Digital signal processing can see below noise

[Mindlab]

Ten or so years ago i was experimenting with data transmission and recovery below the noise floor. The testing we carried out was successful and is commonly used in hf radio. I have carried out the same type of experiments on both pi and vlf metal detectors. The theory holds firm for the testing carried out so far. Depth increases of near 40% are being observed when being tested on common equipment designs. DSP and associated fourier transform analysis seems to also put forward a way to discriminate in heavy mineral soils such as found in parts of Asia and Australia.

[authere]

signals below the noise floor

Hi mindlab,
You definetely have baited us, we all waite in antisipation, you sound as though you have already tried this in metal detectors some results would be good.
Regards Ron

[Carl-NC]

CDMA routinely extracts cell phone signals from below the noise floor, using signal correlation techniques. You can also oversample with an ADC and pick up 3dB per rate doubling, ideally.

[Mindlab]

There are many techniques used in radio communications that can be applied to metal detectors. Due to confidentiality contracts i am not at liberty to go into much detail. Keep a look out for new patents and a new commercial player.

[simonbaker]

Ten or so years ago i was experimenting with data transmission and recovery below the noise floor. The testing we carried out was successful and is commonly used in hf radio. I have carried out the same type of experiments on both pi and vlf metal detectors. The theory holds firm for the testing carried out so far. Depth increases of near 40% are being observed when being tested on common equipment designs. DSP and associated fourier transform analysis seems to also put forward a way to discriminate in heavy mineral soils such as found in parts of Asia and Australia.

I think the synchronous detector in common VLF metal detectors is very similar to FFT/correlation at single frequency and definitely picks out signals below noise floor.

What type of processing beyond that are you referring to?

Regards,

-SB

[mts]

I am particularly interested in hearing about DSP techniques for PI detectors. I'm currently building a detector that uses direct sampling with a microcontroller. I've gotten my air depth on a quarter up to 10" with brute force techniques. Going any further requires that I break through the noise floor and perform signal detection properly. It is very clear from my testing that noise is a serious issue in my system. And cancelling it out or being able to see through it is going to mean the difference between a mediocre detector and a great one.

I'm just getting started and I don't even know what to Google to start reading up on it. If anyone has some terms or ideas to get me on the right path I would very much appreciate it. I'm looking specifically for ideas on noise cancellation and improved target detection. I've pulled out my old digital signal processing textbooks but they only cover simple stuff like basic filters. Nothing on target detection or breaking through the noise floor.

Thanks!

[goldfinder]

1000 to 1 below noise w/ DSP

I worked as an expert of DSP consultant on DSP for military and we developed many techniques for getting signals out of noise. We routinely got 1000 noise to 1 signal extraction w/ over 90% probability.

Goldfinder

[WM6]

I worked as an expert of DSP consultant on DSP for military and we developed many techniques for getting signals out of noise. We routinely got 1000 noise to 1 signal extraction w/ over 90% probability.

Yes, but what to do, if one can extract "target" signal from 1m of depth? I am not willing to dig 30 hole per day at 1m of depth. Nor three 1m deep hole to find finally only piece of copper wire or brass washer at 1m depth.

I need only secure discrimination and identification of target at todays detector depth of say 30cm.
Better S/N is here of little use.

[hotrocks]

Ten or so years ago i was experimenting with data transmission and recovery below the noise floor. The testing we carried out was successful and is commonly used in hf radio. I have carried out the same type of experiments on both pi and vlf metal detectors. The theory holds firm for the testing carried out so far. Depth increases of near 40% are being observed when being tested on common equipment designs. DSP and associated fourier transform analysis seems to also put forward a way to discriminate in heavy mineral soils such as found in parts of Asia and Australia.

This is no big deal its being done already in some new Australian designed detectors
http://goldprospecting.invisionplus....g&showforum=71
hotrocks

[wam]

May You say what type of hardware You work on? - I understood that it is a PSD development PCB. In electronics there are ways to reduce noise, You just need to know in which places. I have tried to combine “Friendly Arm” module driven by ARM9 and it was very noisy experiment. My conclusion was that it needs lots of "noise work".
Presently I am developing solid analog part and then try to
deliver data to separated PC like platform.
At work I deal with ColdFire and it is noisy, yet we manage to detach items operating at signals 0.3 uV. It is fine without any DSP filtering.
I did subject Digital Filtering many years ago – theory is good but practically never had need for it. The only digital filtering I have experienced it was the one built-in with ready specialized chips for low frequency power modems.
In did it was amazing to work on circuits which runs on signals impossible to see or measure. The only way to trace signal it was via statistic methods After all years. I believed that secret of good electronic relays on small problem solving what via years builds for You wonderful product so if You want the best - copy the best and improve it.
Regards,

[simonbaker]

I am particularly interested in hearing about DSP techniques for PI detectors. I'm currently building a detector that uses direct sampling with a microcontroller. I've gotten my air depth on a quarter up to 10" with brute force techniques. Going any further requires that I break through the noise floor and perform signal detection properly. It is very clear from my testing that noise is a serious issue in my system. And cancelling it out or being able to see through it is going to mean the difference between a mediocre detector and a great one.

I'm just getting started and I don't even know what to Google to start reading up on it. If anyone has some terms or ideas to get me on the right path I would very much appreciate it. I'm looking specifically for ideas on noise cancellation and improved target detection. I've pulled out my old digital signal processing textbooks but they only cover simple stuff like basic filters. Nothing on target detection or breaking through the noise floor.

Thanks!

I would google "lock-in amplifier" and also "synchronous detector". They are the principle by which most analog VLF metal detectors achieve signal-to-noise. Fairly easy to program compared to more general digital filters. Then maybe you can move on to more wide spectrum techniques.

-SB