DEEPER PI DETECTION DEPTH

[Tinkerer]

Hmm OK well it would seem that I don't understand at all. Applied to a mono it sought of made sense to me in that it linearized the exponential input potentially allowing earlier sampling. But an IB coil where your RX input isn't exponential and you can sample anywhere you like anyway.. I don't get it.

Anyway I hope I haven't upset you with my lack of tact, you are of course welcome to post anywhere you like, it doesn't have to make sense to me.

For all the analogue electronics noobs out there such as myself, here's my attempt at an incredibly over simplified I'm sure demonstration of what the ideal log amp does given a perfect exponentially decaying input.

Midas

Thanks for your valuable contribution and the interesting graph.

At the time I used the Log amp, I found that the immediate advantage was for detecting short TC targets. For example, a 5us target signal, has lost about 85% of it's amplitude at 10us after switch OFF.
This is where the Log amp brings the advantage, towards the end of the decay curve. I found that I could still detect a 5us target after a delay of 20us, when using the soft clipping combined with the Log function of my simple circuit from post #455.

About IB-PI: this does not automatically mean that one can sample a target at any time. For the time being the "anytime sampling" only applies to the TINKERERS_TEM_IB-PI.

Even with the best balanced coil, there is always some residual TX decay signal. If we amplify this residual signal a 1000 times like it is usual with a Traditional PI, the pre-amp gets saturated. It will take some time to come out of saturation. Clipping helps some, but also distorts the signal.

Some Traditional PI designs use gating to cut out the Flyback. The problem with that, is that gating a several hundred volt signal adds a considerable amount of distortion and switching noise, as well as some time for the gated signal to settle.

In the TINKERERS_TEM_IB-PI design I use a radically different approach, that makes it possible to sample the RX signal at any and all times. It turned out that there is sooooo much more information available that way.

Tinkerer

[Davor]

...But an IB coil where your RX input isn't exponential and you can sample anywhere you like anyway..

In signal processing you tend to tackle coherent phenomena. Everything else is just noise. Many different signals can comply with coherence criteria, which are:
1) autocorrelation - the phenomenon recognition is possible by means of comparison with itself. In everyday life think of a mirror - you can recognise your own image there.
2) cross-correlation - there is such a device that enables autocorrelation (e.g. mirror) also over time (e.g. family photo album). As far in time you can go the more coherent signal you have.

The ultimate infinitely coherent signal would be 4K Cosmic Microwave Background Radiation, however it does not contain much information. You need less coherent one to carry some information, and there you hit the information vs. noise compromise. Mr. Shannon explained that very nicely.

In a nutshell - this is all there is to know about signal processing.

Now back to reality. If you have an exponential exciter, your coherent response will also be exponential, and that's why you surely can apply every filtering or weighting function that works with exponential signal.

Alternatively you may observe your Tx as a Dirac delta function, but only in case of a very delayed response - which is not the case with PI: Tx decay and Rx response are both operating in ~10microsecond range.

If you really need something close to Dirac, a step exciter will provide it much easier than a traditional flyback. However, Rx response will be exponential anyway.

I know the previous lines must sound as techno-Klingon but there you have it.

[Midas]

Just for the spirit of this topic sake. Previously we discussed Tx issues and Rx was not tackled at all. I guess everyone here is familiar with Surf (etc. monocoil) design, so I used it as a platform for further examination. There is no good reason why this approach couldn't be extended to IB coil, differential design or anything else. A whole point is that you'll have a weighted signal shape that will be much easier to deal with. In linear domain you tackle ratios, while in log domain you play with differences. Constant differences are child play in analog world because each sample absolute value may be allowed to float while the difference remains constant, and that alleviates problems of pulse energy variations, ground, or even temperature. That's why.

Hmm OK well it would seem that I don't understand at all. Applied to a mono it sought of made sense to me in that it linearized the exponential input potentially allowing earlier sampling. But an IB coil where your RX input isn't exponential and you can sample anywhere you like anyway.. I don't get it.

Anyway I hope I haven't upset you with my lack of tact, you are of course welcome to post anywhere you like, it doesn't have to make sense to me.

For all the analogue electronics noobs out there such as myself, here's my attempt at an incredibly over simplified I'm sure demonstration of what the ideal log amp does given a perfect exponentially decaying input.

