Direct conversion PI front end

[SadE]

Hi,

I'm new to metal detectors but I want to build one from scratch. I read theses last days a lot of things on PI detectors but I have some technical questions !
Do you think it's possible to avoid some noisy stage preamp/amp and use a high resolution DAC for direct conversion ? For example ADS1675 from TI allows to get ~16 bits noise free at 1M samples /s
and about +17 bits at 125k samples / s . It represents a free noise resolution at the DAC level of 10uV.
Morever is there some benefit to use another channel to sample the high voltage part (and not the < 0.7v one) to get additional data from the complete pulse ?

[liudengyuan]

Come on, dear friends! ! !

[Davor]

You may consider noise level at about 1nV/sqrt(Hz) as the current state of the art. With 100kHz bandwidth, it is ~300nV noise, and to get that level to the LSB level at 10uV, you need some 33x amplification. Find a decent ADC with LSB below 1uV and you are on.

The PI voltage decay comprises of the coil exponential flyback decay governed by Lcoil/Rdamp~1us, target exponential decay, and ground viscous response. Each superimposed on the others. Because unfortunately you can't dismiss avalanche noise and other Tx variations that pass on to early stages of flyback decay, taking too early samples is an exercise in futility. You may expect decent behaviour starting at about 5us or so.

Taking a whole PI pulse ... no. It is pointless. You may get the reactive response at flyback top and thus obtain discrimination information, but period between that and a target sample is dominated by a coil's own response, which is not unknown, and thus not useful.

So if you devise some clever way of sampling a flyback top, you'll be able to make a truly discriminating PI machine.

[SadE]

- You're talking about 100khz of BP , is it enough to extract all necessary data ?
- I read another topic talking about a new(?) driver for the TX coil FET in order to improve greatly the 'slew rate' . Is there real tests results for such a driver ?
Thanks Davor for the precisions, I guess I have to study more deeply the theory behind PI ;-) .

[Davor]

100kHz is the analogue bandwidth of the better preamps in PI, so yes, it is enough.
Good luck

[Teleno]

- You're talking about 100khz of BP , is it enough to extract all necessary data ?
- I read another topic talking about a new(?) driver for the TX coil FET in order to improve greatly the 'slew rate' . Is there real tests results for such a driver ?
Thanks Davor for the precisions, I guess I have to study more deeply the theory behind PI ;-) .

You may want to try this desgn: http://www.geotech1.com/forums/showt...803#post210803

Adjust R3 for your damping requirements, the current in Q1 (via VB1) for the lowest noise fgure, R6 to get Vcc/2 at vout, Vb2 about 2/3 Vcc.

[SadE]

This one looks like a serious contender for the ADC :
http://www.linear.com/product/LTC2380-24

• Guaranteed 24-Bits No Missing Codes
• ±0.5ppm INL (Typ)
• Integrated Digital Filter with Real-Time Averaging
• Low Power: 28mW at 2Msps
• 100dB SNR (Typ) at 1.5Msps
• 145dB Dynamic Range (Typ) at 30.5sps
• –117dB THD (Typ) at f_IN = 2kHz
• 50Hz/60Hz Rejection
• Digital Gain Compression (DGC)
• Guaranteed Operation to 85°C
• Single 2.5V Supply
• Fully Differential Input Range Up to ±5V
• 1.8V to 5V SPI-Compatible Serial I/O with Daisy- Chain Mode
• 16-Lead MSOP and 4mm × 3mm DFN Packages

The digital filter feature is great . The price is not : 50$ for 1 piece , gasp

[Davor]

With that beast you may try direct sampling for sure. Your purse will suffer.

[nickel_n]

This one looks like a serious contender for the ADC :
http://www.linear.com/product/LTC2380-24

• Guaranteed 24-Bits No Missing Codes
• ±0.5ppm INL (Typ)
• Integrated Digital Filter with Real-Time Averaging
• Low Power: 28mW at 2Msps
• 100dB SNR (Typ) at 1.5Msps
• 145dB Dynamic Range (Typ) at 30.5sps
• –117dB THD (Typ) at f_IN = 2kHz
• 50Hz/60Hz Rejection
• Digital Gain Compression (DGC)
• Guaranteed Operation to 85°C
• Single 2.5V Supply
• Fully Differential Input Range Up to ±5V
• 1.8V to 5V SPI-Compatible Serial I/O with Daisy- Chain Mode
• 16-Lead MSOP and 4mm × 3mm DFN Packages

The digital filter feature is great . The price is not : 50$ for 1 piece , gasp

$50.00 perhaps it sounds allot, but with today’s scrap gold price ($30.32 per gram for 14k) you only need to find a 2 gram 14k gold ring to pay for it.
Most rings weighing minimum 2.5-3 grams.

[Max]

Hi
I do agree it's not cheap for just one piece.. but consider all the pain of making a traditional frontend with all required electronics and sampling , layout and time it requires and you're there... don't know in your case but often people time is the major constraint not the parts cost.
I think it's a good idea and at the end you can reuse the part in other projects too, so it's kind of an investment.

btw it seems even too much for what's required, so you can maybe choose something less accurate and get also good results for half the price I think.

Regards
Max

[sled]

LT recently released a new ADC, which is cheaper (17$ for a single unit on digikey)


LTC2512-24 - 24-Bit Over-Sampling ADC with Configurable Flat Passband Digital Filter

Features

• ±1ppm INL (Typ)
• 117dB Dynamic Range (Typ) at 50ksps
• 108dB Dynamic Range (Typ) at 400ksps
• Guaranteed 24-Bits No Missing Codes
• Configurable Digital Filter with Synchronization
• Relaxed Anti-Aliasing Filter Requirements
• Dual Output 24-Bit SAR ADC
• 24-Bit Digitally Filtered Low Noise Output
• 14-Bit Differential + 8-Bit Common Mode
• No Latency Output
• Wide Input Common-Mode Range
• Guaranteed Operation to 85°C
• 1.8V to 5V SPI-Compatible Serial I/O
• Low Power: 30mW at 1.6Msps
• 24-Lead 7mm × 4mm DFN Package

Datasheet: http://www.linear.com/docs/55981