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

[moodz]

Technically yes you would be losing at the breakpoints but over the whole bandpass the loss is not so critical. So if we are trying to squeeze every db then yes f/2 would be a good start point and 11f could be the endpoint.

I have seen many try to do DSP on a PI ( without demodulation ) and it never worked as far as I know. All the interesting stuff happens in a few microseconds out of a hundred or so and DSP does not deal with this well. Also must remember we need gain as well as filtering. The MAGPI now has a 60 db preamp, 75 db diff amp after demod and 12 bits of ADC.

[moodz]

Aren't you losing 6db of gain at the fundamental with that as your low breakpoint hence my suggestion of f/2?
My theory is very rusty but isn't there -90 deg shift at the low and +90 deg at the high breakpoints and somewhere between both until zero at the geometric mean frequency point for a 2nd order bandpass filter?
Would a bandpass filter in DSP solve phase shifting?

[Gunghouk]

Aren't you losing 6db of gain at the fundamental with that as your low breakpoint hence my suggestion of f/2?
My theory is very rusty but isn't there -90 deg shift at the low and +90 deg at the high breakpoints and somewhere between both until zero at the geometric mean frequency point for a 2nd order bandpass filter?
Would a bandpass filter in DSP solve phase shifting?

[moodz]

So where would you put the filter breakpoints, f/2 and 2f perhaps?

The breakpoints should be at f and as high as the bandwith of your opamp will go but less than the SRF ( self resonant frequency ) of your coil

In my previous post above I said 14 Khz for the bode plot ... but it should have been 1.4 Khz.

You want to pass the primary f and odd harmonics ideally up to f9 ( f f3 f5 f7 f9 ).

If you go much below f ( fundamental ) then you are liffing the skirt of the LF rolloff and degrading the LF rejection.

Heres a qeustion ... why is the phase response flat in the LF rolloff shown above in post #49 ?

[Gunghouk]

So where would you put the filter breakpoints, f/2 and 2f perhaps?

[moodz]

Hi moodz,

Please, use sims and show us the phase/frequency plot. This plot is very important for PI MD. Thank you in advance!

Here you go ... this was made for a 14 KHz PI.

[Detectorist#1]

Hi moodz,

Please, use sims and show us the phase/frequency plot. This plot is very important for PI MD. Thank you in advance!

[detectormods]

..and not only that ... but you dont get a set of free steak knives BUT the intermod distortion caused by the mains signal on the TX signal is also resolved. ( this PI is operating at 10 KHZ ).
Green is the new preamp.

With apologies to Detectorist#1 who does not like sims - thought I dont know any engineers who dont use sims nowadays.

As an RF guy, you removed the sidebands, but we all know of a couple of manufactures that would benefit greatly from attenuating 50/60 Hz getting into the domain of the golden goose.

[moodz]

..and not only that ... but you dont get a set of free steak knives BUT the intermod distortion caused by the mains signal on the TX signal is also resolved. ( this PI is operating at 10 KHZ ).
Green is the new preamp.

With apologies to Detectorist#1 who does not like sims - thought I dont know any engineers who dont use sims nowadays.

[moodz]

Well you are getting to the point.

The plot below shows output spectrum from a "traditional" preamp vs output of "new design" preamp. There is a 10 millivolt 50 hertz mains signal hitting the coil in both cases.
The gain of both preamps is 100.

The "new design" clearly attenuates the 50 hertz mains by 40 db ( ie 100 ) AND the noise floor is 20dB lower.

[Carl-NC]

... are you following where this is leading to ?

A new design.

LMAO, good one Woody.

[detectormods]

I ran across the same problem ... you cannot just sub in a highpass ( actually its a band pass ) filter and hope all will be good.

There are three parts or time periods to the waveform coming into the preamp.
Considering a "traditional" PI

The first part is the time during TX ON when the current is ramping up in the TX coil.
The second part is the flyback period ( very short ) .. this is when the front end protection clamping and damping or switching is active.
The third part is the actual RX period starting directly after the fly back.

Basically because the RX period is not continous ( interrupted by TX and FLYBACK ) and non linear response of preamp then the RX is considered to be modulated.

Any form of modulation results in convolution or frequency mixing of the input waveforms.

The RX waveform consists of UNWANTED LOW frequency components ( dc offsets, earth field, mains hum etc ) AND WANTED HIGH frequency target responses ( and ground response .. deal with later )

The preamp non linear response to high level signals causes the LOW frequency unwanted components to be modulated onto the HIGH frequency wanted components.