Midas

[Tinkerer]

I'm not sure why Davor chose to post his new design here. It's really not very related to Tinkerer's IB-PI or the examples you've quoted. Its not IB, it uses a mono-coil. It samples the flyback just like all PI does. Its just further up on the exponential decay curve. I guess its fair to say the TX (and of course RX since they are one and the same) is decaying faster at that point than regular PI but even regular PI has to cope with residual coil decay. I don't see any fundamental difference.
Having said that I think you might be right about increased sensitvity to some undesirable 'super fast ground components'. I seem to remember reading a post here by someone of note (can't remember who, sorry if you reading this) that stated that sampling before ~8usec, although sensitive in the air to very small gold, is useless in real ground conditions.
And as everyone knows all the air-borne gold was snapped up by Jason and the Argonauts.

Midas

Everybody is right here. Looking at different RX methods is certainly a good idea. Mono or IB, continuous wave form or transient only, every method has it's advantages and drawbacks.

About early sampling with traditional PI:
The reactive signal from magnetic minerals is always present while the current flows in the coil. At very early sampling time, there is more current, therefore the reactive signal is stronger.
There is a point in time, where the reactive signal is equal to the resistive signal. If we sample this point, then it can be seen as a pivot, where the reactive signal is positive before and negative after, relative to the pivot.
Looking at the exponential decay curve, one might think that the part of the curve near 0V, seems straight, but in fact it is still exponential until infinity. So there is always some reactive signal, but it gets so weak that it is not noticeable anymore.

With a continuous waveform, the current is continuous, but as it passes the 0 crossover, we again have the pivot, where the reactive and resistive signals are equal. However, the 2 signals do not have the same TC, so there is a phase difference between the 2 signals.

In general, the reactive signal has a short TC, but some hot rocks have a longer TC that is more like the TC of a nugget.

There we have a problem.

Will the log amp help us solve this problem?

Tinkerer

[Davor]

I'm not sure why Davor chose to post his new design here.

Just for the spirit of this topic sake. Previously we discussed Tx issues and Rx was not tackled at all. I guess everyone here is familiar with Surf (etc. monocoil) design, so I used it as a platform for further examination. There is no good reason why this approach couldn't be extended to IB coil, differential design or anything else. A whole point is that you'll have a weighted signal shape that will be much easier to deal with. In linear domain you tackle ratios, while in log domain you play with differences. Constant differences are child play in analog world because each sample absolute value may be allowed to float while the difference remains constant, and that alleviates problems of pulse energy variations, ground, or even temperature. That's why.

[Midas]

You get the same information after any abrupt transition, it's just at a lower voltage during the pulse where it can be easily dealt with. Minelab take information from the pulse on period in their pulse induction gold detectors for fe discrimination, the discrimination isn't as aggressive as with coin detectors and is designed to err on the safe side because we don't wish to leave a monster nugget in the ground.
The BBS and ABS continuous rectangular waveform switches the coil from one polarity to the other. This is better than the switch on period in a conventional PI because the step is much greater at the beginning of each pulse, I.E. the step at the beginning of each pulse in continuous rectangular waveforms is where the flyback normally occurs in conventional pulse induction.

The information for ground balance is present during the complete pulse on period but it isn't available during the off period if you sample the flyback in conventional designs.

Sampling during the pulse or flyback or at any time while the tx coil waveform is changing has all the problems of VLF such as responding to magnetite hot rocks that don't respond to a PI, even a PI without ground balance, so it would pay to figure out how you would deal with this first.

I'm not sure why Davor chose to post his new design here. It's really not very related to Tinkerer's IB-PI or the examples you've quoted. Its not IB, it uses a mono-coil. It samples the flyback just like all PI does. Its just further up on the exponential decay curve. I guess its fair to say the TX (and of course RX since they are one and the same) is decaying faster at that point than regular PI but even regular PI has to cope with residual coil decay. I don't see any fundamental difference.
Having said that I think you might be right about increased sensitvity to some undesirable 'super fast ground components'. I seem to remember reading a post here by someone of note (can't remember who, sorry if you reading this) that stated that sampling before ~8usec, although sensitive in the air to very small gold, is useless in real ground conditions.
And as everyone knows all the air-borne gold was snapped up by Jason and the Argonauts.


