field test unit no 001 "model T"

[Aziz]

Dont forget the bacteria$hit nuggets with this card .... item no 220896231564 on ebay.

384 Khz sampling at 32 bits theres heaps of others too and the price is not too much for a high end card ( around $150 )

moodz.

Thanks Paul,

haven't seen it yet. That would make UD of the UD detectors.
*LOL*
ASIO interface is nice and offers a very low latency interface.
Aziz

[Aziz]

Dont forget the bacteria$hit nuggets with this card .... item no 220896231564 on ebay.

384 Khz sampling at 32 bits theres heaps of others too and the price is not too much for a high end card ( around $150 )

moodz.

It's a DAC only!!! I need ADC too.
Sad.
Aziz

[hdphilip]

moodz,

so with a sampling rate of 192 khz, would equal a sample every 5.2 us, so our sampling window would have to be 26 us wide to sample 5 times, so we'll still need some sort of integrator to help out , would this be correct ?

which adc chip for the 192 Khz?


Philip

[moodz]

It's a DAC only!!! I need ADC too.
Sad.
Aziz

bugga ...back to 24 bits again. LOL

[moodz]

moodz,

so with a sampling rate of 192 khz, would equal a sample every 5.2 us, so our sampling window would have to be 26 us wide to sample 5 times, so we'll still need some sort of integrator to help out , would this be correct ?

which adc chip for the 192 Khz?


Philip

This is correct ... however ... integration is a time domain method ... it can be done post ADC ... but as Aziz said the frequency domain is where the action happens. That is not to say that integration is not a useable method ... it may not be optimal.

moodz.

[Aziz]

This is correct ... however ... integration is a time domain method ... it can be done post ADC ... but as Aziz said the frequency domain is where the action happens. That is not to say that integration is not a useable method ... it may not be optimal.

moodz.

In the frequency domain (FD) we integrate digitally over the FD response. Integration is an essential part of "collecting" all the spectral energy of the response. Particularly, if we have a wide band response.

If we do all the digital demodulation in software, we don't need analog integrators anymore.

Aziz

[Aziz]

Hi Moodz,

trying to figure out, what SAT was or meant to.

What is SAT???
Aziz

[moodz]

In the frequency domain (FD) we integrate digitally over the FD response. Integration is an essential part of "collecting" all the spectral energy of the response. Particularly, if we have a wide band response.

If we do all the digital demodulation in software, we don't need analog integrators anymore.

Aziz

I don't integrate in the dsp ... I calculate the peak and average sum of magnitudes of the filtered baseband spectrum ... The demod low pass has already done the "integration" to remove non target modulation energy. Works the same as a radio ...where you don't integrate either. Integration is a method of filtering the baseband ...ie target signal ...but if you are already at baseband then integration is not required.
Moodz

[moodz]

Hi Moodz,

trying to figure out, what SAT was or meant to.

What is SAT???
Aziz

SAT is Self Adjusting Threshold ......my implementation was a digital version of the differential integrator. It consists of a fast and slow integration of the incoming ADC values followed by a simple subtraction. The output converges to zero so dc offsets are removed .... This means if you don't move the detector you will only be integrating the noise from the front end.However this method integrates all the wideband energy it does not resolve other factors that are absolutely critical like ground response.
Moodz

[moodz]

I don't integrate in the dsp ... I calculate the peak and average sum of magnitudes of the filtered baseband spectrum ... The demod low pass has already done the "integration" to remove non target modulation energy. Works the same as a radio ...where you don't integrate either. Integration is a method of filtering the baseband ...ie target signal ...but if you are already at baseband then integration is not required.
Moodz

Actually the sum of magnitudes on the filtered spectrum could be classed as a form of integration ...up to your own personal interpretation

Moodz

[Aziz]

SAT is Self Adjusting Threshold ......my implementation was a digital version of the differential integrator. It consists of a fast and slow integration of the incoming ADC values followed by a simple subtraction. The output converges to zero so dc offsets are removed .... This means if you don't move the detector you will only be integrating the noise from the front end.However this method integrates all the wideband energy it does not resolve other factors that are absolutely critical like ground response.
Moodz

Thanks. I know now, what you are meaning with SAT.
Aziz

[moodz]

...revisit the blocking FETS

I said I would revisit the blocking FETs and the problem with blowing them up all the time is solved ....
... I will give it a good testing prior to publishing anything ...

... below is a pic of the blocking FETS in action ( scuse my camera shake ) .... neatly snipping the top and bottom off a 500 volt peak flyback with perfect tracking of the base line signal voltage. The coil used here is an 18 inch differential coil. No amplfication .. we are sampling directly off the output of the blocking FETS. The dynamic range and sensitivity is an order better than with crowbar diode blocking The flyback is neatly blocked to a 8 volt peak to peak signal ... so it wont blow the input to the sampling demods .... . the neato thing is that the blocking FETs draw almost no current when blocking .... unlike diode crowbars.

The white trace is the +ve flyback ... the yellow trace is the -ve flyback ... the green trace is the TX mosfet gate drive signal.

moodz

[hdphilip]

great news moodz!

I'll be looking forward to giving it a go!

Philip

[smokingdiode]

The dsPIC30F4012 code worked first time! Pulse waveform easily controllable with the rotary switch setup. Nice. Just waiting for matched FETs to arrive before rigging the differential stage. Interesting project Moodz, can't wait to see how the benchmarks turn out on coins and rings. Does anyone have any preliminary benchmarks (eg. an Australian $2 coin at 19cm in air).

Planning to eventually identify a set of parameters (eg pulse width, gain parameters, etc) and algorithmic control variables and put them in a genetic algorithm / neural network optimizer. Hopefully I'll be able to share some good parameter sets that will be tweaked for specific target sets eg a coin set, a gold set, a ring set etc. The fitness function will comprise of a complete system simulation involving moving parts! Two stepper motors will be connected to large (1 meter) wooden wheels that rotate parallel to each other and are spinning in the same direction at slightly different speeds. At equal spaces on the perimeter of each wheel will be tied a combination of good and junk targets (eg bottle caps, rings, coins, tabs). The coil will be positioned a few inches away from the perimeter of the wheels so that as the wheel spins, new targets will automatically applied (come in range) of the coil. Note, because there will be two wheels different combinations of targets will be presented to a coil to simulate the situation where a good target is coincident on a junk target as well as an isolated good and an isolated junk target. A "fitness" evaluation of a sample parameter set will be an assessment of the ability of the system to positively identify good targets and to reject interference and junk targets. Of course, I'll share all code. Will probably have to run the code on the PC and commicate down the RS232 line as genetic algorithms consume considerable RAM.

Good collaboration.

Simon

[hdphilip]

hello simon

good job!

just wondering, did you have to raise the cpu vcc by .2 volt by any chance?


philip