EMI Noise Reduction Design Ideas for PI Detectors

[Carl-NC]

Prior to the demods, it's just the BW of the preamp. But throttling the preamp cuts into minimum delay time.
After the demods it's the BW of the integrator & subsequent stages. Those are generally set for response recovery speed so throttling them will make the response sluggish.

[green]

I have been trying to get coil resonance and amplifier cut off around 1MHz. Probably needed for short tau targets, around 1us. For rings, coins, >1/2gram nuggets and longer tau targets maybe 200kHz is high enough. Tried an analysis in spice with white noise at 200kHz and 1MHz cutoff. Was expecting to see a greater difference. I like playing with spice but I'm a novice. Don't know how to tell the frequency of the white noise. Don't know if I should be looking at different cutoff frequencies. Planning to look at integrator or differentiator output with 1MHz and 200kHz amplifier cutoff with my bench circuit. Any thoughts on what I'm planning or should I be trying something different?

[moodz]

The best noise mediation I have found is a demodulating feedback loop at the first amp .... works like a charm. No need to sync with the mains either.

A is with noise removed
B is with noise not removed ... ;-)

[green]

The best noise mediation I have found is a demodulating feedback loop at the first amp .... works like a charm. No need to sync with the mains either.

A is with noise removed
B is with noise not removed ... ;-)

[ATTACH]55908[/ATTACH]
[ATTACH]55909[/ATTACH]

Thanks for the circuit. Don't see why circuit would help with EMI or mains. Need to do some studying.

[Carl-NC]

That's a good solution for LF noise. One thing you rarely see in PI designs is any high-pass filtering applied to the preamp, yet it can be done and will help with LF noise. The feedback integrator is more elegant and let's you select what part of the signal gets applied for filtering.

Paul, in practice do you sample the integrator during TX as shown in the sim? I would expect induced EMI to be heavily attenuated by the switch. Seems like you would sample just before TX.

[Altra]

The best noise mediation I have found is a demodulating feedback loop at the first amp .... works like a charm. No need to sync with the mains either.

A is with noise removed
B is with noise not removed ... ;-)

[ATTACH]55908[/ATTACH]
[ATTACH]55909[/ATTACH]

Nice outside of the box thinking! Would this scheme allows use to eliminate the hp or sat circuits in later stages?

Thanks for posting

[green]

That's a good solution for LF noise. One thing you rarely see in PI designs is any high-pass filtering applied to the preamp, yet it can be done and will help with LF noise. The feedback integrator is more elegant and let's you select what part of the signal gets applied for filtering.

Paul, in practice do you sample the integrator during TX as shown in the sim? I would expect induced EMI to be heavily attenuated by the switch. Seems like you would sample just before TX.

I have been thinking the noise problem isn't LF noise(<1/2pps). It's aliasing higher frequency noise(>1/2pps). Maybe I've been thinking wrong?

[Carl-NC]

Would this scheme allows use to eliminate the hp or sat circuits in later stages?

The SAT stage is mostly for removing the demod offset, which can vary with sample delay. LF noise suppression doesn't affect this much.

I have been thinking the noise problem isn't LF noise(<1/2pps). It's aliasing higher frequency noise(>1/2pps). Maybe I've been thinking wrong?

Both will happen.

[green]

Seems like the high pass filter might effect EF cancellation. Does it?

Would changing the amplifier feedback capacitor so amplifier cutoff was around 100Hz allow testing for LF noise(no high pass filter)? If not, what would be a good way?

[moodz]

That's a good solution for LF noise. One thing you rarely see in PI designs is any high-pass filtering applied to the preamp, yet it can be done and will help with LF noise. The feedback integrator is more elegant and let's you select what part of the signal gets applied for filtering.

Paul, in practice do you sample the integrator during TX as shown in the sim? I would expect induced EMI to be heavily attenuated by the switch. Seems like you would sample just before TX.

Thanks Carl

I sample during RX not TX .... the way the TX is connected the sample switch is ON when the coil TX is OFF.

Attached the sim file below.

basicNOISE.zip

[moodz]

Seems like the high pass filter might effect EF cancellation. Does it?

Would changing the amplifier feedback capacitor so amplifier cutoff was around 100Hz allow testing for LF noise(no high pass filter)? If not, what would be a good way?

yep it nails EF totally if you select the right loop filter Green. I dont have any other EF sampling in my design. I generally select components for the feedback so the cutoff is around 1000 Hz not a problem as my PIs run at 10 - 19 Khz... cutoff should be high enough to attenuate EF, mains etc ... but not so high it approaches the TX frequency of the PI or there will be an impact on target signals.

[moodz]

Nice outside of the box thinking! Would this scheme allows use to eliminate the hp or sat circuits in later stages?

Thanks for posting

Thanks Altra ... this idea tries to prevent the low frequency noise being modulated onto the high frequency signal waveform before further sampling ... so you will still need additional sampling / hp / sat etc as Carl mentioned.

[moodz]

Seems like the high pass filter might effect EF cancellation. Does it?

Would changing the amplifier feedback capacitor so amplifier cutoff was around 100Hz allow testing for LF noise(no high pass filter)? If not, what would be a good way?

As the cap increases the cut off gets lower in freq ( really a roll off ) .... however its not a high pass ... the gain gets lower as frequency goes higher.

It is very important to realise that the low frequency noise will be modulated onto the higher frequency components due to the non linearity when the amp overloads due to periodic TX.

Below is a 50 Hertz hum of 10 mv at the coil for A and B and the spectrum plots at the 5 Khz component from the FFT of the A vs the B output .... you can see that the 50 hertz induced sidebands are reduced by around 40 db with this circuit ... better can be achieved.

[Carl-NC]

Seems like the high pass filter might effect EF cancellation. Does it?

Yes, it will help suppress the EF signal.

Would changing the amplifier feedback capacitor so amplifier cutoff was around 100Hz allow testing for LF noise(no high pass filter)? If not, what would be a good way?

I would turn off the TX and try various preamp BWs, from 100Hz to 100kHz.

[green]

I would turn off the TX and try various preamp BWs, from 100Hz to 100kHz.

Some scope pictures. Recordings were made on the kitchen table. Don't see any LF(<100Hz)noise if I did it right. Anyone see anything that doesn't make sense?

capacitor was used to change amplifier cutoff frequency