My stm32 powered pi detector

[ivconic]

Ah I see, gee I have lived a sheltered life lately with pic32 and blackfin
thanks fir the link Ivconic
cheers
Mdtoday

Stay with us!
There is a red pill too!

[waltr]

hi gtdavid,

had a look at specs of st32h7 internal opamps and they would be fine for the purpose of ADC filters /drivers or second gain stage etc but like most combined analog digital chips, they are a compromise. Having said that these particular ones are much better than I expected but really in my books still not enough for a good front end. Just my thoughts and experience.

On another note, it's a pity the internal DAC were not also 16 bit

pretty impressive chips though��

Cheers

Mdtoday

Yes, internal op-amps good enough for those uses but not the Front End.
The hardest part of this project will be a good Front end that has Low Noise and FAST recovery. There are many designs and discussions in this forum.

Maybe the Front End should be on a separate PCB to allow easy modification, changes and experimenting.
The Main PCB can have the STM32, power supplies and other support circuits.

I also like the idea of using one of the small experimental STM32 boards like that 'Blue Pill'. This allows many to use a dense STM processor without soldering STM.
The detector Main PCB can then have headers for the processor board to plug in.

[ivconic]

...The detector Main PCB can then have headers for the processor board to plug in.

[dbanner]

Yes yes yes, projects are good. Problem with pi is no discrimination. Never will be regardless of sophistication of digital processing. Such talents are better wasted on VLf IB.

[dbanner]

An innovative approach would be high speed variable pulse width tx with synchronized electronic damping and comparative sampling. Then one could create a profile of conductive target which would be unique to specific metals or even alloyed metals regardless of its mass. the data can then be extracted and form the basis of target ID.

[waltr]

Tinker has shown a possible method to discriminate Iron with a PI if using an IB coil. I have also looked at this and can see a phase shift during the TX pulse (see my HH2 thread from discussion and scope captures).
I am planning on code to measure this phase shift but haven't yet had the time (winter project).

And as dbanner suggests, there are other methods that could work. We just need a versatile hardware platform and open source code (Github) so we can work on these ideas.

[gtdavid]

The white pill and it's a big one!!
http://www.st.com/en/evaluation-tool...eo-h743zi.html (with programmer)
I could work with that, I am just worried about noise..
The h bridge will still be a bit of a pain but it's not a 100 pin I guess.
I do have hand solder SMD footprints also.
I'm building this pi as my soil is bad real bad.. and there is the old rule of "ya gotta dig it".
I live 20 minutes from gold fields and 40 minutes from beaches.
I have done some work tonight.
ported over to the STM32H753VITx in the schematic.
Using the h7 I have removed the graphics controller as the h7 should be fine to drive the LCD direct via onboard LTDC
looking at the MAX9939AUB+ opamp
removed the 3.5 jack and added a speaker/jack header so rough pay would not damage the board.
re-done the audio to use the h7 SAI interface, resulting in fewer components needed

[green]

An innovative approach would be high speed variable pulse width tx with synchronized electronic damping and comparative sampling. Then one could create a profile of conductive target which would be unique to specific metals or even alloyed metals regardless of its mass. the data can then be extracted and form the basis of target ID.

I don't see how but maybe I'm missing something. The 1inch square of aluminum can, copper clad board and lead sheet chart almost the same decay.

[dbanner]

Is the linear relationship among these three metals constant regardless of tx freq and pulse width. Even small variance in conductivity over a wider range of exposure should produce significantly different decay curves, when combined in totality should create a unique signature for that specific metal, at least with enough variance for DSP to observe.

[dbanner]

Now we enter into the realm of multi freq. PI

[ivconic]

The white pill and it's a big one!!
http://www.st.com/en/evaluation-tool...eo-h743zi.html (with programmer)
I could work with that, I am just worried about noise..
...

Now you talkin...!
Good point. Such boards can be noisy.
Depends on what method of signal processing you'll prefer.
Dedicated ADC can be mounted on smaller front end pcb and than you'll not have that problem anymore.
As far as i will say in public (and than shut up for good) is this;
WM8738 and WM8501 are ultimate choice.
True 24 bits at superb sampling rate for this purpose.
Beats the crap out of any ARM's adc that i know. Consumes almost none power.
I am already using them in quite another kind of device.
And top of the cake is; you don't need more horse power at mcu with those.
The other day i trashed another bijou; WM8731. All in one package.
But its already known from X-Terra.
...
For hobby purpose and for experimenting; white pill will do splendid, i am sure.

44.5 buks here:
https://www.aliexpress.com/item/NEW-...ceBeautifyAB=0

[drninja]

Hi guys! Long time reader of the forum, but first poster...
The F3,L4,H7 series mcus are basically equal to an underpriced but very good ADC with a bunch of free ram, peripherals, and ofc. a processor.
The L4 for example @10MSPS 12bit 3V has about 400nV input referred noise if you manage to toggle a couple of LSBs with out of band dither.
I did not know about the H7, but from datasheet, it should have about 95nV input referred noise @ 16bit 7.2MSPS ofc. with lsb toggling.
I do experiment with the L476 in a bit more complex PI topology, with okayish results...

Back to the ADCs, i just wanted to write a couple of tips, that i found really useful:
- Only use differential input configuration, it gives an order of magnitude better noise performance.
- Do not toggle any gpio at high speed if you want to keep the ADC's performance, putting series resistors in gpio lines helps even more
- Use unused gpio pins as gpio outputs tied to logic H and L, and decouple them with 10-100pF caps to ground plane to lower ground bounce of the chip (keep the traces short)
- Its worth putting some effort into the clock distribution system to get easily filterable tones from different internal clocks
- Never ever use usb peripheral, while adc is sampling

[6666]

https://satoshinm.github.io/blog/171...ogramming.html

Thats amazing, trouble is with all these advances in technology you need to be fresh out of University or work in the design industry, or very smart to be able to understand it
I'm still playing with 12f683's

[waltr]

Back to the ADCs, i just wanted to write a couple of tips, that i found really useful:
- Only use differential input configuration, it gives an order of magnitude better noise performance.
- Do not toggle any gpio at high speed if you want to keep the ADC's performance, putting series resistors in gpio lines helps even more
- Use unused gpio pins as gpio outputs tied to logic H and L, and decouple them with 10-100pF caps to ground plane to lower ground bounce of the chip (keep the traces short)
- Its worth putting some effort into the clock distribution system to get easily filterable tones from different internal clocks
- Never ever use usb peripheral, while adc is sampling

All very important points when using Digital chips on a sensitive analog circuits- we will still need a gain of 500 to 2000 on the RX.
Even then it is really hard to get more then 18 bits of noise free ADC value.

Also:
Ground and power distribution is very important to prevent coupling digital noise into high gain analog.

[gtdavid]

Ok so what I'm getting form this is that we want a modular system with as little SMD as possible.
Nucleo stm32 f7 or h7 pin compatible with each other.
- on board USB programmer
- on board power from 7-12 v input
-$25 to $30 from mouser
Driver module
- PWD13F60 eval board full bridge ($30) this foot print would become driver standard
- second std MOSFET driver in a board config matching above
Audio module
- yet to look but there are plenty pre made options.
Adc/opamp module
- yet to finalise but would have plenty of pins for customisation
LCD header
- simple pre-made ffc 40 pin to 2.54 header breakout

If you think this would be a more acceptable hobby version let me know and I will head in that direction