VLF MD with digital signal processing : Bee-Buzz 1

[Aziz]

So the big problem is still there: EMI noise.
White noise is our friend. My ADC system delivers enough of it. Even the preamp delivers a bit.
Any idea about EMI noise problem without an AI RX coil? Filter won't solve any problem here if the frequency of interest is affected by EMI noise source. Reducing bandwidth of the decoder won't solve it either (would slow down the detection rate).

We need some kind of "flux capacitor".

https://en.m.wikipedia.org/wiki/DeLorean_time_machine#Flux_capacitor

Or something equivalent.

[Aziz]

..those discrete opamps can be alot more than $40. Try $200 ...see Burson range ... apparently they are chemically optimised probably to match the $2000 oxygen free single crystalline metal matrix directional copper speaker cables.

You made my day.

[Altra]

Audiophiles are the only group crazier than metal detector builders.

[moodz]

Hi Paul,

I didn't know, that there is a market for audiophile friends to pay every price. 40 $ for one discrete op-amp.

Direct sampling with a good sound card with a single ended RX-front end (just the RX coil), works fine up to 19 - 20 bits resolution (or -114 dB - 120 dB noise floor level). But if we use an anti-interference (AI) RX-coil, then we have plenty room to go down the noise floor level (at least 20 dB) to get the faint signals. In this case, a pre-amp is very fine to do the job to rise the signal level above the noise floor level of the ADC. Even if we don't use an AI-coil, the EMI noise at the frequency of interest may be lower than the noise floor level of the ADC system. In this case, a pre-amp is required too to get the faint signals.

BTW, I like EMI noise. It's measureable if active. It's predictable if active. It's cancelable if active. A big software effort should be invested in the active EMI noise cancelling using the FFT.
Cheers,
Aziz

..those discrete opamps can be alot more than $40. Try $200 ...see Burson range ... apparently they are chemically optimised probably to match the $2000 oxygen free single crystalline metal matrix directional copper speaker cables.

[Aziz]

I found post #384, but there I see only theories, or is there an example.
I familiar with arduino programming

You have to adapt the reference c-code into your arduino system. That's all. It's only a programming issue.

[Aziz]

Those discrete designs usually dont have good CMRR or PSRR because they only focus on noise preformance. Additionally your spec is barely making 16 bit performance ... its never going to be 24 bit compliant.
For 24 bits you will require fully differential amps which is going to be very messy implemented in discrete.
Besides that you are trying to fix asynchronous noise on a synchronous detector.
Great HIFI discussion though just wrong thread.

Hi Paul,

I didn't know, that there is a market for audiophile friends to pay every price. 40 $ for one discrete op-amp.

Direct sampling with a good sound card with a single ended RX-front end (just the RX coil), works fine up to 19 - 20 bits resolution (or -114 dB - 120 dB noise floor level). But if we use an anti-interference (AI) RX-coil, then we have plenty room to go down the noise floor level (at least 20 dB) to get the faint signals. In this case, a pre-amp is very fine to do the job to rise the signal level above the noise floor level of the ADC. Even if we don't use an AI-coil, the EMI noise at the frequency of interest may be lower than the noise floor level of the ADC system. In this case, a pre-amp is required too to get the faint signals.

BTW, I like EMI noise. It's measureable if active. It's predictable if active. It's cancelable if active. A big software effort should be invested in the active EMI noise cancelling using the FFT.
Cheers,
Aziz

[pito]

I found post #384, but there I see only theories, or is there an example.
I familiar with arduino programming

[Aziz]

can you share goertzel code for 10 kHz (single frequency )

The c-reference code is here for download. Just look at the previous pages of this thread.
There is an Excel file in the zip-package "BinCalculator.xlsx". Open it and play with some values (samplerate SR, block size N, Frequency and k).
The k'th bin (value k) must be an integer number for the required frequency.
Then you have to call the following function with the required parameters.

Goertzel_IQ(N, pSamples, k, iFlags, fEpsilon, pDemod);

If you have problems to understand it, I can't teach you the c-coding. Some level of expertise is required.
Cheers,
Aziz

[Marchel]

Here is a simulation of the Pre-Amp
Equinox Vanquish PA sch.pdf in the Tina-IT program, maybe it will be useful to someone.

https://www.ti.com/tool/TINA-TI

[pito]

Hi all,

I have finished the c-reference code for [B]FFT, Goertzel

Aziz

can you share goertzel code for 10 kHz (single frequency )

[Marchel]

I have no experience with direct sampling so, I can't help. If you look at the schematic JoyJo posted, the pre-amp feeds the PCM1861 thru a common mode choke and anti-alias filters. You should study the PCM1861 datasheet

Yes, I've already looked at it and it seems like it could work, I'll try putting this preamp together and give it a try.

[Altra]

Thank you, that's enough for me, this is exactly what I wanted to see, do you think this circuit could work if connected directly to the ADC processor input or would I need any additional components ?.

I have no experience with direct sampling so, I can't help. If you look at the schematic JoyJo posted, the pre-amp feeds the PCM1861 thru a common mode choke and anti-alias filters. You should study the PCM1861 datasheet

[Marchel]

Here is the only photo I could find. RXN and RXP are equal but 180deg out of phase. The yellow trace is the Tx and the blue is the pre-amp out. Not sure which output it was or what mode the detetector running. This is from 3-4 years ago so I don't recall all the details

Thank you, that's enough for me, this is exactly what I wanted to see, do you think this circuit could work if connected directly to the ADC processor input or would I need any additional components ?.

[moodz]

This is why you dont need amplifier at the front end of the direct sampling ADC metal detector.
[ATTACH=CONFIG]n433354[/ATTACH] ....from the FPGA thread.


For a 10 to 16 bit ADC yes an amplifier is useful ... 24 bits ... No
32 bits ... Definitely not.

Back to the code ...why are you using adjacent bins in the FFT. A linear system can't generate new frequencies it's a phase discrimination problem.

[Altra]

Here is the only photo I could find. RXN and RXP are equal but 180deg out of phase. The yellow trace is the Tx and the blue is the pre-amp out. Not sure which output it was or what mode the detetector running. This is from 3-4 years ago so I don't recall all the details