OpAmp based EMI noise-canceling distorts PI decay curve, help

[radioslicer]

I've built a variation of the PI detector, added Opamp-integrator noise-canceling, got unexpected and undesirable signal distortion. What is causing it? How do I remove it?

1Khz pulses.
Full schematic - full-schematic.pdf
The portion in question:



I had a lot of noise in the amplified signal i.e., noise from my two-stage pre-amp. I think 180Hz noise from saturated main transformers is the worst.

I found this thread where moodz shows an elegant way to integrate and cancel out the noise.

https://www.geotech1.com/forums/foru...ge2#post306066

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]

I've tried it both in LTSpice and in for real. It does significantly eliminates the noise. In the image below, yellow trace is the signal after first pre-amp stage (x10 gain), the blue trace is the B+ net, output of the noise integrator and input into non-inverting pin of the second pre-amp stage.
It looks like 180Hz noise is nicely reflected.




Unfortunately, it also causes decay curve distortion.
Here is how it looks with no target. The voltage of the purple line does drop slightly toward the tail (and EFE pulse)
The yellow trace is the first pre-amp stage
The purple trace is the second stage after the AC coupling cap.



Here is how it looks with the target presented. The "distortions" is visible. I know it's not a distortion, it's the effect of noise integrator U7A, but I can't find a better way to describe it.
The yellow trace is the first pre-amp stage
The purple trace is the second stage after the AC coupling cap.



The bilateral switch U4C feeds the signal into the noise-integrator opamp. The control signal is inverted MOSFET PWM, i.e. when mosfet is open the U4C is closed and vice versa.

How do I get rid of the decay curve hump? I tried playing with R6, C9, and R32 but got marginal flattening or oscillations. Dampening resistor value on schematic is wrong, please ignore.

[radioslicer]

Here is ltspice project
metal-detector-2023-07-02.zip

[Sean_Goddard]

Here's a simple solution which is used in sonar. I don't think I'm breaching the Official Secrets act, but hey, who cares right? Here goes;

If an amplifier (single stage) has a noise factor of say 10dB, what you can do is put a DUPLICATE circuit IN PARALLEL with it. This will reduce the noise by -3dB. 4 amplifiers in parallel -6dB, and 8 amplifiers in parallel by -9dB thus, every doubling of the amplifiers reduces the system noise by 3dB. The reason for this is known as CORRELATION.

Noise, by it's very nature is random so it stands to reason that two noise sources should cancel eachother out to some degree (understand what noise in an electronics system is first, then you will understand WHY this works) so 4 noise sources will further cancel. Like the almost zero inductance cables we designed for Boeing. 400A current draw @ 115Hz caused too high a volts drop across the test equipment we had designed due to skin effect. Using larger cables was NOT allowed due to overall weight of the TE, so, we made up a Litzendradt cable of many smaller cables in pairs. One pair wound anti clockwise, one pair wound clockwise (4 cores). Two of these 4 cores wires were then would one way, and another two the other. With 32 wires, the volts drop at 400A was about zero as we had nulled out the magnetic effects of the wires, AND the overall weight was half of the limit.

Sometimes, you have to think outside the box.

[Mr.Jaick]

your timing is not optimal.
use a 100us pulse just before the next TX period starts, this is for the sampling servo.
or to put it in another way, use the last 100us of your RX period.
reduce the R32 and C9 values as well.

[moodz]

The cct is a band pass filter so you won’t get a classic decay curve which has a DC component.

moodz

[kt315]

use a 100us pulse just before the next TX period starts,

before PWR signal start. look DP schematic, that is.