Perfect PI Frontend

[jon-ecm]

Hey all! It's been a while since I've posted on here but I sure have learned allot from reading all of your posts!

As some of you know I'm working on designing and perfecting XPI a high end kit metal detector.
The unit is made up of 3 boards. The mainboard, coil drive board, and the analog frontend board.

I've tested 4 different analog frontend boards before coming to the (rather simple) design shown below

[Input from Coil] -> [Both inputs go through 5k resistors] -> [are clipped to a low voltage using bi-directional diodes] -> [One leg goes to ground and the other through a 100 Ohm Reistor, we'll call this the signal] -> [Signal goes to analog SPDT, the output from this switch is normally grounded to keep the op amps from saturating and only opened when the decay has reach a point where we can begin to sample] -> [Signal goes to Non-Inverting high speed op amp with gain of 50] -> [Signal goes to high speed Non Inverting op amp with gain of 0 (buffer)] ->[Signal goes through 100 Ohm resistor to reduce ringing] -> [Signal is fed to ADC (16 BIT 1MSPS ADS8330)]

What are your thought's on this arrangement? Especially the clipping part? I've tried other arrangements that completely isolate the frontend and do away with those (possibly noisy) 10 k resistors but it didn't seem to make that much of a difference.
What do you think the perfect PI frontend should look like?


Right now I've been using the detector with the current arrangement and it will go 5.5 times as deep as my infinium on fine sub gram gold nuggets and seems slightly better than a ML 4500.

you can see some info about XPI here. My camera gave up the blue smoke and as soon as I can get it replaced i'll upload pics of the unit, don't forget to read the comments as they provide some more up to date info

Here from you soon!
Jon

[Davor]

It seem a tad too complicated for a monocoil rig.

SPTD at front end seem a right way to go, but 5k on input ... will be better to avoid if possible. I'm not able to fully grasp a buffer and a 100 ohm resistor, but it's your design, not mine.

The rest of it is in signal processing and it is op to you.

I'd say it has a great potential. Sampling at front end is a right thing to do considering signal levels and their wild nature. The most of the nowadays PI designs suffer from too much gain at front end, and slow recovery after PI. This design must be incredibly faster than that.

You could as well put some schematic here so that we can play with. Don't be shy, you can't be robbed your ideas if you publish them. As an author you are benefiting from the copyright laws as soon as you publish something - patents are for big companies and sissies.

[6666]

Good work Jon, any chance of a rough schematic ? thanks
6666

[hdphilip]

jon

i like the idea, it seems everyone throws their (high gain) first stage op amp right after the clamping diodes, but in fact maybe going into a 4066 would be better, then the op-amp would only see the sampled waveform. i've been experimenting with moodz's idea, which sorta does that,you can take a look at it on
http://www.geotech1.com/forums/showp...&postcount=133

But i think your idea, might be spot on!

Philip

[jon-ecm]

Here's a schematic of the latest revision. It slightly differs from the one currently in use and described above but otherwise is very similar. Also the Op amp shown in the schematic (MAX4432) is not the correct one. The one in use is the MAX4434 which is the single supply version. My first design used the 4432 with a positive and negative rail but ive been trying to simplify things as much as possible.


@hdphilip I've kept up with mood'z ideas a bit, but the link you shared was totally new to me. I really like the way he has done the blocking fet's. His ideas are always new and innovative. My plan is to open source the project after it's finished and I have a solid spec. After it's open-sourced it would be my hope that people would develop a number of different daughter boards compatible with the main board, then perhaps we would see lots of analog front-end boards made by a number of different people. Currently the board is programmable by a USB port so flashing new firmwares can be easily done.

@Davor The 100 Ohm resistor on the input was to limit the current to the blocking switch during the negative tx pulse. Because the tx will never be more than 16v I decided that it wasn't needed and just soldered over it. It is removed in the latest revision (see schematic) The other 100 ohm resistor just before the op amp is to reduce ringing/overshoot caused by the input capacitance of the ADC as well as reducing the voltage slightly when combined with the 1k load. The VRef is 4.096 but the op amps are driven by 5 volts so it could potentially damage the adc so that's why it's there.
I've also used this frontend with double-O coils in a noise reduction configuration with good results. I'm still learning about double-D coils so it will be a little while before I have full support for them.

Jon

[Davor]

I see. There is little room for improvement regarding the front end. It is unbalanced, but it may become balanced with not too complicated design change - you have analog switches there to spare.

Have in mind that the on resistance of analog switches is influenced somewhat with Vcc. Perfect impedance match is achieved at geometric mean between on and off resistances. With 5k in series I think there is no problem here.

All in all, nice one. It is simpler than you described it using words