Midas

[crane]

You get the same information after any abrupt transition, it's just at a lower voltage during the pulse where it can be easily dealt with. Minelab take information from the pulse on period in their pulse induction gold detectors for fe discrimination, the discrimination isn't as aggressive as with coin detectors and is designed to err on the safe side because we don't wish to leave a monster nugget in the ground.
The BBS and ABS continuous rectangular waveform switches the coil from one polarity to the other. This is better than the switch on period in a conventional PI because the step is much greater at the beginning of each pulse, I.E. the step at the beginning of each pulse in continuous rectangular waveforms is where the flyback normally occurs in conventional pulse induction.

The information for ground balance is present during the complete pulse on period but it isn't available during the off period if you sample the flyback in conventional designs.

Sampling during the pulse or flyback or at any time while the tx coil waveform is changing has all the problems of VLF such as responding to magnetite hot rocks that don't respond to a PI, even a PI without ground balance, so it would pay to figure out how you would deal with this first.

[Tinkerer]

See Bosnar's patent.
www.geotech1.com/pages/metdet/patents/US6326791.pdf

This was discussed on at least one forum some time ago.

A log amp makes it very difficult to subtract static fields although most of the early low gain designs dealt with low frequency signals by simply capacitively coupling the demods to the preamp but this has problems at high gains.

Sampling the flyback makes no sense because the same information is available and much easier to obtain and process during the pulse as we see in Minelab's BBS and ABS designs. If you do sample the flyback then you can't use a later sample for ground balance because the information contained in the spike isn't there at later times and even if you could address this problem it still suffers the same problems as VLF.

Has anyone considered the problems and figured out how they will deal with them?

Could you clarify your statement?

Are you saying that the same information in the Flyback is contained during the pulse?

Tinkerer

[Davor]

Has anyone considered the problems and figured out how they will deal with them?

I did ... by not expecting them. My reasoning was - if Surf PI works (without at least half the features you mentioned), and apparently it works fine, then a log amp must work even better because it will not saturate as Surf PI does. Nothing else.
My practical experience so far with PI is zero (0) and I simply wish to address the issues often mentioned here on this forum. E.g. "I so wish to sample earlier to be able to detect small gold" - and here you are.
BTW I don't even know about the fancy MDs you listed. Guess they are good.

[Davor]

No. It is a coil capacitance stand-in to be exposed for easy playing with. In reality it will be somewhat larger (+ cable capacitance etc.) but for simulation purposes it is OK. You'll be able to see some interesting consequences of increasing it.

[6666]

Davor is C3 50pf a real component ?

[crane]

See Bosnar's patent.
www.geotech1.com/pages/metdet/patents/US6326791.pdf

This was discussed on at least one forum some time ago.

A log amp makes it very difficult to subtract static fields although most of the early low gain designs dealt with low frequency signals by simply capacitively coupling the demods to the preamp but this has problems at high gains.

Sampling the flyback makes no sense because the same information is available and much easier to obtain and process during the pulse as we see in Minelab's BBS and ABS designs. If you do sample the flyback then you can't use a later sample for ground balance because the information contained in the spike isn't there at later times and even if you could address this problem it still suffers the same problems as VLF.

Has anyone considered the problems and figured out how they will deal with them?

[Aziz]

Hi all,

has anyone tried the log-amp yet?
When I have more time, I can easily mod one of my inverting pre-amp modules to test it. I should see the critical issues, benefits and disadvantages... It will take some time however.

Cheers,
Aziz

[Aziz]

That's now more interesting Davor. This should be interesting to test.
Cheers,
Aziz

[Davor]

You don't need to do any of these things to lower the input impedance, just do it! Try 10k instead of 22k, and you'll notice why D3 is there
I've put everything as is simply to gain insight to what is possible, and to distinguish critical points.

Perhaps the most interesting values are 2k2 instead of 22k, and 220k in feedback instead of 2M2. Please observe how nicely the curve splits, and see why I thought of using comparators instead of bilateral switches, and level reference instead of time reference.

Furthermore, D3 can be replaced with unbiased NPN transistor, thus leaving the 0V reference near opamp pulse-free. See attached...