What is the best frequency response for the preamp in a PI ?

[Teleno]

So the remaining task is to remove the tx decay from the coil but leave the target signal.

Alright but we all know that is the problem.

My question was related to implementing the amplifier bandwidth, if instead of a DC amplifier you connect a high pass filter what happens is,

No filter



RC high pass filter, -3dB at 500 Hz

[moodz]

How to implement this high pass? The response to a step pulse is a decaying exponential that starts at saturation. For 1kHz transmit rate the time constant would be in the 100's of us. Sampling that then try to separate the target from the decay doesn't seem to be advantageous.

The key words are ...response to a step pulse.

The step response of the tx coil being damped masks the target response. The target is isolated from the tx/ rx coil by air ...so intuitively there is already a high pass filter in place as you can't transmit DC through ac magnetic coupling. So the remaining task is to remove the tx decay from the coil but leave the target signal.

[MOR_AL]

I think the following.
Although the signal coming from the back to back diodes has strong components at the frequency of the pulses generated by the 555, this is not a useful signal. It exists all the time. The useful signal will occur when the coil passes close to a metal.
The useful signal will appear in increasing fashion for about half a second to a few seconds.
As this signal is non-cyclical, relative to the 555 signal, the frequency spectrum would be calculated using the Fourier Integral and not the Fourier Series, which refers to cyclical signals.
Leaving theory aside, I believe that the frequency spectrum of the useful signal has to do with how quickly the coil is handled close to the metal.
I think the useful spectrum would start at a few hundredths of a Hertz and extend to a few Hertz.
All AOs have high gain in this frequency range. But due to the presence of the undesirable signal in the diodes, the OA will always saturate. Leaving saturation is important, which is why I consider using a circuit whose OA does not enter its nonlinear region.
MOR_AL

[Teleno]

How to implement this high pass? The response to a step pulse is a decaying exponential that starts at saturation. For 1kHz transmit rate the time constant would be in the 100's of us. Sampling that then try to separate the target from the decay doesn't seem to be advantageous.

[MOR_AL]

Today I already designed the pulse generator part (555), inverter, mosfet and back to back diodes.
I included a circuit that limits the drain current, regardless of the Ton time of the 555/inverter and the value of the coil.
It is imperative that the resistor value in parallel with the coil is optimized for each coil value.
I've been trying for some time now to use an operational amplifier (OA) for the next stage.
I used several of them as tests, always focusing on common components.
I reached the following conclusion.
OAs were designed to function in their linear region. The undesirable signal, coming from the back to back diodes, has a voltage of around 4Vpp or more.
When this signal is amplified by the OA, the OA inevitably reaches its nonlinear region. The exit from this region seems to be the problem.
Due to this fact, I am thinking of designing an amplifier stage such that it does not saturate.
It could be an anti-logarithm amplifier, but for the useful signal, there would be no need for precision, just that it would amplify low voltage levels a lot and a little for high voltage levels.
I even tried to find out what the frequency range of the useful signal would be, but due to the presence of the unwanted signal in diodes with much higher voltage, I was unable to see the difference.
I'm not dealing with that at the moment.
At the moment I'm trying to design a circuit for testing my purchased OA. As I said, I'm not a fan of expensive components and some of my OA are probably fake.
The tests refer to the slew rate and the band gain product (BGP).
For testing, a signal generator and an oscilloscope are required.
MOR_AL

[moodz]

Here is the preamp response for a PI operating at 10 Khz and gain of 1000.

This is a switched preamp and the required response is from the fundamental to at least the 11th harmonic.

By filtering at baseband all the problems with mains EMI earthfield etc ... just disappear .. attenuated by over 100 dB

So has anyone got an idea of wha the target tau for a 10mm x 10mm x .016 mm ( kitchen foil ) ?

[moodz]

Here is the preamp response for a PI operating at 10 Khz and gain of 1000.

This is a switched preamp and the required response is from the fundamental to at least the 11th harmonic.

By filtering at baseband all the problems with mains EMI earthfield etc ... just disappear .. attenuated by over 100 dB

[Gunghouk]

I like it. I'm done

[moodz]

I really do get it but I've been dragged down a rabbit hole, which probably should be in another thread, wondering if the amplitude of the TX harmonics should at least be all the same amplitude as far as possible and do particular harmonics 'excite' particular target types.
This is leading me to wonder about modulating pulse width, frequency sweeping and multi-pulse TX signal harmonic content.
Sorry for derailing your thread.

