Uni-PI - Universal Pulse Induction Metal Detector board for microcontrollers like Ardiuno, PIC, AVR.

[Waikiki_Sweep]

Uni-PI - Universal Pulse Induction Metal Detector board for microcontrollers like Ardiuno, PIC, AVR or whatsoever.

If you have nothing to do at that boring snowy, rainy ... or desert hot day you can tell something about the idea
to create universal analog PI detector board to connect it to any microcontroller like Ardiuno, PIC, AVR or any you like.

How to connect it for example to ATTINY85 you can see on schematic:











Construction is more than classic. Practically it is Surf PI reincarnation with microcontroller and some simplification.

TX Coil drive with P-channel MOSFET receiving logic signal from 5V microcontroller of your choice.
Capacitive coupling to preamplifier build on very low noise single supply OpAmp ISL28291.
Gain from 1 up to 1:2000 can be set with trimmer depending on schematics, coil ans software conditions.
Additional sensitivity pot can be used for fast adjustment by the way.

Then two SPST (NO) switches feeding signal to differential integrator with gain 1:475
to take positive and negative samples according individual algorithm you are preferring to play.

After Low Pass RC filter signal goes out to separated PSB to microcontroller ADC input.

To connect Uni PI board to your favorite microcontroller there is a six wires bus:

TX - Transmit impulse input.
S2 - Ground sample impulse input.
GND - Ground.
S1 - Target sample impulse input.
+5 - +5V power to microcontroller. Careful!!! Some of MC using different power voltages like 3.3V or even 1.8V. Do not connect if not sure.
ADC - Output signal to microcontroller ADC.

Classic timing diagram that widely used you can program:

All inputs are LOW all the time except:
1. 50 microsecond TX HIGH
2. 15-25 microsecond delay (depending on coil)
3. 50 microsecond S1 HIGH. Reading target signal.
4. 250 microsecond delay.
5. 50 microsecond S2 HIGH. Subtracting ground signal.
6. Reading ADC signal by microcontroller ADC, averaging it and comparing with saved overages from previous periods to provide indication of target. Or any algorithm you like.

That cycle usually has 600 repetitions per second. So you need to create an interrupt with one of the timers in your microcontroller to repeat cycle in set time periods.

Sure because of flexibility of microcontroller you can experimenting with more complex timings and frequencies including those for advanced ground balancing or target identification.

Coil is for SurPI (240-300 uH. 1-2 Ohm.) For different coils and timing sequences please be sure you have adjusted dumping resistor 390 Ohm 2W to proper value.

Good Luck!!!

[Waikiki_Sweep]

Because new switches: ISL84541, MAX323, MAX4541 are too expensive ($1.86-$3.50 on Digikey) probably it is good idea to use good old 50 cent CD4066BE.
Sure six more useless holes to drill but it gives $1.5-2.5 economy. So CD4066BE version you can see below:









Size of PCB is 1.675 x 1.9 inches or 42.5 x 48.3 millimeters.

[Beenthereuk]

Were you to use an ENABLE pin type op-amp like the MCP6023, it should be possible to simply switch it on and off to integrate the signal, thereby removing the need for a switch IC. I never use them, personally. There's always a neater way.

You probably also already know my views on the op-amp linear amplifier.

[Waikiki_Sweep]

Were you to use an ENABLE pin type op-amp like the MCP6023, it should be possible to simply switch it on and off to integrate the signal, thereby removing the need for a switch IC. I never use them, personally. There's always a neater way.

You probably also already know my views on the op-amp linear amplifier.

Thanks for asking

Unfortunately MCP6023 has to high noise at 10Hz. - 65nV/√Hz [mcp_6024_21685d.pdf - FIGURE 2-8: Input Noise Voltage Density vs. Frequency.]
ISL28291 has 2.5nV/√Hz for 10Hz.

Also it requires to use two switches because most advanced sampling algorithms need differential integrator and using series of positive and negative samples and both channel switches have to be matching by ON resistance.

