field test unit no 001 "model T"

[moodz]

... I re-tested at 50 us and 100 us Tx pulse with excellent and consistant descrim results ....

moodz.

[mickstv]

Sounds good. I'll put that resistor in as well and see what happens. Is the frontend still immune to ferrite ?



Mick

[moodz]

Sounds good. I'll put that resistor in as well and see what happens. Is the frontend still immune to ferrite ?



Mick

Unfortunately no .. ferrite responds as iron ... but weaker ...... you can still balance it out though. Playing with the timings and balance may do something ... dont have time at the moment ....

moodz.

[mickstv]

Unfortunately no .. ferrite responds as iron ... but weaker ...... you can still balance it out though. Playing with the timings and balance may do something ... dont have time at the moment ....

moodz.

If ferrite responds like iron I can't see that as being a problem. So we'll call it cured, but we wont really know what is going to happen until we have a few units out testing.



Mick

[ODM]

I believe no one has commented yet, but the copper fill on the PCB should be controlled. It should be connected to some potential, perhaps using separate guards for the different areas if necessary. Floating copper fills only act to conduct disturbances between traces instead of shorting them to the connected potential - usually ground.

If a minimal capacitance is wanted, the fills should be left out entirely.

Just a couple cents, this is more in the general electronics area but it might result in noise control improvement for this interesting project.

[mickstv]

Mick .. before I get onto the amp .. I just wanted to make sure that NP FET pack you are using was OK ... I had not actually tried it when I recommended it to you .. only went off specs .....

The result is PERFECT ... works better than the discretes.

Tick that one off ...

I mounted the chip on a little SOIC board with some "pins" that go into the sockets where the blocking fets plug into.

The demod outputs in the pic below are close to, but not balanced ... the descrim balance control setting is 2619.

moodz.

[ATTACH]17290[/ATTACH]
[ATTACH]17291[/ATTACH]

Moodz, the picture of the NP fet you put together it looks like you have pin8 tied to pin3. That can't be right because Pin8 is D1 and pin3 is S2. Or am I missing something.



Mick

[moodz]

I believe no one has commented yet, but the copper fill on the PCB should be controlled. It should be connected to some potential, perhaps using separate guards for the different areas if necessary. Floating copper fills only act to conduct disturbances between traces instead of shorting them to the connected potential - usually ground.

If a minimal capacitance is wanted, the fills should be left out entirely.

Just a couple cents, this is more in the general electronics area but it might result in noise control improvement for this interesting project.

Good point ODM ... I have not actually got to the point of laying out a board yet because the circuit keeps changing ... but you are right this will have to be a consideration with the layout particularly if surface mount is used and closer spacing of pins and traces etc.

moodz.

[moodz]

Moodz, the picture of the NP fet you put together it looks like you have pin8 tied to pin3. That can't be right because Pin8 is D1 and pin3 is S2. Or am I missing something.



Mick

Mick ... good pickup the circuit .... I use now is different but for reasons other than blocking .... all will be revealed in the next schematic.

PS whilst I have been waiting for the post demod ideas about the DC amp to crystallise ... I have been retesting the timing and descrimination response for different targets and made a major discovery ..... the descrimination is dependant on the timing and balance point .....in fact there are optimum timings and balance points where this occurs ... this has other implications which are a bit hard to describe here ( and probably should be described in a patent )

moodz.

[Midas]

A few pretty pictures..

Hi Guys,

I knocked up my own CAT5 based coil on the weekend on the weekend and have assembled the very bare essentials of the coil drive with just a pot for the discrimination balance. Although its all pretty rough as guts, lots of noise, and the sensitivity didn't seem very good. I did at least get to see the discrimination Moodz has been demonstrating for myself. So yeah just in case you were starting worry Moodz, its not an effect entirely confined to your lab The 3 pictures in order are, side of my computer, large chunk of aluminum, no target. Both channels effectively 2v/div (10x\100x probes)

[mickstv]

Hi Guys,

I knocked up my own CAT5 based coil on the weekend on the weekend and have assembled the very bare essentials of the coil drive with just a pot for the discrimination balance. Although its all pretty rough as guts, lots of noise, and the sensitivity didn't seem very good. I did at least get to see the discrimination Moodz has been demonstrating for myself. So yeah just in case you were starting worry Moodz, its not an effect entirely confined to your lab The 3 pictures in order are, side of my computer, large chunk of aluminum, no target. Both channels effectively 2v/div (10x\100x probes)

Hi Midas, good to see the positive results. I've decided my problem it's some sort of strange issue with the analog CRO setup, so I might go and have a look on EBay for a digital replacement.

