DEEPER PI DETECTION DEPTH

[Davor]

Even though the current turns off in 1-2 uS, what does the di/dt turn-off look like? Is that what stimulates the target?

Yes, the term is
v=L*di/dt, where v is an instantaneous voltage across the inductor - hence huge peaks on driving transistors
The least painful way of observing a MD system is by imagining targets as very loosely coupled transformers. And what transformers do with voltage? More voltage in (di/dt), more voltage out. Easy.

Now about the previous questions...

Wouldn't bandwidth of MDs like TGSL be essentially the synchronous detector bandwidth, which is only a few Hz (depending on integration time), maybe for TGSL 100 Hz or so?

Yes and no, depends on where you stick your probe. In case you put it on the very synchronous detector, you'll find a lot of garbage, and a trend which is hardly obvious if signal is weak. Real integration happens at low pass filters. If you stick a probe there, you'll see a nice line and no garbage. LPF BW is roughly 15Hz, and hence your integration period.
With PI you also have a kind of integration during a sample pulse, but in absence of proper weighting function it is more of a window than an integrator. A proper weighting function would be some kind of fade-in, only I’m not yet sure of what kind. IMHO with PI you should sample first, apply the weighting function, and integrate/amplify later.

Would you expand on that some more? I'm not clear on the concepts you are describing.

For this I expect your familiarity with PWM, and ... actually I can do it in simple lingo so that everyone can grasp it.

First off, imagine a perfect square signal, both positive and negative pulses are the same in duration, and Von is, say, 1V. The average voltage is 0V because Voff is -1V. Simple.
Now, imagine that you want a signal which has average voltage somewhat different than 0V. You could make Von variable, but what if you can't? You simple make a small difference in duration of on and off pulses, and there you have it. The principle is called pulse width modulation, or PWM.

Now, think of a mixer. It works as an analogue multiplier. In case you multiply a DC signal with the perfect local oscillator as explained above, you'll get a product that is 0DC and a HF component. In case of a synchronous detector it is just the other way around. In case of PWM shifted local oscillator, you'll get a DC component as well. In special case of quadrature synchronous mixing, you expect a DC component to be proportional to phase shift (and signal amplitude), but not the PWM induced component that lurks about.

Now for the above-mentioned TGSL. It extracts phase shifted local oscillator signal using comparators that are prone to picking up every signal variation and fire sooner or later than perfect. As this signal is derived from a directly coupled Tx oscillator, you may expect a nice PWM local oscillations that are related to Tx coil coupling, soil mineralisation, you name it.

In short, VLF-s could act less erratic if proper care is taken to local oscillator PWM elimination. Less erratic -> better sensitivity -> deeper detection.

[moodz]

...as Davor has pointed out previously a lot of these problems have been addressed in other fields such as Rf engineering. Its just that the application has changed. The problem of maintaining a high Q ( ok let's say optimal ) which ensures maximum power transfer was solved years ago in the induction heating field. By using a PLL to lock the current phase to the voltage phase of a tank circuit the Q will be maintained. The actual Q will vary though due to variable losses from the tank coil to the load however maximum current can be maintained at all times in the coil ..... Easily several amps.

The answer to Aziz question is the VLF. (power density vs bandwidth not to mention receiver bandwidth advantage and synchronous continuous sampling advantage)

Look at the circuit here ....

http://uzzors2k.4hv.org/projectfiles/ihpll1/ETs_4046_IH.GIF

Moods . Free lunch dinner and all your base are mine.

[Carl-NC]

... times the tank Q of say ... 10 ... there you have it. Most of the current is recycled in resonant tanks. In case of resonance the only current you can measure accounts for losses.

There is no need to measure the current, it's easily calculated. Resonance or not.

[Tinkerer]

I'm interested in that potential for hi-Q Tx tanks, but so far I don't know if it is really a good idea (Tx voltage easily affected by ground). I hope to try some experiments (I think others are going to also, maybe you?).

Apparently Carl doesn't know of an MD that actually uses such a high-Q tank/oscillator configuration.

Regards,

-SB

For a deep searching detector we want at least 1A/us of moment. But 1A alone does not say all. We also need to say how many turns in the coil and what the diameter of the coil is.

1A/us at 10 turns of magnetic moment in a coil of one square meter area gives good depth for a target of about 20mm diameter.

1A/us at 20 turns still gives good depth for a 10mm diameter target.

Then we want to take the first sample of the target response when the response is at it's peak.

And then we want to do that at a frequency or repetition rate of 5000Hz, to have many samples to stack.