I have scratched my head for a long time trying to find some simple alternative variant
like to find a way to use multiplexers in microcontrollers (unreal dreams)
or that enabling inputs in opamps. (probable may be used if it has tri-state - high resistance output stage to prevent integrator from leaking. But anyway I need two channels.)
Or use P-channel transistors like - J176 (somehow doable if two transistors will have same parameters)

I have no good solution. That CD4066 is bulky but cheap enough. I hope to use 8 pins switches like MAX323 but they are hell expensive.

I am sure you can easily cook that schematic with few transistors and hope to see it soon as you going to use microcontroller too.

[Beenthereuk]

I will need to simulate one of these advanced detectors to understand the topology better. I'll start with your schematic first when I get a chance.
But you can start by dropping the 47k bias resistors to 4.7k for now and make the input capacitor the largest polystyrene capacitor you can find.
See this for info on capacitor related noise: http://en.wikipedia.org/wiki/Johnson...3Nyquist_noise

[Waikiki_Sweep]

I will need to simulate one of these advanced detectors to understand the topology better. I'll start with your schematic first when I get a chance.
But you can start by dropping the 47k bias resistors to 4.7k for now and make the input capacitor the largest polystyrene capacitor you can find.
See this for info on capacitor related noise: http://en.wikipedia.org/wiki/Johnson...3Nyquist_noise

Thank you. Earlier I have used 1K resistors for bias and it was working good. I have switched to 47K to save power but I see it is so important for the noise so even 470 Ohm should be OK.
That schematic is not much advanced, it is just classic XX years old basic setup of PI detector. For advanced ideas you better read QED schematic http://www.geotech1.com/forums/attac...5&d=1341282613 for it sensitivity and ground balance or Chance http://www.geotech1.com/forums/attac...2&d=1378716722 for ability to recognize metals with using Pulse Induction and monocoil setup.

[Waikiki_Sweep]

Because ISL28291 has "ground sensing" opportunity I think is is even better solution to shift bias voltage on "+" input of preamp
closer to ground (approx +0.5V) and reducing resistance to 100 Ohm from "+" input of preamp to the ground
then use 1K from "+" input to the +5V rail. So 100 Ohm of "+" input impedance will keep noise as low as possible.

[Beenthereuk]

You can use a biased diode to **** it close to the rail. Using a low value resistor will probably give you less noise than a pair of resistors there anyway.

I cannot find a good model for the ISL28291 or I would simulate the noise for it but I have used an LT1028 instead in this simulation and compared it with a single transistor stage with dynamic load (pnp current mirror).

The LT2018 is a £12 chip (Farnell, UK) that has 10 Hz noise = 1nV / RootHz.
The transistor amplifier is the cheap and common BC337 (with the low base resistance).
I could only squeeze 800x gain out of the transistor (can do more but need more time to play) so I set the same gain and BW on the op-amp and compared noise plots.
Considering the cost difference, the results are amazing. See below:

[Waikiki_Sweep]

You can use a biased diode to **** it close to the rail. Using a low value resistor will probably give you less noise than a pair of resistors there anyway.

I cannot find a good model for the ISL28291 or I would simulate the noise for it but I have used an LT1028 instead in this simulation and compared it with a single transistor stage with dynamic load (pnp current mirror).

The LT2018 is a £12 chip (Farnell, UK) that has 10 Hz noise = 1nV / RootHz.
The transistor amplifier is the cheap and common BC337 (with the low base resistance).
I could only squeeze 800x gain out of the transistor (can do more but need more time to play) so I set the same gain and BW on the op-amp and compared noise plots.
Considering the cost difference, the results are amazing. See below:

[ATTACH]28173[/ATTACH]
[ATTACH]28174[/ATTACH]
[ATTACH]28175[/ATTACH]
[ATTACH]28176[/ATTACH]

Results are really amazing!!!
They have to be implemented. I will try it on one of my PCB's.

