frequency with +2.5dB at 5KHz and -20dB at 66KHz.

High frequency stimulation at a different time, +5.75dB at 70Khz and -20dB at 1Mhz.

Probably -20 dB is way beyond any usefulness, maybe we should put the limit of interest a lot higher.

Ideas and comments?

If you plan to implement target detection in the time domain, it's probably just easier to look at each target's time domain response and work on strategies to detect it and maybe make discrimination judgments.

Then vary the things you can control, such as the TX pulse shape and the RX receive circuit (response characteristics) and see what makes your life easier for detection and discrimination.

As Davor indicated, the TX pulse seems to look like a pure delta function passed through a low pass filter, so someone could probably do a fairly good paper and pencil analysis that would be equivalent to simulations. But sims are by far the easiest way to tinker.

A good set of target response graphs would be useful to look at and discuss.

-SB

[Davor]

There are two points to emphasize here. First is magnitude, and the other is phase.
When you see magnitude falling uniformly by 20dB per decade you know it is a first order filter. (no surprise)
Second, you know from the bode plot of any first order low pass filter that all the fun with phase happens one decade lower than with magnitude. Any magnitude ripple will tell you that something also happens with phase, and in this case it is the on-ramp to worry about. Or not. That would depend on how long you wish your PI to be deaf for anything but its Tx coil.

IMHO it would be beneficial to be able to sample as soon as possible.

[Tinkerer]

Please note that many more interesting things happen in frequencies up to 1MHz (micro-seconds).

The way I see it, it is if little use to stimulate the targets with higher than 300KHz. A traditional PI is not capable of detecting targets with a TC if 1us.

With the TEM method I can sample any time so there is maybe an advantage with higher frequency stimulation.

The FFT shows 2 stimulation frequency bands.

Low frequency with +2.5dB at 5KHz and -20dB at 66KHz.

High frequency stimulation at a different time, +5.75dB at 70Khz and -20dB at 1Mhz.

Probably -20 dB is way beyond any usefulness, maybe we should put the limit of interest a lot higher.

Ideas and comments?

[WM6]

Instead of real toroid you may break it in half and use as a "snap-on" ferrite for cables RFI. Of course, it will be prudent to check if snap-ons are any good by themselves because they are dirt cheap and come with a handy enclosure too. At the very moment I'm laboratory-challenged, so I can't test it myself

Davor, thanks for idea. I will try.

[Davor]

Here is the PI_TRADITIONAL switch OFF transient FFT

Please note that many more interesting things happen in frequencies up to 1MHz (micro-seconds).

[Davor]

If you do not like toroid, you can use ferrite pot too.

Instead of real toroid you may break it in half and use as a "snap-on" ferrite for cables RFI. Of course, it will be prudent to check if snap-ons are any good by themselves because they are dirt cheap and come with a handy enclosure too. At the very moment I'm laboratory-challenged, so I can't test it myself

[Tinkerer]

Here is the PI_TRADITIONAL switch OFF transient FFT

[Tinkerer]

And here is the PI TRADITIONAL Full TX cycle FFT

[Tinkerer]

And here is the FFT of the TX ON time.
All these are simulations. The real TX pulses have quite a bit of Mosfet switching noise, specially when high Amps are switched.

Tinkerer

[Tinkerer]

Below is the FFT of the full cycle TX

Tinkerer

[Tinkerer]

Attached is the FFT of the switch OFF transient. This is the target stimulation, sampled after switch off.

What does it tell us?

Tinkerer

[Aziz]

That seems to be a pretty key point worth discussing further. Any idea how to work out the amount of energy allocated to each frequency range? Presumably its strongly related to the dI/dt of the pulse.

Thanks for the info on ML FBS. Only 3 real frequencies, pretty cheeky marketing.

That's a very trivial task.

Make a simple PI switcher spice simulation. Sense (view) the TX coil current and do a FFT for the pulse period time span only (that's important, otherwise, the whole frequency response will be distributed around the harmonics of the pulse frequency). LTspice offers this feature.

To make things easy, use a 1 kHz pulse frequency and make the FFT for let's say 10 ms .. 11 ms (after the circuit stabilizes late time of course).

You can see the frequency spectrum distribution of the TX current pulse. Notice, it contains the pulse-on response too (more low frequency contents). The flyback period current, when decay quickly to zero is a high frequency stimulation.

Aziz