Are the teeth of this nit comb fine enough to catch those pesky sub gramme critters? (rhetorical)

It will almost certainly be an application of the Rumsfeld Matrix ...
The four quadrants of the Rumsfeld Matrix

1. Known knowns: These are facts or variables that we're aware of and understand. They form the basis of our knowledge and provide a solid foundation for decision making.
2. Known unknowns: These are factors we know exist, but don't fully understand. They represent gaps in our knowledge that we must address through research, investigation, or consultation with experts.
3. Unknown knowns: These are elements that we don't realize we know. They're typically buried in our subconscious, overlooked, or dismissed as irrelevant. Uncovering these insights can lead to surprising breakthroughs in decision making.
4. Unknown unknowns: These are factors that we're not aware of and can't predict. They represent the most significant source of uncertainty and risk, as they can catch us off guard and derail our plans.

To illustrate this the current Magpi Design can easily detect a 0.1 gram nugget using a 40 cm coil .... but the first sample occurs at 8 microseconds. Go figure.

[Gunghouk]

I really do get it but I've been dragged down a rabbit hole, which probably should be in another thread, wondering if the amplitude of the TX harmonics should at least be all the same amplitude as far as possible and do particular harmonics 'excite' particular target types.
This is leading me to wonder about modulating pulse width, frequency sweeping and multi-pulse TX signal harmonic content.
Sorry for derailing your thread.

Are the teeth of this nit comb fine enough to catch those pesky sub gramme critters? (rhetorical)

[moodz]

Actually this harmonic content is for equal duty cycle/pure square wave. For a typical PI with, say 40uS TX on in 1000uS @1khz the even harmonics are as pronounced as the odd and are similar in amplitude to the fundamental well past 11f.

See here

https://www.analog.com/media/en/tech..._book_Ch13.pdf

Page 257

I think you might be missing the point of the example ... it does not matter what duty cycle or shape of the TX signal is ... it was merely demonstrating the fact is that there is no useful spectral information below F (the TX fundamental ).
So the "accepted practice" of designing low noise preamps for PIs that extended down to DC is wrong.
Understanding this allows you build a better PI detector than state of the art detectors as they exist today.

[Gunghouk]

Sorry about that ... this board has trouble with PNG images ... I forgot to convert to JPG.

Actually this harmonic content is for equal duty cycle/pure square wave. For a typical PI with, say 40uS TX on in 1000uS @1khz the even harmonics are as pronounced as the odd and are similar in amplitude to the fundamental well past 11f.

See here

https://www.analog.com/media/en/tech..._book_Ch13.pdf

Page 257

[moodz]

So is this with passive filter components around the preamp to give 2nd order filtering and single stage 60dB gain ? (f/10 and f10 would give bang on 1kHz geometric freq and zero phase change @ 1kHz Tx )

A couple of other questions, how critical, if at all, is the magnitude of phase shift introduced in the passband ?

I am not sure about that ... no-one that I know of has modelled a PI RX chain to figure if phase shifts are critical or not ... However I suspect they are and someone needs to investigate it.

Is a gyrator BP design with low Q too severe to be considered in this application?
I have not tried one ... you could .. one important factor is that the input impedance to the ( preamp / filter ) should not impact the damping.

Presumably target composition/descrimination relies on as little phase change during amplification as possible. Is this assumption correct ?
Yes and no ... So far I can separate X and R and indicate ferrous / non ferrous. However it can be shown for some targets no change in the X signal as the targets are conductive but dont identify as ferrous or non- ferrous but still give an R signal.

What are the real sources of noise outside of the home/workshop that require filtering by a BP filter in the field?

In the field there are Earth Magnetic Field, mains EMI 50/60 hertz depending on your location, rocks that have a net magnetic field, earth currents from SWER lines mains ( single wire earth return ), DC offsets in your frontend / dissimiliar metals and or , thermal noise, 1/f noise from the opamp itself, LF noise from resistors in your front end.

Sorry for all the questions, I'm a noob see

Dont worry I have not met any experts in the field yet .. were all noobs IMHO

[Gunghouk]

Of course f/11 and f11 fit your previously stated upper minimum frequency at least

Fmean = SQRT (Flow x Fhigh)

[Gunghouk]

So is this with passive filter components around the preamp to give 2nd order filtering and single stage 60dB gain ? (f/10 and f10 would give bang on 1kHz geometric freq and zero phase change @ 1kHz Tx )

A couple of other questions, how critical, if at all, is the magnitude of phase shift introduced in the passband ?

Is a gyrator BP design with low Q too severe to be considered in this application?

Presumably target composition/descrimination relies on as little phase change during amplification as possible. Is this assumption correct ?

What are the real sources of noise outside of the home/workshop that require filtering by a BP filter in the field?

Sorry for all the questions, I'm a noob see