field test unit no 001 "model T"

[Davor]

OK, but practically speaking, you do not need MHz bandwidths unless you are not into discriminating sea water salinity Every practical target is much slower than that.

[chudster]

well ... you dont need to sample "all the time" and the pic32 will do DMA if you poke it a bit ....

Sorry always trying to solve stuff too far ahead. I realise you haven't done it yet and of course this is one big experiment as per parts of the patent act we were looking at
The UNO32 doesn't have DMA (on the SPI) not that this can't be managed still.


As to the comment from Davor, obvously the coil you attach will self resonate. 2MSPS gives 1MHz via nyquist but in practice you want your frequency response to start to cutout way before 1MHz. I have been figuring on 250KHz or so as being the needed frequency response of (small gold) targets. More is always better but is a frequency response of 250KHz OK or do we need 1MHz? 1MHz probably encompasses the self resonance of the coil. Is that good? Maybe we don't know so the experimental rig has more than we need?

[Aziz]

BTW guys,

everyone interested in direct sampling and decoding can also use the sound card + laptop with appropriate detector controller. It works upto 48 kHz bandwidth. The principles remain the same. On the fast ADC system, you just have more bandwidth for more signal integrity.

The sound card solution could be seen as the poor man's solution.
Aziz

[Davor]

With a coil on the very frontend of the system - yeah, it is bandwidth limited.

[moodz]

The FPGA must use 'store and forward' as the SPI, even with a 40MHz clock would have trouble with real time at 2MSPS? I presume there is no antialiasing filter with the direct sampling? Do you see that as a problem or is the bandwidth limited by other means to prevent aliasing problems?

Chudster

well ... you dont need to sample "all the time" and the pic32 will do DMA if you poke it a bit ....

[chudster]

This one goes straight into the PC ( thats how I get the screenshots ) ... it will easily interface to the PIC32 . eg ..... ADC -> FPGA -> PIC32 .. the comms to the PIC32 is SPI.

The FPGA must use 'store and forward' as the SPI, even with a 40MHz clock would have trouble with real time at 2MSPS? I presume there is no antialiasing filter with the direct sampling? Do you see that as a problem or is the bandwidth limited by other means to prevent aliasing problems?

Chudster

[moodz]

Do you (or how do you) get it into the PIC32?

This one goes straight into the PC ( thats how I get the screenshots ) ... it will easily interface to the PIC32 . eg ..... ADC -> FPGA -> PIC32 .. the comms to the PIC32 is SPI.

[chudster]

$50 FPGA

Do you (or how do you) get it into the PIC32?

[Aziz]

$50 FPGA

F.. Hell!!! Don't give the market leader a fright!!!!!
Expect being sued now....

MUAHAHAHAHAHAHAHAHAHAHAHA
*LOL*

[moodz]

$50 FPGA

[moodz]

As a test of sensitivity we measure the noise difference between a 1K resistor and a 100 ohm resistor on the target channel ..... very nice

[Aziz]

Aziz ... dont forget your projects ... I am working on a "front end" for you to play with ... it will have a multi channel 24 bit adc / dac with no bandwith limiting and access to DSP functions then piping to laptop / whatever through USB.

Yeah, fantastic!!!

I like the USB interfacing feature... Tablet PC is coming...yeah!
Aziz

[Aziz]

Hey Moodz,

do you use a DMA interface for the ADC? It would make heavily sense, rather than bit-banging the ADC interface.
Aziz

[moodz]

BRAVO!!!! I like it.
Forget my projects. The QED2 will be a killer-app. That's a game changer.
(And I can give you a lot of high-tech info as well (WBGB, noise elimination, etc..))
Aziz

Aziz ... dont forget your projects ... I am working on a "front end" for you to play with ... it will have a multi channel 24 bit adc / dac with no bandwith limiting and access to DSP functions then piping to laptop / whatever through USB.

[moodz]

This is direct sampled at 2 msps ... I can sample up to 20 msps but the effective bits will drop to 16 and noise floor to around -80 db ...Lots of testing to do .... LOL .... but now I dont have to tweak hardware .... only software. I am expecting to pick up another 6 bits of effective resolution on top of the 24 bit which together with the "new" front end and adaptive ground balance is going to be a very good system. The overall "real" bit resolution is slightly north of 26 bits and a system gain of around 10,000,000 over a 10 Hz bandwidth. ( ie approx 1000 times better than a 16 bit system ... taking into consideration system noise and processing gain).