Moodz' Awesome Gold Pulse Induction Version 3 - MAGPI V3 Project

[moodz]

I thought it was the TX pulsewidth that changed coil current

From FFT tests I've undertaken, reducing pulsewidth causes the transmitted amplitude of each harmonic to be approximately the same but at a much reduced overall power output of course.

That is right of course but I was referring a symmetrical pulse ( equal time on / off ) so the coil current ( unless it saturates ) is proportional to TX frequency.

For interest I just dialled up a TX on time of 100us and off time of 900 us ( ie 1khz repetition ) and the MAGPI handled it OK however I did not adust the sample timings so sensitivity to sub microsecond targets disappeared.

[Gunghouk]

I thought it was the TX pulsewidth that changed coil current

From FFT tests I've undertaken, reducing pulsewidth causes the transmitted amplitude of each harmonic to approach similar levels but at a much reduced overall power output of course.

I think I went down to 20uS pulse width at 1kHz at which point the power was probably unusable though each harmonic upto about 15th were of similar amplitude.

Wouldn't reducing frequency allow greater gain from your high spec front end amp?

[moodz]

Thanks for that. I thought it was something to do with ground penetration but maybe that was VLF.

A slower PI can penetrate deeper into the ground but the coil current peak will be higher on a slower PI so is the penetration a function of frequency or coil current or both ? ( not sure myself )

[Gunghouk]

Thanks for that. I thought it was something to do with ground penetration but maybe that was VLF.

[moodz]

Thanks Moodz, can you explain the reason for the higher TX frequency compared to the lower found in other PI designs (typically < 2kHz)?

The MAGPI uses a IRF9640 which has a 200 volt avalanche spec. For a 5 volt supply the TX timing is set so that the flyback sits a little below the avalanche voltage ( approx 220 - 240 volts on my devices ).

The flyback voltage is proportional to how fast the coil is damped. Using a conventional damping resistor is much slower than active damping. So because the damping time is much faster on the MAGPI the TX on time has to be shorter than conventional PIs.

I did notice some time ago someone has patented this ( not me ).

[Gunghouk]

Thanks Moodz, can you explain the reason for the higher TX frequency compared to the lower found in other PI designs (typically < 2kHz)?

[moodz]

Good work ... I better let you know about which track might require cutting in the updated version of V3.

The main TX supply is 5 volts ... this supply also powers the front end opamps and the DG411 switch.

In some conditions ( sample settings etc ) the 5 volts is not high enough ( as the op amps are not rail to rail ).
So the 5 volt analogue supply to the frontend (not the TX ) needs to be re-routed to battery voltage VBATT.
This involves cutting a track and jumpering the opamp supply to battery volts. ( input to LM7805 regulator ).

There is a pink circle over the track to cut.

Note this is a last resort ... dont do it if the detector is working ok.

[Gunghouk]

Here's one I prepared earlier, figured this might be of use to someone

[moodz]

Thanks for the explanation. Unfortunately the model didn't appear in your post.

Its the pesky PNG files again. Here is the PI preamp model ... again.

[Gunghouk]

Thanks for the explanation. Unfortunately the model didn't appear in your post.

I think I'll try and play with a 128x128 oled over I2C just for kicks

[moodz]

Sorry Moodz just saw your comments viz V1.5 software. What op amps are you considering as a replacement for the LM6171 and what are the reasons for the change?
Thanks

Its a long story ... but the short version goes like this .... I came to realise that there is no requirement for a DC to daylight ( well about 1 MHZ ) preamp in the front end of pulse induction metal detectors.
All the target information can be obtained starting from the TX fundamental frequency upto the 11th Harmonic or so. ( modulation theory )
So if your PI is transmitting at a pulse rate of 14 Khz approx ( like the MAGPI ) then a bandpass amplifier of say 10Khz to 150 Khz is perfectly adequate for the preamp.
If the roll off of your bandpass is sufficient then this cuts out all the 1/f amp noise / mains noise / earth field etc etc below 1 Khz.
The MAGPI requires an inverting preamp ( for the damping control loop ) BUT if you have high gain the GBW spec of the opamp will catch you out as the higher the gain the lower the frequency response.
So you need an amplifier that has low noise above say 1 - 10 Khz but a high GBW. This is one of the reasons that I chose the LM6171. ( GBW about 100 Mhz )

Because we are limiting the bandwidth of the amplifier ( particularly at low frequencies ) we can now apply much higher gains than you might use in conventional PI preamps ( typically 100 or so for the first amp in alot of PIs).
The MAGPI started with a gain of 100. Currently it is 1000 and I have even tested at 10,000 ( 80 db ) and it did not overload ... but the GBW suffered even on the LM6171.

