PI discrimination???

[Anonymous]

eric, dave, reg, other tech-heads. tell me if this might work or not. it's quiet here so something to think about. we replace the 555 clock pulse feed to the tx fet with a pulse rate determined by the instantaneous inductance of the tx coil using a dc/dc converter chip like tl496, mc34063. as the tx coil moves over non-ferrous target the pulse width will increase and the frequency will go up due to eddy current losses. rusty steel should reduce pulse width and lower frequency. the effect should be audible, and the frequency/pulsewidth could be indicated. a small adjustable load on the converter chip would preset the freq. does this make sense? allan

 

[Anonymous]

Hi Allan,
Have to think about that one <img src="/metal/html/confused.gif" border=0 width=15 height=22 alt=":?"> The inductance changes caused by rings, coins, etc are very small, even when close to the coil.
Another way is to remove the damping resistor and let the coil ring. Then measure the change in ringing frequency by sampling at the crossover point of a late ringing cycle.
Eric.

 

[Anonymous]

Eric, et al:
For pulse/target/inductance calculations why not take on board a computer whizkid and install chips to figure all that out instead of hard circuitry? Early generation pentium and motorola chips are extremely cheap right now and you may be able to jump many evolutionary steps ahead by farming some of these problems out to the chip folks.
Not challenging anyone's abilities -- simply suggesting using the earlier generation chips which are blazing fast to do some of the calculations.

 

[Anonymous]

I agree with Matt's thinking. I have always thought that using a small Palm-like computer device to render a display; do the processing; do visual and audio output, while interfaced to a custom built PI front end circuit is a very efficient way to easily accelerate PI technology into a new era.
Eric, Reg, Matt or some other forum readers (please jump in) who have analog to digital conversion and interfacing experience could help define the functional requirements to drive the interface connection to the Palm-like computer. Then, the software requirements can be defined based on the signals expected to pass through the interface.
Based on the available signals, the speed of processing and the ability to extract and process the appropriate intelligence through the PI interface, the graphic or audio interface to the human operator can be defined.
Somewhere along the way the theory needs to be coupled to the reality in the lab, with folks like Eric, to breadboard and try out some of the theory and do some reality checks. Once the required processing is known to be practical, commercial chips can be surveyed to look for existing chips that will do the "heavy lifting" easily. All of this is an iterative process but starts with conversations like this.
This is a good start!
bbsailor

 

[Anonymous]

Sadly, it won't be me to design and reap the rewards . . . I come from another era. It'll be up to the electronics wizzards to marry their ideas with the chip gurus. A single pentium or motorola or amd chip of today contains more processing power than was in the entire United States 30 years ago . . . I don't know a lot, but I do know that's a heck of a lot of power to figure out if a metal is gold/silver/aluminum/coin/can/or canonball and can do it as fast as any signal that comes into it.

 

[Anonymous]

digital signal processing.

 

[Anonymous]

Eric,
Your following statement got me thinking.
"Another way is to remove the damping resistor and let the coil ring. Then measure the change in ringing frequency by sampling at the crossover point of a late ringing cycle".
One could hunt using the PI machine in the traditional PI mode with the damping resistor to obtain the greatest sensitivity. Then, when a target is located, a button could be pushed or switched to remove the damping resistor, via a small reed relay to allow the analysis of the undamped ringing cycle and possibly switch in other analysis circuitry such as A to D converters, etc.
The key question that only someone like you would know the answer to is: Do the undamped ringing cycles contain any potential indicators that could reveal the nature of the metal? If this type of new circuitry could clearly eliminate ferrous metal, that would be a major leap forward.
Just a thought to ponder.
bbsailor

 

[Anonymous]

I would think that a target that is ferrous would change the inductance of the coil and the effect would depend on how large the target is and how close the coil is to it. This would then change the ringing freq and possibly the amplitude of the signal. I plan on trying it this weekend to see if that is the case. I do not know it gold or silver would have as much an effect but I am going to investigate it my self. Just my 2 cents. I' no expert on it.
Ray

 

[Anonymous]

Just like the BFO detector,ferrous targets would raise the inductance, and non-ferrous lower it. It all depends on whether it is possible to sense small enough shifts in frequency to give a sensible detection range.
As the ringing starts each time the TX pulse switches off, sampling well along the ringing waveform should give an enhanced effect. Like holding one end of a spring fixed; as you move further down the spring the movement gets greater, when it is oscillating.
Eric.

