60Hz noise problem

[Tinkerer]

Hi Tinkerer:

I don't think any filtering of the 3 to 20 Hz band should significantly affect your pulses, because your pulses create an analogy to a "chopper" amplifier, which is able to amplify down to DC with a pass-band amplifier by chopping the input signal to shift the band up into the amplifier pass-band.

But maybe you have observed otherwise. I would like to understand why if that is so.

Regards,

-SB

Thanks for the feedback.
Indeed I should take a closer look. The input is different from the output. What is valid on the one is not valid on the other.

Tinkerer

[Tinkerer]

directly

humhum, thank you for the Corbyn file. It is a good idea to look at it, since it is the original basic idea of my design.

I will study the Corbyn design and then maybe we could discuss some points, or you could help me to better understand some parts.

Tinkerer

[simonbaker]

Hi kt315,
thanks for the help.
Yes, my eyesight is bad and my hearing is bad and my mind is slow and my hands are shaky. These are the things that come wit old age.
This is why I much appreciate your help and patience.

I have now changed the timing so that the cycle is 166.5uS, to coincide with 60Hz. The noise is less.
However, the 60Hz of the local power is not so perfect either, so I should think of making small changes possible in the field. Can anybody help me with the PIC code to do that?

Tinkerer

Not knowing your circuit well, I'm wondering why changing the pulse frequency would help your 60 Hz noise.

It would seem to me that unless you are phase-locked to the noise signal, you'll still get a "beat frequency" noise that eventually comes in and shifts your baseline level -- unless you filter out really low frequencies before sampling. Even if phase-locked, you'd have a random bias.

How do you explain it helping?

Regards,

-SB

[moodz]

Tinkerer ... the best remedy for unwanted signals is not to let them in in the first place. Look at all the single ended unbalanced inputs in your front end. You are 'wasting' the common mode rejection capability of your op amps with the way you have them connected.
Most of this discussion has ignored this fact.

Regards,

Moodz.

[Tinkerer]

Not knowing your circuit well, I'm wondering why changing the pulse frequency would help your 60 Hz noise.

It would seem to me that unless you are phase-locked to the noise signal, you'll still get a "beat frequency" noise that eventually comes in and shifts your baseline level -- unless you filter out really low frequencies before sampling. Even if phase-locked, you'd have a random bias.

How do you explain it helping?

Regards,

-SB

Thanks for the feedback.

You are absolutely right. Yesterday I run the circuit with the synchronized cycle, but not phase locked. The noise was reduced by about 85%. (I also did some other adjustments) However, as I was watching the output on the scope, every now and then I got a big unexplained noise spike. This could well be a "beat".
So the search for improvements goes on.

Tinkerer

[Tinkerer]

Tinkerer ... the best remedy for unwanted signals is not to let them in in the first place. Look at all the single ended unbalanced inputs in your front end. You are 'wasting' the common mode rejection capability of your op amps with the way you have them connected.
Most of this discussion has ignored this fact.

Regards,

Moodz.

Thanks for the feedback.
I totally agree about not letting the unwanted signals in in the first place.

I thought that I had a differential input that should take care of the common mode signals. You tell me that is not so?
Could you please explain?

In this circuit I use a 2 opamp differential amplifier. This I intend to change for a IA once I get the kinks out and move to a solid PCB. the IA is too expensive to use on the breadboard.

Tinkerer

[simonbaker]

Thanks for the feedback.

You are absolutely right. Yesterday I run the circuit with the synchronized cycle, but not phase locked. The noise was reduced by about 85%. (I also did some other adjustments) However, as I was watching the output on the scope, every now and then I got a big unexplained noise spike. This could well be a "beat".
So the search for improvements goes on.

Tinkerer

I would expect your received pulse would slowly drift up and down rather than a sudden noise spike, but I'm not looking at the particulars of your circuit.

I agree with moodz that if this noise is in any way a "common mode" signal, techniques to cancel it out would be the first line of attack.

He's an expert on those configurations -- I'll be interested in what he proposes.

It's great to see your results coming in.

Regards,

-SB

[okantex]

Hi Tinkerer ,

""not related to the TX timing" got me confused now. When I made the timing to sync with 60 Hz, by using a cycle of 166uS, I established a relationship.
Does this make it more difficult for digital filtering?

What cycle time do you suggest I use?"
how did you decide to use 166us cycle,is there any calculation or just trail and error.

btw is this the cycle which enables you to reduce noise %85 .,,

kind regards
Okantex

[Aziz]

Hi Tinkerer,

get either a SSM2019 or INA163. Both are working fine and quiet. And remove all input resistors if you are using an IA (degrading of CMRR).

