PI detection - an alternative approach?

[pebe]

All the detectors I have seen appear to measure the analogue voltage across the coil after specific periods of time following the end of the transmitted pulse.

Do you know if anyone has tried the other way round, ie. measuring how long it takes for the response to fall to, say. 0.3V?

[Qiaozhi]

All the detectors I have seen appear to measure the analogue voltage across the coil after specific periods of time following the end of the transmitted pulse.

Do you know if anyone has tried the other way round, ie. measuring how long it takes for the response to fall to, say. 0.3V?

Yes, that's been done before. The processor turns off the TX pulse, and then starts a timer. When the voltage drops below a certain level (usually measured with a comparator) the timer is stopped.

[pebe]

Thanks for the info. Do you know the relative advantages/disadvantages of the two methods?

[Qiaozhi]

Thanks for the info. Do you know the relative advantages/disadvantages of the two methods?

The normal sampling should be more sensitive, but a lot depends on the resolution you can achieve from either method.
Also (using a timer) would make it difficult to add ground balance and EFE.

[pebe]

Thanks again for the info. I have never built a PI unit before but I understood they could not discriminate between ferrous and non- ferrous metals, so I was surprised when I read this article and saw a graph which implied that discrimination was possible by measuring the time taken for the flyback pulse to reach about 0.35V.
http://www.lammertbies.nl/electronic..._detector.html
I tried to contact the writer via his website, but it appears to be dormant at the moment.

I thought about using a PIC with a 32M oscillator so that I could get a timer with a resolution of 125ns to measure the flyback time, and use 0.3V as the reference voltage. I assumed that ground balance would affect the timing, so I intended to have a pushbutton that would initially copy the measured time into the ‘reference’ register as the norm, when the coil was on the ground with no target. Any measured time longer than that would indicate ferrous; any less time would show non-ferrous metals. The reference could be updated at any time. To reduce noise to a minimum I intended to input the measured time that resulted from of a rolling average of 32 samples.

I have no idea how EFE would affect the results, but I would be very grateful for your comments.

[Qiaozhi]

Thanks again for the info. I have never built a PI unit before but I understood they could not discriminate between ferrous and non- ferrous metals, so I was surprised when I read this article and saw a graph which implied that discrimination was possible by measuring the time taken for the flyback pulse to reach about 0.35V.
http://www.lammertbies.nl/electronic..._detector.html
I tried to contact the writer via his website, but it appears to be dormant at the moment.

I believe that project is long dead.
He is attempting to identify ferrous and non-ferrous targets by examining the way the eddy currents decay. But, the decay curve depends on target conductivity. The GB control of White's TDI can be adjusted so that low-conductive targets give a high tone, and high-conductive targets give a low tone. In this way, small ferrous targets (such as nails) can be rejected. However, this rejection process is affected by the shape, size, and orientation of the target. So the rejection is far from perfect, and using the TDI in this manner means that ground balance is no longer possible. The PI detector in your link will have a similar problem.

Now - having said that - don't let me stop you experimenting with your idea. I'm sure others here would be very interested in the results.

I have no idea how EFE would affect the results, but I would be very grateful for your comments.

Without EFE the detector will give a beep at the end of each sweep (non-vco audio) or a sort of "breathing" effect as you move the coil (vco audio).

[pebe]

So it looks like there is a possibility it all might work. I will get the TX part working first so I can look at the discharge curves with different targets to see the differences. If all goes well I’ll proceed with the detector.

Is there a way I can simulate ground effect when testing at the workbench?

[Qiaozhi]

Is there a way I can simulate ground effect when testing at the workbench?

I depends on how mineralized the ground is where you intend to detect. For normal soil you can use a ferrite core, but this may only be detected if the sample delay is sub-10us. In Scotland you probably won't need to ground balance at all, as a PI detector will inherently ignore light to medium mineralization. This is due to the fact that any eddy currents in the soil will have dissipated before sampling can take place. Or (in your case) before the decay curve has dropped to the specified trigger level.

[pebe]

Many thanks.

[bernte_one]

i think you can´t arrive lammert bies anymore, he moved with his wife to another end of our world you will see if you read his Website

[pebe]

I've got a circuit ready and sent for some wire. One more question; Is the size of the coil important when I am looking for changes in the discharge curve using different metals as targets?

@Bernte. Thanks for your input.

[Qiaozhi]

I've got a circuit ready and sent for some wire. One more question; Is the size of the coil important when I am looking for changes in the discharge curve using different metals as targets?

Just about everything (from the TX pulse width, TX pulse rate, wire size, coil size, coil capacitance, series resistance, etc.) affects the discharge curve in some way. You cannot rely on sampling the discharge curve to identify different metals. The best you will be able to do is separate targets based on conductivity.

[Teleno]

My approach is to short the coil and measure the current. You need an active load connected to a virtual ground. The current is injected at the emitter of a bipolar transistor (the virtual ground) in a common-base configuration, then becomes available at the collector as high impedance for integrating or converting into a voltage via a resistor. Current mirrors all over the place. It's fast, simple and less noisy.

[pebe]

OK, I see the impedance conversion bit, but what do you do with the measured voltage? Do you measure it after a specific time or integrate successive measurements through the entire flyback period and get an average?

[Teleno]

Both measuring an integrating are known techniques. You're looking for alternative approaches. I think the and low impedance, current-to-voltage (trasconductance) front-end amplifier is the most straight forward improvement relative to the classic PI design which is voltage-based and very noisy. Every time I see those anti-parallel clamping diodes at the input in series with a resistor I want to throw up, really.