Android Phone Based Metal Detector - FCMD

[Marchel]

I started testing this STM32 wireless board and it is also very good, just need at least 16bit external ADC. For a metal detector that will work with a mobile phone via Bluetooth, it is a good solution.

https://www.aliexpress.com/item/1005...origin_prod%3A

[moodz]

Hi Paul,

maybe I should buy a few STM32H743VIT6 boards to be able to use your ultimate detector soon.
The STM32 boards are really quite interesting.
Cheers

The sound card is interesting for playing with the math .. but you need low level access to timings and data to achieve real results and the STM32 boards allow that for low cost.

[moodz]

I started testing this STM32 wireless board and it is also very good, just need at least 16bit external ADC. For a metal detector that will work with a mobile phone via Bluetooth, it is a good solution.

https://www.aliexpress.com/item/1005...origin_prod%3A

I would not bother with an external ADC unless it was 24 bit or more. The ADC implemented on the H7 series is a "proper" 16 bit SAR ADC.
On the H743 there are 3 of these onchip ADCs ..
Below is the demodulated ( 0 - 25 hertz ) noise floor of the current firmware running on the chip with the coil connected and the TX running in a noisy lab environment.

we are interested in around the 5 hertz point ( target sweep speeds ) ... the adc is better than -120 db in noise performance. This rivals alot of so called 24 bit converters.
The real benefit is the tight integration of the ADC to the CPU ... and the H743 is a cheap chip ..
you will be hard pressed to build a better combo unit using an external ADC that is anywhere near this price point and performance.

[Aziz]

Hi Paul,

yes, the STM32 platform will allow true real-time DSP processing. Some very special real-time algorithms can be made possible (latency much lesser than 1 ms). For instance a real-time software PLL implementation for demodulating the RX channel via the TXref channel.

As you have 3 ADC channels, you could use one for TX-ref and one for RX channel. With a relative High-Q LC tank in IB-configuration (Q around 10 is enough), you can process the target response on both channels. And a big plus: Active EMI noise cancellation as the TX and RX coils will pick up EMI too, which can be detected and eliminated.

I think such a complex IB design processing has not been done yet.
The STM32 platform will make it easily possible with less problems which I am facing with the sound card solution at the moment.
Cheers

[Marchel]

I would not bother with an external ADC unless it was 24 bit or more. The ADC implemented on the H7 series is a "proper" 16 bit SAR ADC.
On the H743 there are 3 of these onchip ADCs ..
Below is the demodulated ( 0 - 25 hertz ) noise floor of the current firmware running on the chip with the coil connected and the TX running in a noisy lab environment.

we are interested in around the 5 hertz point ( target sweep speeds ) ... the adc is better than -120 db in noise performance. This rivals alot of so called 24 bit converters.
The real benefit is the tight integration of the ADC to the CPU ... and the H743 is a cheap chip ..
you will be hard pressed to build a better combo unit using an external ADC that is anywhere near this price point and performance.

Yes, I also use these STM32H7 boards and they are really great, the STM32WB board is good for wireless applications for transmission via Bluetooth or the LLD method which works similarly to the NRF24L01, the only disadvantage of these chips is that they do not have a 16bit ADC like the STM32H7 has.

[moodz]

Yes, I also use these STM32H7 boards and they are really great, the STM32WB board is good for wireless applications for transmission via Bluetooth or the LLD method which works similarly to the NRF24L01, the only disadvantage of these chips is that they do not have a 16bit ADC like the STM32H7 has.

That is the point I was trying to make.
Yes there are boards that have great bluetooth or whatever ... .but are you trying to build a metal detector or a great wireless box ?

A good and cost effective method of capturing the target signal ( ie ADC in this case ) ...is way higher priority in the design criteria than a wireless system.

[moodz]

..so we have achieved good sensitivity for target detection in FCMD.
We are now getting the corresponding accuracy in discrimination ( without phase rotations or reference phases I might add ! )

The phase accuracy is not impacted by GB either. ( though amplitude will drop a bit for some targets aligned in the phase plane with the ground. )

Below we have copper, silver, gold, al foil, ferrite.
The orange ring is the threshold setting ... anything exceed the threshold gets a phase tag.

Theres no EMI ... because EMI is not a big issue in narrowband systems ( Aziz )

[Aziz]

Nice Paul.