So you need a preamp that does not cross modulate the LOW with the HIGH and problem solved ( earth field and mains hum etc anyway )

Traditionally using high spec opamps partly solves this as the low distortion / low noise / high speed / recovery of the these amps minimises cross modulation and overload.


... are you following where this is leading to ?

A new design.

[moodz]

True, but I posted from experience, by testing, as an example but let’s look at hard facts, yes , lower noise in the 1Khz -100 KHz range is much more important than noise at 10Hz , more important is the slew rate, any laggard behaviour will not capture the dynamics of a target response. But, this is my observance from experimenting, too much bandwidth add noise, but I observed that high noise specification in the lower frequency domain translates into additional noise into the higher frequency spectrum, is it via some mixing component of sub harmonics? I do not know, I have built input stages with the opa1612 ltc2054 that should be exceptional but I get a rise in noise floor, I have even tried building impedance converters but for some reason I do not end up with the performance of the low noise low frequency specified part. I myself would be over moon if someone can explain to me why this happens.

I ran across the same problem ... you cannot just sub in a highpass ( actually its a band pass ) filter and hope all will be good.

There are three parts or time periods to the waveform coming into the preamp.
Considering a "traditional" PI

The first part is the time during TX ON when the current is ramping up in the TX coil.
The second part is the flyback period ( very short ) .. this is when the front end protection clamping and damping or switching is active.
The third part is the actual RX period starting directly after the fly back.

Basically because the RX period is not continous ( interrupted by TX and FLYBACK ) and non linear response of preamp then the RX is considered to be modulated.

Any form of modulation results in convolution or frequency mixing of the input waveforms.

The RX waveform consists of UNWANTED LOW frequency components ( dc offsets, earth field, mains hum etc ) AND WANTED HIGH frequency target responses ( and ground response .. deal with later )

The preamp non linear response to high level signals causes the LOW frequency unwanted components to be modulated onto the HIGH frequency wanted components.

So you need a preamp that does not cross modulate the LOW with the HIGH and problem solved ( earth field and mains hum etc anyway )

Traditionally using high spec opamps partly solves this as the low distortion / low noise / high speed / recovery of the these amps minimises cross modulation and overload.


... are you following where this is leading to ?

[detectormods]

Any waveform ( including all PI waveforms ) can be decomposed to a series of sine waves however the series ( or spectrum ) starts with the fundamental frequency. So if you have a 1 Khz waveform of some shape ( square / triangle / ramp , weird minelab transmit waveform , whatever ) the spectrum STARTS at 1 Khz and is broadband ( ie extending from 1 Khz upwards but NOT downwards in frequency ) and will not contain sub 10 Hz components.

if you think about it .. how stimulating would a sub 10 Hz waveform be to a 0.1 gram gold nugget. ?

True, but I posted from experience, by testing, as an example but let’s look at hard facts, yes , lower noise in the 1Khz -100 KHz range is much more important than noise at 10Hz , more important is the slew rate, any laggard behaviour will not capture the dynamics of a target response. But, this is my observance from experimenting, too much bandwidth add noise, but I observed that high noise specification in the lower frequency domain translates into additional noise into the higher frequency spectrum, is it via some mixing component of sub harmonics? I do not know, I have built input stages with the opa1612 ltc2054 that should be exceptional but I get a rise in noise floor, I have even tried building impedance converters but for some reason I do not end up with the performance of the low noise low frequency specified part. I myself would be over moon if someone can explain to me why this happens.

[moodz]

Oh, thank you for explaining what you are trying post, I thought you were transmitting at 10 KHz and observing higher mixing products. So thanks for explaining what it is. But I do find it strange that do not realise that P.I detectors are wide bandwidth in nature and sub10 hertz noise figures are very important.

Any waveform ( including all PI waveforms ) can be decomposed to a series of sine waves however the series ( or spectrum ) starts with the fundamental frequency. So if you have a 1 Khz waveform of some shape ( square / triangle / ramp , weird minelab transmit waveform , whatever ) the spectrum STARTS at 1 Khz and is broadband ( ie extending from 1 Khz upwards but NOT downwards in frequency ) and will not contain sub 10 Hz components.

if you think about it .. how stimulating would a sub 10 Hz waveform be to a 0.1 gram gold nugget. ?