Plans for my own PI

[TheWizard]

you can remove C5...

[sled]

how fast should the ADC be and which resolution is appropriate?

Perhaps you could propose a ADC chip

[bugwhiskers]

Hi sled,

If you want to use the curve info to disciminate various metals (more than just Fe) then you will need 2 samples in the knee of the curve. If the samples are 4uS apart you will need an AtoD conversion speed of 250 kilo samples per second. The ATMEL chips are about 150KSPS at best regardless of main clock speed. You can either opt for a chip with a faster AtoD or use the multiplexed sample and hold scheme as used in the UPIM (Geotech\Projects).

As for AtoD resolution, the more the better as higher resolution is better than amplification.

regards
bugwhiskers

[sled]

Hi,

are your sure with 250KSP/S ? Look at my first posting, there are some diagrams of the voltage curve.... they were taken with an oscilloscope with 5MSP/S....

I saw that there are some cheap (~20$) ADC with 12bit and 3 MSPS... The problem is: they come in a SMD package.... But I heard that there are "adapter" pcbs for SMDs...

If I use a SMD ADC, I'd prefer this one:

http://www.farnell.com/datasheets/73196.pdf

3 MSPS,10-/12-Bit
ADCs in 8-Lead TSOT

AD7273/AD7274
Throughput rate: 3 MSPS


Specified for VDD of 2.35 V to 3.6 V
Power consumption
11.4 mW at 3 MSPS with 3 V supplies
Wide input bandwidth
70 dB SNR at 1 MHz input frequency
Flexible power/serial clock speed management
No pipeline delays
High speed serial interface
SPI®-/QSPI™-/MICROWIRE™-/DSP-compatible
Temperature range: −40°C to +125°C
Power-down mode: 0.1 μA typ
8-lead TSOT package
8-lead MSOP package

My question is now if the uC is able to read the output of the ADC fast enough.... It has 20MHz, but I don't know how fast it can read an process the serial interface.

Thanks

best regards

sled

[bugwhiskers]

Hi sled,

The AtoD may be fast but then you have to get the data into the micro and stored to a location in RAM. That's where the multiplxed caps help. You store the volatge at the different points into caps and then process them at leisure bewteen TX pulses.

With an earlier prototype running at 20MHz I was able to get snapshots of the decay curve at 100 nS intervals which is equivalent to an internal 10MSPS AtoD.


Regarding the images of the discrimination you posted earlier, silver, copper and steel look nearly identical. My results are totally different with a clear distinction between the 3 metals. It is bizarre that they show silver and steel as being similar when one is ferrous and the other isn't and one is the best conductor and the other a poor conductor.

regards
bugwhiskers

[TheWizard]

100nS.... not need that much fast...
I work with 12 bits ADC 500KS (sample 2uSec) at 3.3 VREF and DSP@40MIPS