 

[Anonymous]

Allan, Here is my input. Yes your idea will work. The downside is that it will not work to anywhere near a PI's maximum depth capability. Add to this that the ground mineralization will upset the results unless some extra methods are employed. With this said, ANY iron discrimination ability is better than none. It is due to the imagination of people like yourself that detector technology will go forward. Go for it, Stick an axe through the TV and get to your work bench. I wish you the very best of luck, Dave. * * *

 

[Anonymous]

Allan,
I find this is a very interesting topic.
If you were to attempt a ground balancing pi using a high current tx circuit such as ML uses then the effect from magnetic minerals such as magnetite becomes a problem if sampling early because of the change in coil inductance, which is sometimes called the magnetic effect. If this change enters the early sample then it upsets a later earth field subtraction or sample taken later for ground balance and can result in a signal from magnetized rocks that have the same make up as non-magnetized rocks. There are quite a few ways that this effect can be obtained and then subtracted but as Dave has suggested, the effect falls off with depth and target size.
If you were to search with the subtraction on then switch the subtraction off over a target then some sort of enhanced linear measurement would be needed for anything other than strong signals but as Dave has said, it would then be also over sensitive to mineralization.
With the tx p/s regulated and the spike clamped with a zener back to the supply then the effect is noticeable as a voltage change at the regulator inputs and this can be then used to form an off-set. Sampling the tx pulse across a low ohm resistor in series with the coil and supply gives a sample for subtraction but the effect then falls off with any added resistance in the coil circuit. Clamping the spike to say 150-210 volts by dumping it thru a diode into a capacitor with a fixed bleed resistor across the cap will show the clamped spike voltage rising or falling with the change in inductance and you can expect around a 4 volt drop for a medium sized hammer or a similar rise for a large aluminium heat sink resting on the coil depending on pulse lengths etc. This effect could also be regulated as in ML's case but instead switched in or out. Unfortunately, a moderately largish bulk sample of magnetite gives a similar response to the hammer. <img src="/metal/html/frown.gif" border=0 width=15 height=15 alt="">
There are other ways of sensing this effect including your suggestion or using two different pulse lengths but at the end of the day you are still stuck with a surface discriminator unless you come up with a way of picking the needed info off the waveforms.
Like I said, interesting topic that I wish had more input from those playing with this.
Rob.

 

[Anonymous]

Ray
Extracting additional intelligence from a PI signal is being examined by the demining folks. They are looking to use both positive as well as negative pulses to help neutralize noise and possibly find a way to extract additional intelligence about the nature of the target composition.
See subject link.
I hope your test went well. Any results?
bbsailor

 

[Anonymous]

Hi everyone,
I am a little late on the scene but a thread started by Allan about PI discrimination led to many interesting replys and comments. One was from Matt on using Pentium proccessors for number crunching along with ADCs etc.
I am currently playing with PI digital front ends, multiple coils and using a 3.5" motherboard from Advantech called a PCM5820 low power version.
While I can get a very quiet ADC front end working, the biggest problem is the inherent noise from the motherboard itself. Even with good shielding methods the microvolt signals from the PI front end are swamped by EMC from the motherboard.
It is possible to design a super fast (as far as PIs are concerned) metal detector using chips other than Atmels and Pics but radiated noise from most of the higher speed proccessors can be a problem if layouts and shielding methods are not spot on. I am currently working on digital filtering software along with complete shielding of the front end ADC and Amp section. The motherboard is around $300 Australian and is based on the NatSemi Geode 300MHZ chipset. It also has a PC104 interface connector which is basically a rugged PCI bus. It allows many types of interface boards from different manufactures (including ADCs) Prices and power consumption are limiting but there are some very low power boards out there. I am using this method to collect data in the field hence I'm not too worried about power consumption just yet. Beats carrying around a laptop.
I believe this sort combination will be an option for PIs albiet with a little more effort.
Will post my progress over the next few months if there is interest shown.

 

[Anonymous]

Brian,
Sounds like on your early designs there needs to be a separation of the module circuits (1)transmit, (2)receive, (3) digital processing and (4)power supply/mother board on individually shielded circuit boards. This would help isolate the radiated noise and allow higher gain circuits. This modular design technique could help speed the building of your prototypes into four distinct parts that can be individually modified, optimized or tweeked.
It might be good to look at some techniques used on Radar circuits to minimize radiated circuit noise. Radar technology uses the same antenna for both transmit and receive, sort of like a PI metal detector. Granted, the frequency ranges are greatly different, but there may be some ideas that can apply.
Another technique that you may want to look at is to dynamically switch in the damping resistor just as the back EMP ringing starts and out when receive sampling is to start. This technique may lead to extracting a little more intelligence from the receive waveform and help track target timing synchronization for better potential ientification.
These are just a few thoughts to stimulate some creative thinking on this forum.
bbsailor

 

[Anonymous]

Yes that is the tack I am taking at present.
95% of the digital noise has been eliminated by careful layout and shielding. I use a CompactFlash drive to hold the operating system and application and this helps keep the overall power consumption and heat down which also helps decrease the noise levels. The ADC module is plugged into the PC104 socket on the motherboard and is only a prototype. Good design of this board will eliminate more noise too. My main aim is to collect field data on multiple coil array designs I have been playing with. Bench testing is good but I find it a bit limiting and so I decided to go down this path. The beauty of this design is that I can change my algorithms in the field simply by plugging in a fold away keyboard. Check out single board computers on the net for more info on what is available.
Cheers
Brian King