This should remove the induced EMI on the differential RX coil leads. But it won't remove EMI, if the RX coil flux area is inducing an external magnetic field (it's not common mode EMI, but it's a differential mode EMI).

To eliminate differential mode EMI:
You can use integrators instead of just sampling the signals via S&H. An integrator is a typical low pass filter and will remove high frequency noise. But you need two sampling windows of same period of time to remove the low frequency noise. This is done by subtracting one integrated window from other integrated window. You have then a good band pass filter.

Even if you don't use an integrator, the subtracting of two samples at adjacent times eliminates quite high amount of low frequency noise.

Good luck.

Aziz

[simonbaker]

Hi Tinkerer,

get either a SSM2019 or INA163. Both are working fine and quiet. And remove all input resistors if you are using an IA (degrading of CMRR).

This should remove the induced EMI on the differential RX coil leads. But it won't remove EMI, if the RX coil flux area is inducing an external magnetic field (it's not common mode EMI, but it's a differential mode EMI).

To eliminate differential mode EMI:
You can use integrators instead of just sampling the signals via S&H. An integrator is a typical low pass filter and will remove high frequency noise. But you need two sampling windows of same period of time to remove the low frequency noise. This is done by subtracting one integrated window from other integrated window. You have then a good band pass filter.

Even if you don't use an integrator, the subtracting of two samples at adjacent times eliminates quite high amount of low frequency noise.

Good luck.

Aziz

Hi Aziz:

Can you show a sample circuit of that? I'm not clear at all on your idea.

Also - why not a simple high-pass filter to eliminate the 60 Hz? He doesn't seem to have a problem with high frequency noise it seems.

Regards,

-SB

[Tinkerer]

Hi Tinkerer ,

""not related to the TX timing" got me confused now. When I made the timing to sync with 60 Hz, by using a cycle of 166uS, I established a relationship.
Does this make it more difficult for digital filtering?

What cycle time do you suggest I use?"
how did you decide to use 166us cycle,is there any calculation or just trail and error.

btw is this the cycle which enables you to reduce noise %85 .,,

kind regards
Okantex

Hi Okantex,

thanks for the feedback.

1Second =1,000,000uS:60Hz =16,666.66666666666666uS.

So with one cycle being 166.6us, there are 6002 cycles or PPS, pulses per second.
I think this is what kt315 meant when he recommended to use 120 PPS with 60Hz.
The other change that gave a lot of improvement, is the integrating of 200 samples. This corresponds to about 15Hz, which corresponds to a fast response and good sweep speed for a 480mm diameter coil.

Another change that I tried without conviction, is a 7.2kHz high pass filter at the input. Surprisingly it made a notable improvement.


At this stage I detect a US$25c with 3mV signal and about 1.5mV noise, so it is the limit of detecting distance. FE discrimination is outstanding at the same distance.

I have not run the tests with all the targets yet, maybe some negative surprises turn up there.

The gain is still low, so now I need to increase the gain without increasing the noise.

Tinkerer

[simonbaker]

Hi Okantex,

thanks for the feedback.

1Second =1,000,000uS:60Hz =16,666.66666666666666uS.

So with one cycle being 166.6us, there are 6002 cycles or PPS, pulses per second.
I think this is what kt315 meant when he recommended to use 120 PPS with 60Hz.
The other change that gave a lot of improvement, is the integrating of 200 samples. This corresponds to about 15Hz, which corresponds to a fast response and good sweep speed for a 480mm diameter coil.

Another change that I tried without conviction, is a 7.2kHz high pass filter at the input. Surprisingly it made a notable improvement.


At this stage I detect a US$25c with 3mV signal and about 1.5mV noise, so it is the limit of detecting distance. FE discrimination is outstanding at the same distance.

I have not run the tests with all the targets yet, maybe some negative surprises turn up there.

The gain is still low, so now I need to increase the gain without increasing the noise.

Tinkerer

Maybe out in the field a lot of that noise will go away. Some noise just can't be separated. But if you can see it and identify it, there's hope. Keep describing your remaining noise.

-SB

[Tinkerer]

Aziz, thanks for the help.

I have a SSM2019N, that I will use once I have the PCB ready. Changing back to +/-5V supply I can try the integrator, to see if there is a difference. My S&H circuit does actually quite a good job at integrating, but it would not work with a S&H that uses an internal capacitor.

I will update the schema with the changes and post it again, there are still many things that need to be improved.

Tinkerer