Noise question.

[Teleno]

I assume you've disconnected the coil and the noise is still there?

That's a great tip. I'll take a look at that.

EDIT: Yes, the noise is still there!

But if you're getting 4mV of noise spikes at the input you should be able to see that on an oscope. If you see them at the output but not the input then look at the ADC reference.

The noise at the input is white. Also, the EF cancellation would do away with flicker noise present in the signal, I believe.

The ADC reference is bypassed with a 100n cap to analog ground and Vanalog of the MCU is connected to Vcc by a ferrite bead.

EDIT: OMG this is what the 5V rail looks like with the coil removed! A hum of about 800 Hz. I'm using 8x AAA batteries.

How can this be? Could it be an artifact of the oscope ground loop?




The baseline at the output of the preamp (AC coupled to oscope). Tee regulator is U1 in the schematic https://www.geotech1.com/forums/show...213#post295213 (sheet 1 top right).

[waltr]

Actually I go often to the beach and I see no difference in the noise, it's still there. It's random rather than periodic
. The coil is shielded.

started thinking of checking without coil but Carl beat me to it.

Just saw your scope pics without coil and they do look pretty bad. I suspect there is not enough by-passing and/or bypass caps are not effective and/or
PCB layout is not as good as it should be.

Very possible the 5V regulators need better by-pass. I have seen some regulators do odd oscillations so recheck their data sheet carefully for recommended caps.

Is this noise still there if you hold the Processor in RESET? If it goes away then it is generated by the processor.

Good luck

[Teleno]

started thinking of checking without coil but Carl beat me to it.

Just saw your scope pics without coil and they do look pretty bad. I suspect there is not enough by-passing and/or bypass caps are not effective and/or
PCB layout is not as good as it should be.

Very possible the 5V regulators need better by-pass. I have seen some regulators do odd oscillations so recheck their data sheet carefully for recommended caps.

Is this noise still there if you hold the Processor in RESET? If it goes away then it is generated by the processor.

Good luck

You're right! According to the datasheet the LM1117 regulator needs a cap with ESR < .5 ohm. I've replaced C5 on the schematic by a 47 uF cap with 0.15 ESR and the picture improved dramatically.

This is the 5V rail now.




Noise is down, I can halve the threshold. detection distance for the 0.75 g gold nugget has increased from 8.5cm to 10 cm.

It can even see this small rock embedded gold at 4cm. Delay is 5.5us. Millimeter scale.

[waltr]

Nice. Noise on scope so much better.

That is very good sensitivity to the tiny fleck of gold.

Are you still getting those random spikes or they they just much smaller.

[Teleno]

I'm not sure I understand. You get 3 bits from oversampling+filtering, but you don't get 3 bits from oversampling and another 3 bits from filtering. I suspect you have ~12 ENOBs in the end.

You may find this interesting. See on page 4 how a low-pass IIR may pick up resolution https://www.infineon.com/dgdl/Infine...7d0d4d0c886c92

Nice. Noise on scope so much better.

That is very good sensitivity to the tiny fleck of gold.

Are you still getting those random spikes or they they just much smaller.

The spikes are still there.

I suspect the invalid 2 LSBs of the ADC produce codes that do not have equal probabities and oversampling and decimation cannot cancel them out. I'm going to reduce the ADC clock speed and see what happens.

EDIT: indeed, the peaks are half the amplitude by halving the ADC clock. The peak went from 100 to 50.

[Skippy]

I took a quick scan at the schematic. One suggestion: put low-value series resistors is the power supply to the op-amp IC's. I've no idea of precise values, that would depend on the current drain of the IC's, but 10 Ohms would be a starting point. If the resistor drops 0.1 Volt it's OK. This significantly isolates the PSU decoupling caps from the rest of the electronics, increasing their effectiveness.

[Teleno]

I took a quick scan at the schematic. One suggestion: put low-value series resistors is the power supply to the op-amp IC's. I've no idea of precise values, that would depend on the current drain of the IC's, but 10 Ohms would be a starting point. If the resistor drops 0.1 Volt it's OK. This significantly isolates the PSU decoupling caps from the rest of the electronics, increasing their effectiveness.

The output of the preamp is clean, I don't think the problem is there.

One problem this design has is that the width of the Tx pulse flickers +-200ns. The sampling delay is timed by the MCU precisely at 6.5us after the falling edge, +- 65.5 ns which is the lock cycle. However, the amplitude of the Rx signal varies accordingly. If the flicker followed a gaussian distribution then oversampling would cancel out the signal changes but if it has a 1/f distribution then it turns into low frequency noise.

One remedy is to replace the U5.A op-amp by a fast comparator. My next version will use an LM311 which has about 200ns rise/fall time. The op-amp I'm using (LM6172) has 2.5 us instead.

This disadvantage on the other hand is what makes iron discrimination possible, although at short distances (4 -5 cm). Good enough to discard trash near the surface or to stop digging or pinpointing the extracted material.

I just uploaded this video showing how this PI reacts to a gold chain.

[Teleno]

Iron discrimination of a bottle cap.