Why not you started your own business to sell DIP-8 adapters with few SMD transistors on it that everyone can use as preamp in his schematic like that expensive epoxied opamp IC's corporations selling us?

[Beenthereuk]

I know, the price of some of these chips is mad. They are designed to meet a series of criteria that are not all applicable in most designs and so we pay for over the top functionality that is not really required. E.g. in this case, the input need not be differential and the output need not be very low impedance. But I dont have the time to make things for sale as my job keeps me too busy.

If you build it, use a 1 MOhm resistor to set the bias as it is beta dependent in that circuit. I can improve this a lot. Need to get the gain to at least 2000x so will add an extra stage that can also buffer the high ouput impedance from the dynamic load. It should also be possible to bias it differently and do away with the large base bias resistor - I need time to play.

[green]

Define coil specs. 300 uh, 100 pf for a fast coil and some cable. Calculates 918 khz resonance. Rd = pi L Fr = 865 ohms. Amplifier gain =160. To damp in circuit the Rd needed to be a little less than 865 ohms. Specs based on what I've read in the forum. Can be changed. The goal was to damp as fast as possible. Comes close to 3/Fr, 3.3 usec.

[Waikiki_Sweep]

Interesting schematic, what R1 (10M) placed for?

I see two amplification stages. I think before to go to next amplification stage may be it is good idea to do some noise filtering.
Because noise can be higher than signal so analog schematics working better this way: filtering..amplification..filtering..amplification ..filtering..amplification
Before decoder (reading samples goes to integrator) we can cut low and high frequencies before every amplification stage let say our signal is in 300-1000Hz range if TX works with 600 pps.
After decoder it is good idea to keep 1-10Hz signal (coil moving over targets) and cut out everything else before every amplification.
Am I right? Correct me.

100pf looks too fast for me. Even 4ft. of good coax cable may have 100pf.

Anyway coils/cables are different so minimum dumping required probably may be somewhere near 390 Ohm and maximum up to 1K.

Dumping 390 can be placed without additional checks probably but higher value which is better for sure requires to look on oscilloscope.

Once I have tried 900 Ohm on overage, not the best coil. Sure it was underdumped but practically working in fields somehow.

[green]

The goal was to damp a coil as fast as I could. The circuit will damp close to zero volts in 3/Fr. The circuits I had seen used a lower value than 10M to decay the volts across the fet. When using LT spice a lower value caused the amplifier volts to drift, maybe doesn't matter. I haven't built a normal non inverting amplifier except with spice to compare. It's faster with spice and it does work in hardware. A 300 uh coil with 200 pf calculates 650 khz resonance and should damp to zero in less than 5 usec with a 600 ohm resistor. I've seen other threads on fast damping and fast coils. A question is, how fast does it need to be. I think it would help to define a coil spec making it easier to compare circuits.

[Beenthereuk]

I dont think you've noticed but the top of your R7 resistor is connected to both poisitve and ground, shorting the supply in that circuit - OK I see you are biaising that branch negatively. But R1 (10 MEG) will weaken the signal excessively and introduce lots of noise.

[Waikiki_Sweep]

I dont think you've noticed but the top of your R7 resistor is connected to both poisitve and ground, shorting the supply in that circuit - OK I see you are biaising that branch negatively. But R1 (10 MEG) will weaken the signal excessively and introduce lots of noise.

Top of R7 connected to ground and yes, to positive supply too. It is just another XX years old setup of classic PI detector where several separated power sources are used. Like +12V (or -12V) and +5V, -5V.
Usually it requires some noisy voltage converter like ICL7660. Most of PI detectors still using such setup.
Problem is that flyback voltage goes over supply range and you have to extend power supply with additional voltage converter to keep signal between your opamp "+" and "-" power lines.
I hoping that capacitive coupling can be a solution but it not used widely yet.

Regarding R1 (10 MEG) I have no idea why it added to schematic too. Probably schematic is functional without R7. May be author knows it better.