Davor, you keep saying that the RX front end is not good. Can you come up with a better RX front end?

Tinkerer

[Carl-NC]

Even though the current turns off in 1-2 uS, what does the di/dt turn-off look like? Is that what stimulates the target?

Let's say 2 amps in 2us, reasonably linear. So the "1,000,000 A/s" slew rate is actually reasonable.

Also, isn't the voltage decay time the confounding characteristic that makes detecting small targets difficult (in a basic PI design)?

The voltage decay is something that just gets in the way. Use an IB coil to get rid of it, and you can then see smaller targets.

[simonbaker]

... times the tank Q of say ... 10 ... there you have it. Most of the current is recycled in resonant tanks. In case of resonance the only current you can measure accounts for losses.

I'm interested in that potential for hi-Q Tx tanks, but so far I don't know if it is really a good idea (Tx voltage easily affected by ground). I hope to try some experiments (I think others are going to also, maybe you?).

Apparently Carl doesn't know of an MD that actually uses such a high-Q tank/oscillator configuration.

Regards,

-SB

[Davor]

I don't know of a single VLF that achieves 1A peak current, maybe 100mA max.

... times the tank Q of say ... 10 ... there you have it. Most of the current is recycled in resonant tanks. In case of resonance the only current you can measure accounts for losses.

[simonbaker]

I don't know of a single VLF that achieves 1A peak current, maybe 100mA max. Also, in a PI, the current turn-off time is not the same as the voltage decay time. Typically, the current shuts off in 1-2 us. Most PI's peak at 2-3 amps.

But the real question is, what defines "better"? Deeper?

Even though the current turns off in 1-2 uS, what does the di/dt turn-off look like? Is that what stimulates the target?

Also, isn't the voltage decay time the confounding characteristic that makes detecting small targets difficult (in a basic PI design)?

Regards,

-SB

[Carl-NC]

Following conditions: VLF vs. PI
VLF: 20 kHz operating frequency, 1 A peak TX coil current
PI: 1 kHz pulse frequency, 10 A peak, 10µs damping decay time

Let's look at the maximum TX coil current change (dI/dt):
VLF: max (dI/dt) = max(1 A*sin(wt)/dt) = 1 A*w*1 = 1 A*2*pi*20kHz = 125,664 A/s
PI: max (dI/dt) = max(10 A/10µs) = 1,000,000 A/s

Which detector is effectively better (delivers more) and why?

I don't know of a single VLF that achieves 1A peak current, maybe 100mA max. Also, in a PI, the current turn-off time is not the same as the voltage decay time. Typically, the current shuts off in 1-2 us. Most PI's peak at 2-3 amps.

But the real question is, what defines "better"? Deeper?

[simonbaker]

In case of PI vs VLF it is time vs frequency, and noise bandwidth is in play. While BW in case of WLF is ~15Hz (e.g. tens of milliseconds), and energy is integrated over larger periods...

Wouldn't bandwidth of MDs like TGSL be essentially the synchronous detector bandwidth, which is only a few Hz (depending on integration time), maybe for TGSL 100 Hz or so?

Both concepts suffer from sub-optimal front-ends and efficiency problems. VLF's inherent trouble is ground reference for mixers, and balanced solutions are only half way better due to hidden PWM clocking problem. Phase information is extracted by synchronous detection, which is fine, but again the reference is floating a bit...

Would you expand on that some more? I'm not clear on the concepts you are describing.

Regards,

-SB

[Davor]

So far I learned that whenever you think that you just found a free buffet there must be something wrong in your reasoning.

In case of PI vs VLF it is time vs frequency, and noise bandwidth is in play. While BW in case of WLF is ~15Hz (e.g. tens of milliseconds), and energy is integrated over larger periods, with PI it is the other way around, you have tens of kHz bandwidth, strong pulses AND ... a dead period.

Both concepts suffer from sub-optimal front-ends and efficiency problems. VLF's inherent trouble is ground reference for mixers, and balanced solutions are only half way better due to hidden PWM clocking problem. Phase information is extracted by synchronous detection, which is fine, but again the reference is floating a bit. Quadrature approach is generally avoided, and it could lead to more sublime solutions, such as Costas filtering etc. Indication, e.g. human interface, is usually something crude, same as with PI.