Have you tested the frontend with any other targets like say some PCB board/copper, lead and steel pipe. Just wondering what sort of results you get.


Mick

[mickstv]

Mick ... good pickup the circuit .... I use now is different but for reasons other than blocking .... all will be revealed in the next schematic.

PS whilst I have been waiting for the post demod ideas about the DC amp to crystallise ... I have been retesting the timing and descrimination response for different targets and made a major discovery ..... the descrimination is dependant on the timing and balance point .....in fact there are optimum timings and balance points where this occurs ... this has other implications which are a bit hard to describe here ( and probably should be described in a patent )

moodz.

That is starting to sound like you might need the frontend to be auto adjusting and a set of timing options for different target types maybe ?



Mick

[Midas]

Have you tested the frontend with any other targets like say some PCB board/copper, lead and steel pipe. Just wondering what sort of results you get.

I just tested it with a big piece of roofing lead I think about 1.5mm thick, it responds in the same way as the aluminium chunk. Then I tested a piece pcb 150 x 150, and an aluminium can, both respond in same direction as iron. It all seems pretty consistent with the discriminating capabilities of multi-coil metal detectors.

[moodz]

updated front end.

FETs now balanced NP on each input.
NP Mosfet pack used ( 300 volts .... but good for 350 )
4066 works better than other more expensive solutions but use mosfet switch to drive ( like a tc4421 ) Make sure you use the high voltage 4066 ( eg 15 volt rating ).
The THAT amplifier is not the best here ... use opamp DC diff amp.

moodz.

[Midas]

Thanks for your continuing contributions Moodz! I see you have the gain cranked to 2000 now, brilliant! Hows the stability ?

I think the diode in series with the descrim balance mosfet is a positive step as my own tests seemed to indicate the over-damping in the one direction through the body diode of the mosfet in the previous circuit was reducing the sensitivity compared to just a pot (equal resistance in both directions). This new version is completely un-damped in that direction, which I guess is better from a sensitivity standpoint, however I would assume (though happy to be proved wrong) is also going to extend settling time to some extent (since its only damping the positive oscillations). I'm wondering if it might be worth implementing a full AC switch, ie. two mosfets source to source with a floating drive. That way you get proper bi-directional control of the resistance. Not sure if its worth it or not, what do you think?

[moodz]

Thanks for your continuing contributions Moodz! I see you have the gain cranked to 2000 now, brilliant! Hows the stability ?

I think the diode in series with the descrim balance mosfet is a positive step as my own tests seemed to indicate the over-damping in the one direction through the body diode of the mosfet in the previous circuit was reducing the sensitivity compared to just a pot (equal resistance in both directions). This new version is completely un-damped in that direction, which I guess is better from a sensitivity standpoint, however I would assume (though happy to be proved wrong) is also going to extend settling time to some extent (since its only damping the positive oscillations). I'm wondering if it might be worth implementing a full AC switch, ie. two mosfets source to source with a floating drive. That way you get proper bi-directional control of the resistance. Not sure if its worth it or not, what do you think?

Thanks Midas ... the diode in the descrim balance control disconnects the damping resistance during receive .. otherwise the coils would be unbalanced during receive and you are correct about the mosfet body diode and output capacitance effects. I tried the floating drive technique but could not stabilise the balance point ... it kept drifting off.

There is another problem ... the method of producing the +VE supply rail via D5 and C4 is too clever by half.... the problem is that a large target or ground effect will substantially effect the magnetic coupling between L1 and L2 this causes a relatively large change in the voltage stored on C4. Because C4 stores a charge that is somewhat proportional to the coupling between L1 and L2 there is a "memory" effect a large target induces a voltage which remains for some time after the target has been removed from the coil....
The effect of this is to unbalance the balance point ... especially at high gain.
.. I will have to use a more conventional synchronised bias generator to generate the +VE.


However all is not lost ... the charge on C4 via D5 is actually a representation of the ground effect ... and could be used for ground balance .... C4 will have to be alot smaller .... so stay tuned .... I have a plan for this piece of information.

The 4066 sampler starts sampling at about 6 - 7 microseconds after TX switch off ... the sampler turns off at about 22 microseconds ... this gives a good descrim point.

I am commencing another rebuild of the board with the alternative +VE power supply.

moodz.