But my sound card is adding noise. And yes, the EMI noise is also present. With Q-factor beyond 100 I am working with with mono coil VLF, there is EMI noise. We can't ignore it. Even if my demodulator bandwidth is very very low (< 1 Hz).

This isn't an issue, if we use standard IB coil configuration with higher TX current however. This is very easy to get there.
But I am still working on mono coil VLF design.
Aziz

[moodz]

Nice Paul.

But my sound card is adding noise. And yes, the EMI noise is also present. With Q-factor beyond 100 I am working with with mono coil VLF, there is EMI noise. We can't ignore it. Even if my demodulator bandwidth is very very low (< 1 Hz).

This isn't an issue, if we use standard IB coil configuration with higher TX current however. This is very easy to get there.
But I am still working on mono coil VLF design.
Aziz

...hmm well you should know there is no physical system to achieve EMI cancellation using external coils or fairy dust within a 1 hz bandwidth.
The other problem is a that with a 1hz bandwidth the detector would be unusable due to lag.
I use 50hz as the token ouput rate on the FCMD detector ... so the bandwith is 25 hz.
The noise floor is around -110db to -120db ... the noise you do measure there is coming from the ADC reference ... not EMI.
You cant fix that with a coil ... it needs better components.
... also remember that EMI is not synchronous ... synchronous EMI can only come from your own system. No coils involved.
If you are picking up non synchrounous EMI then your code is broken.

[Aziz]

...hmm well you should know there is no physical system to achieve EMI cancellation using external coils or fairy dust within a 1 hz bandwidth.
The other problem is a that with a 1hz bandwidth the detector would be unusable due to lag.
I use 50hz as the token ouput rate on the FCMD detector ... so the bandwith is 25 hz.
The noise floor is around -110db to -120db ... the noise you do measure there is coming from the ADC reference ... not EMI.
You cant fix that with a coil ... it needs better components.
... also remember that EMI is not synchronous ... synchronous EMI can only come from your own system. No coils involved.
If you are picking up non synchrounous EMI then your code is broken.

My biggest noise source is the sound card (DAC, output amplifier, ADC, input amplifier, ADC reference, power supply, etc.. ).
So I have ignited the next stage to overcome the issues. Using an amplifier. This should improve the overall SNR by 20 - 30 dB (depending on the gain).
I hope to get with the mono coil VLF close to the performance of a DD coil configuration.

The DD coil configuration delivers similar detection depth performance like yours. But this is trivial.
I want to have the mono coil VLF. Right now. Immediately.
Cheers

[moodz]

I have been using the bipolar pulsing H bridge which uses 6 mosfets + driver chip ..... the bipolar square waves are nice and wideband but produce transitions that cause problems in the RX chain.

So I am switching ( there is a small pun there ) to an E class amplifier to generate the TX signal.
This amp only uses a single mosfet and some passive components.
It was invented back in the 70s.

The secret sauce is in the component values and the timing ... especially the timing
Even though the circuit is simple ... there is alot going on in there.
LT is the transmit coil.

There is also an F class amplifier ... that is for harmonics ... Its not suitable for the FCMD so I did not use it.

Long story short I got the AI to build me a fully auto pushbutton tool that generates the correct values for the E class amplifier and now I am looking at very nice high power efficient transmit signals.
It might look like the components have to be spot on ... but that is not the case. When you find the operating point using the tool the amplifer can work over a range of timings / frequencies.

Below is the tool ... I am still modding it so no point releasing anything yet.
It also builds an ltspice circuit once you find your optimal specs so the simulation can be checked.

[boilcoil]

It would be smarter if you could work in the joint/common region of ZVS and ZSS for maximum efficiency.

[boilcoil]

Something is a duplicate post.

[moodz]

It would be smarter if you could work in the joint/common region of ZVS and ZSS for maximum efficiency.

is 99.4 % efficiency not high enough ???

E class does use ZVS... not sure where ZSS comes in thats more of a motor control parameter.

[boilcoil]

In practice, that's not exactly the case.

Abstract
"The Class E power amplifier has a big popularity because of its high efficiency. The above can be achieved in optimum class E when both soft commutation conditions have been met– zero voltage and zero voltage slope of during turn on. Meeting only the one of the above conditions is described as sub-optimum mode".

The graph below is for three modes:
- optimal s = 0
- suboptimal s < 0
- suboptimal s > 0



The actually achieved efficiency coefficient is almost 95% in optimal mode.