PI on the other hand is flawed by "hit me baby one more time" syndrome and a pursuit of a misleading single coil design, thus losing valuable microseconds of treasure signal. There is a phenomenon equivalent to phase shift in VLF, and it is zero crossing. Using VLF analogies, PI detection is mainly concentrated on partial detection of I component, while Q (zero crossing) is avoided. Detection is achieved by sampling/integrating, and - what do you know - floating reference problem again. To compensate for poor detection, a crude approach of more-juice-better-detection is employed. Nicer detection should lead to less power, which should lead to even better detection yet again ... to some limits, but much longer battery life. Given a choice of battery weight - I'd rather go with smaller.

Both concepts are in pursuit of targets with the same physical properties.

In short, both systems have venues yet to discover, and many small SNAFU-s to iron.

[Aziz]

Hi all,

it's getting almost quiet here. It's time to throw another brain food to keep the ball rolling.

Let's look, whether there is a "free lunch" for us.

Following conditions: VLF vs. PI
VLF: 20 kHz operating frequency, 1 A peak TX coil current
PI: 1 kHz pulse frequency, 10 A peak, 10µs damping decay time

Let's look at the maximum TX coil current change (dI/dt):
VLF: max (dI/dt) = max(1 A*sin(wt)/dt) = 1 A*w*1 = 1 A*2*pi*20kHz = 125,664 A/s
PI: max (dI/dt) = max(10 A/10µs) = 1,000,000 A/s

Which detector is effectively better (delivers more) and why?

Cheers,
Aziz

[Davor]

IMHO there are several opportunities in optimising Rx input for better impedance matching, and true balanced operation. Point is that in radio technology there is something called "link balance" that says (in many words of complicated lingo) that you can get only a certain optimum dynamic range, and that remains the same or becomes worse even if you overpimp your Tx. The lower margin is limited by Rx design (system noise, detection limits) and natural noise. However the noise bandwidth here is much larger than with the VLF IB solutions, it is apparent that most of the problems are induced by TX side, and we are waaaaay far away from noise margins here.
My simple deduction would be to go for nicer Rx with proper balancing and careful impedance matching. Once Rx is working its best the Tx will not have to be as powerful as it is currently a standard, and it will be much closer to ideal because of less problematic components with less parasitics, and less awful transients. With less self-induced problems the dynamic margins will widen and hey presto! - a better PI.

[simonbaker]

Hi SB,

I see, you are still after the discrimination and get worried about the less phase difference at higher frequencies.

But you are just looking at the resistive response cased phase response. Note, that the spice simulation didn't take reactive response effects into account.

Don't be sad, there are possibilities for your required phase difference.


In an induction balanced (IB) coil configuration, magnetic materials do change the coil coupling coefficient between TX and RX and the RX coil will see more signal from the TX coil. (Phase info).

The coil coupling coefficient change can be used further to increase the mutual inductance between TX and RX by placing a load in the RX coil (damping R). (more Phase info).

Well, (de-)magnetising and/or magnetic viscous effects do add more phase info for your convenience too.

And the skin effect:
Ferromagnetic metals cause a high skin effect (less penetration), which will reduce the targets TC dominantly. But the magnetising effects will be superimposed with the TC.

Cheers,
Aziz

I'm not on your wavelength as to what you are saying; maybe a good sim of balanced coils would help show it.

My reasoning is that if at high frequencies these targets' responses all converge to approximately the same phase shift, then it will be harder to discriminate between them because all our MDs can detect is their response. Perhaps what you are saying is that if we look at them as part of a total dynamic system (which they are in our sims), we can vary some other part of the system to enhance the phase differences of some signal measured somewhere. That's what I was hoping, but I'm not seeing exactly how to do that. Especially because the targets are so weakly coupled to everything else, it seems all we can do is measure their responses more or less.

Regards,

-SB

[Aziz]

...
Oops -- I forgot. The phase difference seems to disappear at higher frequencies, making discrimination harder.
...
-SB

Hi SB,

I see, you are still after the discrimination and get worried about the less phase difference at higher frequencies.

But you are just looking at the resistive response cased phase response. Note, that the spice simulation didn't take reactive response effects into account.

Don't be sad, there are possibilities for your required phase difference.


In an induction balanced (IB) coil configuration, magnetic materials do change the coil coupling coefficient between TX and RX and the RX coil will see more signal from the TX coil. (Phase info).

The coil coupling coefficient change can be used further to increase the mutual inductance between TX and RX by placing a load in the RX coil (damping R). (more Phase info).

Well, (de-)magnetising and/or magnetic viscous effects do add more phase info for your convenience too.

And the skin effect:
Ferromagnetic metals cause a high skin effect (less penetration), which will reduce the targets TC dominantly. But the magnetising effects will be superimposed with the TC.

Cheers,
Aziz