So I have ordered some 1 Ghz GBW amps LT1226 2.2 nv root hz noise and this will be the proof of the pudding. ( hopefully no cct change ).

The other really big gotcha which everyone seems to have missed is right there in the spec sheet for opamps. Everyone focuses on the noise figure .... with PIs there the TX cycle causes overloading ( voltage transients at the input to the inputs of the opamp in the preamp ). There is a spec called the maximum differential voltage input. For the NE5534 it is 0.6 volts ... which means that transients will cause currents to start flowing at the opamp inputs as the input protection kicks in. This overloads the amp and then it has to recover .. the waveform is distorted and intermodulation occurs. So much for the noise figure. This is particularly important because we are sampling so close to the flyback now.

The LM6171 and LT1226 have relatively high figures for differential input voltage and so can resist transient distortion currents at thier inputs during overload.

The selection criteria for opamps for PI use should be ...

1. max diff voltage at input spec.
2. GBW ( at least 100 Mhz )
3. Noise above 1 Khz

I am sure everyone will have their own opinion but that is mine ...

Here is my AC model for the PI preamp.
The gain is set by R6 / R7
The low pass cutoff by C3
The high pass cutoff is set by GBW / gain
C4 compensates for opamp input capacitance ( normally only a few puff ) which will form a pole with R6.
It totally blocks DC

This configuration prevents the negative feedback elements of a conventional inverting amp from messing with the input impedance ( esp in overload )

The MAGPI has a demodulator switch to break the feedback R6 from the output during TX cycles this so that TX modulation ( overloads the amp ) will not be "remembered" by the filter cap in the feedback path.

[Gunghouk]

Oh and the encoder is working too

[Gunghouk]

Sorry Moodz just saw your comments viz V1.5 software. What op amps are you considering as a replacement for the LM6171 and what are the reasons for the change?
Thanks

[moodz]

Thank-you Altra for the great I2C info and Moodz for the updated design

Turns out my confusion was caused by the labeling of the ICSP header to say it was for the LCD . It is the J1 header, labelled SERIAL/ROT ENC, that supplies SDA and SCL to the LCD. I looked through the code to determine the reality.

ISCP header J2 should be PROG/ENCODER
1 - MCLR
2 - VL
3 - GND
4 - PGC (PIC pin 5)
5 - PGD/ENCODER PUSHBUTTON (PIC RB1, pin 5)
6 - ENCODER A (PIC RB2, pin 6)
7 - ENCODER B (PIC RB3, pin 7)

LCD header J1 :
1 - GND
2 - SDA (LCD data/TXD, PIC RB9, pin 18 )
3 - SCL (LCD clock/RXD, PIC RB8, pin 17)
4 - KEY (PIC RB7 pin 16) ?????????
5 - VL

The LCD is working now thanks for your all your help.

Now planning the build and acquiring parts for the updated design.

Moodz is there a new software version to supercede v1.4 ?

Thanks guys

I will be updating the 1.5 software but my parts order wont be here till April 19 apparently. I will post an intermediate update in the next few days if I get onto it.
I need to find some time for field testing also .... theres always a chance it only works well in the lab.
It was the original intention of the code to check the LCD port to see if there was an LCD there then revert to serial if nothing found in the initialisation.

[Gunghouk]

Thank-you Altra for the great I2C info and Moodz for the updated design

Turns out my confusion was caused by the labeling of the ICSP header to say it was for the LCD . It is the J1 header, labelled SERIAL/ROT ENC, that supplies SDA and SCL to the LCD. I looked through the code to determine the reality.

ISCP header J2 should be PROG/ENCODER
1 - MCLR
2 - VL
3 - GND
4 - PGC (PIC pin 5)
5 - PGD/ENCODER PUSHBUTTON (PIC RB1, pin 5)
6 - ENCODER A (PIC RB2, pin 6)
7 - ENCODER B (PIC RB3, pin 7)

LCD header J1 :
1 - GND
2 - SDA (LCD data/TXD, PIC RB9, pin 18 )
3 - SCL (LCD clock/RXD, PIC RB8, pin 17)
4 - KEY (PIC RB7 pin 16) ?????????
5 - VL

The LCD is working now thanks for your all your help.

Now planning the build and acquiring parts for the updated design.

Moodz is there a new software version to supercede v1.4 ?

Thanks guys