Tweet
Fully Software driven Metal Detector
Hi ... When I have time for MD design I work on my PI circuits ... but the other half of my time allocated to MD work is is a fully software driven detector.
I did briefly consider doing this project in an FPGA or a custom OEM DSP board however as has been pointed out in other places before you will never find a more ubiquitous DSP platform than the Pentium based PC form factor.
I am sure this will offend some DSP purists but I have not been able to find a DSP platform with the same number of tools, formfactors ( complete PC in a matchbox ) and suppliers for the same price point.
Basically I have proof of concept "bare bones" operational and it does detect metal. But as it is a proof of concept it needs more work and maybe ideas.
The hardware consists of a PC ( actually an eeePC by ASUS ), a simple 5 watt amplifier ( $5 kit ), a preamp ( another $5 kit ) some audio jacks , cabling and two coils ( one for TX and RX ).
The eeePC is fully solid state and quite compact ( especially when I gutted the main board out of it
)
The software uses the onboard audio chip to generate the TX and sample the RX.
PC -> audio out -> audio amp -> TX coil
RX coil -> preamp -> audio in -> PC
The A/D converters are 16 bit so there is around 96db of dynamic range available .... and its stereo.
The sample rate is 96 KHz ( per channel
) and the audio response is quite good to above 48 KHz.
Thats the hardware .... now for the software.
The OS is linux. ( Fedora 8 flavour ) However this will become embedded linux since the port will involve 0 effort.
There are two main parts ... TX part and RX part.
The software is multithreaded ie TX runs and RX can/do run at same time.
The main TX module is the generate module. ( that all there is because thats all TX does ... however thanks to DSP it can generate abitrary waveforms and frequencies. )
The RX consists of
1. Correlator. The correlator syncronises the rx waveform as this is not some deterministic circuit also when we do the FFT later the phase information would not be accurate if the correlator was not present.
The correlator will also align the signal even if it is low level and buried in noise.
2. FFT block this converts the correlated waveform sample to amplitude and phase across mutliple frequencies.
3. Analysis block looks at historic and current FFT data and decides things like ground effect balance, descrimination, target volume etc etc.
This needs more work.
4. Display block at the moment this displays a phasor diagram ( ie frequency amplitude and phase for 64, 128, 256, 512, 1024 freq channels from near DC to 48 Khz etc etc ) and is useful for seeing which frequencies and phases alter as different targets are presented to the search head.
Everything is written ( and I did not author it all .. used some pre existing public domain code like FFT for example) in C and runs in an X window though there is no dependancy here and the code would reasonably port to other OS ,,, though I would draw the line at windoze
a detector with BSOD would be little bit too much to bear.
The CPU utilisation ( and thus the main power consumption ) runs at about 3 % constant ... after all this essentially audio not video.
IMHO this platform has the capability to develop a new era of amatuer detector experimentation as you dont have to get out a soldering iron everytime you want to change a filter, timing, phasing etc etc. You just tweak the code.
Also because it is PC based it would be very simple to incorporate features like motion sensing ( haptics ) and GPS so you dont waste time going over old ground and provide search optimisation patterns. WIFI could network with other detectors in a local network to coordinate sweep teams and optimal search patterns for team detecting. Mapping information could overlay detection data to provide mineralisation / target profiles. Pretty much blue sky.
So is anyone interested enough to start threading a project on this ...
Thanks for reading this far ... Moodz.
I did briefly consider doing this project in an FPGA or a custom OEM DSP board however as has been pointed out in other places before you will never find a more ubiquitous DSP platform than the Pentium based PC form factor.
I am sure this will offend some DSP purists but I have not been able to find a DSP platform with the same number of tools, formfactors ( complete PC in a matchbox ) and suppliers for the same price point.
Basically I have proof of concept "bare bones" operational and it does detect metal. But as it is a proof of concept it needs more work and maybe ideas.
The hardware consists of a PC ( actually an eeePC by ASUS ), a simple 5 watt amplifier ( $5 kit ), a preamp ( another $5 kit ) some audio jacks , cabling and two coils ( one for TX and RX ).
The eeePC is fully solid state and quite compact ( especially when I gutted the main board out of it
)The software uses the onboard audio chip to generate the TX and sample the RX.
PC -> audio out -> audio amp -> TX coil
RX coil -> preamp -> audio in -> PC
The A/D converters are 16 bit so there is around 96db of dynamic range available .... and its stereo.
The sample rate is 96 KHz ( per channel
) and the audio response is quite good to above 48 KHz.Thats the hardware .... now for the software.
The OS is linux. ( Fedora 8 flavour ) However this will become embedded linux since the port will involve 0 effort.
There are two main parts ... TX part and RX part.
The software is multithreaded ie TX runs and RX can/do run at same time.
The main TX module is the generate module. ( that all there is because thats all TX does ... however thanks to DSP it can generate abitrary waveforms and frequencies. )
The RX consists of
1. Correlator. The correlator syncronises the rx waveform as this is not some deterministic circuit also when we do the FFT later the phase information would not be accurate if the correlator was not present.
The correlator will also align the signal even if it is low level and buried in noise.
2. FFT block this converts the correlated waveform sample to amplitude and phase across mutliple frequencies.
3. Analysis block looks at historic and current FFT data and decides things like ground effect balance, descrimination, target volume etc etc.
This needs more work.
4. Display block at the moment this displays a phasor diagram ( ie frequency amplitude and phase for 64, 128, 256, 512, 1024 freq channels from near DC to 48 Khz etc etc ) and is useful for seeing which frequencies and phases alter as different targets are presented to the search head.
Everything is written ( and I did not author it all .. used some pre existing public domain code like FFT for example) in C and runs in an X window though there is no dependancy here and the code would reasonably port to other OS ,,, though I would draw the line at windoze
a detector with BSOD would be little bit too much to bear.The CPU utilisation ( and thus the main power consumption ) runs at about 3 % constant ... after all this essentially audio not video.
IMHO this platform has the capability to develop a new era of amatuer detector experimentation as you dont have to get out a soldering iron everytime you want to change a filter, timing, phasing etc etc. You just tweak the code.
Also because it is PC based it would be very simple to incorporate features like motion sensing ( haptics ) and GPS so you dont waste time going over old ground and provide search optimisation patterns. WIFI could network with other detectors in a local network to coordinate sweep teams and optimal search patterns for team detecting. Mapping information could overlay detection data to provide mineralisation / target profiles. Pretty much blue sky.
So is anyone interested enough to start threading a project on this ...
Thanks for reading this far ... Moodz.
Comment